diff --git a/LICENSE b/LICENSE
new file mode 100644
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,339 @@
+             GNU GENERAL PUBLIC LICENSE
+                Version 2, June 1991
+
+ Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
+ 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                     Preamble
+
+  The licenses for most software are designed to take away your
+freedom to share and change it.  By contrast, the GNU General Public
+License is intended to guarantee your freedom to share and change free
+software--to make sure the software is free for all its users.  This
+General Public License applies to most of the Free Software
+Foundation's software and to any other program whose authors commit to
+using it.  (Some other Free Software Foundation software is covered by
+the GNU Lesser General Public License instead.)  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+this service if you wish), that you receive source code or can get it
+if you want it, that you can change the software or use pieces of it
+in new free programs; and that you know you can do these things.
+
+  To protect your rights, we need to make restrictions that forbid
+anyone to deny you these rights or to ask you to surrender the rights.
+These restrictions translate to certain responsibilities for you if you
+distribute copies of the software, or if you modify it.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must show them these terms so they know their
+rights.
+
+  We protect your rights with two steps: (1) copyright the software, and
+(2) offer you this license which gives you legal permission to copy,
+distribute and/or modify the software.
+
+  Also, for each author's protection and ours, we want to make certain
+that everyone understands that there is no warranty for this free
+software.  If the software is modified by someone else and passed on, we
+want its recipients to know that what they have is not the original, so
+that any problems introduced by others will not reflect on the original
+authors' reputations.
+
+  Finally, any free program is threatened constantly by software
+patents.  We wish to avoid the danger that redistributors of a free
+program will individually obtain patent licenses, in effect making the
+program proprietary.  To prevent this, we have made it clear that any
+patent must be licensed for everyone's free use or not licensed at all.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+             GNU GENERAL PUBLIC LICENSE
+   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+  0. This License applies to any program or other work which contains
+a notice placed by the copyright holder saying it may be distributed
+under the terms of this General Public License.  The "Program", below,
+refers to any such program or work, and a "work based on the Program"
+means either the Program or any derivative work under copyright law:
+that is to say, a work containing the Program or a portion of it,
+either verbatim or with modifications and/or translated into another
+language.  (Hereinafter, translation is included without limitation in
+the term "modification".)  Each licensee is addressed as "you".
+
+Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope.  The act of
+running the Program is not restricted, and the output from the Program
+is covered only if its contents constitute a work based on the
+Program (independent of having been made by running the Program).
+Whether that is true depends on what the Program does.
+
+  1. You may copy and distribute verbatim copies of the Program's
+source code as you receive it, in any medium, provided that you
+conspicuously and appropriately publish on each copy an appropriate
+copyright notice and disclaimer of warranty; keep intact all the
+notices that refer to this License and to the absence of any warranty;
+and give any other recipients of the Program a copy of this License
+along with the Program.
+
+You may charge a fee for the physical act of transferring a copy, and
+you may at your option offer warranty protection in exchange for a fee.
+
+  2. You may modify your copy or copies of the Program or any portion
+of it, thus forming a work based on the Program, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+    a) You must cause the modified files to carry prominent notices
+    stating that you changed the files and the date of any change.
+
+    b) You must cause any work that you distribute or publish, that in
+    whole or in part contains or is derived from the Program or any
+    part thereof, to be licensed as a whole at no charge to all third
+    parties under the terms of this License.
+
+    c) If the modified program normally reads commands interactively
+    when run, you must cause it, when started running for such
+    interactive use in the most ordinary way, to print or display an
+    announcement including an appropriate copyright notice and a
+    notice that there is no warranty (or else, saying that you provide
+    a warranty) and that users may redistribute the program under
+    these conditions, and telling the user how to view a copy of this
+    License.  (Exception: if the Program itself is interactive but
+    does not normally print such an announcement, your work based on
+    the Program is not required to print an announcement.)
+
+These requirements apply to the modified work as a whole.  If
+identifiable sections of that work are not derived from the Program,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works.  But when you
+distribute the same sections as part of a whole which is a work based
+on the Program, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Program.
+
+In addition, mere aggregation of another work not based on the Program
+with the Program (or with a work based on the Program) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+  3. You may copy and distribute the Program (or a work based on it,
+under Section 2) in object code or executable form under the terms of
+Sections 1 and 2 above provided that you also do one of the following:
+
+    a) Accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of Sections
+    1 and 2 above on a medium customarily used for software interchange; or,
+
+    b) Accompany it with a written offer, valid for at least three
+    years, to give any third party, for a charge no more than your
+    cost of physically performing source distribution, a complete
+    machine-readable copy of the corresponding source code, to be
+    distributed under the terms of Sections 1 and 2 above on a medium
+    customarily used for software interchange; or,
+
+    c) Accompany it with the information you received as to the offer
+    to distribute corresponding source code.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form with such
+    an offer, in accord with Subsection b above.)
+
+The source code for a work means the preferred form of the work for
+making modifications to it.  For an executable work, complete source
+code means all the source code for all modules it contains, plus any
+associated interface definition files, plus the scripts used to
+control compilation and installation of the executable.  However, as a
+special exception, the source code distributed need not include
+anything that is normally distributed (in either source or binary
+form) with the major components (compiler, kernel, and so on) of the
+operating system on which the executable runs, unless that component
+itself accompanies the executable.
+
+If distribution of executable or object code is made by offering
+access to copy from a designated place, then offering equivalent
+access to copy the source code from the same place counts as
+distribution of the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+  4. You may not copy, modify, sublicense, or distribute the Program
+except as expressly provided under this License.  Any attempt
+otherwise to copy, modify, sublicense or distribute the Program is
+void, and will automatically terminate your rights under this License.
+However, parties who have received copies, or rights, from you under
+this License will not have their licenses terminated so long as such
+parties remain in full compliance.
+
+  5. You are not required to accept this License, since you have not
+signed it.  However, nothing else grants you permission to modify or
+distribute the Program or its derivative works.  These actions are
+prohibited by law if you do not accept this License.  Therefore, by
+modifying or distributing the Program (or any work based on the
+Program), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Program or works based on it.
+
+  6. Each time you redistribute the Program (or any work based on the
+Program), the recipient automatically receives a license from the
+original licensor to copy, distribute or modify the Program subject to
+these terms and conditions.  You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties to
+this License.
+
+  7. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Program at all.  For example, if a patent
+license would not permit royalty-free redistribution of the Program by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Program.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system, which is
+implemented by public license practices.  Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+  8. If the distribution and/or use of the Program is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Program under this License
+may add an explicit geographical distribution limitation excluding
+those countries, so that distribution is permitted only in or among
+countries not thus excluded.  In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+  9. The Free Software Foundation may publish revised and/or new versions
+of the General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+Each version is given a distinguishing version number.  If the Program
+specifies a version number of this License which applies to it and "any
+later version", you have the option of following the terms and conditions
+either of that version or of any later version published by the Free
+Software Foundation.  If the Program does not specify a version number of
+this License, you may choose any version ever published by the Free Software
+Foundation.
+
+  10. If you wish to incorporate parts of the Program into other free
+programs whose distribution conditions are different, write to the author
+to ask for permission.  For software which is copyrighted by the Free
+Software Foundation, write to the Free Software Foundation; we sometimes
+make exceptions for this.  Our decision will be guided by the two goals
+of preserving the free status of all derivatives of our free software and
+of promoting the sharing and reuse of software generally.
+
+                     NO WARRANTY
+
+  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
+FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
+OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
+PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
+TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
+PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
+REPAIR OR CORRECTION.
+
+  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
+REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
+INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
+OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
+TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
+YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
+PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGES.
+
+              END OF TERMS AND CONDITIONS
+
+     How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+convey the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+Also add information on how to contact you by electronic and paper mail.
+
+If the program is interactive, make it output a short notice like this
+when it starts in an interactive mode:
+
+    Gnomovision version 69, Copyright (C) year name of author
+    Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, the commands you use may
+be called something other than `show w' and `show c'; they could even be
+mouse-clicks or menu items--whatever suits your program.
+
+You should also get your employer (if you work as a programmer) or your
+school, if any, to sign a "copyright disclaimer" for the program, if
+necessary.  Here is a sample; alter the names:
+
+  Yoyodyne, Inc., hereby disclaims all copyright interest in the program
+  `Gnomovision' (which makes passes at compilers) written by James Hacker.
+
+  <signature of Ty Coon>, 1 April 1989
+  Ty Coon, President of Vice
+
+This General Public License does not permit incorporating your program into
+proprietary programs.  If your program is a subroutine library, you may
+consider it more useful to permit linking proprietary applications with the
+library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.
diff --git a/README b/README
new file mode 100644
--- /dev/null
+++ b/README
@@ -0,0 +1,38 @@
+Overview of key modules, roughly from highest- to lowest-level:
+
+* SymmSHE.hs, an implementation of a symmetric-key,
+  somewhat-homomorphic encryption scheme that is essentially
+  equivalent to the one from the toolkit paper [LPR'13].
+
+* Cyc.hs, which defines an interface for using cyclotomic fields, 
+  rings R, and quotient rings Rq=R/qR; as well as many other
+  commonly used operations, e.g., converting
+  between rings, decoding and decomposing elements, modulus
+  reduction/rounding, etc. etc. Cyc is a safe wrapper around the
+  UCyc type, which exposes some representation-dependent operations.
+  UCyc (and hence Cyc) is implemented using a generic Tensor (described below).
+
+* Tensor.hs, which defines a class that encapsulates all the necessary
+  linear transformations for operating on representations of R- and
+  Rq-elements, e.g., the CRT transform, converting between the
+  powerful and decoding bases, generating error terms, etc.
+
+* RepaTensor.hs, which gives an
+  implementation of the Tensor class based on the "repa"
+  package, a highly optimized and parallelizable array library.
+
+* CTensor.hs, which gives an
+  implementation of the Tensor class using a C backend via Haskell's FFI.
+
+* FiniteField.hs, which gives an unoptimized implementation of finite field
+  arithmetic. To use this module, you will need an instance of IrreduciblePoly.
+  These instances provide irreducible polynomials for various degrees and base fields.
+  One instance is provided for characteristic 2 fields of size up to 2^32 in 
+  IrreducibleChar2.hs.
+
+* ZqBasic.hs, which is a basic implementation of Zq=Z/qZ arithmetic.
+
+* Factored.hs, which contains support code for "reifying"
+  runtime-chosen integers as static types (mainly, the types q and m
+  that are floating around as parameters of many other data types),
+  and "reflecting" those types as integers back to the code.
diff --git a/Setup.hs b/Setup.hs
new file mode 100644
--- /dev/null
+++ b/Setup.hs
@@ -0,0 +1,2 @@
+import Distribution.Simple
+main = defaultMain
diff --git a/lol.cabal b/lol.cabal
new file mode 100644
--- /dev/null
+++ b/lol.cabal
@@ -0,0 +1,160 @@
+name:                lol
+-- The package version.  See the Haskell package versioning policy (PVP) 
+-- for standards guiding when and how versions should be incremented.
+-- http://www.haskell.org/haskellwiki/Package_versioning_policy
+-- PVP summary:      +-+------- breaking API changes
+--                   | | +----- non-breaking API additions
+--                   | | | +--- code changes with no API change
+version:             0.0.1.0
+synopsis:            A general-purpose library for lattice cryptography.
+homepage:            https://github.com/cpeikert/Lol
+Bug-Reports:         https://github.com/cpeikert/Lol/issues
+license:             GPL-2
+license-file:        LICENSE
+author:              Eric Crockett, Chris Peikert
+maintainer:          Eric Crockett <ecrockett0@gmail.com>
+copyright:           Eric Crockett, Chris Peikert
+category:            Crypto
+stability:           experimental
+build-type:          Simple
+extra-source-files:  README, 
+                     src/Crypto/Lol/Cyclotomic/Tensor/CTensor/tensorTypes.h,
+                     test-suite/CycTests.hs,
+                     test-suite/SHETests.hs,
+                     test-suite/TensorTests.hs,
+                     test-suite/TestTypes.hs,
+                     test-suite/ZqTests.hs
+cabal-version:       >=1.10
+description:         \\Lambda \\ocirc \\lambda is a general-purpose library for ring-based lattice cryptography.
+source-repository head
+  type: git
+  location: https://github.com/cpeikert/Lol
+
+-- For information on compiling C with cabal: http://blog.ezyang.com/2010/06/setting-up-cabal-the-ffi-and-c2hs/
+
+Flag useICC
+  Description: Use ICC instead of GCC to compile C backend.
+  Default:     False
+
+Flag llvm
+  Description:  Compile via LLVM. This produces much better object code,
+                but you need to have the LLVM compiler installed.
+
+  Default:      False
+
+library
+  hs-source-dirs:     src
+  Include-dirs: src/Crypto/Lol/Cyclotomic/Tensor/CTensor
+  C-sources: src/Crypto/Lol/Cyclotomic/Tensor/CTensor/basic.c, 
+             src/Crypto/Lol/Cyclotomic/Tensor/CTensor/crt.c, 
+             src/Crypto/Lol/Cyclotomic/Tensor/CTensor/g.c, 
+             src/Crypto/Lol/Cyclotomic/Tensor/CTensor/generalfuncs.c, 
+             src/Crypto/Lol/Cyclotomic/Tensor/CTensor/l.c, 
+             src/Crypto/Lol/Cyclotomic/Tensor/CTensor/random.c
+  Includes: src/Crypto/Lol/Cyclotomic/Tensor/CTensor/tensorTypes.h
+  default-language:   Haskell2010
+
+  if flag(useICC)
+    ghc-options: -pgml icc -optc-O3 
+    cc-options: -std=gnu99 -Wall -DSTATS -DCINTRIN
+  else
+    ghc-options: -pgml gcc -fPIC -optc-O3 
+    cc-options: -std=gnu99 -fPIC -Wall
+
+  if flag(llvm)
+    ghc-options: -fllvm -optlo-O3
+
+  -- ghc optimizations
+  ghc-options: -O3 -Odph -funbox-strict-fields -fwarn-dodgy-imports -rtsopts
+  ghc-options: -fno-liberate-case -funfolding-use-threshold1000 -funfolding-keeness-factor1000
+
+  exposed-modules: 
+    Crypto.Lol
+    Crypto.Lol.CRTrans
+    Crypto.Lol.Gadget
+    Crypto.Lol.LatticePrelude
+    
+    Crypto.Lol.Applications.SymmSHE
+    Crypto.Lol.Cyclotomic.Tensor
+    Crypto.Lol.Factored
+
+    Crypto.Lol.Cyclotomic.Cyc
+    Crypto.Lol.Cyclotomic.UCyc
+    Crypto.Lol.Cyclotomic.Utility
+    
+    Crypto.Lol.Cyclotomic.Linear
+
+    Crypto.Lol.Cyclotomic.Tensor.CTensor
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+    
+    Crypto.Lol.Types.FiniteField
+    Crypto.Lol.Types.PrimeField
+    Crypto.Lol.Types.IrreducibleChar2
+    
+    Crypto.Lol.Types.ZPP
+    Crypto.Lol.Types.ZqBasic
+
+    Crypto.Lol.Reflects
+
+  other-modules:
+        
+    Crypto.Lol.Types.ZmStar
+    Crypto.Lol.GaussRandom
+    Crypto.Lol.Types.Complex
+    Crypto.Lol.Types.IZipVector
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor.CRT
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Extension
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Gauss
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor.GL
+    Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon
+
+    Crypto.Lol.Cyclotomic.Tensor.CTensor.Extension
+    Crypto.Lol.Types.Numeric
+
+  build-depends:
+    arithmoi>=0.4.1.3 && <0.5,
+    base==4.8.*,
+    constraints==0.4.*,
+    containers>=0.5.6.2 && < 0.6,
+    data-default>=0.3.0 && < 0.6,
+    deepseq>=1.4.1.1 && <1.5,
+    MonadRandom>=0.2 && < 0.5,
+    mtl>=2.2.1 && < 2.3,
+    numeric-prelude>=0.4.2 && < 0.5,
+    QuickCheck>=2.8 && < 2.9,
+    random>=1.1 && < 1.2,
+    reflection>=1.5.1 && < 2.2,
+    repa==3.4.*,
+    singletons>=1.1.2.1 && < 2.1,
+    storable-record>=0.0.3 && < 0.1,
+    storable-tuple>=0.0.1 && < 0.1,
+    th-desugar>=1.5.4 && < 1.6,
+    type-natural>=0.2.3.2 && < 0.4,
+    tagged-transformer>=0.7 && < 0.9,
+    transformers>=0.4.2.0 && < 0.5,
+    vector==0.11.*,
+    vector-th-unbox>=0.2.1.0 && < 0.3
+
+  other-extensions: TemplateHaskell
+
+test-suite lol-test-suite
+  type:               exitcode-stdio-1.0
+  hs-source-dirs:     test-suite
+  default-language:   Haskell2010
+  main-is:            Main.hs
+
+  -- ghc optimizations
+  ghc-options: -threaded -rtsopts
+
+  build-depends:
+    base,
+    constraints,
+    lol,
+    MonadRandom,
+    QuickCheck>=2.8 && < 2.9,
+    repa,
+    test-framework >= 0.8 && < 0.9,
+    test-framework-quickcheck2 >= 0.3 && < 0.4,
+    time>=1.2 && < 1.6,
+    type-natural,
+    vector
diff --git a/src/Crypto/Lol.hs b/src/Crypto/Lol.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol.hs
@@ -0,0 +1,21 @@
+
+-- | Re-exports primary interfaces.
+
+module Crypto.Lol 
+( module Crypto.Lol.Cyclotomic.Cyc
+, module Crypto.Lol.Gadget
+, module Crypto.Lol.LatticePrelude
+
+, module Crypto.Lol.Types.ZqBasic
+, module Crypto.Lol.Cyclotomic.Tensor.CTensor
+, module Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+, module Crypto.Lol.Types.IrreducibleChar2) where
+
+import Crypto.Lol.Cyclotomic.Cyc
+import Crypto.Lol.Gadget
+import Crypto.Lol.LatticePrelude
+
+import Crypto.Lol.Types.ZqBasic
+import Crypto.Lol.Cyclotomic.Tensor.CTensor
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+import Crypto.Lol.Types.IrreducibleChar2
diff --git a/src/Crypto/Lol/Applications/SymmSHE.hs b/src/Crypto/Lol/Applications/SymmSHE.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Applications/SymmSHE.hs
@@ -0,0 +1,496 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             FlexibleContexts, FlexibleInstances, GADTs,
+             MultiParamTypeClasses, NoImplicitPrelude, ScopedTypeVariables,
+             TypeFamilies, TypeOperators, UndecidableInstances #-}
+
+-- | Symmetric-key somewhat homomorphic encryption.
+
+module Crypto.Lol.Applications.SymmSHE
+(
+-- * Data types
+SK, PT, CT                    -- don't export constructors!
+-- * Keygen, encryption, decryption
+, genSK
+, encrypt
+, errorTerm, errorTermUnrestricted, decrypt, decryptUnrestricted
+-- * Encoding of plaintext
+, toMSD, toLSD
+-- * Arithmetic with public values
+, addScalar, addPublic, mulPublic
+-- * Modulus switching
+, rescaleLinearCT, modSwitchPT
+-- * Key switching
+, keySwitchLinear, keySwitchQuadCirc
+-- * Ring switching
+, embedSK, embedCT, twaceCT
+, tunnelCT
+-- * Constraint synonyms
+, AddPublicCtx, MulPublicCtx, KeySwitchCtx, KSHintCtx, ModSwitchPTCtx
+, ToSDCtx, EncryptCtx, TunnelCtx, GenSKCtx, DecryptCtx
+, ErrorTermCtx
+) where
+
+import qualified Algebra.Additive as Additive (C)
+import qualified Algebra.Ring     as Ring (C)
+
+import Crypto.Lol.Cyclotomic.Cyc
+import Crypto.Lol.Cyclotomic.UCyc (forceDec)
+import Crypto.Lol.Cyclotomic.Linear
+import Crypto.Lol.Gadget
+import Crypto.Lol.LatticePrelude    as LP hiding (sin)
+
+import Control.Applicative  hiding ((*>))
+import Control.DeepSeq
+import Control.Monad        as CM
+import Control.Monad.Random
+import Data.Maybe
+import Data.Traversable     as DT
+import Data.Typeable
+
+import MathObj.Polynomial as P
+
+-- | secret key
+data SK r where
+  SK  :: (ToRational v, NFData v) => v -> r -> SK r
+
+-- | plaintext
+type PT rp = rp
+
+-- | Ciphertext encoding type
+data Encoding = MSD | LSD deriving (Show, Eq)
+
+-- | Ciphertext over @R'_q@, encrypting a plaintext in @R_p (R=O_m)@.
+data CT (m :: Factored) zp r'q =
+  CT
+  !Encoding                     -- MSD/LSD encoding
+  !Int                          -- accumulated power of g_m' in c(s)
+  !zp                           -- factor to mul by upon decryption
+  !(Polynomial r'q)             -- the polynomial c(s)
+  deriving (Typeable, Show)
+
+-- Note: do *not* give an Eq instance for CT, because it's not
+-- meaningful to compare ciphertexts for equality
+
+instance (NFData zp, NFData r'q) => NFData (CT m zp r'q) where
+  rnf (CT _ k sc cs) = rnf k `seq` rnf sc `seq` rnf cs
+
+instance (NFData r) => NFData (SK r) where
+  rnf (SK v s) = rnf v `seq` rnf s
+
+---------- Basic functions: Gen, Enc, Dec ----------
+
+-- | Constraint synonym for generating a secret key.
+type GenSKCtx t m z v =
+  (ToInteger z, Fact m, CElt t z, ToRational v, NFData v)
+
+-- | Generates a secret key with (index-independent) scaled variance
+-- parameter @v@; see 'errorRounded'.
+genSK :: (GenSKCtx t m z v, MonadRandom rnd)
+         => v -> rnd (SK (Cyc t m z))
+genSK v = liftM (SK v) $ errorRounded v
+
+-- | Constraint synonym for encryption.
+type EncryptCtx t m m' z zp zq =
+  (Mod zp, Ring zp, Ring zq, Lift zp (ModRep zp),
+   Reduce z zq, Reduce (LiftOf zp) zq,
+   CElt t zq, CElt t zp, CElt t z, CElt t (LiftOf zp),
+   m `Divides` m')
+
+-- | Encrypt a plaintext under a secret key.
+encrypt :: forall t m m' z zp zq e rnd . (EncryptCtx t m m' z zp zq, MonadRandom rnd)
+           => SK (Cyc t m' z) -> PT (Cyc t m zp) -> rnd (CT m zp (Cyc t m' zq))
+encrypt (SK svar s) =
+  let sq = reduce s
+  in (\pt -> do
+         e <- errorCoset svar (embed pt :: PT (Cyc t m' zp))
+         c1 <- getRandom
+         return $! CT LSD zero one $ fromCoeffs [reduce e - c1 * sq, c1])
+
+-- | Constraint synonym for extracting the error term of a ciphertext.
+type ErrorTermCtx t m' z zp zq =
+  (Reduce z zq, Lift' zq, CElt t z, CElt t (LiftOf zq),
+   ToSDCtx t m' zp zq)
+
+-- | Extract the error term of a ciphertext.
+errorTerm :: (ErrorTermCtx t m' z zp zq)
+             => SK (Cyc t m' z) -> CT m zp (Cyc t m' zq) -> Cyc t m' (LiftOf zq)
+errorTerm (SK _ s) = let sq = reduce s in
+  \ct -> let (CT LSD _ _ c) = toLSD ct
+         in liftCyc Dec $ evaluate c sq
+
+-- for when we know the division must succeed
+divG' :: (Fact m, CElt t r) => Cyc t m r -> Cyc t m r
+divG' = fromJust . divG
+
+-- | Constraint synonym for decryption.
+type DecryptCtx t m m' z zp zq =
+  (ErrorTermCtx t m' z zp zq, Reduce (LiftOf zq) zp,
+   m `Divides` m', CElt t zp)
+
+-- | Decrypt a ciphertext.
+decrypt :: forall t m m' z zp zq . (DecryptCtx t m m' z zp zq)
+           => SK (Cyc t m' z) -> CT m zp (Cyc t m' zq) -> PT (Cyc t m zp)
+decrypt sk ct =
+  let ct'@(CT LSD k l _) = toLSD ct
+  in let e :: Cyc t m' zp = reduce $ errorTerm sk ct'
+     in (scalarCyc l) * twace (iterate divG' e !! k)
+
+--- unrestricted versions ---
+
+-- | More general form of 'errorTerm' that works for unrestricted
+-- output coefficient types.
+errorTermUnrestricted :: 
+  (Reduce z zq, Lift' zq, CElt t z, ToSDCtx t m' zp zq)
+  => SK (Cyc t m' z) -> CT m zp (Cyc t m' zq) -> Cyc t m' (LiftOf zq)
+errorTermUnrestricted (SK _ s) = let sq = reduce s in
+  \ct -> let (CT LSD _ _ c) = toLSD ct
+             eval = evaluate c sq
+         in cyc $ fmap lift $ forceDec $ unsafeUnCyc eval
+
+-- | More general form of 'decrypt' that works for unrestricted output
+-- coefficient types.
+decryptUnrestricted :: 
+ (Fact m, Fact m', CElt t zp, m `Divides` m',
+  Reduce z zq, Lift' zq, CElt t z, ToSDCtx t m' zp zq, Reduce (LiftOf zq) zp)
+  => SK (Cyc t m' z) -> CT m zp (Cyc t m' zq) -> PT (Cyc t m zp)
+
+decryptUnrestricted (SK _ s) = let sq = reduce s in
+  \ct -> let (CT LSD k l c) = toLSD ct
+         in let eval = evaluate c sq
+                e = cyc $ fmap (reduce . lift) $ forceDec $ unsafeUnCyc eval
+                l' = scalarCyc l
+            in l' * twace (iterate divG' e !! k)
+
+---------- LSD/MSD switching ----------
+
+-- | Constraint synonym for converting between ciphertext encodings.
+type ToSDCtx t m' zp zq = (Encode zp zq, Fact m', CElt t zq)
+
+toLSD, toMSD :: ToSDCtx t m' zp zq
+ => CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+
+-- CJP: reduce duplication in these functions?  They differ in only two places
+
+-- | Convert a ciphertext to MSD encoding.
+toMSD = let (zpScale, zqScale) = lsdToMSD
+            rqScale = scalarCyc zqScale
+        in \ct@(CT enc k l c) -> case enc of
+          MSD -> ct
+          LSD -> CT MSD k (zpScale * l) ((rqScale *) <$> c)
+
+-- | Convert a ciphertext to LSD encoding.
+toLSD = let (zpScale, zqScale) = msdToLSD
+            rqScale = scalarCyc zqScale
+        in \ct@(CT enc k l c) -> case enc of
+          LSD -> ct
+          MSD -> CT LSD k (zpScale * l) ((rqScale *) <$> c)
+
+---------- Modulus switching ----------
+
+-- | Rescale a linear polynomial in MSD encoding, for best noise
+-- behavior.
+rescaleLinearMSD :: (RescaleCyc (Cyc t) zq zq', Fact m')
+                    => Polynomial (Cyc t m' zq) -> Polynomial (Cyc t m' zq')
+rescaleLinearMSD c = case coeffs c of
+  [] -> fromCoeffs []
+  [c0] -> fromCoeffs [rescaleCyc Dec c0]
+  [c0,c1] -> let c0' = rescaleCyc Dec c0
+                 c1' = rescaleCyc Pow c1
+             in fromCoeffs [c0', c1']
+  _ -> error $ "rescaleLinearMSD: list too long (not linear): " ++
+       show (length $ coeffs c)
+
+-- | Rescale a linear ciphertext to a new modulus.
+rescaleLinearCT :: (RescaleCyc (Cyc t) zq zq', ToSDCtx t m' zp zq)
+           => CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq')
+rescaleLinearCT ct = let CT MSD k l c = toMSD ct
+                     in CT MSD k l $ rescaleLinearMSD c
+
+-- | Constraint synonym for modulus switching.
+type ModSwitchPTCtx t m' zp zp' zq =
+  (Lift' zp, Reduce (LiftOf zp) zp', ToSDCtx t m' zp zq)
+
+-- | Homomorphically divide a plaintext that is known to be a multiple
+-- of @(p\/p\')@ by that factor, thereby scaling the plaintext modulus
+-- from @p@ to @p\'@.
+modSwitchPT :: (ModSwitchPTCtx t m' zp zp' zq)
+            => CT m zp (Cyc t m' zq) -> CT m zp' (Cyc t m' zq)
+modSwitchPT ct = let CT MSD k l c = toMSD ct in
+    CT MSD k (reduce (lift l)) c
+
+---------- Key switching ----------
+
+type LWECtx t m' z zq =
+  (ToInteger z, Reduce z zq, Ring zq, Fact m', CElt t z, CElt t zq)
+
+-- | An LWE sample for a given secret (corresponding to a linear
+-- ciphertext encrypting 0 in MSD form)
+lweSample :: (LWECtx t m' z zq, MonadRandom rnd)
+             => SK (Cyc t m' z) -> rnd (Polynomial (Cyc t m' zq))
+lweSample (SK svar s) =
+  let sq = adviseCRT $ negate $ reduce s
+  in do
+    e <- errorRounded svar
+    c1 <- getRandom
+    return $ fromCoeffs [c1 * sq + reduce (e `asTypeOf` s), c1]
+
+-- | Constraint synonym for generating key-switch hints.
+type KSHintCtx gad t m' z zq = 
+  (LWECtx t m' z zq, Reduce (DecompOf zq) zq, Gadget gad zq,
+   CElt t (DecompOf zq))
+
+-- | Generate a hint that "encrypts" a value under a secret key, in
+-- the sense required for key-switching.  The hint works for any
+-- plaintext modulus, but must be applied on a ciphertext in MSD form.
+ksHint :: (KSHintCtx gad t m' z zq, MonadRandom rnd)
+          => SK (Cyc t m' z) -> Cyc t m' z
+          -> rnd (Tagged gad [Polynomial (Cyc t m' zq)])
+ksHint skout val = do           -- rnd monad
+  let valq = reduce val
+      valgad = encode valq
+  -- CJP: clunky, but that's what we get without a MonadTagged
+  samples <- DT.mapM (\as -> replicateM (length as) (lweSample skout)) valgad
+  return $ zipWith (+) <$> (map P.const <$> valgad) <*> samples
+
+type KnapsackCtx t (m' :: Factored) z zq' =
+  (Reduce z zq', Fact m', CElt t z, CElt t zq')
+
+-- poor man's module multiplication for knapsack
+(*>>) :: Ring r => r -> Polynomial r -> Polynomial r
+(*>>) r = fmap (r *)
+
+knapsack :: forall t m' z zq' . (KnapsackCtx t m' z zq')
+            => [Polynomial (Cyc t m' zq')] -> [Cyc t m' z] -> Polynomial (Cyc t m' zq')
+knapsack hint xs = sum (zipWith (*>>) (adviseCRT <$> reduce <$> xs) hint)
+
+type InnerKeySwitchCtx gad t m' zq zq' =
+  (RescaleCyc (Cyc t) zq' zq, RescaleCyc (Cyc t) zq zq',
+   Decompose gad zq', KnapsackCtx t m' (DecompOf zq') zq')
+
+-- Helper function: applies key-switch hint to a ring element.
+-- Signature is such that we should rescale input and output.
+switch :: forall gad t m' zq' zq . (InnerKeySwitchCtx gad t m' zq zq')
+  => Tagged gad [Polynomial (Cyc t m' zq')] -> Cyc t m' zq -> Polynomial (Cyc t m' zq)
+switch hint c = rescaleLinearMSD $ untag $ knapsack <$>
+                hint <*> decompose (rescaleCyc Pow c :: Cyc t m' zq')
+
+-- | Constraint synonym for key switching.
+type KeySwitchCtx gad t m' zp zq zq' =
+  (ToSDCtx t m' zp zq,
+   -- EAC: same as InnerKeySwitchCtx, but duplicated for haddock
+   RescaleCyc (Cyc t) zq' zq, RescaleCyc (Cyc t) zq zq',
+   Decompose gad zq', KnapsackCtx t m' (DecompOf zq') zq')
+
+-- | Switch a linear ciphertext under @s_in@ to a linear one under @s_out@
+keySwitchLinear :: forall gad t m' zp zq zq' z rnd m .
+  (KeySwitchCtx gad t m' zp zq zq', KSHintCtx gad t m' z zq', MonadRandom rnd)
+  => SK (Cyc t m' z)                -- sout
+  -> SK (Cyc t m' z)                -- sin
+  -> TaggedT (gad, zq') rnd (CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq))
+keySwitchLinear skout (SK _ sin) = tagT $ do
+  hint :: Tagged gad [Polynomial (Cyc t m' zq')] <- ksHint skout sin
+  return $ hint `deepseq`
+    (\ct -> let CT MSD k l c = toMSD ct
+                [c0,c1] = coeffs c
+            in CT MSD k l $ P.const c0 + switch hint c1)
+
+-- | Switch a quadratic ciphertext (i.e., one with three components)
+-- to a linear one under the /same/ key.
+keySwitchQuadCirc :: forall gad t m' zp zq zq' z m rnd .
+  (KeySwitchCtx gad t m' zp zq zq', KSHintCtx gad t m' z zq', MonadRandom rnd)
+  => SK (Cyc t m' z)
+  -> TaggedT (gad, zq') rnd (CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq))
+keySwitchQuadCirc sk@(SK _ s) = tagT $ do
+  hint :: Tagged gad [Polynomial (Cyc t m' zq')] <- ksHint sk (s*s)
+  return $ hint `deepseq` (\ct ->
+    let CT MSD k l c = toMSD ct
+        [c0,c1,c2] = coeffs c
+    in CT MSD k l $ P.fromCoeffs [c0,c1] + switch hint c2)
+
+---------- Misc homomorphic operations ----------
+
+type AddScalarCtx t m' zp zq =
+  (Lift' zp, Reduce (LiftOf zp) zq, ToSDCtx t m' zp zq)
+
+-- | Homomorphically add a public @Z_p@ value to an encrypted value.  The
+-- ciphertext must not carry any @g@ factors.
+addScalar :: (AddScalarCtx t m' zp zq)
+             => zp -> CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+addScalar b ct =
+  let (l,c) = case toLSD ct of
+        CT LSD 0 l c -> (l,c)
+        CT LSD _ _ _ -> error "cannot add public scalar to ciphertext with 'g' factors"
+        _ -> error "internal error: addScalar"
+      b' = scalarCyc (reduce $ lift $ b * recip l)
+  in CT LSD 0 l $ c + P.const b'
+
+-- | Constraint synonym for adding a public value to an encrypted value
+type AddPublicCtx t m m' zp zq =
+  (Lift' zp, Reduce (LiftOf zp) zq, m `Divides` m',
+   CElt t zp, CElt t (LiftOf zp), ToSDCtx t m' zp zq)
+
+-- | Homomorphically add a public @R_p@ value to an encrypted value.
+addPublic :: forall t m m' zp zq . (AddPublicCtx t m m' zp zq)
+          => Cyc t m zp -> CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+addPublic b ct = let CT LSD k l c = toLSD ct in
+  let linv = scalarCyc $ recip l
+      -- multiply public value by appropriate power of g and divide by the
+      -- scale, to match the form of the ciphertext
+      b' :: Cyc t m zq = reduce $ liftCyc Pow $ linv * (iterate mulG b !! k)
+  in CT LSD k l $ c + P.const (embed b')
+
+-- | Constraint synonym for multiplying a public value with an encrypted value
+type MulPublicCtx t m m' zp zq =
+  (Lift' zp, Reduce (LiftOf zp) zq, Ring zq, m `Divides` m',
+   CElt t zp, CElt t (LiftOf zp), CElt t zq)
+
+-- | Homomorphically multiply an encrypted value by a public @R_p@ value.
+mulPublic :: forall t m m' zp zq . (MulPublicCtx t m m' zp zq)
+             => Cyc t m zp -> CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+mulPublic a (CT enc k l c) =
+  let a' = embed (reduce $ liftCyc Pow a :: Cyc t m zq)
+  in CT enc k l $ (a' *) <$> c
+
+-- | Increment the internal g exponent without changing the encrypted
+-- message.
+mulGCT :: (Fact m', CElt t zq)
+          => CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+mulGCT (CT enc k l c) = CT enc (k+1) l $ mulG <$> c
+
+---------- NumericPrelude instances ----------
+
+instance (Eq zp, m `Divides` m', ToSDCtx t m' zp zq)
+         => Additive.C (CT m zp (Cyc t m' zq)) where
+
+  zero = CT LSD 0 one zero
+
+  -- the scales, g-exponents of ciphertexts, and MSD/LSD types must match.
+  ct1@(CT enc1 k1 l1 c1) + ct2@(CT enc2 k2 l2 c2)
+      -- for simplicity, we don't currently support this. Shouldn't be
+      -- too complicated though.
+      | l1 /= l2 = error "Cannot add ciphertexts with different scale values"
+      | k1 < k2 = iterate mulGCT ct1 !! (k2-k1) + ct2
+      | k1 > k2 = ct1 + iterate mulGCT ct2 !! (k1-k2)
+      | enc1 == LSD && enc2 == MSD = toMSD ct1 + ct2
+      | enc1 == MSD && enc2 == LSD = ct1 + toMSD ct2
+      | otherwise = CT enc1 k1 l1 $ c1 + c2
+
+  negate (CT enc k l c) = CT enc k l $ negate <$> c
+
+instance (ToSDCtx t m' zp zq, Additive (CT m zp (Cyc t m' zq)))
+  => Ring.C (CT m zp (Cyc t m' zq)) where
+
+  one = CT LSD 0 one one
+
+  -- need at least one ct to be in LSD form
+  ct1@(CT MSD _ _ _) * ct2@(CT MSD _ _ _) = toLSD ct1 * ct2
+
+  -- first is in LSD
+  (CT LSD k1 l1 c1) * (CT d2 k2 l2 c2) =
+    -- mul by g so error maintains invariant: error*g is "round"
+    CT d2 (k1+k2+1) (l1*l2) (mulG <$> c1 * c2)
+
+  -- else, second must be in LSD
+  ct1 * ct2 = ct2 * ct1
+
+{- CJP: do we want/need this?  We have mulPublic...
+
+instance (MulPublicCtx t m m' zp zq, Reduce z' zp, CElt t z',
+          l `Divides` m, Additive (CT m zp (c m' zq)), Ring (c l z'))
+  => Module.C (c l z') (CT m zp (c m' zq)) where
+
+  (*>) a = mulPublic (embed $ reduce a)
+-}
+
+---------- Ring switching ----------
+
+type AbsorbGCtx t m' zp zq =
+  (Lift' zp, Reduce (LiftOf zp) zq, Ring zp, Ring zq, Fact m',
+   CElt t (LiftOf zp), CElt t zp, CElt t zq)
+
+-- | "Absorb" the powers of g associated with the ciphertext, at the
+-- cost of some increase in noise. This is usually needed before
+-- changing the index of the ciphertext ring.
+absorbGFactors :: forall t zp zq m m' . (AbsorbGCtx t m' zp zq)
+                  => CT m zp (Cyc t m' zq) -> CT m zp (Cyc t m' zq)
+absorbGFactors ct@(CT enc k l c)
+  | k == 0 = ct
+  | k > 0 = let d :: Cyc t m' zp = iterate divG' one !! k
+                rep = adviseCRT $ reduce $ liftCyc Pow d
+            in CT enc 0 l $ (rep *) <$> c
+  | otherwise = error "k < 0 in absorbGFactors"
+
+-- | Embed a ciphertext in R' encrypting a plaintext in R to a
+-- ciphertext in T' encrypting a plaintext in T. The target ciphertext
+-- ring T' must contain both the the source ciphertext ring R' and the
+-- target plaintext ring T.
+embedCT :: (CElt t zq,
+            r `Divides` r', s `Divides` s', r `Divides` s, r' `Divides` s')
+           => CT r zp (Cyc t r' zq) -> CT s zp (Cyc t s' zq)
+-- We could call absorbGFactors first, insead of error.  Embedding
+-- *essentially* maintains the invariant that noise*g is "round."
+-- While g'/g can be non-spherical, it only stretches by at most a
+-- factor of 2 per new odd prime.  We *cannot* multiply by g, then
+-- embed, then divide by g' because the result would not remain in R'.
+-- So this is the best we can do.
+embedCT (CT d 0 l c) = CT d 0 l (embed <$> c)
+embedCT _ = error "embedCT requires 0 factors of g; call aborbGFactors first"
+
+-- | Embed a secret key from a subring into a superring.
+embedSK :: (CElt t z, m `Divides` m') => SK (Cyc t m z) -> SK (Cyc t m' z)
+embedSK (SK v s) = SK v $ embed s
+
+-- | "Tweaked trace" function for ciphertexts.  Mathematically, the
+-- target plaintext ring @S@ must contain the intersection of the
+-- source plaintext ring @T@ and the target ciphertext ring @S\'@.
+-- Here we make the stricter requirement that @s = gcd(s\', t)@.
+twaceCT :: (CElt t zq, r `Divides` r', s' `Divides` r',
+            s ~ (FGCD s' r))
+           => CT r zp (Cyc t r' zq) -> CT s zp (Cyc t s' zq)
+-- we could call absorbGFactors first, insead of error
+twaceCT (CT d 0 l c) = CT d 0 l (twace <$> c)
+twaceCT _ = error "twaceCT requires 0 factors of g; call absorbGFactors first"
+
+-- | Constraint synonym for ring tunneling.
+type TunnelCtx t e r s e' r' s' z zp zq zq' gad =
+  (ExtendLinIdx e r s e' r' s',     -- liftLin
+   KSHintCtx gad t r' z zq',        -- ksHint
+   Reduce z zq,                     -- Reduce on Linear
+   Lift zp z,                       -- liftLin
+   CElt t zp,                       -- liftLin
+   KeySwitchCtx gad t s' zp zq zq') -- keySwitch
+
+-- | Homomorphically apply the @E@-linear function that maps the
+-- elements of the decoding basis of @R\/E@ to the corresponding
+-- @S@-elements in the input array.
+tunnelCT :: forall gad t e r s e' r' s' z zp zq zq' rnd .
+  (TunnelCtx t e r s e' r' s' z zp zq zq' gad,
+   MonadRandom rnd)
+  => Linear t zp e r s
+  -> SK (Cyc t s' z)
+  -> SK (Cyc t r' z)
+  -> TaggedT (gad,zq') rnd (CT r zp (Cyc t r' zq) -> CT s zp (Cyc t s' zq))
+tunnelCT f skout (SK _ sin) = tagT $ (do -- in rnd
+  -- generate hints
+  let f' = extendLin $ lift f :: Linear t z e' r' s'
+      -- choice of basis here must match coeffsCyc basis below
+      ps = proxy powBasis (Proxy::Proxy e')
+      comps = (evalLin f' . (adviseCRT sin *)) <$> ps
+  hints :: [Tagged gad [Polynomial (Cyc t s' zq')]] <- CM.mapM (ksHint skout) comps
+  return $ hints `deepseq` \ct' ->
+    let CT MSD 0 s c = toMSD $ absorbGFactors ct'
+        [c0,c1] = coeffs c
+        -- apply E-linear function to constant term c0
+        c0' = evalLin (reduce f' :: Linear t zq e' r' s') c0
+        -- apply E-linear function to c1 via key-switching
+        -- this basis must match the basis used above to generate the hints
+        c1s = coeffsCyc Pow c1 :: [Cyc t e' zq]
+        -- CJP: don't embed the c1s before decomposing them (inside
+        -- switch); instead decompose in smaller ring before
+        -- embedding (it matters).
+        -- We may need to generalize switch or define an
+        -- alternative.
+        c1s' = zipWith switch hints (embed <$> c1s)
+        c1' = sum c1s'
+    in CT MSD 0 s $ P.const c0' + c1')
+      \\ lcmDivides (Proxy::Proxy r) (Proxy::Proxy e')
diff --git a/src/Crypto/Lol/CRTrans.hs b/src/Crypto/Lol/CRTrans.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/CRTrans.hs
@@ -0,0 +1,181 @@
+{-# LANGUAGE FlexibleContexts, FlexibleInstances, RebindableSyntax,
+             ScopedTypeVariables, TypeFamilies #-}
+
+-- | Classes and helper methods for the Chinese remainder transform
+-- and ring extensions.
+
+module Crypto.Lol.CRTrans
+( CRTrans(..), CRTEmbed(..)
+, CRTInfo
+, crtInfoFact, crtInfoPPow, crtInfoNatC
+, gEmbPPow, gEmbNatC
+, omegaPowMod, zqHasCRT
+) where
+
+import Crypto.Lol.LatticePrelude
+
+import Math.NumberTheory.Primes.Factorisation (carmichael, factorise)
+
+import           Control.Arrow
+import           Data.Singletons
+import           Data.Singletons.Prelude
+import           Data.Type.Natural       (Sing (SS))
+import qualified Data.Vector             as V
+
+-- | Information that characterizes the (invertible) Chinese remainder
+-- transformation over a ring @r@, namely:
+--
+--     (1) a function that returns the @i@th power of some @m@th root
+--     of unity (for any integer @i@)
+--
+--     (2) the multiplicative inverse of @\\hat{m}@ in @r@.
+
+type CRTInfo r = (Int -> r, r)
+
+-- | A ring that (possibly) supports invertible Chinese remainder
+-- transformations of various indices.
+
+-- | The values of 'crtInfo' for different indices @m@ should be
+-- consistent, in the sense that if @omega@, @omega'@ are respectively
+-- the values returned for @m@, @m'@ where @m'@ divides @m@, then it
+-- should be the case that @omega^(m/m')=omega'@.
+
+class Ring r => CRTrans r where
+
+  -- | 'CRTInfo' for a given index @m@. The method itself may be
+  -- slow, but the function it returns should be fast, e.g., via
+  -- internal memoization.  The default implementation returns
+  -- 'Nothing'.
+  crtInfo :: Int -> Maybe (CRTInfo r)
+  crtInfo = const Nothing
+
+-- | A ring with a ring embedding into some ring @CRTExt r@ that has
+-- an invertible CRT transformation for /every/ positive index @m@.
+class (Ring r, Ring (CRTExt r)) => CRTEmbed r where
+  type CRTExt r
+
+  -- | Embeds from @r@ to @CRTExt r@
+  toExt :: r -> CRTExt r
+  -- | Projects from @CRTExt r@ to @r@
+  fromExt :: CRTExt r -> r
+
+-- CRTrans instance for product rings
+instance (CRTrans a, CRTrans b) => CRTrans (a,b) where
+  crtInfo i = do
+    (apow, aiInv) <- crtInfo i
+    (bpow, biInv) <- crtInfo i
+    return (apow &&& bpow, (aiInv, biInv))
+
+-- CRTEmbed instance for product rings
+instance (CRTEmbed a, CRTEmbed b) => CRTEmbed (a,b) where
+  type CRTExt (a,b) = (CRTExt a, CRTExt b)
+  toExt = toExt *** toExt
+  fromExt = fromExt *** fromExt
+
+-- | Default implementation of 'omegaPow' for 'Mod' types.  The
+-- implementation finds an integer element of maximal multiplicative
+-- order, and raises it to the appropriate power. Therefore, the
+-- functions returned for different values of the first argument are
+-- consistent, i.e., @omega_{m'}^(m'/m) = omega_m@.
+omegaPowMod :: forall r . (Mod r, Enumerable r, Ring r, Eq r)
+               => Int -> Maybe (Int -> r)
+omegaPowMod =                -- use Integers for all intermediate calcs
+
+  -- CJP: there's a mismatch here between the semantics of Mod and the
+  -- use of 'values'.  If r really represents *integers* modulo
+  -- something then we're fine, otherwise we might get weird behavior.
+
+    let -- the exponent of Z_q^*
+        exponent = carmichael $ fromIntegral (proxy modulus (Proxy::Proxy r))
+        -- all prime divisors of exponent
+        primes = map fst $ factorise exponent
+        -- the powers we need to check
+        exps = map (exponent `div`) primes
+        -- whether an element is a unit with maximal order
+        isGood x = (x^exponent == one) && all (\e -> x^e /= one) exps
+    in \m -> let (mq, mr) = exponent `divMod` fromIntegral m
+             in if mr == 0
+                then let omega = head (filter isGood values) ^ mq
+                         omegaPows = V.iterateN m (*omega) one
+                     in Just $ (omegaPows V.!) . (`mod` m)
+                else Nothing
+
+omegaPowC :: (Transcendental a) => Int -> Int -> Complex a
+omegaPowC m i = cis (2*pi*fromIntegral i / fromIntegral m)
+
+-- | 'crtInfo' wrapper for 'Fact' types.
+crtInfoFact :: (Fact m, CRTrans r) => TaggedT m Maybe (CRTInfo r)
+crtInfoFact = (tagT . crtInfo) =<< pureT valueFact
+
+-- | 'crtInfo' wrapper for 'PPow' types.
+crtInfoPPow :: (PPow pp, CRTrans r) => TaggedT pp Maybe (CRTInfo r)
+crtInfoPPow = (tagT . crtInfo) =<< pureT valuePPow
+
+-- | 'crtInfo' wrapper for 'NatC' types.
+crtInfoNatC :: (NatC p, CRTrans r) => TaggedT p Maybe (CRTInfo r)
+crtInfoNatC = (tagT . crtInfo) =<< pureT valueNatC
+
+-- | A function that returns the 'i'th embedding of @g_{p^e} = g_p@ for
+-- @i@ in @Z*_{p^e}@.
+gEmbPPow :: forall pp r . (PPow pp, CRTrans r) => TaggedT pp Maybe (Int -> r)
+gEmbPPow = tagT $ case (sing :: SPrimePower pp) of
+  -- intentionally no match for zero exponents
+  (SPP (STuple2 sp (SS _))) -> withWitnessT gEmbNatC sp
+
+-- | A function that returns the @i@th embedding of @g_p@ for @i@ in @Z*_p@,
+-- i.e., @1-omega_p^i@.
+gEmbNatC :: (NatC p, CRTrans r) => TaggedT p Maybe (Int -> r)
+gEmbNatC = do
+  (f, _) <- crtInfoNatC
+  return $ \i -> one - f i      -- not checking that i /= 0 (mod p)
+
+-- | @zqHasCRT m q@ says whether @Z_q@ has an /invertible/ CRT
+-- transform of index @m@, i.e., @Z_q@ has an element of
+-- multiplicative order @m@, and @mhat@ is invertible in @Z_q@.
+zqHasCRT :: (ToInteger i, PID i) => i -> i -> Bool
+zqHasCRT m q = let exponent = fromIntegral $ carmichael $
+                            fromIntegral q
+                   mhat = if 2 `divides` m then m `div` 2 else m
+               in m `divides` exponent && fst (extendedGCD mhat q) == one
+
+-- the complex numbers have roots of unity of any order
+instance (Transcendental a) => CRTrans (Complex a) where
+  crtInfo m = Just (omegaPowC m, recip $ fromIntegral $ valueHat m)
+
+-- trivial CRTEmbed instance for complex numbers
+instance (Transcendental a) => CRTEmbed (Complex a) where
+  type CRTExt (Complex a) = Complex a
+  toExt = id
+  fromExt = id
+
+-- Default CRTrans instances for real and integer types, which do
+-- not have roots of unity (except in trivial cases). These are needed
+-- to use FastCyc with these integer types.
+instance CRTrans Double
+instance CRTrans Int
+instance CRTrans Int64
+instance CRTrans Integer
+-- can also do for Int8, Int16, Int32 etc.
+
+-- CRTEmbed instances for real and integer types, embedding into
+-- Complex.  These are needed to use FastCyc with these integer types.
+instance CRTEmbed Double where
+  type CRTExt Double = Complex Double
+  toExt = fromReal . realToField
+  fromExt = realToField . real
+
+instance CRTEmbed Int where
+  type CRTExt Int = Complex Double
+  toExt = fromIntegral
+  fromExt = fst . roundComplex
+
+instance CRTEmbed Int64 where
+  type CRTExt Int64 = Complex Double
+  toExt = fromIntegral
+  fromExt = fst . roundComplex
+
+instance CRTEmbed Integer where
+  -- CJP: sufficient precision?  Not in general.
+  type CRTExt Integer = Complex Double
+  toExt = fromIntegral
+  fromExt = fst . roundComplex
diff --git a/src/Crypto/Lol/Cyclotomic/Cyc.hs b/src/Crypto/Lol/Cyclotomic/Cyc.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Cyc.hs
@@ -0,0 +1,175 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts,
+             FlexibleInstances, GADTs, GeneralizedNewtypeDeriving,
+             MultiParamTypeClasses, NoImplicitPrelude, PolyKinds,
+             RankNTypes, ScopedTypeVariables, StandaloneDeriving,
+             TypeFamilies, TypeOperators, UndecidableInstances #-}
+
+-- | An implementation of cyclotomic rings with safe interface:
+-- functions and instances involving 'Cyc' expose nothing about the
+-- internal representations of ring elements (e.g., the basis they are
+-- represented in).  For an experts-only, "unsafe" implementation that
+-- offers limited exposure of internal representation, use
+-- 'Crypto.Lol.Cyclotomic.UCyc.UCyc'.
+
+module Crypto.Lol.Cyclotomic.Cyc
+( 
+-- * Data type
+  Cyc, U.CElt, cyc, unsafeUnCyc
+-- * Basic operations
+, mulG, divG
+, scalarCyc, liftCyc
+, adviseCRT
+-- * Error sampling
+, tGaussian, errorRounded, errorCoset
+-- * Sub/extension rings
+, embed, twace, powBasis, crtSet, coeffsCyc
+, module Crypto.Lol.Cyclotomic.Utility
+) where
+
+import Algebra.Additive as Additive (C)
+import Algebra.Ring     as Ring (C)
+
+import           Crypto.Lol.Cyclotomic.UCyc    (CElt, UCyc)
+import qualified Crypto.Lol.Cyclotomic.UCyc    as U
+import           Crypto.Lol.Cyclotomic.Utility
+import           Crypto.Lol.Gadget
+import           Crypto.Lol.LatticePrelude     as LP
+import           Crypto.Lol.Types.ZPP
+
+import Control.Applicative    hiding ((*>))
+import Control.DeepSeq
+import Control.Monad.Random
+
+import Data.Coerce
+
+import Test.QuickCheck
+
+-- | Wrapper around 'UCyc' that exposes a narrower, safe interface.
+newtype Cyc t m r = Cyc { 
+  -- | Unsafe deconstructor for 'Cyc'.
+  unsafeUnCyc :: UCyc t m r }
+                    deriving (Arbitrary, NFData, Random)
+
+-- See: https://ghc.haskell.org/trac/ghc/ticket/11008
+-- for why I have to use StandaloneDeriving here
+deriving instance Show (UCyc t m a) => Show (Cyc t m a)
+deriving instance Eq (UCyc t m a) => Eq (Cyc t m a)
+deriving instance Additive (UCyc t m a) => Additive.C (Cyc t m a)
+deriving instance Ring (UCyc t m a) => Ring.C (Cyc t m a)
+deriving instance Gadget gad (UCyc t m a) => Gadget gad (Cyc t m a)
+deriving instance Correct gad (UCyc t m a) => Correct gad (Cyc t m a)
+
+-- | Smart constructor (to prevent clients from pattern-matching).
+cyc :: UCyc t m r -> Cyc t m r
+cyc = Cyc
+
+-- (try to) replace all occurrences of 'Cyc' with 'UCyc'
+type family O a where
+  O (Cyc t m a) = UCyc t m a
+  O (a -> b) = O a -> O b
+  O (m a) = m (O a)             -- works for concrete m, but not vars
+  O a = a
+
+-- specialized 'coerce', to aid type inference
+coerceCyc :: (Coercible (O a) a) => O a -> a
+coerceCyc = coerce
+
+-- Can't seem to auto-derive these, due to constraints with GND and 
+-- MPTCs.
+instance (Reduce a b, Fact m, CElt t a, CElt t b)
+         => Reduce (Cyc t m a) (Cyc t m b) where
+  reduce = coerceCyc reduce
+
+-- CJP: will this pick the right overlapping instance for UCyc?  I
+-- think so...
+instance (RescaleCyc (UCyc t) a b) => RescaleCyc (Cyc t) a b where
+  rescaleCyc = coerceCyc rescaleCyc
+
+instance (Decompose gad (UCyc t m zq),
+          Reduce (Cyc t m (DecompOf zq)) (Cyc t m zq))
+         => Decompose gad (Cyc t m zq) where
+
+  type DecompOf (Cyc t m zq) = Cyc t m (DecompOf zq)
+  decompose = coerceCyc decompose
+
+---------- Core cyclotomic operations ----------
+
+-- | Yield an equivalent element that /may/ be in a CRT
+-- representation.  This can serve as an optimization hint. E.g.,
+-- call 'adviseCRT' prior to multiplying the same value by many
+-- other values.
+adviseCRT :: (Fact m, CElt t r) => Cyc t m r -> Cyc t m r
+adviseCRT = coerceCyc U.adviseCRT
+
+-- | Multiply by the special element @g@ of the @m@th cyclotomic.
+mulG :: (Fact m, CElt t r) => Cyc t m r -> Cyc t m r
+mulG = coerceCyc U.mulG
+
+-- | Divide by @g@, returning 'Nothing' if not evenly divisible.
+-- WARNING: this implementation is not a constant-time algorithm, so
+-- information about the argument may be leaked through a timing
+-- channel.
+divG :: (Fact m, CElt t r) => Cyc t m r -> Maybe (Cyc t m r)
+divG = coerceCyc U.divG
+
+-- | Sample from the "tweaked" Gaussian error distribution @t*D@ in
+-- the decoding basis, where @D@ has scaled variance @v@.  Note: This
+-- implementation uses Double precision to generate the Gaussian
+-- sample, which may not be sufficient for rigorous proof-based
+-- security.
+tGaussian :: (Fact m, OrdFloat q, Random q, CElt t q,
+              ToRational v, MonadRandom rnd)
+             => v -> rnd (Cyc t m q)
+tGaussian = (Cyc <$>) . U.tGaussian
+
+-- | Generate an LWE error term with given scaled variance,
+-- deterministically rounded in the decoding basis.
+errorRounded :: (ToInteger z, Fact m, CElt t z,
+                 ToRational v, MonadRandom rnd)
+                => v -> rnd (Cyc t m z)
+errorRounded = (Cyc <$>) . U.errorRounded
+
+-- | Generate an LWE error term with given scaled variance @* p^2@ over
+-- the given coset, deterministically rounded in the decoding basis.
+errorCoset ::
+  (Mod zp, z ~ ModRep zp, Lift zp z, Fact m,
+   CElt t zp, CElt t z, ToRational v, MonadRandom rnd)
+  => v -> Cyc t m zp -> rnd (Cyc t m z)
+errorCoset v = (Cyc <$>) . U.errorCoset v . unsafeUnCyc
+
+-- | Embed into the extension ring.
+embed :: (m `Divides` m', CElt t r) => Cyc t m r -> Cyc t m' r
+embed = coerceCyc U.embed
+
+-- | The "tweaked trace" (twace) function
+-- @Tw(x) = (mhat \/ m'hat) * Tr(g' \/ g * x)@,
+-- which fixes @R@ pointwise (i.e., @twace . embed == id@).
+twace :: (m `Divides` m', CElt t r) => Cyc t m' r -> Cyc t m r
+twace = coerceCyc U.twace
+
+-- | Return the given element's coefficient vector with respect to
+-- the (relative) powerful/decoding basis of the cyclotomic
+-- extension @O_m' / O_m@.
+coeffsCyc :: (m `Divides` m', CElt t r)
+             => Basis -> Cyc t m' r -> [Cyc t m r]
+coeffsCyc = coerceCyc U.coeffsCyc
+
+-- | The relative powerful basis of @O_m' / O_m@.
+powBasis :: (m `Divides` m', CElt t r) => Tagged m [Cyc t m' r]
+powBasis = coerceCyc U.powBasis
+
+-- | The relative mod-@r@ "CRT set" of the extension.
+crtSet :: (m `Divides` m', ZPP r, CElt t r, CElt t (ZPOf r))
+          => Tagged m [Cyc t m' r]
+crtSet = coerceCyc U.crtSet
+
+-- | Lift in the specified basis.
+liftCyc :: (Lift b a, Fact m, CElt t a, CElt t b)
+           => Basis -> Cyc t m b -> Cyc t m a
+liftCyc = coerceCyc U.liftCyc
+
+-- | Embed a scalar from the base ring as a cyclotomic element.
+scalarCyc :: (Fact m, CElt t a) => a -> Cyc t m a
+scalarCyc = Cyc . U.scalarCyc
+
+
diff --git a/src/Crypto/Lol/Cyclotomic/Linear.hs b/src/Crypto/Lol/Cyclotomic/Linear.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Linear.hs
@@ -0,0 +1,90 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts,
+             GeneralizedNewtypeDeriving, KindSignatures,
+             MultiParamTypeClasses, NoImplicitPrelude, RoleAnnotations,
+             ScopedTypeVariables, TypeFamilies, TypeOperators,
+             UndecidableInstances #-}
+
+-- | Functions from one cyclotomic ring to another that are linear
+-- over a common subring.
+
+module Crypto.Lol.Cyclotomic.Linear
+( Linear, ExtendLinIdx
+, linearDec, evalLin, extendLin
+) where
+
+import Crypto.Lol.Cyclotomic.Cyc
+import Crypto.Lol.LatticePrelude
+
+import Algebra.Additive as Additive (C)
+
+import Control.Applicative
+import Control.DeepSeq
+
+-- | An @E@-linear function from @R@ to @S@.
+newtype Linear t z (e::Factored) (r::Factored) (s::Factored) = D [Cyc t s z]
+  deriving (NFData)
+
+-- TODO: have constructor for both relative Pow basis of R/E?
+
+-- some params are phantom but matter for safety
+type role Linear representational nominal representational representational nominal
+
+-- | Construct an @E@-linear function given a list of its output values
+-- (in @S@) on the relative decoding basis of @R/E@.  The number of
+-- elements in the list must not exceed the size of the basis.
+linearDec :: forall t z e r s .
+             (e `Divides` r, e `Divides` s, CElt t z)
+             => [Cyc t s z] -> Linear t z e r s
+linearDec cs = let ps = proxy powBasis (Proxy::Proxy e) `asTypeOf` cs
+               in if length cs <= length ps then D (adviseCRT <$> cs)
+               else error $ "linearDec: too many entries: "
+                           ++ show (length cs) ++ " versus "
+                           ++ show (length ps)
+
+-- | Evaluates the given linear function on the input.
+evalLin :: forall t z e r s .
+           (e `Divides` r, e `Divides` s, CElt t z)
+           => Linear t z e r s -> Cyc t r z -> Cyc t s z
+evalLin (D cs) r = sum (zipWith (*) cs $
+                        embed <$> (coeffsCyc Dec r :: [Cyc t e z]))
+
+instance Additive (Cyc t s z) => Additive.C (Linear t z e r s) where
+  zero = D []
+
+  (D as) + (D bs) = D $ sumall as bs
+    where sumall [] ys = ys
+          sumall xs [] = xs
+          sumall (x:xs) (y:ys) = x+y : sumall xs ys
+
+  negate (D as) = D $ negate <$> as
+
+instance (Reduce z zq, Fact s, CElt t z, CElt t zq)
+         => Reduce (Linear t z e r s) (Linear t zq e r s) where
+  reduce (D cs) = D $ reduce <$> cs
+
+instance (CElt t zp, CElt t z, z ~ LiftOf zp, Lift zp z, Fact s)
+         => Lift' (Linear t zp e r s) where
+  type LiftOf (Linear t zp e r s) = Linear t (LiftOf zp) e r s
+
+  lift (D cs) = D $ liftCyc Dec <$> cs
+
+-- | A convenient constraint synonym for extending a linear function
+-- to larger rings.
+type ExtendLinIdx e r s e' r' s' =
+  (e ~ FGCD r e', r' ~ FLCM r e', -- these imply R'=R\otimes_E E'
+   e' ~ (e * (r' / r)), -- just to help GHC. This is implied by previous two constraints
+   e' `Divides` s', s `Divides` s', -- these imply lcm(s,e')|s' <==> (S+E') \subseteq S'
+   Fact r) -- need Fact r because nothing else gives it
+
+-- | Extend an @E@-linear function @R->S@ to an @E'@-linear function
+-- @R\'->S\'@.  (Mathematically, such extension only requires
+-- @lcm(r,e\') | r\'@ (not equality), but this generality would
+-- significantly complicate the implementation, and for our purposes
+-- there's no reason to use any larger @r'@.)
+extendLin :: (ExtendLinIdx e r s e' r' s', CElt t z)
+           => Linear t z e r s -> Linear t z e' r' s'
+-- CJP: this simple implementation works because R/E and R'/E' have
+-- identical decoding bases, because R' \cong R \otimes_E E'.  If we
+-- relax the constraint on E then we'd have to change the
+-- implementation to something more difficult.
+extendLin (D cs) = D (embed <$> cs)
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor.hs b/src/Crypto/Lol/Cyclotomic/Tensor.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor.hs
@@ -0,0 +1,399 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts,
+             NoImplicitPrelude, RankNTypes, ScopedTypeVariables, 
+             TupleSections, TypeFamilies, TypeOperators, 
+             UndecidableInstances #-}
+
+-- | Interface for cyclotomic tensors, and helper functions for tensor
+-- indexing.
+
+module Crypto.Lol.Cyclotomic.Tensor
+( Tensor(..)
+-- * Top-level CRT functions
+, hasCRTFuncs
+, scalarCRT, mulGCRT, divGCRT, crt, crtInv, twaceCRT, embedCRT
+-- * Tensor indexing
+, Matrix, indexM, twCRTs
+, zmsToIndexFact
+, indexInfo
+, extIndicesPowDec, extIndicesCRT, extIndicesCoeffs
+, baseIndicesPow, baseIndicesDec, baseIndicesCRT
+, digitRev
+)
+where
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.LatticePrelude as LP hiding (lift, (*>))
+import Crypto.Lol.Types.FiniteField
+
+import           Control.Applicative
+import           Control.DeepSeq
+import           Control.Monad.Random
+import           Data.Constraint
+import           Data.Singletons.Prelude hiding ((:-))
+import           Data.Traversable
+import           Data.Tuple           (swap)
+import qualified Data.Vector          as V
+import qualified Data.Vector.Unboxed  as U
+
+-- | 'Tensor' encapsulates all the core linear transformations needed
+-- for cyclotomic ring arithmetic.
+
+-- | The type @t m r@ represents a cyclotomic coefficient tensor of
+-- index @m@ over base ring @r@.  Most of the methods represent linear
+-- transforms corresponding to operations in particular bases.
+-- CRT-related methods are wrapped in 'Maybe' because they are
+-- well-defined only when a CRT basis exists over the ring @r@ for
+-- index @m@.
+
+-- | The superclass constraint is for convenience, to ensure that we
+-- can sample error tensors of 'Double's.
+
+class (TElt t Double, TElt t (Complex Double))
+      => Tensor (t :: Factored -> * -> *) where
+
+  type TElt t r :: Constraint
+
+  -- | Properties that hold for any index. Use with '\\'.
+  entailIndexT :: Tagged (t m r)
+                  (Fact m :- (Applicative (t m), Traversable (t m)))
+  
+  -- | Properties that hold for any (legal) fully-applied tensor. Use
+  -- with '\\'.
+  entailFullT :: Tagged (t m r)
+                 ((Fact m, TElt t r) :- 
+                  (Eq (t m r), ZeroTestable (t m r), Ring (t m r), 
+                   NFData (t m r), Random (t m r)))
+
+  -- | Converts a scalar to a tensor in the powerful basis
+  scalarPow :: (Fact m, TElt t r) => r -> t m r
+
+  -- | 'l' converts from decoding-basis representation to
+  -- powerful-basis representation; 'lInv' is its inverse.
+  l, lInv :: (Fact m, TElt t r) => t m r -> t m r
+
+  -- | Multiply by @g@ in the powerful/decoding basis
+  mulGPow, mulGDec :: (Fact m, TElt t r) => t m r -> t m r
+
+  -- | Divide by @g@ in the powerful/decoding basis.  The 'Maybe'
+  -- output indicates that the operation may fail, which happens
+  -- exactly when the input is not divisible by @g@.
+  divGPow, divGDec :: (Fact m, TElt t r) => t m r -> Maybe (t m r)
+
+  -- | A tuple of all the operations relating to the CRT basis, in a
+  -- single 'Maybe' value for safety.  Clients should typically not
+  -- use this method directly, but instead call the corresponding
+  -- top-level functions: the elements of the tuple correpond to the
+  -- functions 'scalarCRT', 'mulGCRT', 'divGCRT', 'crt', 'crtInv'.
+  crtFuncs :: (Fact m, TElt t r, CRTrans r) =>
+              Maybe (    r -> t m r, -- scalarCRT
+                     t m r -> t m r, -- mulGCRT
+                     t m r -> t m r, -- divGCRT
+                     t m r -> t m r, -- crt
+                     t m r -> t m r) -- crtInv
+
+  -- | Sample from the "skewed" Gaussian error distribution @t*D@
+  -- in the decoding basis, where @D@ has scaled variance @v@.
+  tGaussianDec :: (Fact m, OrdFloat q, Random q, TElt t q,
+                   ToRational v, MonadRandom rnd)
+                  => v -> rnd (t m q)
+
+  -- | The @twace@ linear transformation, which is the same in both the
+  -- powerful and decoding bases.
+  twacePowDec :: (m `Divides` m', TElt t r) => t m' r -> t m r
+
+  -- | The @embed@ linear transformations, for the powerful and
+  -- decoding bases.
+  embedPow, embedDec :: (m `Divides` m', TElt t r)
+                        => t m r -> t m' r
+
+  -- | A tuple of all the extension-related operations involving the
+  -- CRT bases, for safety.  Clients should typically not use this
+  -- method directly, but instead call the corresponding top-level
+  -- functions: the elements of the tuple correpond to the functions
+  -- 'twaceCRT', 'embedCRT'.
+  crtExtFuncs :: (m `Divides` m', TElt t r, CRTrans r) =>
+                 Maybe (t m' r -> t m  r, -- twaceCRT
+                        t m  r -> t m' r) -- embedCRT
+
+  -- | Map a tensor in the powerful\/decoding\/CRT basis, representing
+  -- an @O_m'@ element, to a vector of tensors representing @O_m@
+  -- elements in the same kind of basis.
+  coeffs :: (m `Divides` m', TElt t r) => t m' r -> [t m r]
+
+  -- | The powerful extension basis w.r.t. the powerful basis.
+  powBasisPow :: (m `Divides` m', TElt t r) => Tagged m [t m' r]
+
+  -- | A list of tensors representing the mod-@p@ CRT set of the
+  -- extension.
+  crtSetDec :: (m `Divides` m', PrimeField fp,
+                Coprime (PToF (CharOf fp)) m', TElt t fp)
+               => Tagged m [t m' fp]
+
+  -- | Potentially optimized version of 'fmap' when the input and
+  -- output element types satisfy 'TElt'.
+  fmapT :: (Fact m, TElt t a, TElt t b) => (a -> b) -> t m a -> t m b
+  -- | Potentially optimized monadic 'fmap'.
+  fmapTM :: (Monad mon, Fact m, TElt t a, TElt t b)
+             => (a -> mon b) -> t m a -> mon (t m b)
+
+-- | Convenience value indicating whether 'crtFuncs' exists.
+hasCRTFuncs :: forall t m r . (Tensor t, Fact m, TElt t r, CRTrans r)
+               => TaggedT (t m r) Maybe ()
+hasCRTFuncs = tagT $ do
+  (_ :: r -> t m r,_,_,_,_) <- crtFuncs
+  return ()
+
+-- | Yield a tensor for a scalar in the CRT basis.  (This function is
+-- simply an appropriate entry from 'crtFuncs'.)
+scalarCRT :: (Tensor t, Fact m, TElt t r, CRTrans r) => Maybe (r -> t m r)
+scalarCRT = (\(f,_,_,_,_) -> f) <$> crtFuncs
+
+
+mulGCRT, divGCRT, crt, crtInv :: (Tensor t, Fact m, TElt t r, CRTrans r)
+  => Maybe (t m r -> t m r)
+-- | Multiply by @g@ in the CRT basis. (This function is simply an
+-- appropriate entry from 'crtFuncs'.)
+mulGCRT = (\(_,f,_,_,_) -> f) <$> crtFuncs
+-- | Divide by @g@ in the CRT basis.  (This function is simply an
+-- appropriate entry from 'crtFuncs'.)
+divGCRT = (\(_,_,f,_,_) -> f) <$> crtFuncs
+-- | The CRT transform.  (This function is simply an appropriate entry
+-- from 'crtFuncs'.)
+crt = (\(_,_,_,f,_) -> f) <$> crtFuncs
+-- | The inverse CRT transform.  (This function is simply an
+-- appropriate entry from 'crtFuncs'.)
+crtInv = (\(_,_,_,_,f) -> f) <$> crtFuncs
+
+-- | The "tweaked trace" function for tensors in the CRT basis:
+-- For cyclotomic indices m | m',
+-- @Tw(x) = (mhat\/m\'hat) * Tr(g\'\/g * x)@.
+-- (This function is simply an appropriate entry from 'crtExtFuncs'.)
+twaceCRT :: forall t r m m' . (Tensor t, m `Divides` m', TElt t r, CRTrans r)
+            => Maybe (t m' r -> t m r)
+twaceCRT = proxyT hasCRTFuncs (Proxy::Proxy (t m' r)) *>
+           proxyT hasCRTFuncs (Proxy::Proxy (t m  r)) *>
+           (fst <$> crtExtFuncs)
+
+
+-- | Embed a tensor with index @m@ in the CRT basis to a tensor with
+-- index @m'@ in the CRT basis.
+-- (This function is simply an appropriate entry from 'crtExtFuncs'.)
+embedCRT :: forall t r m m' . (Tensor t, m `Divides` m', TElt t r, CRTrans r)
+            => Maybe (t m r -> t m' r)
+embedCRT = proxyT hasCRTFuncs (Proxy::Proxy (t m' r)) *>
+           proxyT hasCRTFuncs (Proxy::Proxy (t m  r)) *>
+           (snd <$> crtExtFuncs)
+
+fMatrix :: forall m r mon . (Fact m, Monad mon, Ring r)
+           => (forall pp . (PPow pp) => TaggedT pp mon (MatrixC r))
+           -> TaggedT m mon (Matrix r)
+fMatrix mat = tagT $ go $ sUnF (sing :: SFactored m)
+  where go :: Sing (pplist :: [PrimePower]) -> mon (Matrix r)
+        go spps = case spps of
+          SNil -> return MNil
+          (SCons spp rest) -> do
+            rest' <- go rest
+            mat' <- withWitnessT mat spp
+            return $ MKron rest' mat'
+
+-- deeply embedded DSL for Kronecker products of matrices
+
+data MatrixC r = 
+  MC (Int -> Int -> r)           -- yields element i,j
+  Int Int                        -- dims
+
+-- | A Kronecker product of zero of more matrices over @r@.
+data Matrix r = MNil | MKron (Matrix r) (MatrixC r)
+
+-- | Extract the @(i,j)@ element of a 'Matrix'.
+indexM :: Ring r => Matrix r -> Int -> Int -> r
+indexM MNil 0 0 = LP.one
+indexM (MKron m (MC mc r c)) i j =
+  let (iq,ir) = i `divMod` r
+      (jq,jr) = j `divMod` c
+      in indexM m iq jq * mc ir jr
+
+-- | The "tweaked" CRT^* matrix: @CRT^* . diag(sigma(g_m))@.
+twCRTs :: (Fact m, CRTrans r) => TaggedT m Maybe (Matrix r)
+twCRTs = fMatrix twCRTsPPow
+
+-- | The "tweaked" CRT^* matrix (for prime powers): @CRT^* * diag(sigma(g_p))@.
+twCRTsPPow :: (PPow pp, CRTrans r) => TaggedT pp Maybe (MatrixC r)
+twCRTsPPow = do
+  phi    <- pureT totientPPow
+  iToZms <- pureT indexToZmsPPow
+  jToPow <- pureT indexToPowPPow
+  (wPow, _) <- crtInfoPPow
+  gEmb <- gEmbPPow
+
+  return $ MC (\j i -> let i' = iToZms i
+                       in wPow (jToPow j * negate i') * gEmb i') phi phi
+
+-- Reindexing functions
+
+-- | Base-p digit reversal; input and output are in @[p^e]@.
+digitRev :: PP -> Int -> Int
+digitRev (_,0) 0 = 0
+-- CJP: use accumulator to avoid multiple exponentiations?
+digitRev (p,e) j 
+  | e >= 1 = let (q,r) = j `divMod` p
+             in r * (p^(e-1)) + digitRev (p,e-1) q
+
+indexToPowPPow, indexToZmsPPow :: PPow pp => Tagged pp (Int -> Int)
+indexToPowPPow = indexToPow <$> ppPPow
+indexToZmsPPow = indexToZms <$> ppPPow
+
+-- | Convert a @Z_m^*@ index to a linear tensor index in @[m]@.
+zmsToIndexFact :: Fact m => Tagged m (Int -> Int)
+zmsToIndexFact = zmsToIndex <$> ppsFact
+
+-- | For a prime power @p^e@, map a tensor index to the corresponding
+-- power j in @[phi(p^e)]@, as in the powerful basis.
+indexToPow :: PP -> Int -> Int
+-- CJP: use accumulator to avoid multiple exponentiations?
+indexToPow (p,e) j = let (jq,jr) = j `divMod` (p-1)
+                     in p^(e-1)*jr + digitRev (p,e-1) jq
+
+-- | For a prime power @p^e@, map a tensor index to the corresponding
+-- element i in @Z_{p^e}^*@.
+indexToZms :: PP -> Int -> Int
+indexToZms (p,_) i = let (i1,i0) = i `divMod` (p-1)
+                       in p*i1 + i0 + 1 
+
+-- | Convert a Z_m^* index to a linear tensor index.
+zmsToIndex :: [PP] -> Int -> Int
+zmsToIndex [] _ = 0
+zmsToIndex (pp:rest) i = zmsToIndexPP pp (i `mod` valuePP pp)
+                         + (totientPP pp) * zmsToIndex rest i
+
+-- | Inverse of 'indexToZms'.
+zmsToIndexPP :: PP -> Int -> Int
+zmsToIndexPP (p,_) i = let (i1,i0) = i `divMod` p
+                       in (p-1)*i1 + i0 - 1
+
+-- Index correspondences for ring extensions
+
+-- | Correspondences between the linear indexes into a basis of O_m',
+-- and pair indices into (extension basis) \otimes (basis of O_m).
+-- The work the same for Pow,Dec,CRT bases because all these bases
+-- have that factorization.  The first argument is the list of
+-- @(phi(m),phi(m'))@ pairs for the (merged) prime powers of @m@,@m'@.
+toIndexPair :: [(Int,Int)] -> Int -> (Int,Int)
+fromIndexPair :: [(Int,Int)] -> (Int,Int) -> Int
+
+toIndexPair [] 0 = (0,0)
+toIndexPair ((phi,phi'):rest) i' =
+  let (i'q,i'r) = i' `divMod` phi'
+      (i'rq,i'rr) = i'r `divMod` phi
+      (i'q1,i'q0) = toIndexPair rest i'q
+  in (i'rq + i'q1*(phi' `div` phi), i'rr + i'q0*phi)
+
+fromIndexPair [] (0,0) = 0
+fromIndexPair ((phi,phi'):rest) (i1,i0) =
+  let (i0q,i0r) = i0 `divMod` phi
+      (i1q,i1r) = i1 `divMod` (phi' `div` phi)
+      i = fromIndexPair rest (i1q,i0q)
+  in (i0r + i1r*phi) + i*phi'
+
+-- | A collection of useful information for working with tensor
+-- extensions.  The first component is a list of triples @(p,e,e')@
+-- where @e@, @e'@ are respectively the exponents of prime @p@ in @m@,
+-- @m'@.  The next two components are @phi(m)@ and @phi(m')@.  The
+-- final component is a pair @(phi(p^e), phi(p^e'))@ for each triple
+-- in the first component.
+indexInfo :: forall m m' . (m `Divides` m')
+             => Tagged '(m, m') ([(Int,Int,Int)], Int, Int, [(Int,Int)])
+indexInfo = let pps = proxy ppsFact (Proxy::Proxy m)
+                pps' = proxy ppsFact (Proxy::Proxy m')
+                mpps = mergePPs pps pps'
+                phi = totientPPs pps
+                phi' = totientPPs pps'
+                tots = totients mpps
+            in tag (mpps, phi, phi', tots)
+
+-- | A vector of @phi(m)@ entries, where the @i@th entry is the index
+-- into the powerful\/decoding basis of @O_m'@ of the
+-- @i@th entry of the powerful\/decoding basis of @O_m@.
+extIndicesPowDec :: (m `Divides` m') => Tagged '(m, m') (U.Vector Int)
+extIndicesPowDec = do
+  (_, phi, _, tots) <- indexInfo
+  return $ U.generate phi (fromIndexPair tots . (0,))
+
+-- | A vector of @phi(m)@ blocks of @phi(m')\/phi(m)@ consecutive
+-- entries. Each block contains all those indices into the CRT basis
+-- of @O_m'@ that "lie above" the corresponding index into the CRT
+-- basis of @O_m@.
+extIndicesCRT :: forall m m' . (m `Divides` m')
+                 => Tagged '(m, m') (U.Vector Int)
+extIndicesCRT = do
+  (_, phi, phi', tots) <- indexInfo
+  return $ U.generate phi'
+           (fromIndexPair tots . swap . (`divMod` (phi' `div` phi)))
+
+baseWrapper :: forall m m' a . (m `Divides` m', U.Unbox a)
+               => ([(Int,Int,Int)] -> Int -> a)
+               -> Tagged '(m, m') (U.Vector a)
+baseWrapper f = do
+  (mpps, _, phi', _) <- indexInfo
+  return $ U.generate phi' (f mpps)
+
+-- | A lookup table for 'toIndexPair' applied to indices @[phi(m')]@.
+baseIndicesPow :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (U.Vector (Int,Int))
+-- | A lookup table for 'baseIndexDec' applied to indices @[phi(m')]@.
+baseIndicesDec :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (U.Vector (Maybe (Int,Bool)))
+
+-- | Same as 'baseIndicesPow', but only includes the second component
+-- of each pair.
+baseIndicesCRT :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (U.Vector Int)
+
+baseIndicesPow = baseWrapper (toIndexPair . totients)
+
+-- this one is more complicated; requires the prime powers
+baseIndicesDec = baseWrapper baseIndexDec
+
+baseIndicesCRT =
+  baseWrapper (\pps -> snd . toIndexPair (totients pps))
+
+
+-- | The @i0@th entry of the @i1@th vector is 'fromIndexPair' @(i1,i0)@.
+extIndicesCoeffs :: forall m m' . (m `Divides` m')
+                    => Tagged '(m, m') (V.Vector (U.Vector Int))
+extIndicesCoeffs = do
+  (_, phi, phi', tots) <- indexInfo
+  return $ V.generate (phi' `div` phi)
+           (\i1 -> U.generate phi (\i0 -> fromIndexPair tots (i1,i0)))
+
+-- | Convenient reindexing functions
+
+-- | Maps an index of the extension ring array to its corresponding
+-- index in the base ring array (if it exists), with sign, under the
+-- decoding basis.
+baseIndexDec :: [(Int,Int,Int)] -> Int -> Maybe (Int, Bool)
+baseIndexDec [] 0 = Just (0,False)
+baseIndexDec ((p,e,e'):rest) i'
+   = let (i'q, i'r) = i' `divMod` totientPP (p,e')
+         phi = totientPP (p,e)
+         curr
+           | p>2 && e==0 && e' > 0 = case i'r of
+               0 -> Just (0,False)
+               1 -> Just (0,True)
+               _ -> Nothing
+           | otherwise = if i'r < phi then Just (i'r,False) else Nothing
+     in do
+       (i,b) <- curr
+       (j,b') <- baseIndexDec rest i'q
+       return (i + phi*j, b /= b')
+
+-- the first list of pps must "divide" the other.  result is a list of
+-- all (prime, min e, max e).
+mergePPs :: [PP] -> [PP] -> [(Int,Int,Int)]
+mergePPs [] pps = LP.map (\(p,e) -> (p,0,e)) pps
+mergePPs allpps@((p,e):pps) ((p',e'):pps')
+  | p == p' && e <= e' = (p,  e, e') : mergePPs pps pps'
+  | p > p'  = (p', 0, e') : mergePPs allpps pps'
+
+totients :: [(Int, Int, Int)] -> [(Int,Int)]
+totients = LP.map (\(p,e,e') -> (totientPP (p,e), totientPP (p,e')))
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor.hs b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor.hs
@@ -0,0 +1,723 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable, GADTs,
+             FlexibleContexts, FlexibleInstances, TypeOperators, PolyKinds,
+             GeneralizedNewtypeDeriving, InstanceSigs, RoleAnnotations,
+             MultiParamTypeClasses, NoImplicitPrelude, StandaloneDeriving,
+             ScopedTypeVariables, TupleSections, TypeFamilies, RankNTypes,
+             TypeSynonymInstances, UndecidableInstances,
+             RebindableSyntax #-}
+
+-- | Wrapper for a C implementation of the 'Tensor' interface.
+
+module Crypto.Lol.Cyclotomic.Tensor.CTensor
+( CT
+-- Exports below here are due solely to ticket #10338. See CycTests for more details
+, CRNS
+, Dispatch
+) where
+
+import Algebra.Additive as Additive (C)
+import Algebra.Ring     as Ring (C)
+
+import Control.Applicative
+import Control.DeepSeq
+import Control.Monad
+import Control.Monad.Identity
+import Control.Monad.Random
+import Control.Monad.Trans (lift)
+
+import Data.Coerce
+import Data.Constraint
+import Data.Foldable as F
+import Data.Int
+import Data.Maybe
+import Data.Traversable as T
+import Data.Typeable
+import Data.Vector.Generic           as V (zip, unzip)
+import Data.Vector.Storable          as SV (Vector, replicate, replicateM, thaw, convert, foldl',
+                                            unsafeToForeignPtr0, unsafeSlice, mapM, fromList,
+                                            generate, foldl1',
+                                            unsafeWith, zipWith, map, length, unsafeFreeze, thaw)
+import Data.Vector.Storable.Internal (getPtr)
+import Data.Vector.Storable.Mutable  as SM hiding (replicate)
+
+import           Foreign.ForeignPtr
+import           Foreign.Marshal.Array
+import           Foreign.Ptr
+import           Foreign.Storable        (Storable (..))
+import qualified Foreign.Storable.Record as Store
+import           Foreign.Storable.Tuple  ()
+import           System.IO.Unsafe
+import           Test.QuickCheck         hiding (generate)
+import           Unsafe.Coerce
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.LatticePrelude as LP hiding (replicate, unzip, zip, lift)
+import Crypto.Lol.Reflects
+import Crypto.Lol.Cyclotomic.Tensor
+
+import Crypto.Lol.Types.IZipVector
+import Crypto.Lol.Types.ZqBasic
+import Crypto.Lol.GaussRandom
+
+import Crypto.Lol.Cyclotomic.Tensor.CTensor.Extension
+
+import Algebra.ZeroTestable   as ZeroTestable (C)
+
+
+-- | Newtype wrapper around a Vector.
+newtype CT' (m :: Factored) r = CT' { unCT :: Vector r } 
+                              deriving (Show, Eq, NFData, Typeable)
+
+-- the first argument, though phantom, affects representation
+type role CT' representational nominal
+
+-- GADT wrapper that distinguishes between Unbox and unrestricted
+-- element types
+
+-- | An implementation of 'Tensor' backed by C code.
+data CT (m :: Factored) r where 
+  CT :: Storable r => CT' m r -> CT m r
+  ZV :: IZipVector m r -> CT m r
+  deriving (Typeable)
+
+instance Eq r => Eq (CT m r) where
+  (ZV x) == (ZV y) = x == y
+  (CT x) == (CT y) = x == y
+  x@(CT _) == y = x == toCT y
+  y == x@(CT _) = x == toCT y
+
+deriving instance Show r => Show (CT m r)
+
+toCT :: (Storable r) => CT m r -> CT m r
+toCT v@(CT _) = v
+toCT (ZV v) = CT $ zvToCT' v
+
+toZV :: (Fact m) => CT m r -> CT m r
+toZV (CT (CT' v)) = ZV $ fromMaybe (error "toZV: internal error") $
+                    iZipVector $ convert v
+toZV v@(ZV _) = v
+
+zvToCT' :: forall m r . (Storable r) => IZipVector m r -> CT' m r
+zvToCT' v = coerce $ (convert $ unIZipVector v :: Vector r)
+
+wrap :: (Storable r) => (CT' l r -> CT' m r) -> (CT l r -> CT m r)
+wrap f (CT v) = CT $ f v
+wrap f (ZV v) = CT $ f $ zvToCT' v
+
+wrapM :: (Storable r, Monad mon) => (CT' l r -> mon (CT' m r))
+         -> (CT l r -> mon (CT m r))
+wrapM f (CT v) = liftM CT $ f v
+wrapM f (ZV v) = liftM CT $ f $ zvToCT' v
+
+-- convert an CT' *twace* signature to Tagged one
+type family Tw (r :: *) :: * where
+  Tw (CT' m' r -> CT' m r) = Tagged '(m,m') (Vector r -> Vector r)
+  Tw (Maybe (CT' m' r -> CT' m r)) = TaggedT '(m,m') Maybe (Vector r -> Vector r)
+
+type family Em r where
+  Em (CT' m r -> CT' m' r) = Tagged '(m,m') (Vector r -> Vector r)
+  Em (Maybe (CT' m r -> CT' m' r)) = TaggedT '(m,m') Maybe (Vector r -> Vector r)
+
+
+---------- NUMERIC PRELUDE INSTANCES ----------
+instance (Additive r, Storable r, CRNS r, Fact m)
+  => Additive.C (CT m r) where
+  (CT a@(CT' _)) + (CT b@(CT' _)) = CT $ (zipWrapper $ untag $ cZipDispatch dadd) a b  --pack $ SV.zipWith (+) (unpack a) (unpack b) -- Vector code --
+  a + b = (toCT a) + (toCT b)
+  negate (CT (CT' a)) = CT $ CT' $ SV.map negate a -- EAC: This probably should be converted to C code
+  negate a = negate $ toCT a
+
+  zero = CT $ repl zero
+
+instance (Fact m, Ring r, Storable r, CRNS r)
+  => Ring.C (CT m r) where
+  (CT a@(CT' _)) * (CT b@(CT' _)) = CT $ (zipWrapper $ untag $ cZipDispatch dmul) a b  --pack $ SV.zipWith (*) (unpack a) (unpack b) -- Vector code --
+  a * b = (toCT a) * (toCT b)
+
+  fromInteger = CT . repl . fromInteger
+
+instance (ZeroTestable r, Storable r, Fact m)
+         => ZeroTestable.C (CT m r) where
+  --{-# INLINABLE isZero #-} 
+  isZero (CT (CT' a)) = SV.foldl' (\ b x -> b && isZero x) True a
+  isZero (ZV v) = isZero v
+
+---------- "Container" instances ----------
+
+instance Fact m => Functor (CT m) where
+  -- Functor instance is implied by Applicative laws
+  fmap f x = pure f <*> x
+
+instance Fact m => Applicative (CT m) where
+  pure = ZV . pure
+
+  (ZV f) <*> (ZV a) = ZV (f <*> a)
+  f@(ZV _) <*> v@(CT _) = f <*> toZV v
+
+instance Fact m => Foldable (CT m) where
+  -- Foldable instance is implied by Traversable
+  foldMap = foldMapDefault
+
+instance Fact m => Traversable (CT m) where
+  traverse f r@(CT _) = T.traverse f $ toZV r
+  traverse f (ZV v) = ZV <$> T.traverse f v
+
+instance Tensor CT where
+
+  type TElt CT r = (IntegralDomain r, ZeroTestable r, 
+                    Eq r, Random r, NFData r,
+                    Storable r, CRNS r)
+
+  entailIndexT = tag $ Sub Dict
+  entailFullT = tag $ Sub Dict
+
+  scalarPow = CT . scalarPow' -- Vector code
+
+  l = wrap $ lgWrapper $ untag $ lgDispatch dl
+  lInv = wrap $ lgWrapper $ untag $ lgDispatch dlinv
+
+  mulGPow = wrap mulGPow' -- mulGPow' already has lgWrapper
+  mulGDec = wrap $ lgWrapper $ untag $ lgDispatch dmulgdec
+
+  divGPow = wrapM $ divGPow'
+  -- we divide by p in the C code (for divGDec only(?)), do NOT call checkDiv!
+  divGDec = wrapM $ divGWrapper $ Just . (untag $ lgDispatch dginvdec)
+
+  crtFuncs = (,,,,) <$>
+    Just (CT . repl) <*>
+    (liftM wrap $ crtWrapper $ (untag $ cZipDispatch dmul) <$> untagT gCoeffsCRT) <*>
+    (liftM wrap $ crtWrapper $ (untag $ cZipDispatch dmul) <$> untagT gInvCoeffsCRT) <*>
+    (liftM wrap $ untagT $ crt') <*>
+    (liftM wrap $ crtWrapper $ untagT ctCRTInv) 
+
+  twacePowDec = wrap $ runIdentity $ coerceTw twacePowDec'
+  embedPow = wrap $ runIdentity $ coerceEm embedPow'
+  embedDec = wrap $ runIdentity $ coerceEm embedDec'
+
+  tGaussianDec v = liftM CT $ gaussWrapper $ cDispatchGaussian v
+  --tGaussianDec v = liftM CT $ coerceT' $ gaussianDec v
+
+  crtExtFuncs = (,) <$> (liftM wrap $ coerceTw twaceCRT')
+                    <*> (liftM wrap $ coerceEm embedCRT')
+
+  coeffs = wrapM $ coerceCoeffs $ coeffs'
+
+  powBasisPow = (CT <$>) <$> coerceBasis powBasisPow'
+
+  crtSetDec = (CT <$>) <$> coerceBasis crtSetDec'
+
+  fmapT f (CT v) = CT $ coerce (SV.map f) v
+  fmapT f v@(ZV _) = fmapT f $ toCT v
+
+  fmapTM f (CT (CT' arr)) = liftM (CT . CT') $ SV.mapM f arr
+  fmapTM f v@(ZV _) = fmapTM f $ toCT v
+
+coerceTw :: (Functor mon) => (TaggedT '(m, m') mon (Vector r -> Vector r)) -> mon (CT' m' r -> CT' m r)
+coerceTw = (coerce <$>) . untagT
+
+coerceEm :: (Functor mon) => (TaggedT '(m, m') mon (Vector r -> Vector r)) -> mon (CT' m r -> CT' m' r)
+coerceEm = (coerce <$>) . untagT
+
+-- | Useful coersion for defining @coeffs@ in the @Tensor@
+-- interface. Using 'coerce' alone is insufficient for type inference.
+coerceCoeffs :: (Fact m, Fact m') 
+  => Tagged '(m,m') (Vector r -> [Vector r]) -> CT' m' r -> [CT' m r]
+coerceCoeffs = coerce
+
+-- | Useful coersion for defining @powBasisPow@ and @crtSetDec@ in the @Tensor@
+-- interface. Using 'coerce' alone is insufficient for type inference.
+coerceBasis :: 
+  (Fact m, Fact m')
+  => Tagged '(m,m') ([Vector r]) -> Tagged m [CT' m' r]
+coerceBasis = coerce
+
+-- | Class to dispatch to the C backend for various element types.
+class CRNS r where
+
+  zipWrapper :: (Fact m) => 
+    (forall a . (TElt CT a, Dispatch a) => CT' m a -> CT' m a -> CT' m a)
+    -> CT' m r -> CT' m r -> CT' m r
+
+  crtWrapper :: (Fact m, CRTrans r) => 
+    (forall a . (TElt CT a, CRTrans a, Dispatch a) => Maybe (CT' m a -> CT' m a))
+    -> Maybe (CT' m r -> CT' m r)
+
+  lgWrapper :: (Fact m) => 
+    (forall a . (TElt CT a, Dispatch a) => CT' m a -> CT' m a)
+    -> CT' m r -> CT' m r
+
+  divGWrapper :: (Fact m) => 
+    (forall a . (TElt CT a, Dispatch a) => CT' m a -> Maybe (CT' m a))
+    -> CT' m r -> Maybe (CT' m r)
+
+  gaussWrapper :: (Fact m, MonadRandom rnd) => 
+    (forall a . (TElt CT a, Dispatch a, OrdFloat a, MonadRandom rnd) => rnd (CT' m a))
+    -> rnd (CT' m r)
+
+instance CRNS Double where
+  zipWrapper f = f
+  crtWrapper f = f
+  lgWrapper f = f
+  divGWrapper f = f
+  gaussWrapper f = f
+
+instance CRNS Int64 where
+  zipWrapper f = f
+  crtWrapper f = f
+  lgWrapper f = f
+  divGWrapper f = f
+  gaussWrapper = error "Cannot call gaussianDec for Int64"
+
+instance (TElt CT (Complex a), Dispatch (Complex a)) => CRNS (Complex a) where
+  zipWrapper f = f
+  crtWrapper f = f
+  lgWrapper f = f
+  divGWrapper f = f
+  gaussWrapper = error "Cannot call gaussianDec for Complex"
+
+-- EAC: we need PolyKinds in paritcular for this instance
+instance (TElt CT (ZqBasic q i), Dispatch (ZqBasic q i)) => CRNS (ZqBasic q i) where
+  zipWrapper f = f
+  crtWrapper f = f
+  lgWrapper f = f
+  divGWrapper f = f
+  gaussWrapper = error "Cannot call gaussianDec for ZqBasic"
+
+instance (Storable a, Storable b, CRNS a, CRNS b, CRTrans a, CRTrans b) 
+  => CRNS (a,b) where
+  zipWrapper f (CT' x :: CT' m (a,b)) (CT' y) =
+    let (a,b) = unzip x
+        (c,d) = unzip y
+        (CT' ac) = zipWrapper f (CT' a :: CT' m a) (CT' c)
+        (CT' bd) = zipWrapper f (CT' b :: CT' m b) (CT' d)
+    in CT' $ zip ac bd
+
+  crtWrapper f = do
+    fa <- crtWrapper f
+    fb <- crtWrapper f
+    return $ \ (CT' x :: CT' m (a,b)) -> 
+      let (a,b) = unzip x
+          (CT' a') = fa (CT' a :: CT' m a)
+          (CT' b') = fb (CT' b :: CT' m b)
+      in CT' $ zip a' b'
+
+  lgWrapper f (CT' x :: CT' m (a,b)) = 
+    let (a, b) = unzip x
+        (CT' a') = lgWrapper f (CT' a :: CT' m a)
+        (CT' b') = lgWrapper f (CT' b :: CT' m b)
+    in CT' $ zip a' b'
+
+  divGWrapper f (CT' x :: CT' m (a,b)) = 
+    let (a, b) = unzip x
+    in do -- in Maybe
+      (CT' a') <- divGWrapper f (CT' a :: CT' m a)
+      (CT' b') <- divGWrapper f (CT' b :: CT' m b)
+      return $ CT' $ zip a' b'
+
+  gaussWrapper f = do
+    (CT' a) <- gaussWrapper f
+    (CT' b) <- gaussWrapper f
+    return $ CT' $ zip a b
+
+mulGPow' :: (TElt CT r, Fact m) => CT' m r -> CT' m r
+mulGPow' = lgWrapper $ untag $ lgDispatch dmulgpow
+
+divGPow' :: forall m r . (TElt CT r, Fact m) => CT' m r -> Maybe (CT' m r)
+divGPow' = divGWrapper $ untag $ checkDiv $ lgDispatch dginvpow
+
+crt' :: forall m r . (TElt CT r, Fact m, CRTrans r) 
+  => TaggedT m Maybe (CT' m r -> CT' m r)
+crt' = tagT $ crtWrapper $ do
+  f <- proxyT ctCRT (Proxy::Proxy m)
+  return $ CT' . f . unCT
+
+--{-# INLINE lgDispatch #-}
+lgDispatch :: forall m r .
+     (Storable r, Fact m, Additive r)
+      => (Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ())
+         -> Tagged m (CT' m r -> CT' m r)
+lgDispatch f = do
+  factors <- liftM marshalFactors ppsFact
+  totm <- liftM fromIntegral totientFact
+  let numFacts = fromIntegral $ SV.length factors
+  return $ coerce $ \yin -> unsafePerformIO $ do -- in IO
+    yout <- SV.thaw yin
+    SM.unsafeWith yout (\pout ->
+      SV.unsafeWith factors (\pfac ->
+        f pout totm pfac numFacts))
+    unsafeFreeze yout
+
+--{-# INLINE ctCRT #-}
+ctCRT :: forall m r .
+         (Storable r, CRTrans r, Dispatch r,
+          Fact m)
+         => TaggedT m Maybe (Vector r -> Vector r)
+ctCRT = do -- in TaggedT m Maybe
+  ru' <- ru
+  factors <- pureT $ liftM marshalFactors ppsFact
+  totm <- pureT $ liftM fromIntegral totientFact
+  let numFacts = fromIntegral $ SV.length factors
+  return $ \yin -> unsafePerformIO $ do -- in IO
+    yout <- SV.thaw yin
+    SM.unsafeWith yout (\pout ->
+      SV.unsafeWith factors (\pfac ->
+        withPtrArray ru' (\ruptr ->
+          dcrt pout totm pfac numFacts ruptr)))
+    unsafeFreeze yout
+
+-- CTensor CRT^(-1) functions take inverse rus
+--{-# INLINE ctCRTInv #-}
+ctCRTInv :: (Storable r, CRTrans r, Dispatch r,
+          Fact m)
+         => TaggedT m Maybe (CT' m r -> CT' m r)
+ctCRTInv = do -- in Maybe
+  mhatInv <- liftM snd $ crtInfoFact
+  ruinv' <- ruInv
+  factors <- pureT $ liftM marshalFactors ppsFact
+  totm <- pureT $ liftM fromIntegral totientFact
+  let numFacts = fromIntegral $ SV.length factors
+  -- EAC: can't use coerce here?
+  return $ \(CT' yin) -> unsafePerformIO $ do
+    yout <- SV.thaw yin
+    SM.unsafeWith yout (\pout ->
+      SV.unsafeWith factors (\pfac ->
+        withPtrArray ruinv' (\ruptr ->
+          dcrtinv pout totm pfac numFacts ruptr mhatInv)))
+    CT' <$> unsafeFreeze yout
+
+checkDiv :: forall m r . 
+  (IntegralDomain r, Storable r, ZeroTestable r, 
+   Fact m)
+    => Tagged m (CT' m r -> CT' m r) -> Tagged m (CT' m r -> Maybe (CT' m r))
+checkDiv f = do
+  f' <- f
+  oddRad' <- liftM fromIntegral oddRadicalFact
+  return $ \x -> 
+    let (CT' y) = f' x
+    in CT' <$> (SV.mapM (`divIfDivis` oddRad')) y
+
+divIfDivis :: (IntegralDomain r, ZeroTestable r) => r -> r -> Maybe r
+divIfDivis num den = let (q,r) = num `divMod` den
+                     in if isZero r then Just q else Nothing
+
+cZipDispatch :: (Storable r, Fact m, Additive r)
+  => (Ptr r -> Ptr r -> Int64 -> IO ())
+     -> Tagged m (CT' m r -> CT' m r -> CT' m r)
+cZipDispatch f = do -- in Tagged m
+  totm <- liftM fromIntegral $ totientFact
+  return $ coerce $ \a b -> unsafePerformIO $ do
+    yout <- SV.thaw a
+    SM.unsafeWith yout (\pout ->
+      SV.unsafeWith b (\pin ->
+        f pout pin totm))
+    unsafeFreeze yout
+
+cDispatchGaussian :: forall m r var rnd .
+         (Storable r, Transcendental r, Dispatch r, Ord r,
+          Fact m, ToRational var, Random r, MonadRandom rnd)
+         => var -> rnd (CT' m r)
+cDispatchGaussian var = liftM CT' $ flip proxyT (Proxy::Proxy m) $ do -- in TaggedT m rnd
+  -- get rus for (Complex r)
+  ruinv' <- mapTaggedT (return . fromMaybe (error "complexGaussianRoots")) $ ruInv
+  factors <- liftM marshalFactors $ pureT ppsFact
+  totm <- pureT totientFact
+  m <- pureT valueFact
+  rad <- pureT radicalFact
+  yin <- lift $ realGaussians (var * fromIntegral (m `div` rad)) totm
+  let numFacts = fromIntegral $ SV.length factors
+  return $ unsafePerformIO $ do -- in IO
+    --let yin = create $ SM.new totm :: Vector r -- contents will be overwritten, so no need to initialize
+    yout <- SV.thaw yin
+    SM.unsafeWith yout (\pout ->
+      SV.unsafeWith factors (\pfac ->
+       withPtrArray ruinv' (\ruptr ->
+        dgaussdec pout (fromIntegral totm) pfac numFacts ruptr)))
+    unsafeFreeze yout
+
+instance (Arbitrary r, Fact m, Storable r) => Arbitrary (CT' m r) where
+  arbitrary = replM arbitrary
+  shrink = shrinkNothing
+
+instance (Storable r, Arbitrary (CT' m r)) => Arbitrary (CT m r) where
+  arbitrary = CT <$> arbitrary
+
+instance (Storable r, Random r, Fact m) => Random (CT' m r) where
+  --{-# INLINABLE random #-}
+  random = runRand $ replM (liftRand random)
+
+  randomR = error "randomR nonsensical for CT'"
+
+instance (Storable r, Random (CT' m r)) => Random (CT m r) where
+  --{-# INLINABLE random #-}
+  random = runRand $ liftM CT (liftRand random)
+
+  randomR = error "randomR nonsensical for CT"
+
+instance (NFData r) => NFData (CT m r) where
+  rnf (CT v) = rnf v
+  rnf (ZV v) = rnf v
+
+repl :: forall m r . (Fact m, Storable r) => r -> CT' m r
+repl = let n = proxy totientFact (Proxy::Proxy m)
+       in coerce . SV.replicate n
+
+replM :: forall m r mon . (Fact m, Storable r, Monad mon) 
+         => mon r -> mon (CT' m r)
+replM = let n = proxy totientFact (Proxy::Proxy m)
+        in liftM coerce . SV.replicateM n
+
+--{-# INLINE scalarPow' #-}
+scalarPow' :: forall t m r v .
+  (Fact m, Additive r, Storable r)
+  => r -> CT' m r
+-- constant-term coefficient is first entry wrt powerful basis
+scalarPow' = 
+  let n = proxy totientFact (Proxy::Proxy m)
+  in \r -> CT' $ generate n (\i -> if i == 0 then r else zero)
+
+ru, ruInv :: forall r m . 
+   (CRTrans r, Fact m, Storable r)
+   => TaggedT m Maybe [Vector r]
+--{-# INLINE ru #-}
+ru = do
+  mval <- pureT valueFact
+  wPow <- liftM fst $ crtInfoFact
+  liftM (LP.map
+    (\(p,e) -> do
+        let pp = p^e
+            pow = mval `div` pp
+        generate pp (wPow . (*pow)))) $
+      pureT ppsFact
+
+--{-# INLINE ruInv #-}
+ruInv = do
+  mval <- pureT valueFact
+  wPow <- liftM fst $ crtInfoFact
+  liftM (LP.map
+    (\(p,e) -> do
+        let pp = p^e
+            pow = mval `div` pp
+        generate pp (\i -> wPow $ (-i*pow)))) $
+      pureT ppsFact
+
+gCoeffsCRT, gInvCoeffsCRT :: (TElt CT r, CRTrans r, Fact m)
+  => TaggedT m Maybe (CT' m r)
+gCoeffsCRT = crt' <*> (return $ mulGPow' $ scalarPow' LP.one)
+-- It's necessary to call 'fromJust' here: otherwise 
+-- sequencing functions in 'crtFuncs' relies on 'divGPow' having an
+-- implementation in C, which is not true for all types which have a C
+-- implementation of, e.g. 'crt'. In particular, 'Complex Double' has C support
+-- for 'crt', but not for 'divGPow'.
+-- This really breaks the contract of Tensor, so it's probably a bad idea.
+--   Someone can get the "crt" and can even pull the function "divGCRT" from Tensor,
+--   but it will fail when they try to apply it.
+-- As an implementation note if I ever do fix this: the division by rad(m) can be
+-- tricky for Double/Complex Doubles, so be careful! This is why we have a custom
+-- Complex wrapper around NP.Complex.
+gInvCoeffsCRT = ($ fromJust $ divGPow' $ scalarPow' LP.one) <$> crt'
+
+-- we can't put this in Extension with the rest of the twace/embed fucntions because it needs access to 
+-- the C backend
+twaceCRT' :: forall m m' r .
+             (TElt CT r, CRTrans r, m `Divides` m')
+             => TaggedT '(m, m') Maybe (Vector r -> Vector r)
+twaceCRT' = tagT $ do -- Maybe monad
+  (CT' g') <- proxyT gCoeffsCRT (Proxy::Proxy m')
+  (CT' gInv) <- proxyT gInvCoeffsCRT (Proxy::Proxy m)
+  embed <- proxyT embedCRT' (Proxy::Proxy '(m,m'))
+  indices <- pure $ proxy extIndicesCRT (Proxy::Proxy '(m,m'))
+  (_, m'hatinv) <- proxyT crtInfoFact (Proxy::Proxy m')
+  let phi = proxy totientFact (Proxy::Proxy m)
+      phi' = proxy totientFact (Proxy::Proxy m')
+      mhat = fromIntegral $ proxy valueHatFact (Proxy::Proxy m)
+      hatRatioInv = m'hatinv * mhat
+      reltot = phi' `div` phi
+      -- tweak = mhat * g' / (m'hat * g)
+      tweak = SV.map (* hatRatioInv) $ SV.zipWith (*) (embed gInv) g'
+  return $ \ arr -> -- take true trace after mul-by-tweak
+    let v = backpermute' indices (SV.zipWith (*) tweak arr)
+    in generate phi $ \i -> foldl1' (+) $ SV.unsafeSlice (i*reltot) reltot v
+
+
+
+
+
+
+
+
+
+
+
+
+-- C-backend support
+
+marshalFactors :: [PP] -> Vector CPP
+marshalFactors = SV.fromList . LP.map (\(p,e) -> CPP (fromIntegral p) (fromIntegral e))
+
+-- http://stackoverflow.com/questions/6517387/vector-vector-foo-ptr-ptr-foo-io-a-io-a
+withPtrArray :: (Storable a) => [Vector a] -> (Ptr (Ptr a) -> IO b) -> IO b
+withPtrArray v f = do
+  let vs = LP.map SV.unsafeToForeignPtr0 v
+      ptrV = LP.map (\(fp,_) -> getPtr fp) vs
+  res <- withArray ptrV f
+  LP.mapM_ (\(fp,_) -> touchForeignPtr fp) vs
+  return res
+
+data CPP = CPP {p' :: !Int32, e' :: !Int16}
+-- stolen from http://hackage.haskell.org/packages/archive/numeric-prelude/0.4.0.3/doc/html/src/Number-Complex.html#T
+-- the NumericPrelude Storable instance for complex numbers
+instance Storable CPP where
+   sizeOf    = Store.sizeOf store
+   alignment = Store.alignment store
+   peek      = Store.peek store
+   poke      = Store.poke store
+
+store :: Store.Dictionary CPP
+store = Store.run $
+   liftA2 CPP
+      (Store.element p')
+      (Store.element e')
+
+instance Show CPP where
+    show (CPP p e) = "(" LP.++ (show p) LP.++ "," LP.++ (show e) LP.++ ")"
+
+foreign import ccall unsafe "tensorLR" tensorLR ::                  Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorLInvR" tensorLInvR ::            Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorLRq" tensorLRq ::                Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorLInvRq" tensorLInvRq ::          Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorLDouble" tensorLDouble ::       Ptr Double -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorLInvDouble" tensorLInvDouble :: Ptr Double -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorLC" tensorLC ::       Ptr (Complex Double) -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorLInvC" tensorLInvC :: Ptr (Complex Double) -> Int64 -> Ptr CPP -> Int16          -> IO ()
+
+foreign import ccall unsafe "tensorGPowR" tensorGPowR ::         Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorGPowRq" tensorGPowRq ::       Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorGDecR" tensorGDecR ::         Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorGDecRq" tensorGDecRq ::       Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorGInvPowR" tensorGInvPowR ::   Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorGInvPowRq" tensorGInvPowRq :: Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorGInvDecR" tensorGInvDecR ::   Ptr Int64 -> Int64 -> Ptr CPP -> Int16          -> IO ()
+foreign import ccall unsafe "tensorGInvDecRq" tensorGInvDecRq :: Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Int64 -> IO ()
+--foreign import ccall unsafe "tensorGCRTRq" tensorGCRTRq ::       Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (ZqBasic q Int64)) -> Int64   -> IO ()
+--foreign import ccall unsafe "tensorGCRTC" tensorGCRTC ::         Ptr (Complex Double) ->   Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex Double)) -> IO ()
+--foreign import ccall unsafe "tensorGInvCRTRq" tensorGInvCRTRq :: Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (ZqBasic q Int64)) -> Int64   -> IO ()
+--foreign import ccall unsafe "tensorGInvCRTC" tensorGInvCRTC ::   Ptr (Complex Double) ->   Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex Double)) -> IO ()
+
+foreign import ccall unsafe "tensorCRTRq" tensorCRTRq ::         Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (ZqBasic q Int64)) -> Int64 -> IO ()
+foreign import ccall unsafe "tensorCRTC" tensorCRTC ::           Ptr (Complex Double) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex Double)) -> IO ()
+foreign import ccall unsafe "tensorCRTInvRq" tensorCRTInvRq ::   Ptr (ZqBasic q Int64) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (ZqBasic q Int64)) -> Int64 -> Int64 -> IO ()
+foreign import ccall unsafe "tensorCRTInvC" tensorCRTInvC ::     Ptr (Complex Double) -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex Double)) -> Double -> IO ()
+
+foreign import ccall unsafe "tensorGaussianDec" tensorGaussianDec :: Ptr Double -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex Double)) ->  IO ()
+
+foreign import ccall unsafe "mulRq" mulRq :: Ptr (ZqBasic q Int64) -> Ptr (ZqBasic q Int64) -> Int64 -> Int64 -> IO ()
+foreign import ccall unsafe "mulC" mulC :: Ptr (Complex Double) -> Ptr (Complex Double) -> Int64 -> IO ()
+
+foreign import ccall unsafe "addRq" addRq :: Ptr (ZqBasic q Int64) -> Ptr (ZqBasic q Int64) -> Int64 -> Int64 -> IO ()
+foreign import ccall unsafe "addR" addR :: Ptr Int64 -> Ptr Int64 -> Int64 -> IO ()
+foreign import ccall unsafe "addC" addC :: Ptr (Complex Double) -> Ptr (Complex Double) -> Int64 -> IO ()
+foreign import ccall unsafe "addD" addD :: Ptr Double -> Ptr Double -> Int64 -> IO ()
+
+-- | Class to safely match Haskell types with the appropriate C function.
+class Dispatch r where
+  dcrt :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr r) -> IO ()
+  dcrtinv :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr r) -> r -> IO ()
+  dl :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dlinv :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dmulgpow :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dmulgdec :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dginvpow :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dginvdec :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> IO ()
+  dadd :: Ptr r -> Ptr r -> Int64 -> IO ()
+  dmul :: Ptr r -> Ptr r -> Int64 -> IO ()
+  dgcrt :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr r) -> IO ()
+  dginvcrt :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr r) -> IO ()
+  dgaussdec :: Ptr r -> Int64 -> Ptr CPP -> Int16 -> Ptr (Ptr (Complex r)) -> IO ()
+
+instance (Reflects q Int64) => Dispatch (ZqBasic q Int64) where
+  dcrt pout totm pfac numFacts ruptr = 
+    let q = proxy value (Proxy::Proxy q)
+    in tensorCRTRq pout totm pfac numFacts ruptr q
+  dcrtinv pout totm pfac numFacts ruptr minv =
+    let q = proxy value (Proxy::Proxy q)
+    --EAC: GHC doesn't like it if I change the type of minv to ZqBasic in the
+    -- signature of tensorCRTInvRq, and the constructor of ZqBasic isn't exposed
+    -- so using unsafeCoerce for now
+    in tensorCRTInvRq pout totm pfac numFacts ruptr (unsafeCoerce minv) q
+  dl pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorLRq pout totm pfac numFacts q
+  dlinv pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorLInvRq pout totm pfac numFacts q
+  dmulgpow pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorGPowRq pout totm pfac numFacts q
+  dmulgdec pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorGDecRq pout totm pfac numFacts q
+  dginvpow pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorGInvPowRq pout totm pfac numFacts q
+  dginvdec pout totm pfac numFacts =
+    let q = proxy value (Proxy::Proxy q)
+    in tensorGInvDecRq pout totm pfac numFacts q
+  dadd aout bout totm = 
+    let q = proxy value (Proxy::Proxy q)
+    in addRq aout bout totm q
+  dmul aout bout totm =
+    let q = proxy value (Proxy::Proxy q)
+    in mulRq aout bout totm q
+  dgcrt pout totm pfac numFacts gcoeffs' = error "dgcrt zq"
+    --let q = proxy value (Proxy::Proxy q)
+    --in tensorGCRTRq pout totm pfac numFacts gcoeffs' q
+  dginvcrt pout totm pfac numFacts gcoeffs' = error "dginvcrt zq"
+    --let q = proxy value (Proxy::Proxy q)
+    --in tensorGInvCRTRq pout totm pfac numFacts gcoeffs' q
+  dgaussdec = error "cannot call CT gaussianDec on type ZqBasic"
+
+instance Dispatch (Complex Double) where
+  dcrt = tensorCRTC
+  dcrtinv pout totm pfac numFacts ruptr minv = 
+    tensorCRTInvC pout totm pfac numFacts ruptr (real minv)
+  dl = tensorLC
+  dlinv = tensorLInvC
+  dmulgpow = error "cannot call CT mulGPow on type Complex Double"
+  dmulgdec = error "cannot call CT mulGDec on type Complex Double"
+  dginvpow = error "cannot call CT divGPow on type Complex Double"
+  dginvdec = error "cannot call CT divGDec on type Complex Double"
+  dadd = addC
+  dmul = mulC
+  dgcrt = error "tensorGCRTC"
+  dginvcrt = error "tensorGInvCRTC"
+  dgaussdec = error "cannot call CT gaussianDec on type Comple Double"
+
+instance Dispatch Double where
+  dcrt = error "cannot call CT Crt on type Double"
+  dcrtinv = error "cannot call CT CrtInv on type Double"
+  dl = tensorLDouble
+  dlinv = tensorLInvDouble
+  dmulgpow = error "cannot call CT mulGPow on type Double"
+  dmulgdec = error "cannot call CT mulGDec on type Double"
+  dginvpow = error "cannot call CT divGPow on type Double"
+  dginvdec = error "cannot call CT divGDec on type Double"
+  dadd = addD
+  dmul = error "cannot call CT (*) on type Double"
+  dgcrt = error "cannot call CT mulGCRT on type Double"
+  dginvcrt = error "cannot call CT divGCRT on type Double"
+  dgaussdec = tensorGaussianDec
+
+instance Dispatch Int64 where
+  dcrt = error "cannot call CT Crt on type Int64"
+  dcrtinv = error "cannot call CT CrtInv on type Int64"
+  dl = tensorLR
+  dlinv = tensorLInvR
+  dmulgpow = tensorGPowR
+  dmulgdec = tensorGDecR
+  dginvpow = tensorGInvPowR
+  dginvdec = tensorGInvDecR
+  dadd = addR
+  dmul = error "cannot call CT (*) on type Int64"
+  dgcrt = error "cannot call CT mulGCRT on type Int64"
+  dginvcrt = error "cannot call CT divGCRT on type Int64"
+  dgaussdec = error "cannot call CT gaussianDec on type Int64"
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/Extension.hs b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/Extension.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/Extension.hs
@@ -0,0 +1,122 @@
+{-# LANGUAGE ConstraintKinds, FlexibleContexts, MultiParamTypeClasses,
+             NoImplicitPrelude, RebindableSyntax, ScopedTypeVariables,
+             TypeFamilies, TypeOperators, DataKinds #-}
+
+-- | CT-specific functions for embedding/twacing in various bases
+
+module Crypto.Lol.Cyclotomic.Tensor.CTensor.Extension
+( embedPow', embedDec', embedCRT'
+, twacePowDec' -- , twaceCRT'
+, coeffs', powBasisPow'
+, crtSetDec'
+, backpermute'
+) where
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.LatticePrelude as LP hiding (null, lift)
+import Crypto.Lol.Cyclotomic.Tensor as T
+import Crypto.Lol.Types.FiniteField
+import Crypto.Lol.Types.ZmStar
+import Crypto.Lol.Reflects
+
+import Control.Applicative hiding (empty)
+import Control.Monad.Trans (lift)
+
+import           Data.Maybe
+import           Data.Reflection (reify)
+import qualified Data.Vector         as V
+import           Data.Vector.Generic as G (generate, Vector, (!), length)
+import qualified Data.Vector.Unboxed as U
+import qualified Data.Vector.Storable as SV
+
+
+-- | /O(n)/ Yield the vector obtained by replacing each element @i@ of the
+-- index vector by @xs'!'i@. This is equivalent to @'map' (xs'!') is@ but is
+-- often much more efficient.
+--
+-- > backpermute <a,b,c,d> <0,3,2,3,1,0> = <a,d,c,d,b,a>
+backpermute' :: (Vector v a)
+             => U.Vector Int -- ^ @is@ index vector (of length @n@)
+             -> v a   -- ^ @xs@ value vector
+             -> v a
+--{-# INLINE backpermute' #-}
+backpermute' is v = generate (G.length is) (\i -> v ! (is ! i))
+
+embedPow', embedDec' :: (Additive r, Vector v r, m `Divides` m')
+                     => Tagged '(m, m') (v r -> v r)
+-- | Embeds an vector in the powerful basis of the the mth cyclotomic ring
+-- to an vector in the powerful basis of the m'th cyclotomic ring when @m | m'@
+embedPow' = (\indices arr -> generate (U.length indices) $ \idx -> 
+  let (j0,j1) = indices ! idx
+  in if j0 == 0
+     then arr ! j1
+     else zero) <$> baseIndicesPow
+-- | Embeds an vector in the decoding basis of the the mth cyclotomic ring
+-- to an vector in the decoding basis of the m'th cyclotomic ring when @m | m'@
+embedDec' = (\indices arr -> generate (U.length indices)
+  (\idx -> maybe LP.zero
+    (\(sh,b) -> if b then negate (arr ! sh) else arr ! sh)
+    (indices U.! idx))) <$> baseIndicesDec
+
+-- | Embeds an vector in the CRT basis of the the mth cyclotomic ring
+-- to an vector in the CRT basis of the m'th cyclotomic ring when @m | m'@
+embedCRT' :: forall m m' v r . (CRTrans r, Vector v r, m `Divides` m')
+          => TaggedT '(m, m') Maybe (v r -> v r)
+embedCRT' = 
+  (lift (proxyT crtInfoFact (Proxy::Proxy m') :: Maybe (CRTInfo r))) >>
+  (pureT $ backpermute' <$> baseIndicesCRT)
+
+-- | maps a vector in the powerful/decoding basis, representing an
+-- O_m' element, to a vector of arrays representing O_m elements in
+-- the same type of basis
+coeffs' :: (Vector v r, m `Divides` m')
+        => Tagged '(m, m') (v r -> [v r])
+coeffs' = flip (\x -> V.toList . V.map (`backpermute'` x))
+          <$> extIndicesCoeffs
+
+-- | The "tweaked trace" function in either the powerful or decoding
+-- basis of the m'th cyclotomic ring to the mth cyclotomic ring when 
+-- @m | m'@.
+twacePowDec' :: forall m m' r v . (Vector v r, m `Divides` m')
+             => Tagged '(m, m') (v r -> v r)
+twacePowDec' = backpermute' <$> extIndicesPowDec
+
+
+-- EAC: twaceCRT is defined in CTensor because it needs access to C-backend functions
+
+
+-- | The powerful extension basis, wrt the powerful basis.
+-- Outputs a list of vectors in O_m' that are an O_m basis for O_m'
+powBasisPow' :: forall m m' r . (m `Divides` m', Ring r, SV.Storable r)
+                => Tagged '(m, m') [SV.Vector r]
+powBasisPow' = do
+  (_, phi, phi', _) <- indexInfo
+  idxs <- baseIndicesPow
+  return $ LP.map (\k -> generate phi' $ \j -> 
+                           let (j0,j1) = idxs U.! j
+                          in if j0==k && j1==0 then one else zero)
+    [0..phi' `div` phi - 1]
+
+-- | A list of vectors representing the mod-p CRT set of the
+-- extension O_m'/O_m
+crtSetDec' :: forall m m' fp .
+  (m `Divides` m', PrimeField fp, Coprime (PToF (CharOf fp)) m',
+   SV.Storable fp)
+  => Tagged '(m, m') [SV.Vector fp]
+crtSetDec' =
+  let m'p = Proxy :: Proxy m'
+      p = proxy value (Proxy::Proxy (CharOf fp))
+      phi = proxy totientFact m'p
+      d = proxy (order p) m'p
+      h :: Int = proxy valueHatFact m'p
+      hinv = recip $ fromIntegral h
+  in reify d $ \(_::Proxy d) -> do
+      let twCRTs' :: Matrix (GF fp d)
+            = fromMaybe (error "internal error: crtSetDec': twCRTs") $ proxyT twCRTs m'p
+          zmsToIdx = proxy T.zmsToIndexFact m'p
+          elt j i = indexM twCRTs' j (zmsToIdx i)
+          trace' = trace :: GF fp d -> fp -- to avoid recomputing powTraces
+      cosets <- partitionCosets p
+      return $ LP.map (\is -> generate phi
+                          (\j -> hinv * trace'
+                                      (sum $ LP.map (elt j) is))) cosets
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/basic.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/basic.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/basic.c
@@ -0,0 +1,162 @@
+#include "tensorTypes.h"
+#ifdef CINTRIN
+#include <immintrin.h>
+#endif
+
+#ifdef STATS
+int mulCtr = 0;
+struct timespec mulTime = {0,0};
+
+int addCtr = 0;
+struct timespec addTime = {0,0};
+#endif
+
+//a = zipWith (*) a b
+void mulRq (hInt_t* a, hInt_t* b, hDim_t totm, hInt_t q) {
+#ifdef STATS
+    mulCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    for(int i = 0; i < totm; i++) {
+        a[i] = (a[i]*b[i])%q;
+    }
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    mulTime = tsAdd(mulTime, tsSubtract(t1,s1));
+#endif
+}
+
+void mulMq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hByte_t logr, const hInt_t k, const hInt_t q) {
+#ifdef STATS
+    mulCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    hInt_t mask = (1<<logr)-1; // R-1
+
+    for(int i = 0; i < totm; i++) {
+        hInt_t x = a[i]*b[i];
+        hInt_t s = k*(x & mask);
+        hInt_t m = s & mask;
+        a[i] = (x+m*q)>>logr;
+    }
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    mulTime = tsAdd(mulTime, tsSubtract(t1,s1));
+#endif
+}
+
+void mulC (complex_t* a, complex_t* b, hDim_t totm) {
+#ifdef STATS
+    mulCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    for(int i = 0; i < totm; i++)
+    {
+        CMPLX_IMUL(a[i],b[i]);
+    }
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    mulTime = tsAdd(mulTime, tsSubtract(t1,s1));
+#endif
+}
+
+//a = zipWith (+) a b
+void addRq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hInt_t q) {
+#ifdef STATS
+    addCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef CINTRIN
+    __m128i qs = _mm_set1_epi64x(q);
+    for(int i = 0; i < totm; i+=2) {
+        __m128i xs = _mm_load_si128((const __m128i*)(a+i));
+        __m128i ys = _mm_load_si128((const __m128i*)(b+i));
+        __m128i zs = _mm_add_epi64(xs,ys);
+        zs = _mm_rem_epi64(zs,qs);
+        _mm_store_si128((__m128i*)(a+i),zs);
+    }
+#else
+    for(int i = 0; i < totm; i++) {
+        hInt_t temp = a[i]+b[i];
+        if (temp >= q) a[i]=temp-q;
+        else a[i] = temp;
+    }
+#endif
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    addTime = tsAdd(addTime, tsSubtract(t1,s1));
+#endif
+}
+
+void addMq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hByte_t logr, const hInt_t k, const hInt_t q) {
+#ifdef STATS
+    addCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    hInt_t twoq = q<<1;
+    for(int i = 0; i < totm; i++) {
+        hInt_t temp = (a[i]+b[i]);
+        if (temp >= twoq) a[i]=temp-twoq;
+        else a[i] = temp;
+    }
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    addTime = tsAdd(addTime, tsSubtract(t1,s1));
+#endif
+}
+
+//a = zipWith (+) a b
+void addR (hInt_t* a, hInt_t* b, hDim_t totm) {
+#ifdef STATS
+    addCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    for(int i = 0; i < totm; i++)    {
+        a[i] += b[i];
+    }
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    addTime = tsAdd(addTime, tsSubtract(t1,s1));
+#endif
+}
+
+void addC (complex_t* a, complex_t* b, hDim_t totm) {
+#ifdef STATS
+    addCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    for(int i = 0; i < totm; i++)
+    {
+        CMPLX_IADD(a[i],b[i]);
+    }
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    addTime = tsAdd(addTime, tsSubtract(t1,s1));
+#endif
+}
+
+void addD (double* a, double* b, hDim_t totm) {
+#ifdef STATS
+    addCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    for(int i = 0; i < totm; i++)
+    {
+        a[i]+=b[i];
+    }
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    addTime = tsAdd(addTime, tsSubtract(t1,s1));
+#endif
+}
+
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/crt.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/crt.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/crt.c
@@ -0,0 +1,1305 @@
+#include "tensorTypes.h"
+#include <time.h>
+#include <stdlib.h>
+
+// there should be a special cases that do NOT require temp space to be allocated for all primes *smaller* than DFTP_GENERIC_SIZE
+#define DFTP_GENERIC_SIZE 11
+
+#ifdef STATS
+int crtRqCtr = 0;
+int crtInvRqCtr = 0;
+int crtCCtr = 0;
+int crtInvCCtr = 0;
+
+struct timespec crttime1 = {0,0};
+struct timespec crttime2 = {0,0};
+struct timespec crttime3 = {0,0};
+struct timespec crttime4 = {0,0};
+
+struct timespec crtInvRqTime = {0,0};
+struct timespec crtCTime = {0,0};
+struct timespec crtInvCTime = {0,0};
+#endif
+
+hDim_t bitrev (PrimeExponent pe, hDim_t j) {
+    hShort_t e;
+    hDim_t p = pe.prime;
+    hDim_t tempj = j;
+    hDim_t acc = 0;
+
+    for(e = pe.exponent-1; e >= 0; e--) {
+        div_t qr = div(tempj,p);
+        acc += qr.rem * ipow(p,e);
+        tempj = qr.quot;
+    }
+    return acc;
+}
+
+void crtTwiddleRq (hInt_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hInt_t* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    pe.exponent -= 1; // used for an argument to bitrev
+    
+    if(p == 2)
+    {
+        hDim_t mprime = 1<<(e-1);
+        hDim_t blockDim = rts*mprime; // size of block in block diagonal tensor matrix
+
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/(p-1) for i = 0..(m'-1), we can skip i0 = 0
+        {
+            hDim_t temp2 = i0*rts;
+            hInt_t twid = ru[bitrev(pe, i0)];
+
+            for(hDim_t blockIdx = 0; blockIdx < lts; blockIdx++)
+            {
+                hDim_t temp3 = blockIdx*blockDim + temp2;
+                for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                {
+                    hDim_t idx = temp3 + modOffset;
+                    y[idx] = (y[idx]*twid) % q;
+                }
+            }
+        }
+    }
+    else // This loop is faster, probably due to the division in the loop above.
+    // cilk also slows it down
+    {
+        hDim_t mprime = ipow(p,e-1);
+        hDim_t blockDim = rts*(p-1)*mprime; // size of block in block diagonal tensor matrix
+        
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/(p-1) for i = 0..(m'-1), we can skip i0 = 0
+        {
+            hDim_t temp1 = i0*(p-1);
+            for(hDim_t i1 = 0; i1 < (p-1); i1++) // loops over i%(p-1) for i = 0..(m'-1)
+            {        
+                hDim_t temp2 = (temp1+i1)*rts;
+                hInt_t twid = ru[bitrev(pe, i0)*(i1+1)];
+
+                for(hDim_t blockIdx = 0; blockIdx < lts; blockIdx++)
+                {
+                    hDim_t temp3 = blockIdx*blockDim + temp2;
+                    for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                    {
+                        hDim_t idx = temp3 + modOffset;
+                        y[idx] = (y[idx]*twid) % q;
+                    }
+                }
+            }
+        }
+    }
+}
+
+// dim is power of p
+void dftptwidRq (hInt_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t dim, hDim_t rustride, hInt_t* ru, hInt_t q)
+{
+    hDim_t idx;
+    hDim_t p = pe.prime;
+
+    pe.exponent -= 1; // used for an argument to bitrev
+
+    if(p == 2) {
+        hDim_t mprime = dim>>1; // divides evenly
+        hDim_t temp1 = rts*dim; // for use in computing [modified] tensorOffset
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/p for i = 0..(dim-1), but we skip i0=0
+        {
+            hDim_t temp3 = rts*(i0*p+1);
+            hInt_t twid = ru[bitrev(pe,i0)*rustride];
+
+            for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+            {
+                hDim_t temp2 = blockOffset*temp1 + temp3;
+                for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                {
+                    idx = temp2 + modOffset;
+                    y[idx] = (y[idx]*twid) % q;
+                }
+            }
+        }
+    }
+    else
+    {
+        hDim_t mprime = dim/p; // divides evenly
+        hDim_t temp1 = rts*dim; // for use in computing [modified] tensorOffset
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/p for i = 0..(dim-1), but we skip i0=0
+        {
+            for(hDim_t i1 = 1; i1 < p; i1++) // loops over i%p for i = 0..(dim-1), but we skip i1=0
+            {
+                hDim_t temp3 = rts*(i0*p+i1);
+                hInt_t twid = ru[bitrev(pe,i0)*i1*rustride];
+
+                for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+                {
+                    hDim_t temp2 = blockOffset*temp1 + temp3;
+                    for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                    {
+                        idx = temp2 + modOffset;
+                        y[idx] = (y[idx]*twid) % q;
+                    }
+                }
+            }
+        }
+    }
+}
+
+//implied length of ru is rustride*p
+//implied length of tempSpace is p, if p is not a special case
+// temp is allowed to be NULL if p < DFTP_GENERIC_SIZE
+void dftpRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, hInt_t* ru, hInt_t* tempSpace, hInt_t q)
+{
+    hDim_t tensorOffset;
+    
+    if(p == 2)
+    {
+        hDim_t temp1 = rts<<1;
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t u = y[tensorOffset];
+                hInt_t t = y[tensorOffset+rts];
+                y[tensorOffset] = (u + t) % q;
+                y[tensorOffset+rts] = (u - t) % q;
+            }
+        }
+    }
+    else if(p == 3)
+    {
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+        hDim_t temp1 = rts*3;
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2, y3;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                //q is <32 bits, so we can do 3 additions without overflow
+                y[tensorOffset]          = (y1 + y2 + y3) % q;
+                y[tensorOffset+rts]      = (y1 + ((ru1*y2) % q) + ((ru2*y3) % q)) % q;
+                y[tensorOffset+(rts<<1)] = (y1 + ((ru2*y2) % q) + ((ru1*y3) % q)) % q;
+            }   
+        }
+
+    }
+    else if(p == 5)
+    {
+        hDim_t temp1 = rts*5;
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+        hInt_t ru3 = ru[rustride*3];
+        hInt_t ru4 = ru[rustride<<2];
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2, y3, y4, y5;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y[tensorOffset]          = (y1 + y2 + y3 + y4 + y5) % q;
+                y[tensorOffset+rts]      = (y1 + ((ru1*y2) % q) + ((ru2*y3) % q) + ((ru3*y4) % q) + ((ru4*y5) % q)) % q;
+                y[tensorOffset+(rts<<1)] = (y1 + ((ru2*y2) % q) + ((ru4*y3) % q) + ((ru1*y4) % q) + ((ru3*y5) % q)) % q;
+                y[tensorOffset+rts*3]    = (y1 + ((ru3*y2) % q) + ((ru1*y3) % q) + ((ru4*y4) % q) + ((ru2*y5) % q)) % q;
+                y[tensorOffset+(rts<<2)] = (y1 + ((ru4*y2) % q) + ((ru3*y3) % q) + ((ru2*y4) % q) + ((ru1*y5) % q)) % q;
+            }
+        }
+    }
+    else if(p == 7)
+    {
+        hDim_t temp1 = rts*7;
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+        hInt_t ru3 = ru[rustride*3];
+        hInt_t ru4 = ru[rustride<<2];
+        hInt_t ru5 = ru[rustride*5];
+        hInt_t ru6 = ru[rustride*6];
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2, y3, y4, y5, y6, y7;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y6 = y[tensorOffset+rts*5];
+                y7 = y[tensorOffset+rts*6];
+                y[tensorOffset]          = (y1 +     y2 +     y3 +     y4 +     y5 +     y6 +     y7) % q;
+                y[tensorOffset+rts]      = (y1 + ((ru1*y2) % q) + ((ru2*y3) % q) + ((ru3*y4) % q) + ((ru4*y5) % q) + ((ru5*y6) % q) + ((ru6*y7) % q)) % q;
+                y[tensorOffset+(rts<<1)] = (y1 + ((ru2*y2) % q) + ((ru4*y3) % q) + ((ru6*y4) % q) + ((ru1*y5) % q) + ((ru3*y6) % q) + ((ru5*y7) % q)) % q;
+                y[tensorOffset+rts*3]    = (y1 + ((ru3*y2) % q) + ((ru6*y3) % q) + ((ru2*y4) % q) + ((ru5*y5) % q) + ((ru1*y6) % q) + ((ru4*y7) % q)) % q;
+                y[tensorOffset+(rts<<2)] = (y1 + ((ru4*y2) % q) + ((ru1*y3) % q) + ((ru5*y4) % q) + ((ru2*y5) % q) + ((ru6*y6) % q) + ((ru3*y7) % q)) % q;
+                y[tensorOffset+rts*5]    = (y1 + ((ru5*y2) % q) + ((ru3*y3) % q) + ((ru1*y4) % q) + ((ru6*y5) % q) + ((ru4*y6) % q) + ((ru2*y7) % q)) % q;
+                y[tensorOffset+rts*6]    = (y1 + ((ru6*y2) % q) + ((ru5*y3) % q) + ((ru4*y4) % q) + ((ru3*y5) % q) + ((ru2*y6) % q) + ((ru1*y7) % q)) % q;
+            }   
+        }
+    }
+    else
+    {
+        hDim_t temp1 = rts*p;
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;                
+                for(hDim_t row = 0; row < p; row++)
+                {
+                    hInt_t acc = 0;
+                    //p is small (<< 30 bits), so we can do p additions of mod-q values without overflow
+                    for(hDim_t col = 0; col < p; col++)
+                    {
+                        acc += ((y[tensorOffset+col*rts]*ru[((col*row) % p)*rustride])%q);
+                    }
+                    tempSpace[row] = acc % q;
+                }
+                
+                for(hDim_t row = 0; row < p; row++)
+                {
+                    y[tensorOffset+rts*row] = tempSpace[row];
+                }
+            }
+        }
+    }
+}
+
+void crtpRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, hInt_t* ru, hInt_t q)
+{
+    hDim_t tensorOffset;
+    if(p == 2)
+    {
+        return;
+    }
+    else if(p == 3)
+    {
+        hDim_t temp1 = rts*2;
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y[tensorOffset]     = (y1 + ((ru1*y2)%q)) % q;
+                y[tensorOffset+rts] = (y1 + ((ru2*y2)%q)) % q;
+            }   
+        }
+    }
+    else if(p == 5)
+    {
+        hDim_t temp1 = rts*4;
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+        hInt_t ru3 = ru[rustride*3];
+        hInt_t ru4 = ru[rustride<<2];
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2, y3, y4;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+
+                y[tensorOffset]          = (y1 + ((ru1*y2) % q) + ((ru2*y3) % q) + ((ru3*y4) % q)) % q;
+                y[tensorOffset+rts]      = (y1 + ((ru2*y2) % q) + ((ru4*y3) % q) + ((ru1*y4) % q)) % q;
+                y[tensorOffset+(rts<<1)] = (y1 + ((ru3*y2) % q) + ((ru1*y3) % q) + ((ru4*y4) % q)) % q;
+                y[tensorOffset+rts*3]    = (y1 + ((ru4*y2) % q) + ((ru3*y3) % q) + ((ru2*y4) % q)) % q;
+            }   
+        }
+    }
+    else if(p == 7)
+    {
+        hDim_t temp1 = rts*6;
+        hInt_t ru1 = ru[rustride];
+        hInt_t ru2 = ru[rustride<<1];
+        hInt_t ru3 = ru[rustride*3];
+        hInt_t ru4 = ru[rustride<<2];
+        hInt_t ru5 = ru[rustride*5];
+        hInt_t ru6 = ru[rustride*6];
+
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hInt_t y1, y2, y3, y4, y5, y6;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y6 = y[tensorOffset+rts*5];
+                y[tensorOffset]          = (y1 + ((ru1*y2) % q) + ((ru2*y3) % q) + ((ru3*y4) % q) + ((ru4*y5) % q) + ((ru5*y6) % q)) % q;
+                y[tensorOffset+rts]      = (y1 + ((ru2*y2) % q) + ((ru4*y3) % q) + ((ru6*y4) % q) + ((ru1*y5) % q) + ((ru3*y6) % q)) % q;
+                y[tensorOffset+(rts<<1)] = (y1 + ((ru3*y2) % q) + ((ru6*y3) % q) + ((ru2*y4) % q) + ((ru5*y5) % q) + ((ru1*y6) % q)) % q;
+                y[tensorOffset+rts*3]    = (y1 + ((ru4*y2) % q) + ((ru1*y3) % q) + ((ru5*y4) % q) + ((ru2*y5) % q) + ((ru6*y6) % q)) % q;
+                y[tensorOffset+(rts<<2)] = (y1 + ((ru5*y2) % q) + ((ru3*y3) % q) + ((ru1*y4) % q) + ((ru6*y5) % q) + ((ru4*y6) % q)) % q;
+                y[tensorOffset+rts*5]    = (y1 + ((ru6*y2) % q) + ((ru5*y3) % q) + ((ru4*y4) % q) + ((ru3*y5) % q) + ((ru2*y6) % q)) % q;
+            }
+        }
+    }
+    else
+    {
+        hInt_t* tempSpace = (hInt_t*)malloc((p-1)*sizeof(hInt_t));
+        hDim_t temp1 = rts*(p-1);
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                
+                for(hDim_t row = 1; row < p; row++)
+                {
+                    hInt_t acc = 0;
+                    for(hDim_t col = 0; col < p-1; col++)
+                    {
+                        acc += ((y[tensorOffset+col*rts]*ru[((col*row) % p)*rustride]) % q);
+                    }
+                    tempSpace[row-1] = acc % q;
+                }
+                
+                for(hDim_t row = 0; row < p-1; row++)
+                {
+                    y[tensorOffset+rts*row] = tempSpace[row];
+                }
+            }
+        }
+        free(tempSpace);
+    }
+}
+
+//takes inverse rus
+void crtpinvRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, hInt_t* ruinv, hInt_t q)
+{
+    if(p ==2)
+    {
+        // need this case so that we can divide overall by mhat^(-1)
+        return;
+    }
+    else
+    {
+        hDim_t tensorOffset,i;
+        hInt_t* tempSpace = (hInt_t*)malloc((p-1)*sizeof(hInt_t));
+        hDim_t temp1 = rts*(p-1);
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                
+                for(i = 0; i < p-1; i++)
+                {
+                    hInt_t sum = 0;
+                    int j;
+                    for(j = 0; j < p-1; j++)
+                    {
+                        int ruIdx = ((j+1)*i) % p;
+                        sum += ((y[tensorOffset+j*rts] * ruinv[ruIdx*rustride]) % q);
+                    }
+                    tempSpace[i] = sum % q;
+                }
+
+                hInt_t shift = 0;
+                for(i = 0; i < p-1; i++)
+                {
+                    // we were given the inverse rus, so we need to negate the indices
+                    shift += ((y[tensorOffset+i*rts] * ruinv[rustride*(p-(i+1))]) % q);
+                }
+
+                for(i = 0; i < p-1; i++)
+                {
+                    y[tensorOffset+i*rts] = (tempSpace[i] - shift) % q; 
+                }
+            }
+        }
+    }
+}
+
+void ppDFTRq (hInt_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t rustride, hInt_t* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+    if(e == 0)
+    {
+        return;
+    }
+    
+    hDim_t primeRuStride = rustride*ipow(p,e-1);
+    hInt_t* temp = 0;
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        temp = (hInt_t*)malloc(p*sizeof(hInt_t));
+    }
+    hShort_t i;
+    
+    hDim_t ltsScale = ipow(p,e-1);
+    hDim_t rtsScale = 1;
+    hDim_t twidRuStride = rustride;
+    for(i = 0; i < e; i++)
+    {
+        hDim_t rtsDim = rts*rtsScale;
+        dftpRq (y, lts*ltsScale, rtsDim, p, primeRuStride, ru, temp, q);
+        dftptwidRq (y, lts, rtsDim, pe, ltsScale*p, twidRuStride, ru, q);
+        
+        ltsScale /= p;
+        rtsScale *= p;
+        twidRuStride *= p;
+        pe.exponent -= 1;
+    }
+    
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        free(temp);
+    }
+}
+
+void ppDFTInvRq (hInt_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t rustride, hInt_t* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+    if(e == 0)
+    {
+        return;
+    }
+    hDim_t primeRuStride = rustride*ipow(p,e-1);
+    hInt_t* temp = 0;
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        temp = (hInt_t*)malloc(p*sizeof(hInt_t));
+    }
+    hShort_t i;
+    
+    hDim_t ltsScale = 1;
+    hDim_t rtsScale = ipow(p,e-1);
+    hDim_t twidRuStride = primeRuStride;
+    pe.exponent = 1;
+    for(i = 0; i < e; i++)
+    {
+        hDim_t rtsDim = rts*rtsScale;
+        hDim_t ltsScaleP = ltsScale*p;
+        dftptwidRq (y, lts, rtsDim, pe, ltsScaleP, twidRuStride, ru, q);
+        dftpRq (y, lts*ltsScale, rtsDim, p, primeRuStride, ru, temp, q);
+        
+        ltsScale = ltsScaleP;
+        rtsScale /= p;
+        twidRuStride /= p;
+        pe.exponent += 1;
+    }
+    
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        free(temp);
+    }
+}
+
+void ppcrtRq (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hDim_t e = pe.exponent;
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    hDim_t mprime = ipow(p,e-1);
+    
+#ifdef DEBUG_MODE
+    printf("lts is %" PRId32 "\trts is %" PRId32 "\n", lts, rts);
+    printf("rus for p=%" PRId32 ", e=%" PRId16 "\t[", pe.prime, pe.exponent);
+    hDim_t i;
+    for(i = 0; i < ipow(p,e); i++) {
+        printf("%" PRId64 ",", ((hInt_t*)ru)[i]);
+    }
+    printf("]\n");
+#endif
+    
+    crtpRq ((hInt_t*)y, lts*mprime, rts, p, mprime, (hInt_t*)ru, q);
+    crtTwiddleRq ((hInt_t*)y, lts, rts, pe, (hInt_t*)ru, q);
+    pe.exponent -= 1;
+    ppDFTRq ((hInt_t*)y, lts, rts*(p-1), pe, p, (hInt_t*)ru, q);
+}
+
+void ppcrtinvRq (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hDim_t e = pe.exponent;
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    hDim_t mprime = ipow(p,e-1);
+#ifdef DEBUG_MODE
+    printf("lts is %" PRId32 "\trts is %" PRId32 "\n", lts, rts);
+    printf("rus for p=%" PRId32 ", e=%" PRId16 "\t[", pe.prime, pe.exponent);
+    hDim_t i;
+    for(i = 0; i < ipow(p,e); i++) {
+        printf("%" PRId64 ",", ((hInt_t*)ru)[i]);
+    }
+    printf("]\n");
+#endif
+    pe.exponent -= 1;
+    ppDFTInvRq ((hInt_t*)y, lts, rts*(p-1), pe, p, (hInt_t*)ru, q);
+    pe.exponent += 1;
+    crtTwiddleRq ((hInt_t*)y, lts, rts, pe, (hInt_t*)ru, q);
+    crtpinvRq ((hInt_t*)y, lts*mprime, rts, p, mprime, (hInt_t*)ru, q);
+}
+
+// EAC: Somebody who knows C/C++ should find a better way to handle pointers-to-pointers in a generic way
+void tensorCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** ru, hInt_t q)
+{
+    
+    hDim_t i;
+#ifdef STATS
+    struct timespec s1,s2,s3,s4,t1,t2,t3,t4;
+
+    crtRqCtr++;
+
+    clock_gettime(CLOCK_REALTIME, &s1);
+    clock_gettime(CLOCK_MONOTONIC, &s2);
+    clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &s3);
+    clock_gettime(CLOCK_THREAD_CPUTIME_ID, &s4);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorCRTRq\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\tq=%" PRId64 "\n[", totm, sizeOfPE, q);
+
+    for(i = 0; i < totm; i++) {
+        printf("%" PRId64 ",", y[i]);
+    }
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+    void** rus = (void**)malloc(sizeOfPE*sizeof(void*));
+    
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        rus[i] = (void*) (ru[i]);
+    }
+	tensorFuserCRT (y, ppcrtRq, totm, peArr, sizeOfPE, rus, q);
+	
+	for(hDim_t j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        y[j]+=q;
+	    }
+#ifdef DEBUG_MODE
+	    if(y[j]<0)
+	    {
+	        printf("TENSOR CRT^T INV\n");
+	    }
+#endif
+	}
+
+	free(rus);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    clock_gettime(CLOCK_MONOTONIC, &t2);
+    clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &t3);
+    clock_gettime(CLOCK_THREAD_CPUTIME_ID, &t4);
+
+    crttime1 = tsAdd(crttime1, tsSubtract(t1,s1));
+    crttime2 = tsAdd(crttime2, tsSubtract(t2,s2));
+    crttime3 = tsAdd(crttime3, tsSubtract(t3,s3));
+    crttime4 = tsAdd(crttime4, tsSubtract(t4,s4));
+#endif
+}
+
+//takes inverse rus
+void tensorCRTInvRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** ruinv, hInt_t mhatInv, hInt_t q)
+{
+	hDim_t i;
+#ifdef STATS
+    struct timespec s1,t1;
+    crtInvRqCtr++;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorCRTInvRq\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\tminv=%" PRId64 "\tq=%" PRId64 "\n[", totm, sizeOfPE, mhatInv, q);
+    for(i = 0; i < totm; i++) {
+        printf("%" PRId64 ",", y[i]);
+    }
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+
+	void** rus = (void**)malloc(sizeOfPE*sizeof(void*));
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        rus[i] = (void*) (ruinv[i]);
+    }
+	
+	tensorFuserCRT (y, ppcrtinvRq, totm, peArr, sizeOfPE, rus, q);
+
+	for (hDim_t j = 0; j < totm; j++)
+	{
+	    y[j] = (y[j]*mhatInv)%q;
+	    if(y[j] < 0)
+	    {
+	        y[j] +=q;
+	    }
+#ifdef DEBUG_MODE
+	    if(y[j]<0)
+	    {
+	        printf("TENSOR CRT INV\n");
+	    }
+#endif
+	}
+
+	free(rus);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    crtInvRqTime = tsAdd(crtInvRqTime, tsSubtract(t1,s1));
+#endif
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+void crtTwiddleC (complex_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, complex_t* ru)
+{
+    hDim_t idx;
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+
+    pe.exponent -= 1; // used for an argument to bitrev
+    
+    if(p == 2)
+    {
+        hDim_t mprime = 1<<(e-1);
+        hDim_t blockDim = rts*mprime; // size of block in block diagonal tensor matrix
+
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/(p-1) for i = 0..(m'-1), we can skip i0 = 0
+        {
+            hDim_t temp2 = i0*rts;
+            complex_t twid = ru[bitrev(pe,i0)];
+
+            for(hDim_t blockIdx = 0; blockIdx < lts; blockIdx++)
+            {
+                hDim_t temp3 = blockIdx*blockDim + temp2;
+                for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                {
+                    idx = temp3 + modOffset;
+                    CMPLX_IMUL(y[idx],twid);
+                }
+            }
+        }
+    }
+    else
+    {
+        hDim_t mprime = ipow(p,e-1);
+        hDim_t blockDim = rts*(p-1)*mprime; // size of block in block diagonal tensor matrix
+
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/(p-1) for i = 0..(m'-1), we can skip i0 = 0
+        {
+            hDim_t temp1 = i0*(p-1);
+            for(hDim_t i1 = 0; i1 < (p-1); i1++) // loops over i%(p-1) for i = 0..(m'-1)
+            {        
+                hDim_t temp2 = (temp1+i1)*rts;
+                complex_t twid = ru[bitrev(pe,i0)*(i1+1)];
+
+                for(hDim_t blockIdx = 0; blockIdx < lts; blockIdx++)
+                {
+                    hDim_t temp3 = blockIdx*blockDim + temp2;
+                    for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                    {
+                        idx = temp3 + modOffset;
+                        CMPLX_IMUL(y[idx],twid);
+                    }
+                }
+            }
+        }
+    }
+}
+    
+// dim is power of p
+void dftptwidC (complex_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t dim, hDim_t rustride, complex_t* ru)
+{
+    hDim_t idx;
+    hDim_t p = pe.prime;
+    pe.exponent -= 1; // used for an argument to bitrev
+    
+    if(p == 2)
+    {
+        hDim_t mprime = dim>>1; // divides evenly
+        hDim_t temp1 = rts*dim; // for use in computing [modified] tensorOffset
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/p for i = 0..(dim-1), but we skip i0=0
+        {
+            hDim_t temp3 = rts*(i0*p+1);
+            complex_t twid = ru[bitrev(pe,i0)*rustride];
+
+            for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+            {
+                hDim_t temp2 = blockOffset*temp1 + temp3;
+                for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                {
+                    idx = temp2 + modOffset;
+                    CMPLX_IMUL(y[idx],twid);
+                }
+            }
+        }
+    }
+    else
+    {
+        hDim_t mprime = dim/p; // divides evenly
+        hDim_t temp1 = rts*dim; // for use in computing [modified] tensorOffset
+        for(hDim_t i0 = 1; i0 < mprime; i0++) // loops over i/p for i = 0..(dim-1), but we skip i0=0
+        {
+            for(hDim_t i1 = 1; i1 < p; i1++) // loops over i%p for i = 0..(dim-1), but we skip i1=0
+            {
+                hDim_t temp3 = rts*(i0*p+i1);
+                complex_t twid = ru[bitrev(pe,i0)*i1*rustride];
+
+                for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+                {
+                    hDim_t temp2 = blockOffset*temp1 + temp3;
+                    for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+                    {
+                        idx = temp2 + modOffset;
+                        CMPLX_IMUL(y[idx],twid);
+                    }
+                }
+            }
+        }
+    }
+}
+
+//implied length of ru is rustride*p
+//implied length of tempSpace is p, if p is not a special case
+void dftpC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, complex_t* ru, complex_t* tempSpace)
+{
+    hDim_t blockOffset, modOffset, tensorOffset;
+    
+    if(p == 2)
+    {
+        hDim_t temp1 = rts<<1;
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t u = y[tensorOffset];
+                complex_t t = y[tensorOffset+rts];
+                y[tensorOffset] = CMPLX_ADD(u,t);
+                y[tensorOffset+rts] = CMPLX_SUB(u,t);
+            }
+        }
+    }
+    else if(p == 3)
+    {
+        hDim_t temp1 = rts*3;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2, y3;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y[tensorOffset]          = CMPLX_ADD3(y1,               y2,                y3);
+                y[tensorOffset+rts]      = CMPLX_ADD3(y1, CMPLX_MUL(ru1,y2), CMPLX_MUL(ru2,y3));
+                y[tensorOffset+(rts<<1)] = CMPLX_ADD3(y1, CMPLX_MUL(ru2,y2), CMPLX_MUL(ru1,y3));
+            }   
+        }
+    }
+    else if(p == 5)
+    {
+        hDim_t temp1 = rts*5;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+        complex_t ru3 = ru[rustride*3];
+        complex_t ru4 = ru[rustride<<2];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2, y3, y4, y5;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y[tensorOffset]          = CMPLX_ADD5(y1,               y2,                y3,                y4,                y5);
+                y[tensorOffset+rts]      = CMPLX_ADD5(y1, CMPLX_MUL(ru1,y2), CMPLX_MUL(ru2,y3), CMPLX_MUL(ru3,y4), CMPLX_MUL(ru4,y5));
+                y[tensorOffset+(rts<<1)] = CMPLX_ADD5(y1, CMPLX_MUL(ru2,y2), CMPLX_MUL(ru4,y3), CMPLX_MUL(ru1,y4), CMPLX_MUL(ru3,y5));
+                y[tensorOffset+rts*3]    = CMPLX_ADD5(y1, CMPLX_MUL(ru3,y2), CMPLX_MUL(ru1,y3), CMPLX_MUL(ru4,y4), CMPLX_MUL(ru2,y5));
+                y[tensorOffset+(rts<<2)] = CMPLX_ADD5(y1, CMPLX_MUL(ru4,y2), CMPLX_MUL(ru3,y3), CMPLX_MUL(ru2,y4), CMPLX_MUL(ru1,y5));
+            }   
+        }
+    }
+    else if(p == 7)
+    {
+        hDim_t temp1 = rts*7;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+        complex_t ru3 = ru[rustride*3];
+        complex_t ru4 = ru[rustride<<2];
+        complex_t ru5 = ru[rustride*5];
+        complex_t ru6 = ru[rustride*6];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2, y3, y4, y5, y6, y7;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y6 = y[tensorOffset+rts*5];
+                y7 = y[tensorOffset+rts*6];
+                y[tensorOffset]          = CMPLX_ADD7(y1,               y2,                y3,                y4,                y5,                y6,                y7);
+                y[tensorOffset+rts]      = CMPLX_ADD7(y1, CMPLX_MUL(ru1,y2), CMPLX_MUL(ru2,y3), CMPLX_MUL(ru3,y4), CMPLX_MUL(ru4,y5), CMPLX_MUL(ru5,y6), CMPLX_MUL(ru6,y7));
+                y[tensorOffset+(rts<<1)] = CMPLX_ADD7(y1, CMPLX_MUL(ru2,y2), CMPLX_MUL(ru4,y3), CMPLX_MUL(ru6,y4), CMPLX_MUL(ru1,y5), CMPLX_MUL(ru3,y6), CMPLX_MUL(ru5,y7));
+                y[tensorOffset+rts*3]    = CMPLX_ADD7(y1, CMPLX_MUL(ru3,y2), CMPLX_MUL(ru6,y3), CMPLX_MUL(ru2,y4), CMPLX_MUL(ru5,y5), CMPLX_MUL(ru1,y6), CMPLX_MUL(ru4,y7));
+                y[tensorOffset+(rts<<2)] = CMPLX_ADD7(y1, CMPLX_MUL(ru4,y2), CMPLX_MUL(ru1,y3), CMPLX_MUL(ru5,y4), CMPLX_MUL(ru2,y5), CMPLX_MUL(ru6,y6), CMPLX_MUL(ru3,y7));
+                y[tensorOffset+rts*5]    = CMPLX_ADD7(y1, CMPLX_MUL(ru5,y2), CMPLX_MUL(ru3,y3), CMPLX_MUL(ru1,y4), CMPLX_MUL(ru6,y5), CMPLX_MUL(ru4,y6), CMPLX_MUL(ru2,y7));
+                y[tensorOffset+rts*6]    = CMPLX_ADD7(y1, CMPLX_MUL(ru6,y2), CMPLX_MUL(ru5,y3), CMPLX_MUL(ru4,y4), CMPLX_MUL(ru3,y5), CMPLX_MUL(ru2,y6), CMPLX_MUL(ru1,y7));
+            }   
+        }
+    }
+    else
+    {
+        hDim_t temp1 = rts*p;
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hDim_t row, col;
+                
+                for(row = 0; row < p; row++)
+                {
+                    complex_t acc = ((complex_t){0,0});
+                    for(col = 0; col < p; col++)
+                    {
+                        CMPLX_IADD(acc, CMPLX_MUL(y[tensorOffset+col*rts], ru[((col*row) % p)*rustride]));
+                    }
+                    tempSpace[row] = acc;
+                }
+                
+                for(row = 0; row < p; row++)
+                {
+                    y[tensorOffset+rts*row] = tempSpace[row];   
+                }
+            }
+        }
+    }
+}
+
+void crtpC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, complex_t* ru)
+{
+    hDim_t blockOffset, modOffset, tensorOffset;
+    
+    if(p == 2)
+    {
+        return;
+    }
+    else if(p == 3)
+    {
+        hDim_t temp1 = rts*2;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y[tensorOffset]     = CMPLX_ADD(y1, CMPLX_MUL(ru1,y2));
+                y[tensorOffset+rts] = CMPLX_ADD(y1, CMPLX_MUL(ru2,y2));
+            }   
+        }
+    }
+    else if(p == 5)
+    {
+        hDim_t temp1 = rts*4;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+        complex_t ru3 = ru[rustride*3];
+        complex_t ru4 = ru[rustride<<2];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2, y3, y4;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y[tensorOffset]          = CMPLX_ADD4(y1, CMPLX_MUL(ru1,y2), CMPLX_MUL(ru2,y3), CMPLX_MUL(ru3,y4));
+                y[tensorOffset+rts]      = CMPLX_ADD4(y1, CMPLX_MUL(ru2,y2), CMPLX_MUL(ru4,y3), CMPLX_MUL(ru1,y4));
+                y[tensorOffset+(rts<<1)] = CMPLX_ADD4(y1, CMPLX_MUL(ru3,y2), CMPLX_MUL(ru1,y3), CMPLX_MUL(ru4,y4));
+                y[tensorOffset+rts*3]    = CMPLX_ADD4(y1, CMPLX_MUL(ru4,y2), CMPLX_MUL(ru3,y3), CMPLX_MUL(ru2,y4));
+            }   
+        }
+    }
+    else if(p == 7)
+    {
+        hDim_t temp1 = rts*6;
+        complex_t ru1 = ru[rustride];
+        complex_t ru2 = ru[rustride<<1];
+        complex_t ru3 = ru[rustride*3];
+        complex_t ru4 = ru[rustride<<2];
+        complex_t ru5 = ru[rustride*5];
+        complex_t ru6 = ru[rustride*6];
+
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                complex_t y1, y2, y3, y4, y5, y6;
+                y1 = y[tensorOffset];
+                y2 = y[tensorOffset+rts];
+                y3 = y[tensorOffset+(rts<<1)];
+                y4 = y[tensorOffset+3*rts];
+                y5 = y[tensorOffset+(rts<<2)];
+                y6 = y[tensorOffset+rts*5];
+                y[tensorOffset]          = CMPLX_ADD6(y1, CMPLX_MUL(ru1,y2), CMPLX_MUL(ru2,y3), CMPLX_MUL(ru3,y4), CMPLX_MUL(ru4,y5), CMPLX_MUL(ru5,y6));
+                y[tensorOffset+rts]      = CMPLX_ADD6(y1, CMPLX_MUL(ru2,y2), CMPLX_MUL(ru4,y3), CMPLX_MUL(ru6,y4), CMPLX_MUL(ru1,y5), CMPLX_MUL(ru3,y6));
+                y[tensorOffset+(rts<<1)] = CMPLX_ADD6(y1, CMPLX_MUL(ru3,y2), CMPLX_MUL(ru6,y3), CMPLX_MUL(ru2,y4), CMPLX_MUL(ru5,y5), CMPLX_MUL(ru1,y6));
+                y[tensorOffset+rts*3]    = CMPLX_ADD6(y1, CMPLX_MUL(ru4,y2), CMPLX_MUL(ru1,y3), CMPLX_MUL(ru5,y4), CMPLX_MUL(ru2,y5), CMPLX_MUL(ru6,y6));
+                y[tensorOffset+(rts<<2)] = CMPLX_ADD6(y1, CMPLX_MUL(ru5,y2), CMPLX_MUL(ru3,y3), CMPLX_MUL(ru1,y4), CMPLX_MUL(ru6,y5), CMPLX_MUL(ru4,y6));
+                y[tensorOffset+rts*5]    = CMPLX_ADD6(y1, CMPLX_MUL(ru6,y2), CMPLX_MUL(ru5,y3), CMPLX_MUL(ru4,y4), CMPLX_MUL(ru3,y5), CMPLX_MUL(ru2,y6));
+            }   
+        }
+    }
+    else
+    {
+        complex_t* tempSpace = (complex_t*)malloc((p-1)*sizeof(complex_t));
+        hDim_t temp1 = rts*(p-1);
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+                hDim_t row, col;
+                
+                for(row = 1; row < p; row++)
+                {
+                    complex_t acc = ((complex_t){0,0});
+                    for(col = 0; col < p-1; col++)
+                    {
+                        CMPLX_IADD(acc, CMPLX_MUL(y[tensorOffset+col*rts], ru[((col*row) % p)*rustride]));
+                    }
+                    tempSpace[row-1] = acc;
+                }
+                
+                for(row = 0; row < p-1; row++)
+                {
+                    y[tensorOffset+rts*row] = tempSpace[row];   
+                }
+            }
+        }
+        free(tempSpace);
+    }
+}
+
+//takes inverse rus
+void crtpinvC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, complex_t* ruinv)
+{
+    if(p ==2)
+    {
+        // need this case so that we can divide overall by mhat^(-1)
+        return;
+    }
+    else
+    {
+        hDim_t tensorOffset,i;
+        complex_t* tempSpace = (complex_t*)malloc(p*sizeof(complex_t));
+        hDim_t temp1 = rts*(p-1);
+        for(hDim_t blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1;
+            for(hDim_t modOffset = 0; modOffset < rts; modOffset++)
+            {
+                tensorOffset = temp2 + modOffset;
+
+                for(i = 0; i < p-1; i++)
+                {
+                    complex_t sum = ((complex_t){0,0});
+                    int j;
+                    for(j = 0; j < p-1; j++)
+                    {
+                        int ruIdx = (((j+1)*i) % p)*rustride;
+                        CMPLX_IADD(sum, CMPLX_MUL(y[tensorOffset+j*rts],ruinv[ruIdx]));
+                    }
+                    tempSpace[i] = sum;
+                }
+
+                complex_t shift = ((complex_t){0,0});
+                for(i = 0; i < p-1; i++)
+                {
+                    // we were given the inverse rus, so we need to negate the indices
+                    int ruIdx = p-(i+1);
+                    CMPLX_IADD(shift, CMPLX_MUL(y[tensorOffset+i*rts], ruinv[rustride*ruIdx]));
+                }
+
+                for(i = 0; i < p-1; i++)
+                {
+                    y[tensorOffset+i*rts] = CMPLX_SUB(tempSpace[i], shift); 
+                }
+            }
+        }
+    }
+}
+
+void ppDFTC (complex_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t rustride, complex_t* ru)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+    if(e == 0)
+    {
+        return;
+    }
+    
+    hDim_t primeRuStride = rustride*ipow(p,e-1);    
+    complex_t* temp = 0;
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        temp = (complex_t*)malloc(p*sizeof(complex_t));
+    }
+    hShort_t i;
+    
+    hDim_t ltsScale = ipow(p,e-1);
+    hDim_t rtsScale = 1;
+    hDim_t twidRuStride = rustride;
+    for(i = 0; i < e; i++)
+    {
+        hDim_t rtsDim = rts*rtsScale;
+        dftpC (y, lts*ltsScale, rtsDim, p, primeRuStride, ru, temp);
+        dftptwidC (y, lts, rtsDim, pe, ltsScale*p, twidRuStride, ru);
+        
+        ltsScale /= p;
+        rtsScale *= p;
+        twidRuStride *= p;
+        pe.exponent -= 1;
+    }
+    
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        free(temp);
+    }
+}
+
+void ppDFTInvC (complex_t* y, hDim_t lts, hDim_t rts, PrimeExponent pe, hDim_t rustride, complex_t* ru)
+{
+
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+    if(e == 0)
+    {
+        return;
+    }
+    hDim_t primeRuStride = rustride*ipow(p,e-1);
+    complex_t* temp = 0;
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        temp = (complex_t*)malloc(p*sizeof(complex_t));
+    }
+    hShort_t i;
+    
+    hDim_t ltsScale = 1;
+    hDim_t rtsScale = ipow(p,e-1);
+    hDim_t twidRuStride = primeRuStride;
+    pe.exponent = 1;
+    for(i = 0; i < e; i++)
+    {
+        hDim_t rtsDim = rts*rtsScale;
+        hDim_t ltsScaleP = ltsScale*p;
+        dftptwidC (y, lts, rtsDim, pe, ltsScaleP, twidRuStride, ru);
+        dftpC (y, lts*ltsScale, rtsDim, p, primeRuStride, ru, temp);
+        
+        ltsScale = ltsScaleP;
+        rtsScale /= p;
+        twidRuStride /= p;
+        pe.exponent += 1;
+    }
+    
+    if(p >= DFTP_GENERIC_SIZE)
+    {
+        free(temp);
+    }
+}
+
+void ppcrtC (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hDim_t e = pe.exponent;
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    hDim_t mprime = ipow(p,e-1);
+
+#ifdef DEBUG_MODE
+    printf("rus for p=%" PRId32 ", e=%" PRId16 "\t[", pe.prime, pe.exponent);
+    hDim_t i;
+    for(i = 0; i < ipow(p,e); i++) {
+        printf("(%f,%f),", ((complex_t*)ru)[i].real, ((complex_t*)ru)[i].imag);
+    }
+    printf("]\n");
+#endif
+
+    crtpC ((complex_t*)y, lts*mprime, rts, p, mprime, (complex_t*)ru);
+    crtTwiddleC ((complex_t*)y, lts, rts, pe, (complex_t*)ru);
+    pe.exponent -= 1;
+    ppDFTC ((complex_t*)y, lts, rts*(p-1), pe, p, (complex_t*)ru);
+}
+
+void ppcrtinvC (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* ru, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hDim_t e = pe.exponent;
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    hDim_t mprime = ipow(p,e-1);
+    
+    pe.exponent -= 1;
+    ppDFTInvC ((complex_t*)y, lts, rts*(p-1), pe, p, (complex_t*)ru);
+    pe.exponent += 1;
+    crtTwiddleC ((complex_t*)y, lts, rts, pe, (complex_t*)ru);
+    crtpinvC ((complex_t*)y, lts*mprime, rts, p, mprime, (complex_t*)ru);
+}
+
+void tensorCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** ru)
+{
+#ifdef STATS
+    struct timespec s1,t1;
+    crtCCtr++;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorCRTC\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\n[", totm, sizeOfPE);
+    hDim_t j;
+    for(j = 0; j < totm; j++) {
+        printf("(%f,%f),", y[j].real, y[j].imag);
+    }
+    printf("]\n[");
+    for(j = 0; j < sizeOfPE; j++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[j].prime, peArr[j].exponent);
+    }
+    printf("]\n");
+#endif
+    void** rus = (void**)malloc(sizeOfPE*sizeof(void*));
+    hShort_t i;
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        rus[i] = (void*) (ru[i]);
+    }
+	tensorFuserCRT (y, ppcrtC, totm, peArr, sizeOfPE, rus, 0);
+	free(rus);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    crtCTime = tsAdd(crtCTime, tsSubtract(t1,s1));
+#endif
+}
+
+//takes inverse rus
+void tensorCRTInvC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** ruinv, double mhatInv)
+{
+#ifdef STATS
+    struct timespec s1,t1;
+    crtInvCCtr++;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	hDim_t i;
+	
+	void** rus = (void**)malloc(sizeOfPE*sizeof(void*));
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        rus[i] = (void*) (ruinv[i]);
+    }
+	
+	tensorFuserCRT (y, ppcrtinvC, totm, peArr, sizeOfPE, rus, 0);
+	complex_t minvcmplx = ((complex_t){mhatInv,0});
+
+	for (hDim_t j = 0; j < totm; j++)
+	{
+	    CMPLX_IMUL(y[j], minvcmplx);
+	}
+	
+	free(rus);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    crtInvCTime = tsAdd(crtInvCTime, tsSubtract(t1,s1));
+#endif
+}
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/g.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/g.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/g.c
@@ -0,0 +1,685 @@
+#include "tensorTypes.h"
+
+
+void gPowR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p)
+{
+  hDim_t tmp1 = rts*(p-1);
+  hDim_t tmp2 = tmp1 - rts;
+  hDim_t blockOffset, modOffset;
+  hDim_t i;
+  for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+  {
+    hDim_t tmp3 = blockOffset * tmp1;
+    for (modOffset = 0; modOffset < rts; ++modOffset)
+    {
+      hDim_t tensorOffset = tmp3 + modOffset;
+      hInt_t last = y[tensorOffset + tmp2];
+      for (i = p-2; i != 0; --i)
+      {
+        hDim_t idx = tensorOffset + i * rts;
+        y[idx] += last - y[idx-rts];
+      }
+      y[tensorOffset] += last;
+    }
+  }
+}
+
+void gPowRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q)
+{
+	hDim_t tmp1 = rts*(p-1);
+	hDim_t tmp2 = tmp1 - rts;
+	hDim_t blockOffset, modOffset;
+	hDim_t i;
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp3 = blockOffset * tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp3 + modOffset;
+			hInt_t last = y[tensorOffset + tmp2];
+			for (i = p-2; i != 0; --i)
+			{
+				hDim_t idx = tensorOffset + i * rts;
+				y[idx] = (y[idx] + last - y[idx-rts]) % q;
+			}
+			y[tensorOffset] = (y[tensorOffset] + last) % q;
+		}
+	}
+}
+
+
+void ppGPowR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+     
+	if (p != 2)
+	{
+		gPowR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+	}
+}
+
+
+void ppGPowRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gPowRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+	}
+}
+
+
+
+void gDecR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p)
+{
+	hDim_t tmp1 = rts*(p-1);
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	hDim_t i;
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset * tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t acc = y[tensorOffset];
+			for (i = p-2; i != 0; --i)
+			{
+				hDim_t idx = tensorOffset + i * rts;
+				acc += y[idx];
+				y[idx] -= y[idx-rts];
+			}
+			y[tensorOffset] += acc;
+		}
+	}
+}
+
+void gDecRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q)
+{
+	hDim_t tmp1 = rts*(p-1);
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	hDim_t i;
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset * tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t acc = y[tensorOffset];
+			for (i = p-2; i != 0; --i)
+			{
+				hDim_t idx = tensorOffset + i * rts;
+        // acc is at most p*q << 64 bits, so no need to mod
+				acc = acc + y[idx];
+				y[idx] = (y[idx] - y[idx-rts]) % q;
+			}
+			y[tensorOffset] = (y[tensorOffset] + acc) % q;
+		}
+	}
+}
+
+void ppGDecR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gDecR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+	}
+}
+
+void ppGDecRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gDecRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+	}
+}
+
+
+void gInvPowR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p)
+{
+	hDim_t tmp1 = rts * (p-1);
+	hDim_t blockOffset, modOffset;
+	hDim_t i;
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset * tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t lelts = 0;
+			for (i = 0; i < p-1; ++i)
+			{
+				lelts += y[tensorOffset + i*rts];
+			}
+			hInt_t relts = 0;
+			for (i = p-2; i >= 0; --i)
+			{
+				hDim_t idx = tensorOffset + i*rts;
+				hInt_t z = y[idx];
+				y[idx] = (p-1-i) * lelts - (i+1)*relts;
+				lelts -= z;
+				relts += z;
+			}
+		}
+	}
+}
+
+void gInvPowRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q)
+{
+	hDim_t tmp1 = rts * (p-1);
+	hDim_t blockOffset, modOffset;
+	hDim_t i;
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset * tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t lelts = 0;
+      //lelts is at most p*q, so we can mod once at the end
+			for (i = 0; i < p-1; ++i)
+			{
+				lelts = lelts + y[tensorOffset + i*rts];
+			}
+      lelts = lelts % q;
+      //in the next loop, lelts <= p*q and relts <= p*q
+      //products are <= p*p*q, and diff is <= 2*p*p*q
+      //so we assume 2*p^2 << 31 bits
+			hInt_t relts = 0;
+			for (i = p-2; i >= 0; --i)
+			{
+				hDim_t idx = tensorOffset + i*rts;
+				hInt_t z = y[idx];
+				y[idx] = (((p-1-i) * lelts) - ((i+1)*relts)) % q;
+				lelts -= z;
+				relts += z;
+			}
+		}
+	}
+}
+
+
+void ppGInvPowR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gInvPowR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+	}
+}
+
+void ppGInvPowRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gInvPowRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+	}
+}
+
+//do not call for p=2!
+void gCRTRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t* gcoeffs, hInt_t q)
+{
+    hDim_t gindex;
+    hDim_t blockOffset, modOffset, idx;
+    hDim_t temp1 = rts*(p-1);
+    
+    for(gindex = 0; gindex < p-1; gindex++)
+    {
+        hInt_t coeff = gcoeffs[gindex];
+        hDim_t temp3 = gindex*rts;
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1 + temp3;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                idx = temp2 + modOffset;
+                y[idx] = (y[idx]*coeff)%q;
+            }
+        }
+    }
+}
+
+//do not call for p=2!
+void gCRTC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p, complex_t* gcoeffs)
+{
+    hDim_t gindex;
+    hDim_t blockOffset, modOffset, idx;
+    hDim_t temp1 = rts*(p-1);
+    
+    for(gindex = 0; gindex < p-1; gindex++)
+    {
+        complex_t coeff = gcoeffs[gindex];
+        hDim_t temp3 = gindex*rts;
+        for(blockOffset = 0; blockOffset < lts; blockOffset++)
+        {
+            hDim_t temp2 = blockOffset*temp1 + temp3;
+            for(modOffset = 0; modOffset < rts; modOffset++)
+            {
+                idx = temp2 + modOffset;
+                CMPLX_IMUL(y[idx],coeff);
+            }
+        }
+    }
+}
+
+void ppGCRTRq (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* gcoeffs, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+#ifdef DEBUG_MODE
+    printf("gcoeffs for p=%" PRId32 ", e=%" PRId16 "\t[", pe.prime, pe.exponent);
+    int i;
+    for(i = 0; i < ((p-1)*ipow(p,e-1)); i++) {
+        printf("%" PRId64 ",", ((hInt_t*)gcoeffs)[i]);
+    }
+    printf("]\n");
+#endif
+    
+	if (p != 2)
+	{
+		gCRTRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, (hInt_t*)gcoeffs, q);
+	}
+}
+
+void ppGCRTC (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* gcoeffs, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+    
+#ifdef DEBUG_MODE
+    printf("gcoeffs for p=%" PRId32 ", e=%" PRId16 "\t[", pe.prime, pe.exponent);
+    int i;
+    for(i = 0; i < ((p-1)*ipow(p,e-1)); i++) {
+        printf("(%f,%f),", ((complex_t*)gcoeffs)[i].real, ((complex_t*)gcoeffs)[i].imag);
+    }
+    printf("]\n");
+#endif
+    
+	if (p != 2)
+	{
+		gCRTC ((complex_t*)y, lts*ipow(p,e-1), rts, p, (complex_t*)gcoeffs);
+	}
+}
+
+void gInvDecR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p)
+{
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	hDim_t i;
+	hDim_t tmp1 = rts*(p-1);
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t lastOut = 0;
+			for (i=1; i < p; ++i)
+			{
+				lastOut += i * y[tensorOffset + (i-1)*rts];
+			}
+			hInt_t acc = lastOut / p;
+			ASSERT (acc * p == lastOut);  // this line asserts that lastOut % p == 0, without calling % operator
+			for (i = p-2; i > 0; --i)
+			{
+				hDim_t idx = tensorOffset + i*rts;
+				hInt_t tmp = acc;
+				acc -= y[idx]; // we already divided acc by p, do not multiply y[idx] by p
+				y[idx] = tmp;
+			}
+			y[tensorOffset] = acc;
+		}
+	}
+}
+
+void gInvDecRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q)
+{
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	hDim_t i;
+	hDim_t tmp1 = rts*(p-1);
+	hInt_t reciprocalOfP = reciprocal (q,p);
+
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)
+	{
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset)
+		{
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hInt_t lastOut = 0;
+			for (i=1; i < p; ++i)
+			{
+				lastOut += (i * y[tensorOffset + (i-1)*rts]);
+			}
+      //in the previous loop, |lastOut| <= p*p*q
+      lastOut = lastOut % q;
+			hInt_t acc = (lastOut * reciprocalOfP) % q;
+      // |acc| <= p*q
+			for (i = p-2; i > 0; --i)
+			{
+				hDim_t idx = tensorOffset + i*rts;
+				hInt_t tmp = acc;
+				acc = acc - y[idx];
+				y[idx] = tmp % q;
+			}
+			y[tensorOffset] = acc % q;
+		}
+	}
+}
+
+void ppGInvDecR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gInvDecR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+	}
+}
+
+void ppGInvDecRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	if (p != 2)
+	{
+		gInvDecRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+	}
+}
+
+#ifdef STATS
+int gprCtr = 0;
+int gprqCtr = 0;
+int gdrCtr = 0;
+int gdrqCtr = 0;
+int giprCtr = 0;
+int giprqCtr = 0;
+int gidrCtr = 0;
+int gidrqCtr = 0;
+int gcrqCtr = 0;
+int gccCtr = 0;
+int gicrqCtr = 0;
+int giccCtr = 0;
+
+struct timespec gprTime = {0,0};
+struct timespec gprqTime = {0,0};
+struct timespec gdrTime = {0,0};
+struct timespec gdrqTime = {0,0};
+struct timespec giprTime = {0,0};
+struct timespec giprqTime = {0,0};
+struct timespec gidrTime = {0,0};
+struct timespec gidrqTime = {0,0};
+struct timespec gcrqTime = {0,0};
+struct timespec gccTime = {0,0};
+#endif
+
+void tensorGPowR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE)
+{
+#ifdef STATS
+    gprCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGPowR, totm, peArr, sizeOfPE, 0);
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gprTime = tsAdd(gprTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGPowRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q)
+{
+#ifdef STATS
+    gprqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGPowRq, totm, peArr, sizeOfPE, q);
+
+  hDim_t j;
+	for(j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        y[j]+=q;
+	    }
+	}
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gprqTime = tsAdd(gprqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGDecR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE)
+{
+#ifdef STATS
+    gdrCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGDecR, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gdrTime = tsAdd(gdrTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGDecRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q)
+{
+#ifdef STATS
+    gdrqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGDecRq, totm, peArr, sizeOfPE, q);
+
+  hDim_t j;
+	for(j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        y[j]+=q;
+	    }
+	}
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gdrqTime = tsAdd(gdrqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGInvPowR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE)
+{
+#ifdef STATS
+    giprCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGInvPowR, totm, peArr, sizeOfPE, 0);
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    giprTime = tsAdd(giprTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGInvPowRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q)
+{
+#ifdef STATS
+    giprqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGInvPowRq, totm, peArr, sizeOfPE, q);
+
+  hDim_t j;
+	for(j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        y[j]+=q;
+	    }
+	}
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    giprqTime = tsAdd(giprqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGInvDecR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE)
+{
+#ifdef STATS
+    gidrCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppGInvDecR, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gidrTime = tsAdd(gidrTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGInvDecRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q)
+{
+#ifdef STATS
+    gidrqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+    tensorFuser (y, ppGInvDecRq, totm, peArr, sizeOfPE, q);
+
+  hDim_t j;
+	for(j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        y[j]+=q;
+	    }
+	}
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gidrqTime = tsAdd(gidrqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorGCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** gcoeffs, hInt_t q)
+{
+#ifdef STATS
+    gcrqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorGCRTRq\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\tq=%" PRId64 "\n[", totm, sizeOfPE, q);
+    hDim_t j;
+    for(j = 0; j < totm; j++) {
+        printf("%" PRId64 ",", y[j]);
+    }
+    printf("]\n[");
+    for(j = 0; j < sizeOfPE; j++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[j].prime, peArr[j].exponent);
+    }
+    printf("]\n");
+#endif
+    void** vgcoeffs = (void**)malloc(sizeOfPE*sizeof(void*));
+    hDim_t i;
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        vgcoeffs[i] = (void*) (gcoeffs[i]);
+    }
+
+    tensorFuserCRT (y, ppGCRTRq, totm, peArr, sizeOfPE, vgcoeffs, q);
+
+#ifdef DEBUG_MODE
+    for(j = 0; j < totm; j++)
+	{
+	    if(y[j]<0)
+	    {
+	        printf("tensorGCRTRq\n");
+	    }
+	}
+#endif
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gcrqTime = tsAdd(gcrqTime, tsSubtract(t1,s1));
+#endif
+}
+void tensorGCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** gcoeffs)
+{
+#ifdef STATS
+    gccCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorGCRTC\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\n[", totm, sizeOfPE);
+    hDim_t j;
+    for(j = 0; j < totm; j++) {
+        printf("(%f,%f),", (y[j]).real, (y[j]).imag);
+    }
+    printf("]\n[");
+    for(j = 0; j < sizeOfPE; j++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[j].prime, peArr[j].exponent);
+    }
+    printf("]\n");
+#endif
+    void** vgcoeffs = (void**)malloc(sizeOfPE*sizeof(void*));
+    hDim_t i;
+    for(i = 0; i < sizeOfPE; i++)
+    {
+        vgcoeffs[i] = (void*) (gcoeffs[i]);
+    }
+
+    tensorFuserCRT (y, ppGCRTC, totm, peArr, sizeOfPE, vgcoeffs, 0);
+
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    gccTime = tsAdd(gccTime, tsSubtract(t1,s1));
+#endif
+}
+void tensorGInvCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** gcoeffs, hInt_t q)
+{
+#ifdef STATS
+    gicrqCtr++;
+#endif
+    tensorGCRTRq (y, totm, peArr, sizeOfPE, gcoeffs, q); //output is already shifted
+}
+void tensorGInvCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** gcoeffs)
+{
+#ifdef STATS
+    giccCtr++;
+#endif
+    tensorGCRTC (y, totm, peArr, sizeOfPE, gcoeffs);
+}
+
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/generalfuncs.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/generalfuncs.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/generalfuncs.c
@@ -0,0 +1,293 @@
+#include "tensorTypes.h"
+
+hDim_t ipow(hDim_t base, hShort_t exp)
+{
+#ifdef DEBUG_MODE
+    ASSERT(exp >= 0);
+#endif
+	hDim_t result = 1;
+    while (exp)
+    {
+        if (exp & 1)
+        {
+            result *= base;
+        }
+        exp >>= 1;
+        base *= base;
+    }
+    return result;
+}
+
+complex_t cmplxpow(complex_t base, hShort_t exp)
+{
+	complex_t result = (complex_t){1,0};
+    while (exp)
+    {
+        if (exp & 1)
+        {
+            CMPLX_IMUL(result,base);
+        }
+        exp >>= 1;
+        CMPLX_IMUL(base,base);
+    }
+    return result;
+}
+
+hInt_t qpow(hInt_t base, hShort_t exp, hInt_t q)
+{
+	hInt_t result = 1;
+    while (exp)
+    {
+        if (exp & 1)
+        {
+            result = (result*base)%q;
+        }
+        exp >>= 1;
+        base = (base*base)%q;
+    }
+    return result;
+}
+
+// a is the field size. we are looking for reciprocal of b
+hInt_t reciprocal (hInt_t a, hInt_t b)
+{
+	hInt_t fieldSize = a;
+
+	hInt_t y = 1;
+	hInt_t lasty = 0;
+	while (b != 0)
+	{
+		hInt_t quotient = a / b;
+		hInt_t tmp = a % b;
+		a = b;
+		b = tmp;
+		tmp = y;
+		y  = lasty - quotient*y;
+		lasty = tmp;
+	}
+	ASSERT (a==1);  // if this one fails, then b is not invertible mod a
+
+	// this actually returns EITHER the reciprocal OR reciprocal + fieldSize
+	hInt_t res = lasty + fieldSize;
+#ifdef DEBUG_MODE
+	ASSERT (0);
+	ASSERT ((res >= 0) && (res < fieldSize + fieldSize));
+	hInt_t test = res * b % fieldSize;
+	ASSERT (test == 1);
+#endif
+	return res;
+
+}
+
+//for square transforms
+void tensorFuser (void* y, funcPtr f, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q)
+{
+    hDim_t lts = totm;
+    hDim_t rts = 1;
+    hShort_t i;
+
+    for (i = 0; i < sizeOfPE; ++i)
+    {
+        PrimeExponent pe = peArr[i];
+        hDim_t ipow_pe = ipow(pe.prime, (pe.exponent-1));
+        hDim_t dim = (pe.prime-1) * ipow_pe;  // the totient of pe
+        lts /= dim;
+        (*f) (y, pe, lts, rts, q);
+        rts  *= dim;
+    }
+}
+
+void tensorFuserCRT (void* y, crtFuncPtr f, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, void** ru, hInt_t q)
+{
+    hDim_t lts = totm;
+    hDim_t rts = 1;
+    hShort_t i;
+
+    for (i = 0; i < sizeOfPE; ++i)
+    {
+        PrimeExponent pe = peArr[i];
+        hDim_t ipow_pe = ipow(pe.prime, (pe.exponent-1));
+        hDim_t dim = (pe.prime-1) * ipow_pe;  // the totient of pe
+        lts /= dim;
+        (*f) (y, lts, rts, pe, ru[i], q);
+        rts  *= dim;
+    }
+}
+
+struct  timespec  tsSubtract (struct  timespec  time1, struct  timespec  time2)
+{    /* Local variables. */
+    struct  timespec  result ;
+
+/* Subtract the second time from the first. */
+
+    if ((time1.tv_sec < time2.tv_sec) ||
+        ((time1.tv_sec == time2.tv_sec) &&
+         (time1.tv_nsec <= time2.tv_nsec))) {		/* TIME1 <= TIME2? */
+        result.tv_sec = result.tv_nsec = 0 ;
+    } else {						/* TIME1 > TIME2 */
+        result.tv_sec = time1.tv_sec - time2.tv_sec ;
+        if (time1.tv_nsec < time2.tv_nsec) {
+            result.tv_nsec = time1.tv_nsec + 1000000000L - time2.tv_nsec ;
+            result.tv_sec-- ;				/* Borrow a second. */
+        } else {
+            result.tv_nsec = time1.tv_nsec - time2.tv_nsec ;
+        }
+    }
+
+    return (result) ;
+}
+
+struct  timespec  tsAdd (struct  timespec  time1, struct  timespec  time2)
+{    /* Local variables. */
+    struct  timespec  result ;
+
+/* Add the two times together. */
+
+    result.tv_sec = time1.tv_sec + time2.tv_sec ;
+    result.tv_nsec = time1.tv_nsec + time2.tv_nsec ;
+    if (result.tv_nsec >= 1000000000L) {		/* Carry? */
+        result.tv_sec++ ;  result.tv_nsec = result.tv_nsec - 1000000000L ;
+    }
+
+    return (result) ;
+}
+
+const  char  *tsShow (struct  timespec  binaryTime, bool  inLocal, const  char  *format)
+{    /* Local variables. */
+    struct  tm  calendarTime ;
+#define  MAX_TIMES  4
+    static  char  asciiTime[MAX_TIMES][64] ;
+    static  int  current = 0 ;
+
+/* Convert the TIMESPEC to calendar time: year, month, day, etc. */
+
+#ifdef VXWORKS
+    if (inLocal)
+        localtime_r ((time_t *) &binaryTime.tv_sec, &calendarTime) ;
+    else
+        gmtime_r ((time_t *) &binaryTime.tv_sec, &calendarTime) ;
+#else
+    if (inLocal)
+        calendarTime = *(localtime ((time_t *) &binaryTime.tv_sec)) ;
+    else
+        calendarTime = *(gmtime ((time_t *) &binaryTime.tv_sec)) ;
+#endif
+
+/* Format the time in ASCII. */
+
+    current = (current + 1) % MAX_TIMES ;
+
+    if (format == NULL) {
+        strftime (asciiTime[current], 64, "%Y-%j-%H:%M:%S", &calendarTime) ;
+        sprintf (asciiTime[current] + strlen (asciiTime[current]),
+                 ".%06ld", (binaryTime.tv_nsec % 1000000000L) / 1000L) ;
+    } else {
+        strftime (asciiTime[current], 64, format, &calendarTime) ;
+        sprintf (asciiTime[current] + strlen (asciiTime[current]),
+                 ".%06ld", (binaryTime.tv_nsec % 1000000000L) / 1000L) ;
+    }
+
+    return (asciiTime[current]);
+}
+
+
+
+const char* timeformat = "%M:%S";
+
+void getStats() { 
+
+#ifdef STATS
+    struct timespec total;
+    printf("CRT Stats:\n");
+    printf("CRT_Rq times: Real:%s\tMono:%s\tProc:%s\tThread:%s\n", tsShow(crttime1, false, timeformat),tsShow(crttime2, false, timeformat),tsShow(crttime3, false, timeformat),tsShow(crttime4, false, timeformat));
+    printf("CTR_Rq: %d\t%s\t%d\t%s\n", crtRqCtr, tsShow(crttime1, false, timeformat), crtInvRqCtr, tsShow(crtInvRqTime, false, timeformat));
+    printf("CTR_C: %d\t%s\t%d\t%s\n", crtCCtr, tsShow(crtCTime, false, timeformat), crtInvCCtr, tsShow(crtInvCTime, false, timeformat));
+    
+    printf("\nG Stats:\n");
+    printf("GPow_R: %d\t%s\t%d\t%s\n", gprCtr, tsShow(gprTime, false, timeformat), giprCtr, tsShow(giprTime, false, timeformat));
+    printf("GPow_Rq: %d\t%s\t%d\t%s\n", gprqCtr, tsShow(gprqTime, false, timeformat), giprqCtr, tsShow(giprqTime, false, timeformat));
+    printf("GDec_R: %d\t%s\t%d\t%s\n", gdrCtr, tsShow(gdrTime, false, timeformat), gidrCtr, tsShow(gidrTime, false, timeformat));
+    printf("GDec_Rq: %d\t%s\t%d\t%s\n", gdrqCtr, tsShow(gdrqTime, false, timeformat), gidrqCtr, tsShow(gidrqTime, false, timeformat));
+    printf("GCRT_Rq: %d\t%d\t%s\n", gcrqCtr, gicrqCtr, tsShow(gcrqTime, false, timeformat));
+    printf("GCRT_C: %d\t%d\t%s\n", gccCtr, giccCtr, tsShow(gccTime, false, timeformat));
+
+    printf("\nL Stats:\n");
+    printf("L_R: %d\t%s\t%d\t%s\n", lrCtr, tsShow(lrTime, false, timeformat), lirCtr, tsShow(lirTime, false, timeformat));
+    printf("L_Rq: %d\t%s\t%d\t%s\n", lrqCtr, tsShow(lrqTime, false, timeformat), lirqCtr, tsShow(lirqTime, false, timeformat));
+    printf("L_D: %d\t%s\t%d\t%s\n", ldCtr, tsShow(ldTime, false, timeformat), lidCtr, tsShow(lidTime, false, timeformat));
+    printf("L_C: %d\t%s\t%d\t%s\n", lcCtr, tsShow(lcTime, false, timeformat), licCtr, tsShow(licTime, false, timeformat));
+
+    printf("\nBasic Stats:\n");
+    printf("Mul: %d\t%s\n", mulCtr, tsShow(mulTime, false, timeformat));
+    printf("Add: %d\t%s\n", addCtr, tsShow(addTime, false, timeformat));
+
+    total = tsAdd(crttime1, tsAdd(crtInvRqTime, tsAdd(crtCTime, tsAdd(crtInvCTime, tsAdd(gprTime, tsAdd(giprTime, tsAdd(gdrTime, tsAdd(gidrTime, tsAdd(gprqTime, tsAdd(giprqTime, tsAdd(gdrqTime, tsAdd(gidrqTime, tsAdd(gcrqTime, tsAdd(gccTime, tsAdd(lrTime, tsAdd(lirTime, tsAdd(lrqTime, tsAdd(lirqTime, tsAdd(ldTime, tsAdd(lidTime, tsAdd(lcTime, tsAdd(licTime, tsAdd(mulTime,addTime)))))))))))))))))))))));
+
+    printf("\nTotal C Time: %s\n\n", tsShow(total, false, timeformat));
+
+    crtRqCtr = 0;
+    crtInvRqCtr = 0;
+    crtCCtr = 0;
+    crtInvCCtr = 0;
+
+    gprCtr = 0;
+    gprqCtr = 0;
+    gdrCtr = 0;
+    gdrqCtr = 0;
+    giprCtr = 0;
+    giprqCtr = 0;
+    gidrCtr = 0;
+    gidrqCtr = 0;
+    gcrqCtr = 0;
+    gccCtr = 0;
+    gicrqCtr = 0;
+    giccCtr = 0;
+
+    lrqCtr = 0;
+    lrCtr = 0;
+    ldCtr = 0;
+    lcCtr = 0;
+    lirqCtr = 0;
+    lirCtr = 0;
+    lidCtr = 0;
+    licCtr = 0;
+
+    mulCtr = 0;
+    addCtr = 0;
+
+    mulTime = (struct timespec){0,0};
+    addTime = (struct timespec){0,0};
+
+    lrqTime = (struct timespec){0,0};
+    lrTime = (struct timespec){0,0};
+    ldTime = (struct timespec){0,0};
+    lcTime = (struct timespec){0,0};
+    lirqTime = (struct timespec){0,0};
+    lirTime = (struct timespec){0,0};
+    lidTime = (struct timespec){0,0};
+    licTime = (struct timespec){0,0};
+
+    gprTime = (struct timespec){0,0};
+    gprqTime = (struct timespec){0,0};
+    gdrTime = (struct timespec){0,0};
+    gdrqTime = (struct timespec){0,0};
+    giprTime = (struct timespec){0,0};
+    giprqTime = (struct timespec){0,0};
+    gidrTime = (struct timespec){0,0};
+    gidrqTime = (struct timespec){0,0};
+    gcrqTime = (struct timespec){0,0};
+    gccTime = (struct timespec){0,0};
+
+    crttime1 = (struct timespec){0,0};
+    crttime2 = (struct timespec){0,0};
+    crttime3 = (struct timespec){0,0};
+    crttime4 = (struct timespec){0,0};
+
+    crtInvRqTime = (struct timespec){0,0};
+    crtCTime = (struct timespec){0,0};
+    crtInvCTime = (struct timespec){0,0};
+#endif
+    fflush(stdout);
+}
+
+
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/l.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/l.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/l.c
@@ -0,0 +1,410 @@
+#include "tensorTypes.h"
+
+void lpRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset) {
+			hDim_t idx = tmp2 + modOffset + rts;
+			for (i = 1; i < p-1; ++i) {
+        hInt_t temp = y[idx-rts] + y[idx];
+        if (temp >= q) y[idx]=temp-q;
+        else y[idx] = temp;
+				idx += rts;
+			}
+		}
+	}
+}
+
+void lpR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)	{
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset) {
+			hDim_t idx = tmp2 + modOffset + rts;
+			for (i = 1; i < p-1; ++i) {
+				y[idx] += y[idx-rts];
+				idx += rts;
+			}
+		}
+	}
+}
+
+void lpDouble (double* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset) {
+			hDim_t idx = tmp2 + modOffset + rts;
+			for (i = 1; i < p-1; ++i) {
+				y[idx] += y[idx-rts];
+				idx += rts;
+			}
+		}
+	}
+}
+
+void lpC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++modOffset) {
+			hDim_t idx = tmp2 + modOffset + rts;
+			for (i = 1; i < p-1; ++i) {
+				CMPLX_IADD (y[idx], y[idx-rts]);
+				idx += rts;
+			}
+		}
+	}
+}
+
+void lpInvRq (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p, hInt_t q) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset)	{
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++ modOffset) {
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hDim_t idx = tensorOffset + (p-2) * rts;
+			for (i = p-2; i != 0; --i) {
+        hInt_t temp = y[idx] - y[idx-rts] + q;
+        if (temp >= q) y[idx]=temp-q;
+        else y[idx] = temp;
+				idx -= rts;
+			}
+		}
+	}
+}
+
+void lpInvR (hInt_t* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++ modOffset) {
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hDim_t idx = tensorOffset + (p-2) * rts;
+			for (i = p-2; i != 0; --i) {
+				y[idx] -= y[idx-rts] ;
+				idx -= rts;
+			}
+		}
+	}
+}
+
+void lpInvDouble (double* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++ modOffset) {
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hDim_t idx = tensorOffset + (p-2) * rts;
+			for (i = p-2; i != 0; --i) {
+				y[idx] -= y[idx-rts] ;
+				idx -= rts;
+			}
+		}
+	}
+}
+
+void lpInvC (complex_t* y, hDim_t lts, hDim_t rts, hDim_t p) {
+	hDim_t blockOffset;
+	hDim_t modOffset;
+	int i;
+
+	hDim_t tmp1 = rts*(p-1);
+	for (blockOffset = 0; blockOffset < lts; ++blockOffset) {
+		hDim_t tmp2 = blockOffset*tmp1;
+		for (modOffset = 0; modOffset < rts; ++ modOffset) {
+			hDim_t tensorOffset = tmp2 + modOffset;
+			hDim_t idx = tensorOffset + (p-2) * rts;
+			for (i = p-2; i != 0; --i) {
+				CMPLX_ISUB (y[idx], y[idx-rts]);
+				idx -= rts;
+			}
+		}
+	}
+}
+
+void ppLRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+}
+
+void ppLR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+}
+
+void ppLDouble (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpDouble ((double*)y, lts*ipow(p,e-1), rts, p);
+}
+
+void ppLC (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpC ((complex_t*)y, lts*ipow(p,e-1), rts, p);
+}
+
+
+void ppLInvRq (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpInvRq ((hInt_t*)y, lts*ipow(p,e-1), rts, p, q);
+}
+
+void ppLInvR (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpInvR ((hInt_t*)y, lts*ipow(p,e-1), rts, p);
+}
+
+void ppLInvDouble (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpInvDouble ((double*)y, lts*ipow(p,e-1), rts, p);
+}
+
+void ppLInvC (void* y, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q) {
+#ifdef DEBUG_MODE
+	ASSERT (q==0);
+#endif
+    hDim_t p = pe.prime;
+    hShort_t e = pe.exponent;
+	lpInvC ((complex_t*)y, lts*ipow(p,e-1), rts, p);
+}
+
+#ifdef STATS
+int lrqCtr = 0;
+int lrCtr = 0;
+int ldCtr = 0;
+int lcCtr = 0;
+int lirqCtr = 0;
+int lirCtr = 0;
+int lidCtr = 0;
+int licCtr = 0;
+
+struct timespec lrqTime = {0,0};
+struct timespec lrTime = {0,0};
+struct timespec ldTime = {0,0};
+struct timespec lcTime = {0,0};
+struct timespec lirqTime = {0,0};
+struct timespec lirTime = {0,0};
+struct timespec lidTime = {0,0};
+struct timespec licTime = {0,0};
+#endif
+
+
+void tensorLRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q) {
+#ifdef STATS
+    lrqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    hDim_t i;
+    printf("\n\nEntered tensorLRq\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\tq=%" PRId64 "\n[", totm, sizeOfPE, q);
+    /*for(i = 0; i < totm; i++) {
+        printf("%" PRId64 ",", y[i]);
+    }*/
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+	tensorFuser (y, ppLRq, totm, peArr, sizeOfPE, q); // don't need to shift here
+#ifdef DEBUG_MODE
+	for(i = 0; i < totm; i++) {
+	    if(y[i]<0) {
+	        printf("tensorLRq\n");
+	    }
+	}
+#endif
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lrqTime = tsAdd(lrqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    lrCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorLR\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\n[", totm, sizeOfPE);
+    hDim_t i;
+    for(i = 0; i < totm; i++) {
+        printf("%" PRId64 ",", y[i]);
+    }
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+	tensorFuser (y, ppLR, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lrTime = tsAdd(lrTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLDouble (double* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    ldCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorLDouble\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\n[", totm, sizeOfPE);
+    hDim_t i;
+    for(i = 0; i < totm; i++) {
+        printf("%f,", y[i]);
+    }
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+	tensorFuser (y, ppLDouble, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    ldTime = tsAdd(ldTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    lcCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+#ifdef DEBUG_MODE
+    printf("\n\nEntered tensorLC\ttotm=%" PRId32 "\tnumFacts=%" PRId16 "\n[", totm, sizeOfPE);
+    hDim_t i;
+    for(i = 0; i < totm; i++) {
+        printf("(%f,%f),", y[i].real, y[i].imag);
+    }
+    printf("]\n[");
+    for(i = 0; i < sizeOfPE; i++) {
+        printf("(%" PRId32 ",%" PRId16 "),", peArr[i].prime, peArr[i].exponent);
+    }
+    printf("]\n");
+#endif
+	tensorFuser (y, ppLC, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lcTime = tsAdd(lcTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLInvRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q) {
+#ifdef STATS
+    lirqCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppLInvRq, totm, peArr, sizeOfPE, q);  // don't need to shift here
+#ifdef DEBUG_MODE
+	hDim_t i;
+	for(i = 0; i < totm; i++)
+	{
+	    if(y[i]<0)
+	    {
+	        printf("tensorLInvRq\n");
+	    }
+	}
+#endif
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lirqTime = tsAdd(lirqTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLInvR (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    lirCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppLInvR, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lirTime = tsAdd(lirTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLInvDouble (double* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    lidCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppLInvDouble, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    lidTime = tsAdd(lidTime, tsSubtract(t1,s1));
+#endif
+}
+
+void tensorLInvC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE) {
+#ifdef STATS
+    licCtr++;
+    struct timespec s1,t1;
+    clock_gettime(CLOCK_REALTIME, &s1);
+#endif
+	tensorFuser (y, ppLInvC, totm, peArr, sizeOfPE, 0);
+#ifdef STATS
+    clock_gettime(CLOCK_REALTIME, &t1);
+    licTime = tsAdd(licTime, tsSubtract(t1,s1));
+#endif
+}
+
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/random.c b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/random.c
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/random.c
@@ -0,0 +1,72 @@
+
+#include <math.h>
+#include <stdlib.h>
+#include "tensorTypes.h"
+
+// this function takes *inverse* RUs, so no negation is needed on the indexing
+// I had been negating the ru-idx, but this was causing a *negative* mod, resulting in a hard-to-find bug
+void primeD (double *y, hDim_t lts, hDim_t rts, hDim_t p, hDim_t rustride, complex_t* ruinv)
+{
+	if(p == 2)
+  {
+      return;
+  }
+  hDim_t blockOffset, modOffset, tensorOffset;
+	double *tempSpace = (double*)malloc((p-1)*sizeof(double));
+  hDim_t temp1 = rts*(p-1);
+  for(blockOffset = 0; blockOffset < lts; blockOffset++)
+  {
+    hDim_t temp2 = blockOffset*temp1;
+    for(modOffset = 0; modOffset < rts; modOffset++)
+    {
+      tensorOffset = temp2 + modOffset;
+      hDim_t row, col;
+      
+      for(row = 0; row < p-1; row++)
+      {
+        double acc = 0;
+        for(col = 1; col <= (p>>1); col++)
+        {
+          acc += 2 * ruinv[((row*col) % p)*rustride].real * y[tensorOffset+rts*(col-1)];
+        }
+        for(col = (p>>1)+1; col <= p-1; col++)
+        {
+          acc += 2 * ruinv[((row*col) % p)*rustride].imag * y[tensorOffset+rts*(col-1)];
+        }
+        tempSpace[row] = acc/sqrt(2);
+      }
+      
+      for(row = 0; row < p-1; row++)
+      {
+        y[tensorOffset+rts*row] = tempSpace[row];
+      }
+    }
+  }
+  free(tempSpace);
+}
+
+void ppD (void *y, hDim_t lts, hDim_t rts, PrimeExponent pe, void *ruinv, hInt_t q)
+{
+    hDim_t p = pe.prime;
+    hDim_t e = pe.exponent;
+#ifdef DEBUG_MODE
+    ASSERT(e != 0);
+#endif
+    hDim_t mprime = ipow(p,e-1);
+    primeD (y, lts*mprime, rts, p, mprime, (complex_t*)ruinv);
+}
+
+//the contents of y will be destroyed, but should be initialized in Haskell-land to independent Guassians over the reals
+void tensorGaussianDec (double *y, hDim_t totm, PrimeExponent *peArr, hShort_t sizeOfPE, complex_t** ruinv)
+{
+  void** ruinvs = (void**)malloc(sizeOfPE*sizeof(void*));
+  hShort_t i;
+  for(i = 0; i < sizeOfPE; i++)
+  {
+      ruinvs[i] = (void*) (ruinv[i]);
+  }
+    
+	tensorFuserCRT (y, ppD, totm, peArr, sizeOfPE, ruinvs, 0);
+	
+	free(ruinvs);
+}
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/tensorTypes.h b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/tensorTypes.h
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/CTensor/tensorTypes.h
@@ -0,0 +1,224 @@
+
+#ifndef TENSORTYPES_H_
+#define TENSORTYPES_H_
+
+
+// remove next line for more efficient code
+//#define DEBUG_MODE
+
+
+#include <stdbool.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+
+#define ASSERT(EXP) { \
+	if (!(EXP)) { \
+		fprintf (stderr, "Assertion in file '%s' line %d : " #EXP "  is false\n", __FILE__, __LINE__); \
+		exit(-1); \
+	} \
+}
+
+
+//timers and counters
+#ifdef STATS
+extern int crtRqCtr;
+extern int crtInvRqCtr;
+extern int crtCCtr;
+extern int crtInvCCtr;
+
+extern int gprCtr;
+extern int gprqCtr;
+extern int gdrCtr;
+extern int gdrqCtr;
+extern int giprCtr;
+extern int giprqCtr;
+extern int gidrCtr;
+extern int gidrqCtr;
+extern int gcrqCtr;
+extern int gccCtr;
+extern int gicrqCtr;
+extern int giccCtr;
+
+extern int lrqCtr;
+extern int lrCtr;
+extern int ldCtr;
+extern int lcCtr;
+extern int lirqCtr;
+extern int lirCtr;
+extern int lidCtr;
+extern int licCtr;
+
+extern int mulCtr;
+extern struct timespec mulTime;
+extern int addCtr;
+extern struct timespec addTime;
+
+extern struct timespec lrqTime;
+extern struct timespec lrTime;
+extern struct timespec ldTime;
+extern struct timespec lcTime;
+extern struct timespec lirqTime;
+extern struct timespec lirTime;
+extern struct timespec lidTime;
+extern struct timespec licTime;
+
+extern struct timespec crttime1;
+extern struct timespec crttime2;
+extern struct timespec crttime3;
+extern struct timespec crttime4;
+extern struct timespec crtInvRqTime;
+extern struct timespec crtCTime;
+extern struct timespec crtInvCTime;
+
+extern struct timespec gprTime;
+extern struct timespec gprqTime;
+extern struct timespec gdrTime;
+extern struct timespec gdrqTime;
+extern struct timespec giprTime;
+extern struct timespec giprqTime;
+extern struct timespec gidrTime;
+extern struct timespec gidrqTime;
+extern struct timespec gcrqTime;
+extern struct timespec gccTime;
+#endif
+
+typedef int64_t hInt_t ;
+typedef int32_t hDim_t ;
+typedef int16_t hShort_t ;
+typedef int8_t hByte_t ;
+
+typedef struct
+{
+	hDim_t prime;
+	hShort_t exponent;
+}  PrimeExponent;
+
+
+typedef struct
+{
+	double real;
+	double imag;
+} complex_t;
+
+//complex_t _add (complex_t a, complex_t b);
+//complex_t _mul (complex_t a, complex_t b);
+
+#define CMPLX_ADD(a,b)  ((complex_t){((a).real + (b).real), ((a).imag + (b).imag)})
+#define CMPLX_ADD3(a,b,c)  ((complex_t){((a).real + (b).real + (c).real), ((a).imag + (b).imag + (c).imag)})
+#define CMPLX_ADD4(a,b,c,d)  ((complex_t){((a).real + (b).real + (c).real + (d).real), ((a).imag + (b).imag + (c).imag + (d).imag)})
+#define CMPLX_ADD5(a,b,c,d,e)  ((complex_t){((a).real + (b).real + (c).real + (d).real + (e).real), ((a).imag + (b).imag + (c).imag + (d).imag + (e).imag)})
+#define CMPLX_ADD6(a,b,c,d,e,f)  ((complex_t){((a).real + (b).real + (c).real + (d).real + (e).real + (f).real), ((a).imag + (b).imag + (c).imag + (d).imag + (e).imag + (f).imag)})
+#define CMPLX_ADD7(a,b,c,d,e,f,g)  ((complex_t){((a).real + (b).real + (c).real + (d).real + (e).real + (f).real + (g).real), ((a).imag + (b).imag + (c).imag + (d).imag + (e).imag + (f).imag + (g).imag)})
+
+#define CMPLX_SUB(a,b)  ((complex_t){((a).real - (b).real), ((a).imag - (b).imag)})
+#define CMPLX_MUL(a,b)  ((complex_t){((a).real*(b).real - (a).imag*(b).imag), \
+	                                  (a).real*(b).imag + (a).imag*(b).real})
+#define CMPLX_DIV(a,b)  ((complex_t){((a).real*(b).real + (a).imag*(b).imag)/((b).real*(b).real+(b).imag*(b).imag), \
+                                     ((a).imag*(b).real - (a).real*(b).imag)/((b).real*(b).real+(b).imag*(b).imag)})
+
+// 'inside' operators
+#define CMPLX_IADD(a,b)  { (a).real += (b).real;  (a).imag += (b).imag; }
+#define CMPLX_ISUB(a,b)  { (a).real -= (b).real;  (a).imag -= (b).imag; }
+#define CMPLX_IMUL(a,b)  { double temp = ((a).real*(b).real - (a).imag*(b).imag); \
+	                       (a).imag = ((a).real*(b).imag + (a).imag*(b).real); \
+	                       (a).real = temp; }
+
+
+/*
+// check if both vectors are identical
+bool eqInt (hInt_t a[], hInt_t b[], size_t n);
+// operations on integer vectors point-wise
+void addInt (hInt_t result[], hInt_t a[], hInt_t b[], size_t n);
+void mulInt (hInt_t result[], hInt_t a[], hInt_t b[], size_t n);
+*/
+
+// calculates base ** exp
+hDim_t ipow(hDim_t base, hShort_t exp);
+complex_t cmplxpow(complex_t base, hShort_t exp);
+hInt_t qpow(hInt_t base, hShort_t exp, hInt_t q);
+
+hInt_t reciprocal (hInt_t a, hInt_t b);
+
+struct  timespec  tsSubtract (struct  timespec  time1, struct  timespec  time2);
+struct  timespec  tsAdd (struct  timespec  time1, struct  timespec  time2);
+const  char  *tsShow (struct  timespec  binaryTime, bool  inLocal, const  char  *format);
+
+void getStats();
+
+void mulRq (hInt_t* a, hInt_t* b, hDim_t totm, hInt_t q);
+void mulMq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hByte_t logr, const hInt_t k, const hInt_t q);
+void mulC (complex_t* a, complex_t* b, hDim_t totm);
+
+void addR (hInt_t* a, hInt_t* b, hDim_t totm);
+void addRq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hInt_t q);
+void addMq (hInt_t* a, const hInt_t* b, const hDim_t totm, const hByte_t logr, const hInt_t k, const hInt_t q);
+void addC (complex_t* a, complex_t* b, hDim_t totm);
+void addD (double* a, double* b, hDim_t totm);
+
+typedef void (*funcPtr) (void* outputVec, PrimeExponent pe, hDim_t lts, hDim_t rts, hInt_t q);
+void tensorFuser (void* y, funcPtr f, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+typedef void (*crtFuncPtr) (void* y, hDim_t lts, hDim_t rts, PrimeExponent pe, void* ru, hInt_t q);
+void tensorFuserCRT (void* y, crtFuncPtr f, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, void** ru, hInt_t q);
+
+void tensorGPowR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorGPowRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorGDecR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorGDecRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorGInvPowR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorGInvPowRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorGInvDecR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorGInvDecRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorGCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** gcoeffs, hInt_t q);
+
+void tensorGInvCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** gcoeffs, hInt_t q);
+
+void tensorGCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** gcoeffs);
+
+void tensorGInvCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** gcoeffs);
+
+
+
+void tensorLRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorLR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorLDouble (double* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorLC (complex_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorLInvRq (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t q);
+
+void tensorLInvR (hInt_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorLInvDouble (double* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+void tensorLInvC (complex_t* x, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE);
+
+
+
+void tensorCRTRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** ru, hInt_t q);
+
+void tensorCRTC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** ru);
+
+void tensorCRTInvRq (hInt_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, hInt_t** ru, hInt_t minv, hInt_t q);
+
+void tensorCRTInvC (complex_t* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** ru, double minv);
+
+void tensorGaussianDec (double* y, hDim_t totm, PrimeExponent* peArr, hShort_t sizeOfPE, complex_t** ru);
+
+
+#endif /* TENSORTYPES_H_ */
+
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor.hs
@@ -0,0 +1,197 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             FlexibleContexts, FlexibleInstances, GADTs, InstanceSigs,
+             MultiParamTypeClasses, NoImplicitPrelude, RebindableSyntax,
+             RoleAnnotations, ScopedTypeVariables, StandaloneDeriving,
+             TypeFamilies, TypeOperators, UndecidableInstances #-}
+
+-- | A pure, repa-based implementation of the Tensor interface.
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+( RT ) where
+
+import Crypto.Lol.Cyclotomic.Tensor                      as T
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.CRT
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Extension
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Gauss
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.GL
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon  as RT
+import Crypto.Lol.LatticePrelude                         as LP hiding
+                                                                ((!!))
+import Crypto.Lol.Types.IZipVector
+
+import Algebra.Additive     as Additive (C)
+import Algebra.Ring         as Ring (C)
+import Algebra.ZeroTestable as ZeroTestable (C)
+
+import Control.Applicative
+import Control.DeepSeq       (NFData (rnf))
+import Control.Monad         (liftM)
+import Control.Monad.Random
+import Data.Coerce
+import Data.Constraint
+import Data.Foldable         as F
+import Data.Maybe
+import Data.Traversable      as T
+import Data.Typeable
+import Data.Vector.Unboxed   as U hiding (force)
+import Test.QuickCheck
+
+-- | An implementation of 'Tensor' backed by repa.
+data RT (m :: Factored) r where
+  RT :: Unbox r => !(Arr m r) -> RT m r
+  ZV :: IZipVector m r -> RT m r
+  deriving (Typeable)
+
+deriving instance Show r => Show (RT m r)
+
+instance Eq r => Eq (RT m r) where
+  (ZV a) == (ZV b) = a == b
+  (RT a) == (RT b) = a == b
+  a@(RT _) == b = a == toRT b
+  a == b@(RT _) = toRT a == b
+
+zvToArr :: Unbox r => IZipVector m r -> Arr m r
+zvToArr v = let vec = convert $ unIZipVector v
+            in Arr $ fromUnboxed (Z :. U.length vec) vec
+
+-- converts to RT constructor
+toRT :: Unbox r => RT m r -> RT m r
+toRT v@(RT _) = v
+toRT (ZV v) = RT $ zvToArr v
+
+toZV :: Fact m => RT m r -> RT m r
+toZV (RT (Arr v)) = ZV $ fromMaybe (error "toZV: internal error") $
+                    iZipVector $ convert $ toUnboxed v
+toZV v@(ZV _) = v
+
+wrap :: Unbox r => (Arr l r -> Arr m r) -> RT l r -> RT m r
+wrap f (RT v) = RT $ f v
+wrap f (ZV v) = RT $ f $ zvToArr v
+
+wrapM :: (Unbox r, Monad mon) => (Arr l r -> mon (Arr m r))
+         -> RT l r -> mon (RT m r)
+wrapM f (RT v) = liftM RT $ f v
+wrapM f (ZV v) = liftM RT $ f $ zvToArr v
+
+instance Tensor RT where
+
+  type TElt RT r = (IntegralDomain r, ZeroTestable r,
+                    Eq r, Random r, NFData r,
+                    Unbox r, Elt r)
+
+  entailIndexT  = tag $ Sub Dict
+  entailFullT   = tag $ Sub Dict
+
+  scalarPow = RT . scalarPow'
+
+  l = wrap fL
+  lInv = wrap fLInv
+
+  mulGPow = wrap fGPow
+  mulGDec = wrap fGDec
+
+  divGPow = wrapM  fGInvPow
+  divGDec = wrapM  fGInvDec
+
+  crtFuncs = (,,,,) <$>
+             (liftM (RT .) scalarCRT') <*>
+             (wrap <$> mulGCRT') <*>
+             (wrap <$> divGCRT') <*>
+             (wrap <$> fCRT) <*>
+             (wrap <$> fCRTInv)
+
+  -- instance sigs are the cleanest way to handle many weird types
+  -- coming up
+
+  tGaussianDec :: forall v rnd m q .
+                  (Fact m, OrdFloat q, Random q, TElt RT q,
+                   ToRational v, MonadRandom rnd) => v -> rnd (RT m q)
+  tGaussianDec = liftM RT . tGaussianDec'
+
+  twacePowDec = wrap twacePowDec'
+
+  embedPow = wrap embedPow'
+  embedDec = wrap embedDec'
+
+  crtExtFuncs = (,) <$> (liftM wrap twaceCRT')
+                    <*> (liftM wrap embedCRT')
+
+  coeffs = wrapM coeffs'
+
+  powBasisPow = (RT <$>) <$> powBasisPow'
+
+  crtSetDec = (RT <$>) <$> crtSetDec'
+
+  fmapT f (RT v) = RT $ (coerce $ force . RT.map f) v
+  fmapT f v@(ZV _) = fmapT f $ toRT v
+
+  -- Repa arrays don't have mapM, so apply to underlying Unboxed
+  -- vector instead
+  fmapTM f (RT (Arr arr)) = liftM (RT . Arr . fromUnboxed (extent arr)) $
+                            U.mapM f $ toUnboxed arr
+  fmapTM f v@(ZV _) = fmapTM f $ toRT v
+
+---------- "Container" instances ----------
+
+instance Fact m => Functor (RT m) where
+  -- Functor instance is implied by Applicative
+  fmap f x = pure f <*> x
+
+instance Fact m => Applicative (RT m) where
+  pure = ZV . pure
+
+  -- RT can never hold an a -> b
+  (ZV f) <*> (ZV a) = ZV (f <*> a)
+  f@(ZV _) <*> v@(RT _) = f <*> toZV v
+
+instance Fact m => Foldable (RT m) where
+  -- Foldable instance is implied by Traversable
+  foldMap = foldMapDefault
+
+instance Fact m => Traversable (RT m) where
+  traverse f r@(RT _) = T.traverse f $ toZV r
+  traverse f (ZV v) = ZV <$> T.traverse f v
+
+
+---------- Numeric Prelude instances ----------
+
+-- CJP: should Elt, Unbox be constraints on these instances?  It's
+-- possible to zipWith on IZipVector, so it's not *necessary* to
+-- convert toRT.
+
+instance (Fact m, Additive r, Unbox r, Elt r) => Additive.C (RT m r) where
+  (RT a) + (RT b) = RT $ coerce (\x -> force . RT.zipWith (+) x) a b
+  a + b = toRT a + toRT b
+
+  negate (RT a) = RT $ (coerce $ force . RT.map negate) a
+  negate a = negate $ toRT a
+
+  zero = RT $ repl zero
+
+instance (Fact m, Ring r, Unbox r, Elt r) => Ring.C (RT m r) where
+  (RT a) * (RT b) = RT $ coerce (\x -> force . RT.zipWith (*) x) a b
+  a * b = (toRT a) * (toRT b)
+
+  fromInteger = RT . repl . fromInteger
+
+instance (Fact m, ZeroTestable r, Unbox r, Elt r) => ZeroTestable.C (RT m r) where
+  -- not using 'zero' to avoid Additive r constraint
+  isZero (RT (Arr a)) = isZero $ foldAllS (\ x y -> if isZero x then y else x) (a RT.! (Z:.0)) a
+  isZero (ZV v) = isZero v
+
+---------- Miscellaneous instances ----------
+
+-- CJP: shouldn't these instances be defined in RTCommon, where the
+-- Arr data type is defined?  Here they are orphans.
+
+instance (Unbox r, Random (Arr m r)) => Random (RT m r) where
+  random = runRand $ liftM RT (liftRand random)
+
+  randomR = error "randomR nonsensical for RT"
+
+instance (Unbox r, Arbitrary (Arr m r)) => Arbitrary (RT m r) where
+  arbitrary = RT <$> arbitrary
+
+instance (NFData r) => NFData (RT m r) where
+  rnf (RT v) = rnf v
+  rnf (ZV v) = rnf v
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/CRT.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/CRT.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/CRT.hs
@@ -0,0 +1,185 @@
+{-# LANGUAGE ConstraintKinds, FlexibleContexts, GADTs, NoImplicitPrelude,
+             ScopedTypeVariables #-}
+
+-- | Functions to support the chinese remainder transform on Repa arrays
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor.CRT
+( scalarCRT'
+, fCRT, fCRTInv
+, mulGCRT', divGCRT'
+, gCRT, gInvCRT
+) where
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.Cyclotomic.Tensor
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.GL
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon as RT
+import Crypto.Lol.LatticePrelude                        as LP
+
+import Control.Applicative
+import Data.Coerce
+import Data.Singletons.Prelude
+import Data.Type.Natural       as N hiding (Z, one, zero)
+
+-- | Embeds a scalar into the CRT-basis when such basis exists
+scalarCRT' :: forall m r . (Fact m, CRTrans r, Unbox r)
+              => Maybe (r -> Arr m r)
+scalarCRT'
+  = let pps = proxy ppsFact (Proxy::Proxy m)
+        sz = Z :. totientPPs pps
+    in pure $ Arr . force . fromFunction sz . const
+
+-- | Multiplies an array in the CRT basis by 'g', when the CRT basis exists
+mulGCRT' :: forall m r . (Fact m, CRTrans r, Unbox r, Elt r)
+            => Maybe (Arr m r -> Arr  m r)
+mulGCRT' = (coerce (\x -> force . RT.zipWith (*) x) `asTypeOf` asTypeOf) <$> gCRT
+
+-- | Divides an array in the CRT basis by 'g', when the CRT basis exists.
+divGCRT' :: (Fact m, CRTrans r, IntegralDomain r, ZeroTestable r,
+             Unbox r, Elt r) => Maybe (Arr m r -> Arr m r)
+divGCRT' =  (coerce (\x -> force . RT.zipWith (*) x) `asTypeOf` asTypeOf) <$> gInvCRT
+
+-- | The CRT-basis representation of 'g'
+gCRT :: (Fact m, CRTrans r, Unbox r, Elt r)
+        => Maybe (Arr m r)
+gCRT = fCRT <*> pure (fGPow $ scalarPow' LP.one)
+
+-- EAC: This was defined using (a safe call to) fromJust
+
+-- | The CRT-basis representation of 'g^{ -1 }'
+gInvCRT:: (Fact m, CRTrans r, IntegralDomain r,
+           ZeroTestable r, Unbox r, Elt r)
+          => Maybe (Arr m r)
+gInvCRT = fCRT <*> fGInvPow (scalarPow' LP.one)
+
+
+fCRT, fCRTInv ::
+  forall m r . (Fact m, CRTrans r, Unbox r, Elt r)
+  => Maybe (Arr m r -> Arr m r)
+-- | The chinese remainder transform on Repa arrays.
+-- Exists if and only if crt exists for all prime powers
+fCRT = evalM $ fTensor ppCRT
+
+-- divide by mhat after doing crtInv'
+-- | The inverse chinese remainder transform on Repa arrays.
+-- Exists if and only if crt exists for all prime powers
+fCRTInv = do -- in Maybe
+  (_, mhatInv) :: (CRTInfo r) <- proxyT crtInfoFact (Proxy :: Proxy m)
+  let totm = proxy totientFact (Proxy :: Proxy m)
+      divMhat = trans totm $ RT.map (*mhatInv)
+  evalM $ (divMhat .*) <$> fTensor ppCRTInv'
+
+ppDFT, ppDFTInv', ppCRT, ppCRTInv' ::
+  forall pp r . (PPow pp, CRTrans r, Unbox r, Elt r)
+  => TaggedT pp Maybe (Trans r)
+
+ppDFT = case (sing :: SPrimePower pp) of
+  (SPP (STuple2 _ SZ)) -> return $ Id 1
+  spp@(SPP (STuple2 sp (SS se1))) ->
+    tagT $ do
+      let spp' = SPP (STuple2 sp se1)
+      pp'dft <- withWitnessT ppDFT spp'
+      pptwid <- withWitnessT (ppTwid False) spp
+      pdft <- withWitnessT pDFT sp
+      return $ (pp'dft @* Id (dim pdft)) .* pptwid .* (Id (dim pp'dft) @* pdft)
+
+ppDFTInv' = case (sing :: SPrimePower pp) of
+  (SPP (STuple2 _ SZ)) -> return $ Id 1
+  spp@(SPP (STuple2 sp (SS se1))) ->
+    tagT $ do
+      let spp' = SPP (STuple2 sp se1)
+      pp'dftInv' <- withWitnessT ppDFTInv' spp'
+      pptwidInv <- withWitnessT (ppTwid True) spp
+      pdftInv' <- withWitnessT pDFTInv' sp
+      return $
+        (Id (dim pp'dftInv') @* pdftInv') .* pptwidInv .*
+        (pp'dftInv' @* Id (dim pdftInv'))
+
+ppCRT = case (sing :: SPrimePower pp) of
+  (SPP (STuple2 _ SZ)) -> return $ Id 1
+  spp@(SPP (STuple2 sp (SS se1))) ->
+    tagT $ do
+      let spp' = SPP (STuple2 sp se1)
+      pp'dft <- withWitnessT ppDFT spp'
+      pptwid <- withWitnessT (ppTwidHat False) spp
+      pcrt <- withWitnessT pCRT sp
+      return $
+        (pp'dft @* Id (dim pcrt)) .* pptwid .*
+        -- save some work when p=2
+        (if dim pcrt > 1 then Id (dim pp'dft) @* pcrt else Id (dim pp'dft))
+
+ppCRTInv' = case (sing :: SPrimePower pp) of
+  (SPP (STuple2 _ SZ)) -> return $ Id 1
+  spp@(SPP (STuple2 sp (SS se1))) ->
+    tagT $ do
+      let spp' = SPP (STuple2 sp se1)
+      pp'dftInv' <- withWitnessT ppDFTInv' spp'
+      pptwidInv <- withWitnessT (ppTwidHat True) spp
+      pcrtInv' <- withWitnessT pCRTInv' sp
+      return $ -- special case for p=2 (necessary for scaling!)
+        (if dim pcrtInv' > 1
+         then Id (dim pp'dftInv') @* pcrtInv' else Id (dim pp'dftInv')) .*
+        pptwidInv .* (pp'dftInv' @* Id (dim pcrtInv'))
+
+-- DFT_p, CRT_p, (scaled) DFT_p^-1, etc.
+pDFT, pDFTInv', pCRT, pCRTInv' ::
+  forall p r . (NatC p, CRTrans r, Unbox r, Elt r)
+  => TaggedT p Maybe (Trans r)
+
+pDFT = let pval = proxy valueNatC (Proxy::Proxy p)
+       in do (omegaPPow, _) <- crtInfoNatC
+             return $ trans pval $ mulMat $ force $
+                                   fromFunction (Z :. pval :. pval)
+                                   (\(Z:.i:.j) -> omegaPPow (i*j))
+
+pDFTInv' = let pval = proxy valueNatC (Proxy::Proxy p)
+           in do (omegaPPow, _) <- crtInfoNatC
+                 return $ trans pval $ mulMat $ force $
+                                       fromFunction (Z :. pval :. pval)
+                                       (\(Z:.i:.j) -> omegaPPow (-i*j))
+
+pCRT = let pval = proxy valueNatC (Proxy::Proxy p)
+       in do (omegaPPow, _) <- crtInfoNatC
+             return $ trans (pval-1) $ mulMat $ force $
+                                     fromFunction (Z :. pval-1 :. pval-1)
+                                     (\(Z:.i:.j) -> omegaPPow ((i+1)*j))
+
+-- crt_p * this = pI, for all values of p.  For p=2 this isn't the
+-- matrix we "want," but it doesn't matter because we don't use it in
+-- ppCRTInv'
+pCRTInv' =
+  let pval = proxy valueNatC (Proxy::Proxy p)
+  in do (omegaPPow, _) <- crtInfoNatC
+        return $ trans (pval-1) $  mulMat $ force $
+                                fromFunction (Z :. pval-1 :. pval-1)
+                                (\(Z:.i:.j) -> omegaPPow (negate i*(j+1)) -
+                                               omegaPPow (j+1))
+
+-- twiddle factors for DFT_pp and CRT_pp decompositions
+ppTwid, ppTwidHat ::
+  forall pp r . (PPow pp, CRTrans r, Unbox r, Elt r)
+  => Bool -> TaggedT pp Maybe (Trans r)
+
+ppTwid inv =
+  let pp@(p,e) = proxy ppPPow (Proxy :: Proxy pp)
+      ppval = valuePP pp
+  in do
+    (omegaPPPow, _) <- crtInfoPPow
+    return $ trans ppval $ mulDiag $ force $
+                           fromFunction (Z :. ppval)
+                           (\(Z:.i) -> let (iq,ir) = i `divMod` p
+                                           pow = (if inv then negate else id)
+                                                 ir * digitRev (p,e-1) iq
+                                       in omegaPPPow pow)
+
+ppTwidHat inv =
+  let pp@(p,e) = proxy ppPPow (Proxy :: Proxy pp)
+      pptot = totientPP pp
+  in do
+    (omegaPPPow, _) <- crtInfoPPow
+    return $ trans pptot $ mulDiag $ force $
+                           fromFunction (Z :. pptot)
+                           (\(Z:.i) -> let (iq,ir) = i `divMod` (p-1)
+                                           pow = (if inv then negate else id)
+                                                 (ir+1) * digitRev (p,e-1) iq
+                                       in omegaPPPow pow)
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Extension.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Extension.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Extension.hs
@@ -0,0 +1,193 @@
+{-# LANGUAGE BangPatterns, ConstraintKinds, DataKinds, FlexibleContexts,
+             FlexibleInstances, MultiParamTypeClasses, NoImplicitPrelude,
+             ScopedTypeVariables, TemplateHaskell, TypeFamilies,
+             TypeOperators #-}
+
+-- | RT-specific functions for embedding/twacing in various bases
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Extension
+( twacePowDec', twaceCRT', embedPow', embedDec', embedCRT'
+, coeffs', powBasisPow', crtSetDec' --, fromCoeffs'
+) where
+
+import           Crypto.Lol.LatticePrelude              as LP hiding (lift, (!!))
+import           Crypto.Lol.CRTrans
+import           Crypto.Lol.Reflects
+import qualified Crypto.Lol.Cyclotomic.Tensor                      as T
+import           Crypto.Lol.Cyclotomic.Tensor.RepaTensor.CRT
+import           Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon as RT
+import           Crypto.Lol.Types.FiniteField
+import           Crypto.Lol.Types.ZmStar
+
+import Control.Applicative
+import Control.Arrow       (first, (***))
+
+import           Data.Coerce
+import           Data.Default
+import           Data.Maybe
+import           Data.Reflection (reify)
+import qualified Data.Vector                  as V
+import qualified Data.Vector.Unboxed          as U
+import           Data.Vector.Unboxed.Deriving
+
+-- Default instances
+instance Default Z where def = Z
+instance (Default a, Default b) => Default (a:.b) where def = def:.def
+
+-- derived Unbox instances
+derivingUnbox "DIM1"
+  [t| (Z:.Int) -> Int |]
+  [| \(Z:.i) -> i |]
+  [| (Z :.) |]
+
+-- | The "tweaked trace" function in either the powerful or decoding
+-- basis of the m'th cyclotomic ring to the mth cyclotomic ring when 
+-- @m | m'@.
+twacePowDec' :: forall m m' r . (m `Divides` m', Unbox r)
+                 => Arr m' r -> Arr m r
+twacePowDec'
+  = let indices = proxy extIndicesPowDec (Proxy::Proxy '(m, m'))
+    in coerce $ \ !arr -> force $ backpermute (extent indices) (indices !) arr
+
+-- | The "tweaked trace" function in the CRT
+-- basis of the m'th cyclotomic ring to the mth cyclotomic ring when 
+-- @m | m'@.
+twaceCRT' :: forall m m' r .
+             (m `Divides` m', CRTrans r, IntegralDomain r,
+              ZeroTestable r, Unbox r, Elt r)
+             => Maybe (Arr m' r -> Arr m r)
+twaceCRT' = do           -- Maybe monad
+  g' :: Arr m' r <- gCRT
+  gInv <- gInvCRT
+  embed :: Arr m r -> Arr m' r <- embedCRT'
+  (_, m'hatinv) <- proxyT crtInfoFact (Proxy::Proxy m')
+  let hatRatioInv = m'hatinv * fromIntegral (proxy valueHatFact (Proxy::Proxy m))
+      -- tweak = mhat * g' / (m'hat * g)
+      tweak = (coerce $ \x -> force . RT.map (* hatRatioInv) . RT.zipWith (*) x) (embed gInv) g' :: Arr m' r
+      indices = proxy extIndicesCRT (Proxy::Proxy '(m, m'))
+  return $ 
+    -- take true trace after mul-by-tweak
+    coerce (\ !arr -> sumS . backpermute (extent indices) (indices !) . RT.zipWith (*) arr) tweak
+
+embedPow', embedDec' :: forall m m' r .
+             (m `Divides` m', Unbox r, Additive r)
+             => Arr m r -> Arr m' r
+-- | Embeds an array in the powerful basis of the the mth cyclotomic ring
+-- to an array in the powerful basis of the m'th cyclotomic ring when @m | m'@
+embedPow'
+  = let indices = proxy baseIndicesPow (Proxy::Proxy '(m, m'))
+    in coerce $ \ !arr -> force $ fromFunction (extent indices)
+                       (\idx -> let (j0,j1) = (indices ! idx)
+                                in if j0 == 0 then arr ! j1 else zero)
+-- | Embeds an array in the decoding basis of the the mth cyclotomic ring
+-- to an array in the decoding basis of the m'th cyclotomic ring when @m | m'@
+embedDec'
+  = let indices = proxy baseIndicesDec (Proxy::Proxy '(m, m'))
+    in coerce $ \ !arr -> force $
+                       fromFunction (extent indices)
+                         (\idx -> maybe zero
+                                  (\(sh,b) -> if b then negate (arr ! sh)
+                                              else arr ! sh)
+                                  (indices ! idx))
+
+-- | Embeds an array in the CRT basis of the the mth cyclotomic ring
+-- to an array in the CRT basis of the m'th cyclotomic ring when @m | m'@
+embedCRT' :: forall m m' r . (m `Divides` m', CRTrans r, Unbox r)
+             => Maybe (Arr m r -> Arr m' r)
+embedCRT' = do -- in Maybe
+  -- first check existence of CRT transform of index m'
+  proxyT crtInfoFact (Proxy::Proxy m') :: Maybe (CRTInfo r)
+  let idxs = proxy baseIndicesCRT (Proxy::Proxy '(m,m'))
+  return $ coerce $ \ !arr -> (force $ backpermute (extent idxs) (idxs !) arr)
+
+-- | maps an array in the powerful/decoding basis, representing an
+-- O_m' element, to an array of arrays representing O_m elements in
+-- the same type of basis
+coeffs' :: forall m m' r . (m `Divides` m', Unbox r)
+             => Arr m' r -> [Arr m r]
+coeffs' =
+  let indices = proxy extIndicesCoeffs (Proxy::Proxy '(m, m'))
+  in coerce $ \ !arr -> V.toList $
+  V.map (\idxs -> force $ backpermute (extent idxs) (idxs !) arr) indices
+
+-- | The powerful extension basis, wrt the powerful basis.
+-- Outputs a list of arrays in O_m' that are an O_m basis for O_m'
+powBasisPow' :: forall m m' r . (m `Divides` m', Ring r, Unbox r)
+                => Tagged m [Arr m' r]
+powBasisPow' = return $  
+  let (_, phi, phi', _) = proxy T.indexInfo (Proxy::Proxy '(m,m'))
+      idxs = proxy T.baseIndicesPow (Proxy::Proxy '(m,m'))
+  in LP.map (\k -> Arr $ force $ fromFunction (Z :. phi')
+                         (\(Z:.j) -> let (j0,j1) = idxs U.! j
+                                     in if j0==k && j1==0 then one else zero))
+      [0..phi' `div` phi - 1]
+
+-- | A list of arrays representing the mod-p CRT set of the
+-- extension O_m'/O_m
+crtSetDec' :: forall m m' fp .
+              (m `Divides` m', PrimeField fp, 
+               Coprime (PToF (CharOf fp)) m', Unbox fp)
+              => Tagged m [Arr m' fp]
+crtSetDec' = return $ 
+  let m'p = Proxy :: Proxy m'
+      p = proxy value (Proxy::Proxy (CharOf fp))
+      phi = proxy totientFact m'p
+
+      d = proxy (order p) m'p
+      h :: Int = proxy valueHatFact m'p
+      hinv = recip $ fromIntegral h
+  in reify d $ \(_::Proxy d) ->
+       let twCRTs' :: T.Matrix (GF fp d)
+             = fromMaybe (error "internal error: crtSetDec': twCRTs") $ proxyT T.twCRTs m'p
+           zmsToIdx = proxy T.zmsToIndexFact m'p
+           elt j i = T.indexM twCRTs' j (zmsToIdx i)
+           trace' = trace :: GF fp d -> fp
+           cosets = proxy (partitionCosets p) (Proxy::Proxy '(m,m'))
+       in LP.map (\is -> Arr $ force $ fromFunction (Z :. phi) 
+                          (\(Z:.j) -> hinv * trace'
+                                      (sum $ LP.map (elt j) is))) cosets
+
+
+-- convert memoized reindexing Vectors to Arrays, for convenience and speed
+
+extIndicesPowDec :: forall m m' . (m `Divides` m')
+                    => Tagged '(m, m') (Array U DIM1 DIM1)
+extIndicesPowDec = do
+  idxs <- T.extIndicesPowDec
+  return $ fromUnboxed (Z :. U.length idxs) $ U.map (Z:.) idxs
+
+extIndicesCRT :: forall m m' . (m `Divides` m')
+                 => Tagged '(m, m') (Array U DIM2 DIM1)
+extIndicesCRT =
+  let phi = proxy totientFact (Proxy::Proxy m)
+      phi' = proxy totientFact (Proxy::Proxy m')
+  in do
+    idxs <- T.extIndicesCRT
+    return $ fromUnboxed (Z :. phi :. phi' `div` phi) $ U.map (Z:.) idxs
+
+baseIndicesPow :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (Array U DIM1 (Int,DIM1))
+
+baseIndicesDec :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (Array U DIM1 (Maybe (DIM1, Bool)))
+
+baseIndicesCRT :: forall m m' . (m `Divides` m')
+                  => Tagged '(m, m') (Array U DIM1 DIM1)
+
+baseIndicesPow = do
+  idxs <- T.baseIndicesPow
+  return $ fromUnboxed (Z :. U.length idxs) $ U.map (id *** (Z:.)) idxs
+
+baseIndicesDec = do
+  idxs <- T.baseIndicesDec
+  return $ fromUnboxed (Z :. U.length idxs) $ U.map (liftA (first (Z:.))) idxs
+
+baseIndicesCRT = do
+  idxs <- T.baseIndicesCRT
+  return $ fromUnboxed (Z :. U.length idxs) $ U.map (Z:.) idxs
+
+extIndicesCoeffs :: forall m m' . (m `Divides` m')
+                    => Tagged '(m, m') (V.Vector (Array U DIM1 DIM1))
+extIndicesCoeffs = 
+  V.map (\arr -> fromUnboxed (Z :. U.length arr) $ 
+                 U.map (Z:.) arr) <$> T.extIndicesCoeffs
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/GL.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/GL.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/GL.hs
@@ -0,0 +1,112 @@
+{-# LANGUAGE BangPatterns, ConstraintKinds, FlexibleContexts, GADTs,
+             MultiParamTypeClasses, NoImplicitPrelude, RankNTypes,
+             RebindableSyntax, ScopedTypeVariables #-}
+
+-- | The G and L transforms for Repa arrays
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor.GL
+( fL, fLInv, fGPow, fGDec, fGInvPow, fGInvDec
+) where
+
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon as RT
+import Crypto.Lol.LatticePrelude                        as LP
+import Data.Coerce
+
+fL, fLInv, fGPow, fGDec :: (Fact m, Additive r, Unbox r, Elt r)
+  => Arr m r -> Arr m r
+
+fGInvPow, fGInvDec ::
+ (Fact m, IntegralDomain r, ZeroTestable r, Unbox r, Elt r)
+  => Arr m r -> Maybe (Arr m r)
+-- | Arbitrary-index L transform to convert a dec-basis Repa array to its powerful-basis representation
+fL = eval $ fTensor $ ppTensor pL
+-- | Arbitrary-index L^{ -1 } transform to convert a powerful-basis Repa array to its dec-basis representation
+fLInv = eval $ fTensor $ ppTensor pLInv
+-- | Arbitrary-index multiplication by the ring element g in the powerful basis
+fGPow = eval $ fTensor $ ppTensor pGPow
+-- | Arbitrary-index multiplication by the ring element g in the dec basis
+fGDec = eval $ fTensor $ ppTensor pGDec
+-- | Arbitrary-index division by the ring element g in the powerful basis. May fail if the input is not a multiple of g.
+fGInvPow = wrapGInv' pGInvPow'
+-- | Arbitrary-index multiplication by the ring element g in the dec basis. May fail if the input is not a multiple of g.
+fGInvDec = wrapGInv' pGInvDec'
+
+wrapGInv' :: forall m r .
+  (Fact m, IntegralDomain r, ZeroTestable r, Unbox r, Elt r)
+  => (forall p . (NatC p) => Tagged p (Trans r))
+  -> Arr m r -> Maybe (Arr m r)
+wrapGInv' ginv =
+  let fGInv = eval $ fTensor $ ppTensor ginv
+      oddrad = fromIntegral $ proxy oddRadicalFact (Proxy::Proxy m)
+  in (`divCheck` oddrad) . fGInv
+
+-- | This is not a constant-time algorithm!  Depending on its usage,
+-- it might provide a timing side-channel.
+divCheck :: (IntegralDomain r, ZeroTestable r, Unbox r)
+            => Arr m r -> r -> Maybe (Arr m r)
+divCheck = coerce $  \ !arr den ->
+  let qrs = force $ RT.map (`divMod` den) arr
+      pass = foldAllS (&&) True $ RT.map (isZero . snd) qrs
+      out = force $ RT.map fst qrs
+  in if pass then Just out else Nothing
+
+pWrap :: forall p r . (NatC p)
+         => (forall rep . Source rep r => Int -> Array rep DIM2 r -> Array D DIM2 r)
+         -> Tagged p (Trans r)
+pWrap f = let pval = proxy valueNatC (Proxy::Proxy p)
+              -- special case: return identity function for p=2
+          in return $ if pval > 2
+                      then trans  (pval-1) $ f pval
+                      else Id 1
+
+
+pL, pLInv, pGPow, pGDec :: (NatC p, Additive r, Unbox r, Elt r)
+  => Tagged p (Trans r)
+
+pGInvPow', pGInvDec' :: (NatC p, Ring r, Unbox r, Elt r)
+  => Tagged p (Trans r)
+
+pL = pWrap (\_ !arr ->
+             fromFunction (extent arr) $
+             \ (i':.i) -> sumAllS $ extract (Z:.0) (Z:.(i+1)) $ slice arr (i':.All))
+
+pLInv = pWrap (\_ !arr ->
+                let f (i' :. 0) = arr! (i' :. 0)
+                    f (i' :. i) = arr! (i' :. i) - arr! (i' :. i-1)
+                in fromFunction (extent arr) f)
+
+
+-- multiplicaton by g_p=1-zeta_p in power basis.
+-- this is "wrong" for p=2 but we never use that case thanks to pWrap.
+pGPow = pWrap (\p !arr ->
+                let f (i':.0) = arr! (i':.p-2) + arr! (i':.0)
+                    f (i':.i) = arr! (i':.p-2) + arr! (i':.i) - arr! (i':.i-1)
+                in fromFunction (extent arr) f)
+
+-- multiplication by g_p=1-zeta_p in decoding basis
+pGDec = pWrap (\_ !arr ->
+                let f (i':.0) = arr! (i':.0) + sumAllS (slice arr (i':.All))
+                    f (i':.i) = arr! (i':.i) - arr! (i':.i-1)
+                in fromFunction (extent arr) f)
+
+-- CJP: profiling suggests that this does two read passes through the
+-- array; see if we can rewrite to make it one
+
+-- doesn't do division by (odd) p
+pGInvPow' =
+  pWrap (\p !arr ->
+          let f (i':.i) =
+                let col = slice arr (i':.All)
+                in fromIntegral (p-i-1) * sumAllS (extract (Z:.0) (Z:.i+1) col) +
+                   fromIntegral (-i-1) * sumAllS (extract (Z:.i+1) (Z:.p-i-2) col)
+          in fromFunction (extent arr) f)
+
+-- doesn't do division by (odd) p
+pGInvDec' =
+  pWrap (\p !arr ->
+          let f (i':.i) =
+                let col = slice arr (i':.All)
+                    nats = fromFunction (Z:.p-1) (\(Z:.j) -> fromIntegral j+1)
+                in (sumAllS $ RT.zipWith (*) col nats) -
+                   fromIntegral p * sumAllS (extract (Z:.i+1) (Z:.p-i-2) col)
+          in fromFunction (extent arr) f)
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Gauss.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Gauss.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/Gauss.hs
@@ -0,0 +1,51 @@
+{-# LANGUAGE ConstraintKinds, FlexibleContexts, NoImplicitPrelude,
+             RebindableSyntax, ScopedTypeVariables #-}
+
+-- | (Continuous) Gaussian sampling for Repa arrays
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor.Gauss
+( tGaussianDec' ) where
+
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon
+import Crypto.Lol.GaussRandom
+import Crypto.Lol.LatticePrelude
+
+import Control.Monad.Random
+
+-- | A function tagged by the cyclotomic index which,
+-- given a (scaled) variance, outputs a Gaussian-distributed
+-- vector in the decoding basis
+tGaussianDec' :: forall m v r rnd .
+                 (Fact m, OrdFloat r, Random r, Unbox r, Elt r,
+                  ToRational v, MonadRandom rnd)
+                 => v -> rnd (Arr m r)
+tGaussianDec' =
+  let pm = Proxy::Proxy m
+      m = proxy valueFact pm
+      n = proxy totientFact pm
+      rad = proxy radicalFact pm
+  in \v -> do             -- rnd monad
+    x <- realGaussians (v * fromIntegral (m `div` rad)) n
+    let arr = Arr $ fromUnboxed (Z:.n) x
+    return $ fD arr
+
+-- | The fully tensored D transformation
+fD :: (Fact m, Transcendental r, Unbox r, Elt r) => Arr m r -> Arr m r
+fD = eval $ fTensor $ ppTensor pD
+
+-- | The D transformation for a prime
+pD :: forall p r . (NatC p, Transcendental r, Unbox r, Elt r)
+      => Tagged p (Trans r)
+pD = let pval = proxy valueNatC (Proxy::Proxy p)
+     in tag $
+        if pval==2
+        then Id 1
+        else trans (pval-1) $ mulMat $ force $
+                            fromFunction (Z :. pval-1 :. pval-1)
+                            (\(Z:.i:.j) ->
+                              -- mtx is sqrt(2)*[ cos(2pi*i*(j+1)/p) | sin(same) ]
+                              -- (signs of columns doesn't matter for our purposes.)
+                              let theta = 2 * pi * fromIntegral (i*(j+1)) /
+                                          fromIntegral pval
+                              in sqrt 2 * if j < pval `div` 2
+                                          then cos theta else sin theta)
diff --git a/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/RTCommon.hs b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/RTCommon.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Tensor/RepaTensor/RTCommon.hs
@@ -0,0 +1,233 @@
+{-# LANGUAGE BangPatterns, ConstraintKinds, DataKinds, FlexibleContexts,
+             FlexibleInstances, GADTs, GeneralizedNewtypeDeriving,
+             KindSignatures, MultiParamTypeClasses, NoImplicitPrelude,
+             RankNTypes, RebindableSyntax, RoleAnnotations,
+             ScopedTypeVariables, TypeOperators #-}
+
+-- | A simple DSL for tensoring Repa arrays and other common functionality
+-- on Repa arrays
+
+module Crypto.Lol.Cyclotomic.Tensor.RepaTensor.RTCommon
+( module R
+, module Data.Array.Repa.Eval
+, module Data.Array.Repa.Repr.Unboxed
+, Arr(..), repl, replM, eval, evalM, fTensor, ppTensor
+, Trans(Id), trans, dim, (.*), (@*), force
+, mulMat, mulDiag
+, scalarPow'
+, sumS, sumAllS
+) where
+
+import Crypto.Lol.LatticePrelude as LP hiding ((!!))
+
+import Control.DeepSeq              (NFData (..))
+import Control.Monad.Identity
+import Control.Monad.Random
+import Data.Array.Repa              as R hiding (sumAllP, sumAllS, sumP,
+                                          sumS, (*^), (+^), (-^), (/^))
+import Data.Array.Repa.Eval         hiding (one, zero)
+import Data.Array.Repa.Repr.Unboxed
+import Data.Coerce
+import Data.Singletons
+import Data.Singletons.Prelude      hiding ((:.))
+import Data.Type.Natural            hiding (Z)
+import Data.Typeable
+import qualified Data.Vector.Unboxed as U
+import Test.QuickCheck
+
+-- always unboxed (manifest); intermediate calculations can use
+-- delayed arrays
+
+-- | Indexed newtype for 1-dimensional Unbox repa arrays
+newtype Arr (m :: Factored) r = Arr (Array U DIM1 r)
+                              deriving (Eq, Show, Typeable, NFData)
+
+-- the first argument, though phantom, affects representation
+-- CJP: why must the second arg be nominal?
+-- EAC: From https://ghc.haskell.org/trac/ghc/wiki/Roles#Thesolution:
+--   "The exception to the above algorithm is for classes: all parameters for a class default to a nominal role."
+-- Arr is a synonym for Array, which is an associated data type to the class Source. The parameter `r` above
+-- corresponds to the parameter `e` in the definition of class Source, so it's role must be nominal.
+type role Arr nominal nominal
+
+-- | An 'Arr' filled with the argument.
+repl :: forall m r . (Fact m, Unbox r) => r -> Arr m r
+repl = let n = proxy totientFact (Proxy::Proxy m)
+       in Arr . fromUnboxed (Z:.n) . U.replicate n
+
+-- | Monadic version of 'repl'.
+replM :: forall m r mon . (Fact m, Unbox r, Monad mon)
+         => mon r -> mon (Arr m r)
+replM = let n = proxy totientFact (Proxy::Proxy m)
+        in liftM (Arr . fromUnboxed (Z:.n)) . U.replicateM n
+
+instance (Unbox r) => NFData (Array U DIM1 r) where
+  -- EAC: Repa doesn't define any NFData instances,
+  -- I'm hoping deepSeqArray is a reasonable approx
+  rnf x = deepSeqArray x ()
+
+instance (Unbox r, Random r, Fact m) => Random (Arr m r) where
+  random = runRand $ replM (liftRand random)
+
+  randomR = error "randomR nonsensical for Arr"
+
+instance (Arbitrary r, Unbox r, Fact m) => Arbitrary (Arr m r) where
+    arbitrary = replM arbitrary
+    shrink = shrinkNothing
+
+-- | For a factored index, tensors up any function defined for (and
+-- tagged by) any prime power
+fTensor :: forall m r mon . (Fact m, Monad mon, Unbox r)
+  => (forall pp . (PPow pp) => TaggedT pp mon (Trans r))
+  -> TaggedT m mon (Trans r)
+
+fTensor func = tagT $ go $ sUnF (sing :: SFactored m)
+  where
+    go :: Sing (pplist :: [PrimePower]) -> mon (Trans r)
+    go spps = case spps of
+          SNil -> return $ Id 1
+          (SCons spp rest) -> do
+            rest' <- go rest
+            func' <- withWitnessT func spp
+            return $ rest' @* func'
+
+-- | For a prime power pp > 1, tensors up any function f defined for
+-- (and tagged by) a prime to (I_(pp/p) \otimes f)
+ppTensor :: forall pp r mon . (PPow pp, Monad mon)
+            => (forall p . (NatC p) => TaggedT p mon (Trans r))
+            -> TaggedT pp mon (Trans r)
+
+ppTensor func = tagT $ case (sing :: SPrimePower pp) of
+  -- intentionally no match for zero exponents, because that is
+  -- ill-formed and indicates an internal error
+  (SPP (STuple2 sp (SS se1))) -> do
+    func' <- withWitnessT func sp
+    let lts = withWitness valuePPow (SPP (STuple2 sp se1))
+    return $ Id lts @* func'
+
+
+-- deeply embedded DSL for transformations and their various
+-- compositions
+
+-- (dim(f), f) where f operates on innermost dimension of array
+data Tensorable r = Tensorable
+  Int (forall rep . Source rep r => Array rep DIM2 r -> Array D DIM2 r)
+
+-- transform component: a Tensorable with particular I_l, I_r
+type TransC r = (Tensorable r, Int, Int)
+
+-- full transform: sequence of zero or more components
+-- | a DSL for tensor transforms on Repa arrays
+data Trans r = Id Int           -- ^| identity sentinel
+             | TSnoc (Trans r) (TransC r) -- ^| (function) composition of transforms
+
+dimC :: TransC r -> Int
+dimC (Tensorable d _, l, r) = l*d*r
+
+-- | Returns the (linear) dimension of a transform
+dim :: Trans r -> Int
+dim (Id n) = n
+dim (TSnoc _ f) = dimC f        -- just use dimension of head
+
+-- | smart constructor from a Tensorable
+trans :: Int -> (forall rep . Source rep r => Array rep DIM2 r -> Array D DIM2 r) -> Trans r
+trans d f = TSnoc (Id d) (Tensorable d f, 1, 1)
+
+-- | compose transforms
+(.*) :: Trans r -> Trans r -> Trans r
+f .* g | dim f == dim g = f ..* g
+       | otherwise = error $ "(.*): transform dimensions don't match "
+                     LP.++ show (dim f) LP.++ ", " LP.++ show (dim g)
+  where
+    f' ..* (Id _) = f'          -- drop sentinel
+    f' ..* (TSnoc rest g') = TSnoc (f' ..* rest) g'
+
+-- | tensor/Kronecker product (otimes)
+(@*) :: Trans r -> Trans r -> Trans r
+-- merge identity transforms
+(Id n) @* (Id m) = Id (n*m)
+-- Id on left or right
+i@(Id n) @* (TSnoc g' (g, l, r)) = TSnoc (i @* g') (g, n*l, r)
+(TSnoc f' (f, l, r)) @* i@(Id n) = TSnoc (f' @* i) (f, l, r*n)
+-- no Ids: compose
+f @* g = (f @* Id (dim g)) .* (Id (dim f) @* g)
+
+evalC :: (Unbox r) => TransC r -> Array U DIM1 r -> Array U DIM1 r
+evalC (Tensorable d f, _, r) arr =
+  arr `deepSeqArray` force $ unexpose r $ f $ expose d r arr
+
+-- | Creates an evaluatable Haskell function from a tensored transform
+eval :: (Unbox r) => Tagged m (Trans r) -> Arr m r -> Arr m r
+eval x = coerce $ eval' $ untag x
+  where eval' (Id _) = id
+        eval' (TSnoc rest f) = eval' rest . evalC f
+
+-- | Monadic version of 'eval'
+evalM :: (Unbox r, Monad mon) => TaggedT m mon (Trans r) -> mon (Arr m r -> Arr m r)
+evalM = liftM (eval . return) . untagT
+
+
+-- | maps the innermost dimension to a 2-dim array with innermost dim d,
+-- for performing a I_l \otimes f_d \otimes I_r transformation
+expose !d !r !arr =
+  let (sh :. sz) = extent arr
+      f (s :. i :. j) = let imodr = i `mod` r
+                            idx = (i-imodr)*d + j*r + imodr
+                        in arr ! (s :. idx)
+  in fromFunction (sh :. sz `div` d :. d) f
+
+-- | inverse of expose
+unexpose !r !arr =
+  let (sh:.sz:.d) = extent arr
+      f (s :. i) = let (idivr,imodr) = i `divMod` r
+                       (idivrd,j) = idivr `divMod` d
+                   in arr ! (s :. r*idivrd + imodr :. j)
+  in fromFunction (sh :. sz*d) f
+
+-- | general matrix multiplication along innermost dim of v
+mulMat :: (Source r1 r, Source r2 r, Ring r, Unbox r, Elt r)
+          => Array r1 DIM2 r -> Array r2 DIM2 r -> Array D DIM2 r
+mulMat !m !v
+  = let (Z :. mrows :. mcols) = extent m
+        (sh :. vrows) = extent v
+        f (sh' :. i) = sumAllS $ R.zipWith (*) (slice m (Z:.i:.All)) $ slice v (sh':.All)
+    in if mcols == vrows then fromFunction (sh :. mrows) f
+       else error "mulMatVec: mcols != vdim"
+
+-- | multiplication by a diagonal matrix along innermost dim
+mulDiag :: (Source r1 r, Source r2 r, Ring r, Unbox r, Elt r)
+           => Array r1 DIM1 r -> Array r2 DIM2 r -> Array D DIM2 r
+mulDiag !diag !arr = fromFunction (extent arr) f
+  where f idx@(_ :. i) = arr! idx * diag! (Z:.i)
+
+-- misc Tensor functions
+
+-- | Embeds a scalar into a powerful-basis representation of a Repa array,
+-- tagged by the cyclotomic index
+scalarPow' :: forall m r . (Fact m, Additive r, Unbox r) => r -> Arr m r
+scalarPow' = coerce . (go $ proxy totientFact (Proxy::Proxy m))
+  where go n !r = let fct (Z:.0) = r
+                      fct _ = LP.zero
+                  in force $ fromFunction (Z:.n) fct
+
+-- | Forces a delayed array to a manifest array.
+force :: (Shape sh, Unbox r) => Array D sh r -> Array U sh r
+-- CJP: computeS just until we figure out how to avoid nested parallel
+-- computation!
+--force = computeS
+force = runIdentity . computeP
+
+-- copied implementations of functions we need that normally require
+-- Num
+
+-- | Sum the inner-most dimension of an array sequentially
+sumS :: (Source r a, Elt a, Unbox a, Additive a, Shape sh)
+  => Array r (sh :. Int) a
+  -> Array U sh a
+sumS = foldS (+) LP.zero
+
+-- | Sum all array indices to a scalar sequentially
+sumAllS :: (Shape sh, Source r a, Elt a, Unbox a, Additive a)
+  => Array r sh a
+  -> a
+sumAllS = foldAllS (+) LP.zero
diff --git a/src/Crypto/Lol/Cyclotomic/UCyc.hs b/src/Crypto/Lol/Cyclotomic/UCyc.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/UCyc.hs
@@ -0,0 +1,655 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             FlexibleContexts, FlexibleInstances, GADTs, InstanceSigs,
+             MultiParamTypeClasses, NoImplicitPrelude, PolyKinds,
+             RankNTypes, RebindableSyntax, ScopedTypeVariables,
+             TypeFamilies, TypeOperators, UndecidableInstances #-}
+
+-- | An implementation of cyclotomic rings.  WARNING: this module
+-- provides an experts-only, "unsafe" interface that may result in
+-- runtime errors if not used correctly!
+-- 'Crypto.Lol.Cyclotomic.Cyc.Cyc' provides a safe interface, and
+-- should be used in applications whenever possible.
+--
+-- 'UCyc' transparently handles all necessary conversions between
+-- internal representations to support fast ring operations, and
+-- efficiently stores and operates upon elements that are known to
+-- reside in subrings.
+--
+-- The 'Functor', 'Applicative', 'Foldable', and 'Traversable'
+-- instances of 'UCyc', as well as the 'fmapC' and 'fmapCM' functions,
+-- work over the element's /current/ @r@-basis representation (or
+-- 'pure' scalar representation as a special case, to satisfy the
+-- 'Applicative' laws), and the output remains in that representation.
+-- If the input's representation is not one of these, the behavior is
+-- a runtime error.  To ensure a valid representation when using the
+-- methods from these classes, first call 'forceBasis' or one of its
+-- specializations ('forcePow', 'forceDec', 'forceAny').
+
+module Crypto.Lol.Cyclotomic.UCyc
+(
+-- * Data type
+  UCyc, CElt
+-- * Basic operations
+, mulG, divG
+, scalarCyc, liftCyc
+, adviseCRT
+-- * Error sampling
+, tGaussian, errorRounded, errorCoset
+-- * Sub/extension rings
+, embed, twace, coeffsCyc, powBasis, crtSet
+-- * Representations
+, forceBasis, forcePow, forceDec, forceAny
+-- * Specialized maps
+, fmapC, fmapCM
+, U.Basis(..), U.RescaleCyc
+) where
+
+import           Crypto.Lol.CRTrans
+import           Crypto.Lol.Cyclotomic.Tensor  as T
+import qualified Crypto.Lol.Cyclotomic.Utility as U
+import           Crypto.Lol.Gadget
+import           Crypto.Lol.LatticePrelude     as LP
+import           Crypto.Lol.Types.FiniteField
+import           Crypto.Lol.Types.ZPP
+
+import Algebra.Additive     as Additive (C)
+import Algebra.Ring         as Ring (C)
+import Algebra.ZeroTestable as ZeroTestable (C)
+
+import Control.Applicative    hiding ((*>))
+import Control.DeepSeq
+import Control.Monad.Identity
+import Control.Monad.Random
+import Data.Coerce
+import Data.Foldable          as F
+import Data.Maybe
+import Data.Traversable
+import Data.Typeable
+import Test.QuickCheck
+
+import qualified Debug.Trace as DT
+
+-- | A data type for representing cyclotomic rings such as @Z[zeta]@,
+-- @Zq[zeta]@, and @Q(zeta)@: @t@ is the 'Tensor' type for storing
+-- coefficients; @m@ is the cyclotomic index; @r@ is the base ring of
+-- the coefficients (e.g., @Z@, @Zq@).
+data UCyc t (m :: Factored) r where
+  Pow  :: !(t m r) -> UCyc t m r -- representation wrt powerful basis
+  Dec  :: !(t m r) -> UCyc t m r -- decoding basis
+
+  -- Invariant: use CRTr if and only if crtFuncs exists for (t m r);
+  -- otherwise use CRTe (because crtFuncs is guaranteed to exist for
+  -- (t m (CRTExt r))
+  CRTr :: !(t m r) -> UCyc t m r -- wrt CRT basis over r, if it exists
+  CRTe :: !(t m (CRTExt r)) -> UCyc t m r -- wrt CRT basis over r-extension
+
+  -- super-optimized storage of scalars
+  Scalar :: !r -> UCyc t m r
+
+  -- optimized storage of subring elements
+  Sub  :: (l `Divides` m) => !(UCyc t l r) -> UCyc t m r
+
+
+  --EAC: Consider this representation for product rings, but beware of combinatorial explosion of cases.
+  --Product :: !(UCyc t m a) -> !(UCyc t m b) -> UCyc t m (a,b)
+  deriving (Typeable)
+
+-- | Shorthand for frequently reused constraints that are needed for
+--  change of basis.
+type UCCtx t r = (Tensor t, CRTrans r, CRTrans (CRTExt r), CRTEmbed r,
+                  ZeroTestable r, TElt t r, TElt t (CRTExt r))
+
+-- | Shorthand for frequently reused constraints that are needed for
+-- most functions involving 'UCyc' and 'Crypto.Lol.Cyclotomic.Cyc.Cyc'.
+
+-- EAC: duplicated UCCtx for haddock
+type CElt t r = (Tensor t, CRTrans r, CRTrans (CRTExt r), CRTEmbed r,
+                 ZeroTestable r, TElt t r, TElt t (CRTExt r), Eq r, NFData r)
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.scalarCyc', but for 'UCyc'.
+scalarCyc :: (Fact m, CElt t a) => a -> UCyc t m a
+scalarCyc = Scalar
+
+-- Eq instance
+instance (UCCtx t r, Fact m, Eq r) => Eq (UCyc t m r) where
+  -- handle same bases when fidelity allows (i.e., *not* CRTe basis)
+  (Scalar v1) == (Scalar v2) = v1 == v2
+  (Pow v1) == (Pow v2) = v1 == v2 \\ witness entailFullT v1
+  (Dec v1) == (Dec v2) = v1 == v2 \\ witness entailFullT v1
+  (CRTr v1) == (CRTr v2) = v1 == v2 \\ witness entailFullT v1
+
+  (Sub (c1 :: UCyc t l1 r)) == (Sub (c2 :: UCyc t l2 r)) =
+    (embed' c1 :: UCyc t (FLCM l1 l2) r) == embed' c2
+    \\ lcmDivides (Proxy::Proxy l1) (Proxy::Proxy l2)
+
+  -- otherwise compare in power basis for fidelity, which involves
+  -- the most efficient transforms in all cases
+  p1 == p2 = toPow' p1 == toPow' p2
+
+---------- Numeric Prelude instances ----------
+
+-- ZeroTestable instance
+instance (UCCtx t r, Fact m) => ZeroTestable.C (UCyc t m r) where
+  isZero (Scalar v) = isZero v
+  isZero (Pow v) = isZero v \\ witness entailFullT v
+  isZero (Dec v) = isZero v \\ witness entailFullT v
+  isZero (CRTr v) = isZero v \\ witness entailFullT v
+  isZero x@(CRTe _) = isZero $ toPow' x
+  isZero (Sub c) = isZero c
+
+-- Additive instance
+instance (UCCtx t r, Fact m) => Additive.C (UCyc t m r) where
+
+  zero = Scalar zero
+
+  -- optimized addition of zero
+  (Scalar c1) + v2 | isZero c1 = v2
+  v1 + (Scalar c2) | isZero c2 = v1
+
+  -- SAME CONSTRUCTORS
+  (Scalar c1) + (Scalar c2) = Scalar (c1+c2)
+  (Pow v1) + (Pow v2) = Pow $ v1 + v2 \\ witness entailFullT v1
+  (Dec v1) + (Dec v2) = Dec $ v1 + v2 \\ witness entailFullT v1
+  (CRTr v1) + (CRTr v2) = CRTr $ v1 + v2 \\ witness entailFullT v1
+  -- CJP: is this OK for precision?
+  (CRTe v1) + (CRTe v2) = CRTe $ v1 + v2 \\ witness entailFullT v1
+  -- Sub plus Sub: work in compositum
+  (Sub (c1 :: UCyc t m1 r)) + (Sub (c2 :: UCyc t m2 r)) =
+    (Sub $ (embed' c1 :: UCyc t (FLCM m1 m2) r) + embed' c2)
+    \\ lcm2Divides (Proxy::Proxy m1) (Proxy::Proxy m2) (Proxy::Proxy m)
+
+  -- SCALAR PLUS SOMETHING ELSE
+
+  p1@(Scalar _) + p2@(Pow _) = toPow' p1 + p2
+  p1@(Scalar _) + p2@(Dec _) = toDec' p1 + p2
+  p1@(Scalar _) + p2@(CRTr _) = toCRT' p1 + p2
+  p1@(Scalar _) + p2@(CRTe _) = toCRT' p1 + p2
+  (Scalar v1) + (Sub c2) = Sub $ Scalar v1 + c2
+
+  p1@(Pow _) + p2@(Scalar _) = p1 + toPow' p2
+  p1@(Dec _) + p2@(Scalar _) = p1 + toDec' p2
+  p1@(CRTr _) + p2@(Scalar _) = p1 + toCRT' p2
+  p1@(CRTe _) + p2@(Scalar _) = p1 + toCRT' p2
+  (Sub c1) + (Scalar v2) = Sub $ c1 + Scalar v2
+
+  -- SUB PLUS SOMETHING ELSE (NON-SCALAR): work in full ring
+  (Sub c1) + c2 = embed' c1 + c2
+  c1 + (Sub c2) = c1 + embed' c2
+
+  -- mixed Dec and Pow: use linear time conversions
+  p1@(Dec _) + p2@(Pow _) = toPow' p1 + p2
+  p1@(Pow _) + p2@(Dec _) = p1 + toPow' p2
+
+  -- one CRTr: convert other to CRTr
+  p1@(CRTr _) + p2 = p1 + toCRT' p2
+  p1 + p2@(CRTr _) = toCRT' p1 + p2
+
+  -- else, one is CRTe: convert both to Pow for fidelity and best
+  -- efficiency
+  p1 + p2 = toPow' p1 + toPow' p2
+
+  negate (Scalar c) = Scalar (negate c)
+  negate (Pow v) = Pow $ fmapT negate v
+  negate (Dec v) = Dec $ fmapT negate v
+  negate (CRTr v) = CRTr $ fmapT negate v
+  negate (CRTe v) = CRTe $ fmapT negate v
+  negate (Sub c) = Sub $ negate c
+
+-- Ring instance
+instance (UCCtx t r, Fact m) => Ring.C (UCyc t m r) where
+
+  one = Scalar one
+
+  -- optimized mul-by-zero
+  v1@(Scalar c1) * _ | isZero c1 = v1
+  _ * v2@(Scalar c2) | isZero c2 = v2
+
+  -- BOTH IN A CRT BASIS
+  (CRTr v1) * (CRTr v2) = CRTr $ v1 * v2 \\ witness entailFullT v1
+  (CRTe v1) * (CRTe v2) = toPow' $ CRTe $ v1 * v2 \\ witness entailFullT v1
+
+  -- AT LEAST ONE SCALAR
+  (Scalar c1) * (Scalar c2) = Scalar $ c1 * c2
+
+  (Scalar c) * (Pow v) = Pow $ fmapT (*c) v
+  (Scalar c) * (Dec v) = Dec $ fmapT (*c) v
+  (Scalar c) * (CRTr v) = CRTr $ fmapT (*c) v
+  s@(Scalar _) * c'@(CRTe _) = s * toPow' c'
+  (Scalar c) * (Sub c2) = Sub $ Scalar c * c2
+
+  (Pow v) * (Scalar c) = Pow $ fmapT (*c) v
+  (Dec v) * (Scalar c) = Dec $ fmapT (*c) v
+  (CRTr v) * (Scalar c) = CRTr $ fmapT (*c) v
+  c'@(CRTe _) * s@(Scalar _) = toPow' c' * s
+  (Sub c1) * (Scalar c) = Sub $ c1 * Scalar c
+
+  -- AT LEAST ONE SUB
+
+  -- two Subs: work in compositum
+  (Sub (c1 :: UCyc t m1 r)) * (Sub (c2 :: UCyc t m2 r)) =
+    (Sub $ (embed' c1 :: UCyc t (FLCM m1 m2) r) * embed' c2)
+    \\ lcm2Divides (Proxy::Proxy m1) (Proxy::Proxy m2) (Proxy::Proxy m)
+
+  -- Sub times something else (non-Scalar): work in full ring
+  (Sub c1) * p2 = embed' c1 * p2
+  p1 * (Sub c2) = p1 * embed' c2
+
+  -- ELSE: work in appropriate CRT basis
+  p1 * p2 = toCRT' p1 * toCRT' p2
+
+  fromInteger = Scalar . fromInteger
+
+-- reduces in any basis
+instance (Reduce a b, Fact m, CElt t a, CElt t b)
+  => Reduce (UCyc t m a) (UCyc t m b) where
+
+  reduce = fmapC reduce . forceAny
+
+-- promote Gadget from base ring
+instance (Gadget gad zq, Fact m, CElt t zq) => Gadget gad (UCyc t m zq) where
+  gadget = (scalarCyc <$>) <$> gadget
+  -- specialization of 'encode', done efficiently (via 'adviseCRT').
+  encode s = ((* adviseCRT s) <$>) <$> gadget
+
+-- promote Decompose, using the powerful basis
+instance (Decompose gad zq, Fact m, CElt t zq,
+          Reduce (UCyc t m (DecompOf zq)) (UCyc t m zq))
+  => Decompose gad (UCyc t m zq) where
+
+  type DecompOf (UCyc t m zq) = UCyc t m (DecompOf zq)
+
+  -- traverse: Traversable (c m) and Applicative (Tagged gad ZL)
+  decompose = fromZL . traverse (toZL . decompose) . forcePow
+    where toZL :: Tagged s [a] -> TaggedT s ZipList a
+          toZL = coerce
+          fromZL :: TaggedT s ZipList a -> Tagged s [a]
+          fromZL = coerce
+
+-- promote Correct, using the decoding basis
+instance (Correct gad zq, Fact m, CElt t zq)
+         => Correct gad (UCyc t m zq) where
+  -- sequenceA: Applicative (c m) and Traversable (TaggedT [])
+  correct bs = (correct . pasteT) <$> (sequenceA $ forceDec <$> peelT bs)
+
+-- generic RescaleCyc instance
+
+instance {-# OVERLAPS #-} (Rescale a b, CElt t a, CElt t b)
+         => U.RescaleCyc (UCyc t) a b where
+  rescaleCyc b = fmapC rescale . forceBasis (Just b)
+
+-- specialized instance for product rings: ~2x faster algorithm
+instance (Mod a, Field b, Lift a z, Reduce z b,
+          CElt t a, CElt t b, CElt t (a,b), CElt t z)
+         => U.RescaleCyc (UCyc t) (a,b) b where
+  rescaleCyc bas =
+    let aval = proxy modulus (Proxy::Proxy a)
+  -- CJP: could use unzipC here to get (a,b) in one pass, but it
+  -- requires adding that method, and unzipT to Tensor and all its
+  -- instances. Probably not worth it.
+    in \x -> let y = forceAny x
+                 a = fmapC fst y
+                 b = fmapC snd y
+                 z = liftCyc bas a
+             in (pure (recip (fromIntegral aval))) * (b - reduce z)
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.liftCyc', but for 'UCyc'.
+liftCyc :: (Lift b a, Fact m, CElt t a, CElt t b)
+           => U.Basis -> UCyc t m b -> UCyc t m a
+liftCyc U.Pow = fmapC lift . forceBasis (Just U.Pow)
+liftCyc U.Dec = fmapC lift . forceBasis (Just U.Dec)
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.adviseCRT', but for 'UCyc'.
+adviseCRT :: (Fact m, CElt t r) => UCyc t m r -> UCyc t m r
+adviseCRT x@(Scalar _) = x
+adviseCRT (Sub c) = Sub $ adviseCRT c
+adviseCRT x = toCRT' x
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.mulG', but for 'UCyc'.
+mulG :: (Fact m, CElt t r) => UCyc t m r -> UCyc t m r
+mulG (Scalar c) = Pow $ mulGPow $ scalarPow c -- must go to full ring
+mulG (Sub c) = mulG $ embed' c                -- must go to full ring
+mulG (Pow v) = Pow $ mulGPow v
+mulG (Dec v) = Dec $ mulGDec v
+-- fromJust is safe here because we're already in CRTr
+mulG (CRTr v) = CRTr $ fromMaybe (error "FC.mulG CRTr") mulGCRT v
+mulG (CRTe v) = CRTe $ fromMaybe (error "FC.mulG CRTe") mulGCRT v
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.divG', but for 'UCyc'.
+divG :: (Fact m, CElt t r) => UCyc t m r -> Maybe (UCyc t m r)
+divG (Scalar c) = liftM Pow (divGPow $ scalarPow c) -- full ring
+divG (Sub c) = divG $ embed' c                      -- full ring
+divG (Pow v) = Pow <$> divGPow v
+divG (Dec v) = Dec <$> divGDec v
+-- fromJust is safe here because we're already in CRTr
+divG (CRTr v) = Just $ CRTr $ fromMaybe (error "FC.divG CRTr") divGCRT v
+divG (CRTe v) = Just $ CRTe $ fromMaybe (error "FC.divG CRTe") divGCRT v
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.tGaussian', but for 'UCyc'.
+tGaussian :: (Fact m, OrdFloat q, Random q, CElt t q,
+              ToRational v, MonadRandom rnd)
+             => v -> rnd (UCyc t m q)
+tGaussian = liftM Dec . tGaussianDec
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.errorRounded', but for 'UCyc'.
+errorRounded :: forall v rnd t m z .
+                (ToInteger z, Fact m, CElt t z, ToRational v, MonadRandom rnd)
+                => v -> rnd (UCyc t m z)
+errorRounded svar = 
+  fmapC (roundMult one) <$> (tGaussian svar :: rnd (UCyc t m Double))
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.errorCoset', but for 'UCyc'.
+errorCoset :: forall t m zp z v rnd .
+  (Mod zp, z ~ ModRep zp, Lift zp z, Fact m,
+   CElt t zp, CElt t z, ToRational v, MonadRandom rnd)
+  => v -> UCyc t m zp -> rnd (UCyc t m z)
+errorCoset =
+  let pval = fromIntegral $ proxy modulus (Proxy::Proxy zp)
+  -- we don't force* here because tGaussian is always in Dec
+  in \ svar c ->
+    roundCosetDec c <$> (tGaussian (svar * pval * pval) :: rnd (UCyc t m Double))
+
+-- | Deterministically round to the given coset @c+pR@, using the
+-- decoding basis.
+roundCosetDec ::
+    (Mod zp, z ~ ModRep zp, Lift zp z, RealField q,
+     Fact m, CElt t q, CElt t zp, CElt t z)
+    => UCyc t m zp -> UCyc t m q -> UCyc t m z
+roundCosetDec c x = roundCoset <$> forceDec c <*> forceDec x
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.embed', but for 'UCyc'.
+embed :: forall t r m m' . (m `Divides` m') => UCyc t m r -> UCyc t m' r
+embed (Scalar c) = Scalar c
+embed (Sub (c :: UCyc t l r)) = Sub c
+  \\ transDivides (Proxy::Proxy l) (Proxy::Proxy m) (Proxy::Proxy m')
+embed c = Sub c
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.twace', but for 'UCyc'.
+twace :: forall t r m m' . (UCCtx t r, m `Divides` m')
+         => UCyc t m' r -> UCyc t m r
+twace (Scalar c) = Scalar c
+-- twace on Sub goes to the largest common subring of O_l and O_m
+twace (Sub (c :: UCyc t l r)) =
+  Sub (twace c :: UCyc t (FGCD l m) r)
+  \\ gcdDivides (Proxy::Proxy l) (Proxy::Proxy m)
+twace (Pow v) = Pow $ twacePowDec v
+twace (Dec v) = Dec $ twacePowDec v
+-- stay in CRTr only if it's possible, otherwise go to Pow
+twace x@(CRTr v) =
+  fromMaybe (twace $ toPow' x) (CRTr <$> (twaceCRT <*> pure v))
+-- CJP: stay in CRTe: precision OK?
+twace (CRTe v) = CRTe $ fromMaybe (error "FC.twace CRTe") twaceCRT v
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.coeffsCyc', but for 'UCyc'.
+coeffsCyc :: (m `Divides` m', CElt t r) 
+             => U.Basis -> UCyc t m' r -> [UCyc t m r]
+coeffsCyc U.Pow (Pow v) = LP.map Pow $ coeffs v
+coeffsCyc U.Dec (Dec v) = LP.map Dec $ coeffs v
+coeffsCyc U.Pow x = coeffsCyc U.Pow $ toPow' x
+coeffsCyc U.Dec x = coeffsCyc U.Dec $ toDec' x
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.powBasis', but for 'UCyc'.
+powBasis :: (m `Divides` m', CElt t r) => Tagged m [UCyc t m' r]
+powBasis = map Pow <$> powBasisPow
+
+-- | Same as 'Crypto.Lol.Cyclotomic.Cyc.crtSet', but for 'UCyc'.
+crtSet :: forall t m m' r p mbar m'bar .
+           (m `Divides` m', ZPP r, p ~ CharOf (ZPOf r), 
+            mbar ~ PFree p m, m'bar ~ PFree p m',
+            CElt t r, CElt t (ZPOf r))
+           => Tagged m [UCyc t m' r]
+crtSet =
+  -- CJP: consider using traceEvent or traceMarker
+  --DT.trace ("UCyc.fcCrtSet: m = " ++
+  --          show (proxy valueFact (Proxy::Proxy m)) ++ ", m'= " ++
+  --          show (proxy valueFact (Proxy::Proxy m'))) $
+  let (p,e) = proxy modulusZPP (Proxy::Proxy r)
+      pp = Proxy::Proxy p
+      pm = Proxy::Proxy m
+      pm' = Proxy::Proxy m'
+  in retag (fmap (embed . (^(p^(e-1))) . Dec . fmapT liftZp) <$>
+            (crtSetDec :: Tagged mbar [t m'bar (ZPOf r)]))
+     \\ pFreeDivides pp pm pm'
+     \\ pSplitTheorems pp pm \\ pSplitTheorems pp pm'
+
+----- "Unsafe" functions that expose or rely upon internal representation
+
+-- | Yield an equivalent element whose internal representation /must/
+-- be in the indicated basis: powerful or decoding (for 'Just' 'Pow'
+-- and 'Just' 'Dec' arguments, respectively), or any @r@-basis of the
+-- implementation's choice (for 'Nothing' argument).  (See also the
+-- convenient specializations 'forcePow', 'forceDec', 'forceAny'.)
+forceBasis :: (Fact m, CElt t r) => Maybe U.Basis -> UCyc t m r -> UCyc t m r
+forceBasis (Just U.Pow) x = toPow' x
+forceBasis (Just U.Dec) x = toDec' x
+forceBasis Nothing x@(Scalar _) = toPow' x
+forceBasis Nothing (Sub c) = forceBasis Nothing $ embed' c
+forceBasis Nothing x@(CRTe _) = toPow' x
+forceBasis Nothing x = x
+
+forcePow, forceDec, forceAny :: (Fact m, CElt t r) => UCyc t m r -> UCyc t m r
+-- | Force a cyclotomic element into the powerful basis.
+forcePow = forceBasis (Just U.Pow)
+-- | Force a cyclotomic element into the decoding basis.
+forceDec = forceBasis (Just U.Dec)
+-- | Force a cyclotomic into any @r@-basis of the implementation's
+-- choice.
+forceAny = forceBasis Nothing
+
+-- | A more specialized version of 'fmap': apply a function
+-- coordinate-wise in the current representation.  The caller must
+-- ensure that the current representation is an @r@-basis (one of
+-- powerful, decoding, or CRT, if it exists), usually by using
+-- 'forceBasis' or its specializations ('forcePow', 'forceDec',
+-- 'forceAny').  Otherwise, behavior is undefined.
+fmapC :: (Fact m, CElt t a, CElt t b) => (a -> b) -> UCyc t m a -> UCyc t m b
+
+-- must be in an r-basis for correct semantics, e.g., f 0 = 1
+fmapC _ (Scalar _) = error "can't fmapC on Scalar.  Must forceBasis first!"
+fmapC _ (Sub _) = error "can't fmapC on Sub.  Must forceBasis first!"
+fmapC _ (CRTe _) = error "can't fmapC on CRTe.  Must forceBasis first!"
+
+fmapC f (Pow v) = Pow $ fmapT f v
+fmapC f (Dec v) = Dec $ fmapT f v
+fmapC f (CRTr v) = CRTr $ fmapT f v
+
+-- | Monadic version of 'fmapC'.
+fmapCM :: (Fact m, CElt t a, CElt t b, Monad mon)
+  => (a -> mon b) -> UCyc t m a -> mon (UCyc t m b)
+
+-- must embed into full ring
+fmapCM _ (Scalar _) = error "can't fmapCM on Scalar. Must forceBasis first!"
+fmapCM _ (Sub _) = error "can't fmapCM on Sub. Must forceBasis first!"
+fmapCM _ (CRTe _) =  error "can't fmapCM on CRTe.  Must forceBasis first!"
+
+fmapCM f (Pow v) = liftM Pow $ fmapTM f v
+fmapCM f (Dec v) = liftM Dec $ fmapTM f v
+fmapCM f (CRTr v) = liftM CRTr $ fmapTM f v
+
+
+
+
+---------- HELPER FUNCTIONS (NOT FOR EXPORT) ----------
+
+-- | Force embed, to a non-Sub constructor.
+embed' :: forall t r l m .
+          (UCCtx t r, l `Divides` m) => UCyc t l r -> UCyc t m r
+embed' (Scalar v) = Scalar v
+embed' (Pow v) = Pow $ embedPow v
+embed' (Dec v) = Dec $ embedDec v
+-- stay in CRTr only if it's possible, otherwise go to Pow
+embed' x@(CRTr v) =
+    fromMaybe (embed' $ toPow' x) (CRTr <$> (embedCRT <*> pure v))
+-- Staying in CRTe might not be safe, because the target tensor
+-- might have implemented a CRTr even if the source tensor
+-- hasn't.  Mathematically this is impossible (because target has
+-- CRTr only if source does), so this is purely about implementation.
+embed' x@(CRTe _) = embed' $ toPow' x
+embed' (Sub (c :: UCyc t k r)) = embed' c
+  \\ transDivides (Proxy::Proxy k) (Proxy::Proxy l) (Proxy::Proxy m)
+
+
+--------- Basis conversion methods ------------------
+
+toPow', toDec' :: (UCCtx t r, Fact m) => UCyc t m r -> UCyc t m r
+-- forces the argument into the powerful basis
+toPow' (Scalar c) = Pow $ scalarPow c
+toPow' (Sub c) = toPow' $ embed' c
+toPow' x@(Pow _) = x
+toPow' (Dec v) = Pow $ l v
+toPow' (CRTr v) = Pow $ fromMaybe (error "FC.toPow'") crtInv v
+toPow' (CRTe v) = Pow $ fmapT fromExt $ fromMaybe (error "FC.toPow'") crtInv v
+
+-- forces the argument into the decoding basis
+toDec' x@(Scalar _) = toDec' $ toPow' x -- use scalarDec instead
+toDec' (Sub c) = toDec' $ embed' c
+toDec' (Pow v) = Dec $ lInv v
+toDec' x@(Dec _) = x
+toDec' (CRTr v) = Dec $ lInv $ fromMaybe (error "FC.toDec'") crtInv v
+toDec' (CRTe v) = Dec $ lInv $ fmapT fromExt $ fromMaybe (error "FC.toDec'") crtInv v
+
+-- forces the argument into a CRT basis, according to the invariant
+-- about which one should be used
+toCRT' :: forall t m r . (UCCtx t r, Fact m) => UCyc t m r -> UCyc t m r
+toCRT' (Sub c) = toCRT' $ embed' c
+toCRT' x@(CRTr _) = x
+toCRT' x@(CRTe _) = x
+toCRT' x = fromMaybe (toCRTe x) (toCRTr <*> pure x)
+  -- CJP: defining these helpers internally so they can't be called
+  -- from anywhere else.  Therefore, the only way to convert to a
+  -- CRT basis is through the toCRT' method.
+  where
+    toCRTr = do -- Maybe monad
+      crt' <- crt
+      scalarCRT' <- scalarCRT
+      return $ \x -> case x of
+        (Scalar c) -> CRTr $ scalarCRT' c
+        (Pow v) -> CRTr $ crt' v
+        (Dec v) -> CRTr $ crt' $ l v
+          -- deliberately omit CRTe case, which should
+          -- never happen by internal invariant, so trigger
+          -- error if it does
+    toCRTe = let m = proxy valueFact (Proxy::Proxy m)
+                 crt' = fromMaybe (error $ "FC.toCRT': no crt': " ++ (show m)) crt :: t m (CRTExt r) -> t m (CRTExt r) -- must exist
+                 scalarCRT' = fromMaybe (error "FC.toCRT': no scalar crt'") scalarCRT :: CRTExt r -> t m (CRTExt r)
+             in \x -> case x of
+               (Scalar c) -> CRTe $ scalarCRT' $ toExt c
+               (Pow v) -> CRTe $ crt' $ fmapT toExt v
+               (Dec v) -> CRTe $ crt' $ fmapT toExt $ l v
+
+---------- "Container" instances ----------
+
+instance (Tensor t, Fact m) => Functor (UCyc t m) where
+  -- Functor instance is implied by Applicative laws
+  fmap f x = pure f <*> x
+
+errApp name = error $ "UCyc.Applicative: can't/won't handle " ++ name ++
+              "; call forcePow|Dec first"
+
+instance (Tensor t, Fact m) => Applicative (UCyc t m) where
+
+  -- This implementation is restricted to the Scalar, Pow, Dec, or
+  -- CRTr constructors, in order to force the client to choose a
+  -- concrete @r@-basis and avoid unanticipated non-failure behavior.
+  -- Encountering a CRTe, or Sub constructor almost certainly means
+  -- that the client expressed something it did not intend (since it
+  -- cannot force such constructors to be used), so we want to raise
+  -- an exception early instead of doing unintended work.
+
+  -- This implementation has one corner case that may
+  -- yield unexpected non-failure behavior: consider
+  --   fmap f (pure a) = (pure f) <*> (pure a) = (pure $ f a)
+  -- which is required by the Applicative homomorphism law.
+
+  -- If the (pure a) is intended as an element of the base ring (which
+  -- is the custom), then its Pow coeffs are *not* all a's, so the
+  -- (likely intended) expression
+  --   fmap f $ forcePow (pure a)
+  -- may be a different result.  If the client forgets the force, we
+  -- can't recognize it here and throw an error.  (This is certainly the
+  -- client's fault; if it's not specifying a basis before fmap'ing
+  -- then it shouldn't expect the results to make sense.  We just
+  -- can't catch the error here.)
+
+  -- A solution is to introduce an explicit Pure constructor that's
+  -- only ever applied in 'pure', and throw an error if we encounter a
+  -- Scalar here.  Arithmetically we'd treat Pures as Scalars, but in
+  -- a one-way fashion (outputs of arith ops are never Pure).
+
+  pure = Scalar
+
+  -- homomorphism (of pure)
+  (Scalar f) <*> (Scalar a) = Scalar $ f a
+
+  -- constructors must match
+  (Pow v1) <*> (Pow v2) = Pow $ v1 <*> v2 \\ witness entailIndexT v1
+  (Dec v1) <*> (Dec v2) = Dec $ v1 <*> v2 \\ witness entailIndexT v1
+  (CRTr v1) <*> (CRTr v2) = CRTr $ v1 <*> v2 \\ witness entailIndexT v1
+
+  -- ... but we can also match Scalar with (almost) anything
+  (Scalar f) <*> (Pow v) = Pow $ pure f <*> v \\ witness entailIndexT v
+  (Scalar f) <*> (Dec v) = Dec $ pure f <*> v \\ witness entailIndexT v
+  (Scalar f) <*> (CRTr v) = CRTr $ pure f <*> v \\ witness entailIndexT v
+
+  (Pow v) <*> (Scalar a) = Pow $ v <*> pure a \\ witness entailIndexT v
+  (Dec v) <*> (Scalar a) = Dec $ v <*> pure a \\ witness entailIndexT v
+  (CRTr v) <*> (Scalar a) = CRTr $ v <*> pure a \\ witness entailIndexT v
+
+  -- cases we can't/won't handle
+  (Pow _) <*> (Dec _) = error "UCyc.Applicative: Pow/Dec combo"
+  (Dec _) <*> (Pow _) = error "UCyc.Applicative: Pow/Dec combo"
+  (Sub _) <*> _  = errApp "Sub"
+  _ <*> (Sub _)  = errApp "Sub"
+  (CRTe _) <*> _ = errApp "CRTe"
+  _ <*> (CRTe _) = errApp "CRTe"
+
+instance (Tensor t, Fact m) => Foldable (UCyc t m) where
+  foldr f b (Scalar r) = f r b
+  foldr f b (Sub c) = F.foldr f b c
+  foldr f b (Pow v) = F.foldr f b v \\ witness entailIndexT v
+  foldr f b (Dec v) = F.foldr f b v \\ witness entailIndexT v
+  foldr f b (CRTr v) = F.foldr f b v \\ witness entailIndexT v
+  foldr _ _ (CRTe _) = error "UCyc.Foldable: can't handle CRTe"
+
+instance (Tensor t, Fact m) => Traversable (UCyc t m) where
+  traverse f (Scalar r) = Scalar <$> f r
+  traverse f (Sub c) = Sub <$> traverse f c
+  traverse f (Pow v) = Pow <$> traverse f v \\ witness entailIndexT v
+  traverse f (Dec v) = Dec <$> traverse f v \\ witness entailIndexT v
+  traverse f (CRTr v) = CRTr <$> traverse f v \\ witness entailIndexT v
+  traverse _ (CRTe _) = error "UCyc.Traversable: can't handle CRTe"
+
+---------- Utility instances ----------
+
+instance (Tensor t, Fact m, TElt t r, CRTrans r) => Random (UCyc t m r) where
+
+  -- create in CRTr basis if legal, otherwise in powerful
+  random = let cons = fromMaybe Pow
+                      (proxyT hasCRTFuncs (Proxy::Proxy (t m r)) >> Just CRTr)
+           in \g -> let (v,g') = random g
+                                 \\ proxy entailFullT (Proxy::Proxy (t m r))
+                    in (cons v, g')
+
+  randomR _ = error "randomR non-sensical for cyclotomic rings"
+
+instance (Show r, Show (t m r), Show (t m (CRTExt r)))
+  => Show (UCyc t m r) where
+
+  show (Scalar c) = "scalar " ++ show c
+  show (Sub _) = "subring (not showing due to missing constraints)"
+  show (Pow v) = "powerful basis coeffs " ++ show v
+  show (Dec v) = "decoding basis coeffs " ++ show v
+  show (CRTr v) = "CRTr basis coeffs " ++ show v
+  show (CRTe v) = "CRTe basis coeffs " ++ show v
+
+instance (Arbitrary (t m r)) => Arbitrary (UCyc t m r) where
+  arbitrary = liftM Pow arbitrary
+  shrink = shrinkNothing
+
+instance (Tensor t, Fact m, NFData r, TElt t r, TElt t (CRTExt r))
+         => NFData (UCyc t m r) where
+  rnf (Pow x)    = rnf x \\ witness entailFullT x
+  rnf (Dec x)    = rnf x \\ witness entailFullT x
+  rnf (CRTr x)   = rnf x \\ witness entailFullT x
+  rnf (CRTe x)   = rnf x \\ witness entailFullT x
+  rnf (Scalar x) = rnf x
+  rnf (Sub x)    = rnf x
diff --git a/src/Crypto/Lol/Cyclotomic/Utility.hs b/src/Crypto/Lol/Cyclotomic/Utility.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Cyclotomic/Utility.hs
@@ -0,0 +1,22 @@
+{-# LANGUAGE MultiParamTypeClasses #-}
+
+module Crypto.Lol.Cyclotomic.Utility where
+
+import Crypto.Lol.Factored
+
+import Control.DeepSeq
+
+-- | Represents the powerful or decoding basis.
+data Basis = Pow | Dec
+
+instance NFData Basis where
+  rnf Pow = ()
+  rnf Dec = ()
+
+-- | Represents cyclotomic rings that are rescalable over their base
+-- rings.  (This is a class because it allows for more efficient
+-- specialized implementations.)
+
+class RescaleCyc c a b where
+  -- | Rescale in the given basis.
+  rescaleCyc :: Fact m => Basis -> c m a -> c m b
diff --git a/src/Crypto/Lol/Factored.hs b/src/Crypto/Lol/Factored.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Factored.hs
@@ -0,0 +1,481 @@
+{-# LANGUAGE ConstraintKinds, DataKinds,
+             GADTs, KindSignatures, PolyKinds,
+             ScopedTypeVariables, TemplateHaskell, TypeFamilies,
+             TypeOperators, UndecidableInstances #-}
+
+-- | This file defines types and operations for type-level
+-- representation and manipulation of factored integers. It relies on
+-- TH, so the documentation will be difficult to read. See comments
+-- for better documentation.
+
+module Crypto.Lol.Factored
+(
+-- * Factored natural numbers
+  Factored(..), SFactored, Fact
+-- * Prime powers
+, PrimePower(..), SPrimePower, PPow, Sing (SPP)
+-- * Naturals
+, Nat, NatC, PrimeNat, Prime
+-- * Constructors
+, toPP, sToPP, ToPP, ppToF, sPpToF, PpToF, PToF
+-- * Unwrappers
+, unF, sUnF, UnF, unPP, sUnPP, UnPP, primePP, PrimePP, exponentPP, ExponentPP
+-- * Arithmetic operations
+, fPPMul, FPPMul, fMul, FMul, type (*)
+, fDivides, FDivides, Divides, fDiv, FDiv, type (/)
+, fGCD, FGCD, fLCM, FLCM, Coprime
+, fOddRadical, FOddRadical
+, pFree, PFree
+-- * Reflections
+, ppsFact, valueFact, totientFact, valueHatFact, radicalFact, oddRadicalFact
+, ppPPow, primePPow, exponentPPow, valuePPow, totientPPow
+, valueNatC
+-- * Number-theoretic laws
+, transDivides, gcdDivides, lcmDivides, lcm2Divides
+, pSplitTheorems, pFreeDivides
+, (\\) -- re-export from Data.Constraint for convenience
+-- * Utility operations (on prime powers)
+, valueHat
+, PP, ppToPP, valuePP, totientPP, radicalPP, oddRadicalPP
+, valuePPs, totientPPs, radicalPPs, oddRadicalPPs
+-- * Type synonyms (not type families)
+, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10
+, F11, F12, F13, F14, F15, F16, F17, F18, F19, F20
+, F21, F22, F24, F25, F26, F27, F28, F30
+, F32, F33, F34, F35, F36, F38, F39
+, F40, F42, F44, F45, F48, F49
+, F50, F51, F52, F54, F55, F56, F57
+, F60, F63, F64, F65, F66, F68
+, F70, F72, F75, F76, F77, F78, F80, F81, F84, F85, F88
+, F90, F91, F95, F96, F98, F99
+, F128, F256, F512, F1024, F2048
+) where
+
+import Data.Constraint hiding ((***))
+import Data.Functor.Trans.Tagged
+import Data.Singletons.Prelude hiding (sMin, sMax, MinSym0, MaxSym0, (:-))
+import Data.Singletons.TH
+import Data.Type.Natural         as N hiding ((:-))
+import Data.Typeable
+
+import Control.Arrow ((***))
+import Unsafe.Coerce
+
+-- | Copied from Data.Type.Natural because the data-level version
+-- is not exported there.
+(<<=) :: Nat -> Nat -> Bool
+Z   <<= _   = True
+S _ <<= Z   = False
+S n <<= S m = n <<= m
+
+singletons [d|
+
+            -- Invariant: first component is prime, second component
+            -- (the exponent) is positive (nonzero)
+            newtype PrimePower = PP (Nat,Nat) deriving (Eq,Show,Typeable)
+
+            -- List invariant: primes appear in strictly increasing
+            -- order (no duplicates)
+            newtype Factored = F [PrimePower] deriving (Eq,Show,Typeable)
+
+            -- unwrap 'Factored'
+            unF :: Factored -> [PrimePower]
+            unF (F pps) = pps
+
+            -- unwrap 'PrimePower'
+            unPP :: PrimePower -> (Nat,Nat)
+            unPP (PP pp) = pp
+
+            -- grab individual components of a 'PrimePower'
+            primePP, exponentPP :: PrimePower -> Nat
+            primePP = fst . unPP
+            exponentPP = snd . unPP
+
+            |]
+
+-- SMART CONSTRUCTORS
+singletons [d|
+
+            fPPMul :: PrimePower -> Factored -> Factored
+            fMul :: Factored -> Factored -> Factored
+
+            -- constructor implementations
+            -- multiply a new 'PrimePower' into a 'Factored' number
+            fPPMul (PP(_,Z)) y = y -- throw away trivial prime power
+            fPPMul pp@(PP(_,S _)) (F pps) = F (ppMul pp pps)
+
+            -- multiply two 'Factored' numbers
+            fMul (F pps1) (F pps2) = F (ppsMul pps1 pps2)
+
+            -- helper functions (not for export)
+
+            -- keeps primes in sorted order; merges duplicates
+
+            -- EAC: Singletons(?) doesn't play well with pattern synonyms (e.g. x@(PP(p,e)))
+            -- when compiling with -O2
+            -- reported as #10924
+            -- singletons-2.0 doesn't work well with guards: https://github.com/goldfirere/singletons/issues/131
+            ppMul :: PrimePower -> [PrimePower] -> [PrimePower]
+            ppMul x [] = [x]
+            ppMul (PP(p,e)) (PP (p',e'):pps') =
+              if p == p' then PP(p,e + e'):pps'
+              else if p <<= p' then (PP(p,e)):(PP (p',e'):pps')
+              else (PP(p',e')):ppMul (PP(p,e)) pps'
+
+            ppsMul :: [PrimePower] -> [PrimePower] -> [PrimePower]
+            ppsMul [] ys = ys
+            ppsMul (pp:pps) ys = ppsMul pps (ppMul pp ys)
+
+            |]
+
+-- ARITHMETIC OPERATIONS
+singletons [d|
+            -- Smart constructor that checks that the first arg is
+            -- prime (< 20) and the second arg is positive
+            toPP :: Nat -> Nat -> PrimePower
+            toPP p e | primeNat p && (n1 <<= e) = PP (p,e)
+
+            -- EAC: isn't there a singletons promotion for 'F'
+            -- that could replace this function?
+            ppToF :: PrimePower -> Factored
+            ppToF pp = F [pp]
+
+            primeToF :: Nat -> Factored
+            primeToF p | primeNat p = ppToF $ PP (p, n1)
+            
+            fGCD, fLCM :: Factored -> Factored -> Factored
+            fDivides :: Factored -> Factored -> Bool
+            fDiv :: Factored -> Factored -> Factored
+            fOddRadical :: Factored -> Factored
+            
+            -- can't pattern-match on n*, but can test equality
+            primeNat n
+              | n==n2 = True
+              | n==n3 = True
+              | n==n5 = True
+              | n==n7 = True
+              | n==n11 = True
+              | n==n13 = True
+              | n==n17 = True
+              | n==n19 = True
+            fGCD (F pps1) (F pps2) = F (ppsGCD pps1 pps2)
+            fLCM (F pps1) (F pps2) = F (ppsLCM pps1 pps2)
+
+            fDivides (F pps1) (F pps2) = ppsDivides pps1 pps2
+            fDiv (F pps1) (F pps2) = F (ppsDiv pps1 pps2)
+            fOddRadical (F pps) = F (ppsOddRad pps)
+
+            -- Helper functions (not for export) on PrimePowers and
+            -- lists.  Can assume that input lists obey the invariant
+            -- of Factored lists, and need to ensure that output lists
+            -- also obey the invariant.
+            ppsGCD :: [PrimePower] -> [PrimePower] -> [PrimePower]
+            ppsGCD [] [] = []
+            ppsGCD [] (_:_) = []
+            ppsGCD (_:_) [] = []
+            ppsGCD (PP (p,e) : xs') (PP (p',e') : ys') =
+              if p == p' then PP (p,N.min e e') : ppsGCD xs' ys'
+              else if p <<= p' then ppsGCD xs' (PP (p',e') : ys')
+              else ppsGCD (PP (p,e) : xs') ys'
+
+            ppsLCM :: [PrimePower] -> [PrimePower] -> [PrimePower]
+            ppsLCM [] [] = []
+            ppsLCM [] ys@(_:_) = ys
+            ppsLCM xs@(_:_) [] = xs
+            ppsLCM ((PP (p,e)) : xs') ((PP (p',e')) : ys') =
+              if p == p' then PP (p,N.max e e') : ppsLCM xs' ys'
+              else if p <<= p' then (PP (p,e)) : ppsLCM xs' ((PP (p',e')) : ys')
+              else (PP (p',e')) : ppsLCM ((PP (p,e)) : xs') ys'
+
+            ppsDivides :: [PrimePower] -> [PrimePower] -> Bool
+            ppsDivides [] _ = True
+            ppsDivides (_:_) [] = False
+            ppsDivides (PP (p,e) : xs') (PP (p',e') : ys') =
+              if p == p' then (e <<= e') && ppsDivides xs' ys'
+              else not (p <<= p') && ppsDivides (PP (p,e) : xs') ys'
+
+            ppsDiv :: [PrimePower] -> [PrimePower] -> [PrimePower]
+            ppsDiv xs [] = xs
+            ppsDiv ((PP (p,e)) : xs') (PP (p',e') : ys') =
+              if p == p' && e' == e then ppsDiv xs' ys'
+              else if p == p' && e' <<= e then PP (p,e-e') : ppsDiv xs' ys'
+              else if p <<= p' then (PP (p,e)) : ppsDiv xs' (PP (p',e') : ys')
+              else error "type error in ppsDiv"                -- if p' <<= p then it's an error
+
+            ppsOddRad :: [PrimePower] -> [PrimePower]
+            ppsOddRad [] = []
+            ppsOddRad (PP ((S (S Z)),_) : xs') = ppsOddRad xs'
+            -- need to expand to avoid overlapping with previous case
+            ppsOddRad (PP (p@(S (S (S _))),_) : xs') = PP (p,n1) : ppsOddRad xs'
+
+            |]
+
+singletons [d|
+            -- removes all @p@-factors from a 'Factored'
+            pFree :: Nat -> Factored -> Factored
+            pFree n (F pps) = F (go pps)
+              where go [] = []
+                    go (pp@(PP (p,_)) : ps) =
+                      if n == p then ps
+                      else pp : (go ps)
+            |]
+
+singletons [d|
+
+            f1 = F []
+            f2 = primeToF n2
+            f3 = primeToF n3
+            f4 = f2 `fMul` f2
+            f5 = primeToF n5
+            f6 = f2 `fMul` f3
+            f7 = primeToF n7
+            f8 = f2 `fMul` f4
+            f9 = f3 `fMul` f3
+            f10 = f2 `fMul` f5
+            f11 = primeToF n11
+            f12 = f4 `fMul` f3
+            f13 = primeToF n13
+            f14 = f2 `fMul` f7
+            f15 = f3 `fMul` f5
+            f16 = f2 `fMul` f8
+            f17 = primeToF n17
+            f18 = f2 `fMul` f9
+            f19 = primeToF n19
+            f20 = f2 `fMul` f10
+            f21 = f3 `fMul` f7
+            f22 = f2 `fMul` f11
+            f24 = f2 `fMul` f12
+            f25 = f5 `fMul` f5
+            f26 = f2 `fMul` f13
+            f27 = f3 `fMul` f9
+            f28 = f2 `fMul` f14
+            f30 = f2 `fMul` f15
+            f32 = f2 `fMul` f16
+            f33 = f3 `fMul` f11
+            f34 = f2 `fMul` f17
+            f35 = f5 `fMul` f7
+            f36 = f2 `fMul` f18
+            f38 = f2 `fMul` f19
+            f39 = f3 `fMul` f13
+            f40 = f2 `fMul` f20
+            f42 = f2 `fMul` f21
+            f44 = f2 `fMul` f22
+            f45 = f3 `fMul` f15
+            f48 = f2 `fMul` f24
+            f49 = f7 `fMul` f7
+            f50 = f2 `fMul` f25
+            f51 = f3 `fMul` f17
+            f52 = f2 `fMul` f26
+            f54 = f2 `fMul` f27
+            f55 = f5 `fMul` f11
+            f56 = f2 `fMul` f28
+            f57 = f3 `fMul` f19
+            f60 = f2 `fMul` f30
+            f63 = f3 `fMul` f21
+            f64 = f2 `fMul` f32
+            f65 = f5 `fMul` f13
+            f66 = f2 `fMul` f33
+            f68 = f2 `fMul` f34
+            f70 = f2 `fMul` f35
+            f72 = f2 `fMul` f36
+            f75 = f3 `fMul` f25
+            f76 = f2 `fMul` f38
+            f77 = f7 `fMul` f11
+            f78 = f2 `fMul` f39
+            f80 = f2 `fMul` f40
+            f81 = f3 `fMul` f27
+            f84 = f2 `fMul` f42
+            f85 = f5 `fMul` f17
+            f88 = f2 `fMul` f44
+            f90 = f2 `fMul` f45
+            f91 = f7 `fMul` f13
+            f95 = f5 `fMul` f19
+            f96 = f2 `fMul` f48
+            f98 = f2 `fMul` f49
+            f99 = f9 `fMul` f11
+            f128 = f2 `fMul` f64
+            f256 = f2 `fMul` f128
+            f512 = f2 `fMul` f256
+            f1024 = f2 `fMul` f512
+            f2048 = f2 `fMul` f1024
+            |]
+
+-- | Type (family) synonym for division of 'Factored' types
+type a / b = FDiv a b
+
+-- | Type (family) synonym for multiplication of 'Factored' types
+type a * b = FMul a b
+
+-- | Type (family) synonym to create a Factored from a prime Nat
+type PToF p = PpToF (ToPP p N1)
+
+-- convenience aliases: enforce kind, hide SingI
+
+-- | Kind-restricted synonym for 'SingI'. Use this in constraints 
+-- for types requiring a 'Factored' type.
+type Fact (m :: Factored) = SingI m
+
+-- | Kind-restricted synonym for 'SingI'. Use this in constraints 
+-- for types requiring a 'PrimePower' type.
+type PPow (pp :: PrimePower) = SingI pp
+
+-- | Kind-restricted synonym for 'SingI'. Use this in constraints 
+-- for types requiring a 'Nat' type.
+type NatC (p :: Nat) = SingI p
+
+type Prime p = (NatC p, PrimeNat p ~ 'True)
+
+-- | Constraint synonym for divisibility of 'Factored' types
+type Divides m m' = (Fact m, Fact m', FDivides m m' ~ 'True)
+
+-- | Constraint synonym for coprimality of 'Factored' types
+type Coprime m m' = (FGCD m m' ~ F1)
+
+-- coercions: using proxy arguments here due to compiler bugs in usage
+
+-- coerce any divisibility relationship we want
+coerceFDivs :: p m -> p' m' -> (() :- (FDivides m m' ~ True))
+coerceFDivs _ _ = Sub $ unsafeCoerce (Dict :: Dict ())
+
+-- coerce any GCD we want
+coerceGCD :: p a -> p' a' -> p'' a'' -> (() :- (FGCD a a' ~ a''))
+coerceGCD _ _ _ = Sub $ unsafeCoerce (Dict :: Dict ())
+
+-- | Entails constraint for transitivity of division, i.e.
+-- if @k|l@ and @l|m@, then @k|m@.
+transDivides :: forall k l m . Proxy k -> Proxy l -> Proxy m ->
+                ((k `Divides` l, l `Divides` m) :- (k `Divides` m))
+transDivides k _ m = Sub Dict \\ coerceFDivs k m
+
+-- | Entails constraint for divisibility by GCD, i.e.
+-- if @g=GCD(m1,m2)@, then @g|m1@ and @g|m2@.
+gcdDivides :: forall m1 m2 g . Proxy m1 -> Proxy m2 ->
+              ((Fact m1, Fact m2, g ~ FGCD m1 m2) :-
+               (g `Divides` m1, g `Divides` m2))
+gcdDivides m1 m2 =
+  Sub $ withSingI (sFGCD (sing :: SFactored m1) (sing :: SFactored m2))
+  Dict \\ coerceFDivs (Proxy::Proxy g) m1
+       \\ coerceFDivs (Proxy::Proxy g) m2
+
+-- | Entails constraint for LCM divisibility, i.e.
+-- if @l=LCM(m1,m2)@, then @m1|l@ and @m2|l@.
+lcmDivides :: forall m1 m2 l . Proxy m1 -> Proxy m2 ->
+              ((Fact m1, Fact m2, l ~ FLCM m1 m2) :-
+               (m1 `Divides` l, m2 `Divides` l))
+lcmDivides m1 m2 = 
+  Sub $ withSingI (sFLCM (sing :: SFactored m1) (sing :: SFactored m2))
+  Dict \\ coerceFDivs m1 (Proxy::Proxy l)
+       \\ coerceFDivs m2 (Proxy::Proxy l)
+
+-- | Entails constraint for LCM divisibility, i.e.
+-- the LCM of two divisors of @m@ also divides @m@.
+lcm2Divides :: forall m1 m2 l m . Proxy m1 -> Proxy m2 -> Proxy m ->
+               ((m1 `Divides` m, m2 `Divides` m, l ~ FLCM m1 m2) :-
+                (m1 `Divides` l, m2 `Divides` l, (FLCM m1 m2) `Divides` m))
+lcm2Divides m1 m2 m = 
+  Sub $ withSingI (sFLCM (sing :: SFactored m1) (sing :: SFactored m2))
+  Dict \\ coerceFDivs (Proxy::Proxy (FLCM m1 m2)) m \\ lcmDivides m1 m2
+
+-- | Entails basic facts for @p@-free division, i.e.
+-- if @f@ is @m@ after removing all @p@-factors, then @f|m@ and
+-- @gcd(f,p)=1@
+pSplitTheorems :: forall p m f . Proxy p -> Proxy m ->
+                  ((NatC p, Fact m, f ~ PFree p m) :-
+                   (f `Divides` m, Coprime (PToF p) f))
+pSplitTheorems _ m =
+  Sub $ withSingI (sPFree (sing :: SNat p) (sing :: SFactored m))
+  Dict \\ coerceFDivs (Proxy::Proxy f) m 
+  \\ coerceGCD (Proxy::Proxy (PToF p)) (Proxy::Proxy f) (Proxy::Proxy F1)
+
+-- | Entails basic facts for @p@-free division, i.e.,
+-- if @m|m'@, then @p-free(m) | p-free(m')@
+pFreeDivides :: forall p m m' . Proxy p -> Proxy m -> Proxy m' ->
+                ((NatC p, m `Divides` m') :-
+                 ((PFree p m) `Divides` (PFree p m')))
+pFreeDivides _ _ _ =
+  Sub $ withSingI (sPFree (sing :: SNat p) (sing :: SFactored m)) $
+        withSingI (sPFree (sing :: SNat p) (sing :: SFactored m')) $
+        Dict \\ coerceFDivs (Proxy::Proxy (PFree p m)) (Proxy::Proxy (PFree p m'))
+
+-- | Type synonym for @(prime :: Int, exponent :: Int)@ pair
+type PP = (Int, Int)
+
+-- | Value-level prime-power factorization tagged by a 'Factored' type.
+ppsFact :: forall m . (Fact m) => Tagged m [PP]
+ppsFact = tag $ map ppToPP $ unF $ fromSing (sing :: SFactored m)
+
+valueFact, totientFact, valueHatFact, radicalFact, oddRadicalFact ::
+  (Fact m) => Tagged m Int
+
+-- | @Int@ representing the value of a 'Factored' type
+valueFact = valuePPs <$> ppsFact
+
+-- | @Int@ representing the totient of a 'Factored' type's value
+totientFact = totientPPs <$> ppsFact
+
+-- | @Int@ representing the "hat" of a 'Factored' type's value @m@:
+-- @m@, if @m@ is odd, or @m/2@ otherwise.
+valueHatFact = valueHat <$> valueFact
+
+-- | @Int@ representing the radical (product of prime divisors)
+-- of a 'Factored' type
+radicalFact = radicalPPs <$> ppsFact
+
+-- | @Int@ representing the odd radical (product of odd prime divisors)
+-- of a 'Factored' type
+oddRadicalFact = oddRadicalPPs <$> ppsFact
+
+-- | Reflects a 'PrimePower' type to a 'PP' value
+ppPPow :: forall pp . (PPow pp) => Tagged pp PP
+ppPPow = tag $ ppToPP $ fromSing (sing :: SPrimePower pp)
+
+primePPow, exponentPPow, valuePPow, totientPPow :: (PPow pp) => Tagged pp Int
+-- | Reflects the prime component of a 'PrimePower' type
+primePPow = fst <$> ppPPow
+-- | Reflects the exponent component of a 'PrimePower' type
+exponentPPow = snd <$> ppPPow
+-- | @Int@ representing the value of a 'PrimePower' type
+valuePPow = valuePP <$> ppPPow
+-- | @Int@ representing the totient of a 'PrimePower' type's value
+totientPPow = totientPP <$> ppPPow
+
+-- | @Int@ representing the value of a 'Nat'
+valueNatC :: forall p . (NatC p) => Tagged p Int
+valueNatC = tag $ sNatToInt (sing :: SNat p)
+
+-- | Returns @m@, if @m@ is odd, or @m/2@ otherwise
+valueHat :: (Integral i) => i -> i
+valueHat m = if m `mod` 2 == 0 then m `div` 2 else m
+
+-- | Converts a 'Nat' prime-power pair to an @Int@ prime-power pair
+ppToPP :: PrimePower -> PP
+ppToPP = (natToInt *** natToInt) . unPP
+
+valuePP, totientPP, radicalPP, oddRadicalPP :: PP -> Int
+-- | Evaluates a prime-power pair @(p,e)@ to @p^e@
+valuePP (p,e) = p^e
+
+-- | Euler's totient function of a prime-power pair
+totientPP (_,0) = 1
+totientPP (p,e) = (p-1)*(p^(e-1))
+
+-- | The prime component of a prime-power pair
+radicalPP (_,0) = 1
+radicalPP (p,_) = p
+
+-- | The odd radical of a prime-power pair (p,_):
+-- p if p is odd,
+-- 1 if p==2
+oddRadicalPP (_,0) = 1
+oddRadicalPP (2,_) = 1
+oddRadicalPP (p,_) = p
+
+valuePPs, totientPPs, radicalPPs, oddRadicalPPs :: [PP] -> Int
+-- | Product of values of individual 'PP's
+valuePPs = product . map valuePP
+-- | Product of totients of individual 'PP's
+totientPPs = product . map totientPP
+-- | Product of radicals of individual 'PP's
+radicalPPs = product . map radicalPP
+-- | Product of odd radicals of individual 'PP's
+oddRadicalPPs = product . map oddRadicalPP
diff --git a/src/Crypto/Lol/Gadget.hs b/src/Crypto/Lol/Gadget.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Gadget.hs
@@ -0,0 +1,93 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             FlexibleContexts, FlexibleInstances, MultiParamTypeClasses,
+             NoImplicitPrelude, PolyKinds, ScopedTypeVariables,
+             TupleSections, TypeFamilies, UndecidableInstances #-}
+
+-- | Interfaces for "gadgets," decomposition, and error correction.
+
+module Crypto.Lol.Gadget
+( Gadget(..), Decompose(..), Correct(..)
+, TrivGad, BaseBGad
+) where
+
+import Crypto.Lol.LatticePrelude
+
+import Control.Applicative
+import Data.Typeable
+
+-- | Dummy type representing the gadget @[1]@.
+data TrivGad deriving (Typeable)
+-- | Dummy type representing the gadget @[1,b,b^2,...]@.
+data BaseBGad b deriving (Typeable)
+
+-- | "Gadget" vectors, parameterized by an index type.
+
+class Ring u => Gadget gad u where
+  -- | The gadget vector over @u@.
+  gadget :: Tagged gad [u]
+
+  -- | Yield an error-tolerant encoding of an element with respect to
+  -- the gadget.  (Mathematically, this should just be the product of
+  -- the input with the gadget, but it is a class method to allow for
+  -- optimized implementations.)
+  encode :: u -> Tagged gad [u]
+  encode s = ((* s) <$>) <$> gadget
+
+-- | Decomposition relative to a gadget.
+
+class (Gadget gad u, Reduce (DecompOf u) u) => Decompose gad u where
+  -- | The ring that @u@ decomposes over.
+  type DecompOf u
+
+  -- | Yield a short vector @x@ such that @\<g, x\> = u@.
+  decompose :: u -> Tagged gad [DecompOf u]
+
+-- | Error correction relative to a gadget.
+
+class Gadget gad u => Correct gad u where
+
+  -- | Correct a "noisy" encoding of an element (see 'encode').
+  correct :: Tagged gad [u] -> u
+
+
+
+-- instances for products
+
+instance (Gadget gad a, Gadget gad b) => Gadget gad (a,b) where
+
+  gadget = (++) <$> (map (,zero) <$> gadget) <*> (map (zero,) <$> gadget)
+
+instance (Decompose gad a, Decompose gad b, DecompOf a ~ DecompOf b)
+         => Decompose gad (a,b) where
+
+  type DecompOf (a,b) = DecompOf a
+
+  decompose (a,b) = (++) <$> decompose a <*> decompose b
+
+
+-- TODO: need some extra constraints on a,b, like Mod and maybe Rescale.
+-- instance (Correct gad a, Correct gad b) => Correct gad (a,b) where
+
+
+
+{- CJP: strawman class for the more general view of LWE secrets as
+"module characters," i.e., module homomorphisms into a particular
+range.  This is probably wrong, though.
+
+class Character u where       -- Module superclass(es)?
+  type CharRange u
+  data Char u                   -- need data for injectivity
+
+  evalChar :: Char u -> u -> CharRange u
+
+class (Gadget gad u, Character u) => Correct gad u where
+
+  -- | Correct a "noisy" encoding of an LWE secret (i.e., a
+  -- 'ModuleHomom' on 'u').
+  correct :: Tagged gad [CharRange u] -> Char u
+
+encode :: (Correct gad u) => Char u -> Tagged gad [CharRange u]
+encode s = pasteT $ evalMH s <$> peelT gadget
+
+-}
+
diff --git a/src/Crypto/Lol/GaussRandom.hs b/src/Crypto/Lol/GaussRandom.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/GaussRandom.hs
@@ -0,0 +1,84 @@
+{-# LANGUAGE NoImplicitPrelude, RebindableSyntax, ScopedTypeVariables #-}
+
+-- | Functions for sampling from a continuous Gaussian distribution
+
+module Crypto.Lol.GaussRandom
+( realGaussian, realGaussians ) where
+
+import Crypto.Lol.LatticePrelude
+
+import qualified Data.Vector.Generic as V
+
+import Control.Monad
+import Control.Monad.Random
+
+-- | Using polar form of Box-Muller transform, returns a pair of
+-- centered, Gaussian-distributed real numbers with scaled variance
+-- @svar = true variance * (2*pi)@. See
+-- <http://www.alpheratz.net/murison/Maple/GaussianDistribution/GaussianDistribution.pdf
+-- this link> for details.
+
+realGaussian :: forall v q m .
+                (ToRational v, OrdFloat q, Random q, MonadRandom m)
+                => v -> m (q,q)
+realGaussian svar =
+    let var = realToField svar / pi :: q -- twice true variance
+    in do (u,v) <- iterateWhile uvGuard getUV
+          let t = u*u+v*v
+              com = sqrt (-var * log t / t)
+          return (u * com, v * com)
+    where getUV = do u <- getRandomR (zero,one)
+                     v <- getRandomR (zero,one)
+                     return (u,v)
+          uvGuard (u,v) = (u*u+v*v >= one) || (u*u+v*v == zero)
+
+-- | Generate @n@ real, independent gaussians of scaled variance @svar
+-- = true variance * (2*pi)@.
+realGaussians ::
+    (ToRational svar, OrdFloat i, Random i, V.Vector v i, MonadRandom m)
+    => svar -> Int -> m (v i)
+realGaussians var n
+    | odd n = liftM V.tail (realGaussians var (n+1)) -- O(1) tail
+    | otherwise = liftM (V.fromList . uncurry (++) . unzip) $
+                  replicateM (n `div` 2) (realGaussian var)
+
+
+
+
+
+
+-- Taken from monad-loops-0.4.3
+
+-- | Execute an action repeatedly until its result fails to satisfy a predicate,
+-- and return that result (discarding all others).
+iterateWhile :: (Monad m) => (a -> Bool) -> m a -> m a
+iterateWhile p x = x >>= iterateUntilM (not . p) (const x)
+
+-- | Analogue of @('Prelude.until')@
+-- Yields the result of applying f until p holds.
+iterateUntilM :: (Monad m) => (a -> Bool) -> (a -> m a) -> a -> m a
+iterateUntilM p f v 
+    | p v       = return v
+    | otherwise = f v >>= iterateUntilM p f
+
+{-
+-- | Returns a Gaussian-distributed sample over 'pZ' with given
+-- (scaled) variance parameter @v=var/(2*pi)@ and center, using
+-- rejection sampling
+
+gaussRound :: (RealTranscendental v, Random v,
+               RealRing c, ToRational c,
+               Ring i, ToInteger i, Random i, MonadRandom m)
+               => v -> c -> m i
+gaussRound svar c =
+    let dev = ceiling $ 6 * sqrt svar -- 6 gives stat dist < 2^-163
+        lower = floor c - dev
+        upper = ceiling c + dev
+        sampler = do
+           z <- getRandomR (lower, upper)
+           u <- getRandomR (zero, one)
+           let dist = fromIntegral z - realToField c
+           let prob = exp (-pi * (dist*dist / svar))
+           if u <= prob then return z else sampler
+    in sampler
+-}
diff --git a/src/Crypto/Lol/LatticePrelude.hs b/src/Crypto/Lol/LatticePrelude.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/LatticePrelude.hs
@@ -0,0 +1,244 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts,
+             FlexibleInstances, FunctionalDependencies,
+             GeneralizedNewtypeDeriving, MultiParamTypeClasses,
+             NoImplicitPrelude, PolyKinds, RankNTypes, RebindableSyntax,
+             ScopedTypeVariables, StandaloneDeriving, TemplateHaskell,
+             TypeFamilies, TypeOperators, UndecidableInstances #-}
+
+-- | A substitute for the Prelude that is more suitable for Lol.  This
+-- module exports most of the Numeric Prelude and other frequently
+-- used modules, plus some low-level classes, missing instances, and
+-- assorted utility functions.
+
+module Crypto.Lol.LatticePrelude
+( 
+-- * Classes
+  Enumerable(..)
+, Mod(..)
+, Reduce(..), Lift, Lift'(..), Rescale(..), Encode(..), msdToLSD
+-- * Numeric
+, module Crypto.Lol.Types.Numeric
+-- * Complex
+, module Crypto.Lol.Types.Complex
+-- * Factored
+, module Crypto.Lol.Factored
+-- * Miscellaneous
+, rescaleMod, roundCoset
+, pureT, peelT, pasteT, withWitness, withWitnessT
+, module Data.Functor.Trans.Tagged
+, module Data.Proxy
+) where
+
+import Crypto.Lol.Factored
+import Crypto.Lol.Types.Complex
+import Crypto.Lol.Types.Numeric
+
+import Algebra.Field          as Field (C)
+import Algebra.IntegralDomain as IntegralDomain (C)
+import Algebra.Ring           as Ring (C)
+
+import Control.Applicative
+import Control.Arrow
+import Control.DeepSeq
+import Control.Monad.Identity
+import Control.Monad.Random
+import Data.Coerce
+import Data.Default
+import Data.Functor.Trans.Tagged
+import Data.Maybe
+import Data.Proxy
+import Data.Singletons
+
+-- for Unbox instance of Maybe a
+import qualified Data.Vector.Unboxed          as U
+import           Data.Vector.Unboxed.Deriving
+
+instance NFData (Proxy (a :: k)) where rnf Proxy = ()
+
+deriving instance NFData (m a) => NFData (TaggedT s m a)
+deriving instance (MonadRandom m) => MonadRandom (TaggedT (tag :: k) m)
+
+derivingUnbox "Maybe"
+  [t| forall a . (Default a, U.Unbox a) => Maybe a -> (Bool, a) |]
+  [| maybe (False, def) (\ x -> (True, x)) |]
+  [| \ (b, x) -> if b then Just x else Nothing |]
+
+instance Default Bool where def = False
+
+-- | Poor man's 'Enum'.
+class Enumerable a where
+  values :: [a]
+
+-- | Represents a quotient group modulo some integer.
+class (ToInteger (ModRep a), Additive a) => Mod a where
+  type ModRep a
+  modulus :: Tagged a (ModRep a)
+
+-- | Represents that @b@ is a quotient group of @a@.
+class (Additive a, Additive b) => Reduce a b where
+  reduce :: a -> b
+
+-- | Represents that @b@ can be lifted to a "short" @a@ congruent to @b@.
+type Lift b a = (Lift' b, LiftOf b ~ a)
+
+-- | Fun-dep version of Lift.
+class (Reduce (LiftOf b) b) => Lift' b where
+  type LiftOf b
+  lift :: b -> LiftOf b
+
+-- | Represents that @a@ can be rescaled to @b@, as an "approximate"
+-- additive homomorphism.
+class (Additive a, Additive b) => Rescale a b where
+  rescale :: a -> b
+
+-- | Represents that the target ring can "noisily encode" values from
+-- the source ring, in either "most significant digit" (MSD) or "least
+-- significant digit" (LSD) encodings, and provides conversion factors
+-- between the two types of encodings.
+
+class (Field src, Field tgt) => Encode src tgt where
+    -- | The factor that converts an element from LSD to MSD encoding
+    -- in the target field, with associated scale factor to apply to
+    -- correct the resulting encoded value.
+    lsdToMSD :: (src, tgt)
+
+-- | Inverted entries of 'lsdToMSD'.
+msdToLSD :: (Encode src tgt) => (src, tgt)
+msdToLSD = (recip *** recip) lsdToMSD
+
+-- | A default implementation of rescaling for 'Mod' types.
+rescaleMod :: forall a b .
+              (Mod a, Mod b, (ModRep a) ~ (ModRep b),
+               Lift a (ModRep b), Ring b)
+              => a -> b
+rescaleMod =
+    let qval = proxy modulus (Proxy :: Proxy a)
+        q'val = proxy modulus (Proxy :: Proxy b)
+    in \x -> let (quot',_) = divModCent (q'val * lift x) qval
+             in fromIntegral quot'
+
+-- | Deterministically round to a nearby value in the desired coset
+roundCoset :: forall zp z r .
+              (Mod zp, z ~ ModRep zp, Lift zp z, RealField r) => zp -> r -> z
+roundCoset = let pval = proxy modulus (Proxy::Proxy zp)
+             in \ zp x -> let rep = lift zp
+                          in rep + roundMult pval (x - fromIntegral rep)
+
+---------- Instances for product groups/rings ----------
+
+instance (Mod a, Mod b, Lift' a, Lift' b, Reduce Integer (a,b),
+          ToInteger (LiftOf a), ToInteger (LiftOf b))
+         => Lift' (a,b) where
+
+  type LiftOf (a,b) = Integer
+
+  lift (a,b) =
+    let moda = toInteger $ proxy modulus (Proxy::Proxy a)
+        modb = toInteger $ proxy modulus (Proxy::Proxy b)
+        q = moda * modb
+        ainv = fromMaybe (error "Lift' (a,b): moduli not coprime") $ moda `modinv` modb
+        lifta = toInteger $ lift a
+        liftb = toInteger $ lift b
+        -- put in [-q/2, q/2)
+        (_,r) = (moda * (liftb - lifta) * ainv + lifta) `divModCent` q
+    in r
+
+
+-- NP should define Ring and Field instances for pairs, but doesn't.
+-- So we do it here.
+instance (Ring r1, Ring r2) => Ring.C (r1, r2) where
+
+  (x1, x2) * (y1, y2) = (x1*y1, x2*y2)
+  one = (one,one)
+  fromInteger x = (fromInteger x, fromInteger x)
+
+instance (Field f1, Field f2) => Field.C (f1, f2) where
+  (x1, x2) / (y1, y2) = (x1 / y1, x2 / y2)
+  recip = recip *** recip
+
+instance (IntegralDomain a, IntegralDomain b) => IntegralDomain.C (a,b) where
+  (a1,b1) `divMod` (a2,b2) =
+    let (da,ra) = (a1 `divMod` a2)
+        (db,rb) = (b1 `divMod` b2)
+    in ((da,db), (ra,rb))
+
+instance (Mod a, Mod b) => Mod (a,b) where
+  type ModRep (a,b) = Integer
+
+  modulus = tag $ fromIntegral (proxy modulus (Proxy::Proxy a)) *
+            fromIntegral (proxy modulus (Proxy::Proxy b))
+
+instance (Reduce a b1, Reduce a b2) => Reduce a (b1, b2) where
+  reduce x = (reduce x, reduce x)
+
+-- instances of Rescale for a product
+instance (Mod a, Field b, Lift a (ModRep a), Reduce (LiftOf a) b)
+         => Rescale (a,b) b where
+  rescale = let q1val = proxy modulus (Proxy::Proxy a)
+                q1inv = recip $ reduce q1val
+            in \(x1,x2) -> q1inv * (x2 - reduce (lift x1))
+
+instance (Mod b, Field a, Lift b (ModRep b), Reduce (LiftOf b) a)
+         => Rescale (a,b) a where
+  rescale = let q2val = proxy modulus (Proxy::Proxy b)
+                q2inv = recip $ reduce q2val
+            in \(x1,x2) -> q2inv * (x1 - reduce (lift x2))
+
+-- some multi-step scaledowns; could do this forever
+instance (Rescale (a,(b,c)) (b,c), Rescale (b,c) c)
+         => Rescale (a,(b,c)) c where
+  rescale = (rescale :: (b,c) -> c) . rescale
+
+instance (Rescale ((a,b),c) (a,b), Rescale (a,b) a)
+         => Rescale ((a,b),c) a where
+  rescale = (rescale :: (a,b) -> a) . rescale
+
+-- scaling up to a product
+instance (Ring a, Mod b, Reduce (ModRep b) a) => Rescale a (a,b) where
+  -- multiply by q2
+  rescale = let q2val = reduce $ proxy modulus (Proxy::Proxy b)
+            in \x -> (q2val * x, zero)
+
+instance (Ring b, Mod a, Reduce (ModRep a) b) => Rescale b (a,b) where
+  -- multiply by q1
+  rescale = let q1val = reduce $ proxy modulus (Proxy::Proxy a)
+            in \x -> (zero, q1val * x)
+
+-- Instance of 'Encode' for product ring.
+instance (Encode s t1, Encode s t2, Field (t1, t2)) => Encode s (t1, t2) where
+
+  lsdToMSD = let (s1, t1conv) = lsdToMSD
+                 (s2, t2conv) = lsdToMSD
+             in (negate s1 * s2, (t1conv,t2conv))
+
+-- Random could have defined this instance, but didn't, so we do it
+-- here.
+instance (Random a, Random b) => Random (a,b) where
+  random g = let (a,g') = random g
+                 (b, g'') = random g'
+             in ((a,b), g'')
+
+  randomR ((loa,lob), (hia,hib)) g = let (a,g') = randomR (loa,hia) g
+                                         (b,g'') = randomR (lob,hib) g'
+                                     in ((a,b),g'')
+
+-- | Apply any applicative to a Tagged value.
+pureT :: Applicative f => TaggedT t Identity a -> TaggedT t f a
+pureT = mapTaggedT (pure . runIdentity)
+
+-- | Expose the monad of a tagged value.
+peelT :: Tagged t (f a) -> TaggedT t f a
+peelT = coerce
+
+-- | Hide the monad of a tagged value.
+pasteT :: TaggedT t f a -> Tagged t (f a)
+pasteT = coerce
+
+-- | Use a singleton as a witness to extract a value from a tagged value.
+withWitness :: forall n r . (SingI n => Tagged n r) -> Sing n -> r
+withWitness t wit = withSingI wit $ proxy t (Proxy::Proxy n)
+
+-- | Monadic version of 'withWitness'.
+withWitnessT :: forall n mon r . (Monad mon) =>
+                (SingI n => TaggedT n mon r) -> Sing n -> mon r
+withWitnessT t wit = withSingI wit $ proxyT t (Proxy::Proxy n)
diff --git a/src/Crypto/Lol/Reflects.hs b/src/Crypto/Lol/Reflects.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Reflects.hs
@@ -0,0 +1,39 @@
+{-# LANGUAGE DataKinds, FlexibleContexts, FlexibleInstances,
+             KindSignatures, MultiParamTypeClasses, PolyKinds,
+             ScopedTypeVariables, UndecidableInstances #-}
+
+-- | Generic interface for reflecting types to values.
+
+module Crypto.Lol.Reflects
+( Reflects(..)
+) where
+
+import Crypto.Lol.Factored
+
+import Data.Functor.Trans.Tagged
+import Data.Proxy
+import Data.Reflection
+import GHC.TypeLits              as TL
+
+-- | Reflection without fundep, and with tagged value. Intended only
+-- for low-level code; build specialized wrappers around it for
+-- specific functionality.
+
+class Reflects a i where
+  -- | Reflect the value assiated with the type @a@.
+  value :: Tagged a i
+
+instance (KnownNat a, Integral i) => Reflects (a :: TL.Nat) i where
+  value = return $ fromIntegral $ natVal (Proxy::Proxy a)
+
+instance (NatC a, Integral i) => Reflects a i where
+  value = fmap fromIntegral valueNatC
+
+instance (PPow pp, Integral i) => Reflects pp i where
+  value = fmap fromIntegral valuePPow
+
+instance (Fact m, Integral i) => Reflects m i where
+  value = fmap fromIntegral valueFact
+
+instance {-# OVERLAPS #-} (Reifies rei a) => Reflects (rei :: *) a where
+  value = tag $ reflect (Proxy::Proxy rei)
diff --git a/src/Crypto/Lol/Types/Complex.hs b/src/Crypto/Lol/Types/Complex.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/Complex.hs
@@ -0,0 +1,89 @@
+{-# LANGUAGE DataKinds, DeriveDataTypeable, FlexibleContexts,
+             FlexibleInstances, GeneralizedNewtypeDeriving,
+             MultiParamTypeClasses, NoImplicitPrelude, RebindableSyntax,
+             ScopedTypeVariables, StandaloneDeriving, TemplateHaskell,
+             TypeFamilies, UndecidableInstances #-}
+
+-- | Data type, functions, and instances for complex numbers.
+
+module Crypto.Lol.Types.Complex (
+  Complex
+, roundComplex
+, cis, real, imag, fromReal
+) where
+
+import           Algebra.Additive       as Additive (C)
+import           Algebra.Field          as Field (C)
+import           Algebra.IntegralDomain as IntegralDomain
+import           Algebra.Ring           as Ring (C)
+import           Algebra.ZeroTestable   as ZeroTestable (C)
+import qualified Number.Complex         as C hiding (exp, signum)
+
+import Crypto.Lol.Types.Numeric as LP
+
+import Control.DeepSeq
+import Data.Array.Repa.Eval         as R
+import Data.Vector.Storable         (Storable)
+import Data.Vector.Unboxed          (Unbox)
+import Data.Vector.Unboxed.Deriving
+import System.Random
+import Test.QuickCheck
+
+-- | Newtype wrapper (with slightly different instances) for
+-- <https://hackage.haskell.org/package/numeric-prelude-0.4.2/docs/Number-Complex.html numeric-prelude Complex>.
+newtype Complex a = Complex (C.T a) deriving (Additive.C, Ring.C, ZeroTestable.C, Field.C, Storable, Eq, Show, Arbitrary)
+
+derivingUnbox "Complex"
+  [t| forall a . (Unbox a) => Complex a -> (a, a) |]
+  [| \ (Complex x) -> (C.real x, C.imag x) |]
+  [| \ (r, i) -> Complex $ r C.+: i |]
+
+-- a custom IntegralDomain instance, replacing the one provided by NP.
+-- it always returns 0 as the remainder of a division.  If we were to
+-- use the NP instance, sometimes precision issues yield nonzero
+-- remainders, which makes, e.g., 'divGPow' think that division has
+-- failed, when it has not.  This in turn causes 'divGCRT' to yield
+-- Nothing, among other problems.
+instance (Field a) => IntegralDomain.C (Complex a) where
+  (Complex a) `divMod` (Complex b) = (Complex $ a / b, LP.zero)
+
+-- we can't use Generics for NFData because NP doesn't export the
+-- (deep) constructor for Complex.T
+instance (NFData a) => NFData (Complex a) where
+  rnf (Complex x) = let r = C.real x
+                        i = C.imag x
+                    in rnf r `seq` rnf i `seq` ()
+
+instance (Random a) => Random (Complex a) where
+    random g = let (a,g') = random g
+                   (b,g'') = random g'
+               in (Complex $ a C.+: b, g'')
+
+    randomR = error "randomR not defined for (Complex t)"
+
+instance (R.Elt a) => R.Elt (Complex a) where
+    touch (Complex c) = do
+        touch $ C.real c
+        touch $ C.imag c
+    zero = Complex $ R.zero C.+: R.zero
+    one = Complex $ R.one C.+: R.zero
+
+-- | Rounds the real and imaginary components to the nearest integer.
+roundComplex :: (RealRing a, ToInteger b) => Complex a -> (b,b)
+roundComplex (Complex x) = (round $ C.real x, round $ C.imag x)
+
+-- | 'cis' @t@ is a complex value with magnitude 1 and phase t (modulo @2*Pi@).
+cis :: Transcendental a => a -> Complex a
+cis = Complex . C.cis
+
+-- | Real component of a complex number.
+real :: Complex a -> a
+real (Complex a) = C.real a
+
+-- | Imaginary component of a complex number.
+imag :: Complex a -> a
+imag (Complex a) = C.imag a
+
+-- | Embeds a scalar as the real component of a complex number.
+fromReal :: Additive a => a -> Complex a
+fromReal = Complex . C.fromReal
diff --git a/src/Crypto/Lol/Types/FiniteField.hs b/src/Crypto/Lol/Types/FiniteField.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/FiniteField.hs
@@ -0,0 +1,122 @@
+{-# LANGUAGE ConstraintKinds, FlexibleContexts,
+             GeneralizedNewtypeDeriving, 
+             NoImplicitPrelude, PolyKinds,
+             RebindableSyntax, RoleAnnotations, ScopedTypeVariables #-}
+
+-- CJP: need PolyKinds to allow deg to have non-* kind
+
+-- | Basic (unoptimized) finite field arithmetic.
+
+module Crypto.Lol.Types.FiniteField
+( PrimeField, CharOf, GF   -- export type but not constructor
+, trace
+, size
+) where
+
+import           Crypto.Lol.CRTrans
+import           Crypto.Lol.Factored
+import           Crypto.Lol.LatticePrelude
+import           Crypto.Lol.Reflects
+import           Crypto.Lol.Types.PrimeField hiding ((^))
+import qualified Crypto.Lol.Types.PrimeField as PF
+
+import Algebra.Additive     as Additive (C)
+import Algebra.Field        as Field (C)
+import Algebra.Ring         as Ring (C)
+import Algebra.ZeroTestable as ZeroTestable (C)
+import MathObj.Polynomial
+
+import Math.NumberTheory.Primes.Factorisation
+
+import           Control.Applicative
+import           Control.DeepSeq
+import           Control.Monad
+import qualified Data.Vector              as V
+
+--import qualified Debug.Trace as DT
+
+-- | A finite field of given degree over @F_p@.
+newtype GF fp deg = GF (Polynomial fp)
+                  deriving (Eq, Show, Additive.C, ZeroTestable.C, NFData)
+
+-- the second argument, though phantom, affects representation
+type role GF representational representational
+
+type GFCtx fp deg = (PrimeField fp, Reflects deg Int)
+
+instance (GFCtx fp deg) => Enumerable (GF fp deg) where
+  values = GF <$> fromCoeffs <$>
+           -- d-fold cartesian product of Fp values
+           replicateM (proxy value (Proxy::Proxy deg)) values
+
+instance (GFCtx fp deg) => Ring.C (GF fp deg) where
+
+  one = GF one
+
+  (*) = let poly = proxy irreduciblePoly (Proxy :: Proxy deg)
+        in \(GF f) (GF g) -> GF $ (f*g) `mod` poly
+
+  fromInteger = GF . fromInteger
+
+instance (GFCtx fp deg) => Field.C (GF fp deg) where
+
+  recip = let g = proxy irreduciblePoly (Proxy :: Proxy deg)
+          in \(GF f) -> let (_,(a,_)) = extendedGCD f g
+                           in GF a
+
+instance (GFCtx fp deg) => CRTrans (GF fp deg) where
+
+  crtInfo m = (,) <$> omegaPow <*> scalarInv
+    where
+      omegaPow =
+        let size' = proxy size (Proxy :: Proxy (GF fp deg))
+            (q,r) = (size'-1) `quotRem` m
+            gen = head $ filter isPrimitive values
+            omega = gen^q
+            omegaPows = V.iterateN m (*omega) one
+        in if r == 0
+           then Just $ (omegaPows V.!) . (`mod` m)
+           else Nothing
+      scalarInv = Just $ recip $ fromIntegral $ valueHat m
+
+sizePP :: forall fp deg . (GFCtx fp deg) => Tagged (GF fp deg) PP
+sizePP = tag (proxy value (Proxy::Proxy (CharOf fp)),
+              proxy value (Proxy::Proxy deg))
+
+-- | The order of the field: @size (GF fp deg) = p^deg@
+size :: (GFCtx fp deg) => Tagged (GF fp deg) Int
+size = uncurry (^) <$> sizePP
+
+isPrimitive :: forall fp deg . (GFCtx fp deg) => GF fp deg -> Bool
+isPrimitive = let q = proxy size (Proxy :: Proxy (GF fp deg))
+                  ps = map (fromIntegral . fst) $ factorise $
+                       fromIntegral $ q-1
+                  exps = map ((q-1) `div`) ps
+              in \g -> not (isZero g) && all (\e -> g^e /= 1) exps
+
+dotp :: (Ring a) => [a] -> [a] -> a
+dotp a b = sum $ zipWith (*) a b
+
+-- | Trace into the prime subfield.
+trace :: forall fp deg . (GFCtx fp deg) => GF fp deg -> fp
+trace = let ts = proxy powTraces (Proxy::Proxy (GF fp deg))
+        in \(GF f) -> dotp ts (coeffs f)
+
+-- | Traces of the power basis elements 1, x, x^2, ..., x^(deg-1).
+powTraces :: forall fp deg . (GFCtx fp deg) => Tagged (GF fp deg) [fp]
+powTraces = 
+  --DT.trace ("FiniteField.powTraces: p = " ++ 
+  --          show (proxy value (Proxy::Proxy (CharOf fp)) :: Int) ++
+  --          ", d = " ++ show (proxy value (Proxy::Proxy deg) :: Int)) $
+  let d = proxy value (Proxy :: Proxy deg)
+  in tag $ map trace' $ take d $
+     iterate (* (GF (X PF.^ 1))) (one :: GF fp deg)
+
+-- helper that computes trace via brute force: sum frobenius
+-- automorphisms
+trace' :: (GFCtx fp deg) => GF fp deg -> fp
+trace' e = let (p,d) = witness sizePP e
+               (GF t) = sum $ take d $ iterate (^p) e
+               -- t is a constant polynomial
+           in head $ coeffs t
+
diff --git a/src/Crypto/Lol/Types/IZipVector.hs b/src/Crypto/Lol/Types/IZipVector.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/IZipVector.hs
@@ -0,0 +1,58 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             DeriveTraversable, FlexibleContexts,
+             GeneralizedNewtypeDeriving, KindSignatures,
+             MultiParamTypeClasses, RoleAnnotations, ScopedTypeVariables,
+             TypeFamilies, UndecidableInstances #-}
+
+-- | Provides applicative-like functions for indexed vectors
+
+module Crypto.Lol.Types.IZipVector
+( IZipVector, iZipVector, unIZipVector
+) where
+
+import Crypto.Lol.Factored
+
+import Algebra.ZeroTestable as ZeroTestable
+
+import Control.DeepSeq
+import Data.Data
+import Data.Functor.Trans.Tagged
+import Data.Vector               as V
+
+-- | Indexed Zip Vector: a wrapper around a (boxed) 'Vector' that has
+-- zip-py 'Applicative' behavior, analogous to
+-- 'Control.Applicative.ZipList' for lists.  The index @m@ enforces
+-- proper lengths (and is necessary to implement 'pure').
+
+newtype IZipVector (m :: Factored) a =
+  IZipVector { -- | Deconstructor for IZipVector
+               unIZipVector :: Vector a}
+  -- not deriving Read, Monoid, Alternative, Monad[Plus], IsList
+  -- because of different semantics and/or length restriction
+  deriving (Show, Eq, Data, NFData, Typeable, Functor,
+            Foldable, Traversable, ZeroTestable.C)
+
+-- the first argument, though phantom, affects representation
+type role IZipVector representational representational
+
+-- | Smart constructor that checks whether length of input is right
+-- (should be totient of @m@).
+iZipVector :: forall m a . (Fact m) => Vector a -> Maybe (IZipVector m a)
+iZipVector = let n = proxy totientFact (Proxy::Proxy m)
+            in \vec -> if n == V.length vec
+                       then Just $ IZipVector vec
+                       else Nothing
+
+-- don't export
+repl :: forall m a . (Fact m) => a -> IZipVector m a
+repl = let n = proxy totientFact (Proxy::Proxy m)
+       in IZipVector . V.replicate n
+
+-- Zip-py 'Applicative' instance.
+instance (Fact m) => Applicative (IZipVector m) where
+  pure = repl
+  (IZipVector f) <*> (IZipVector a) = IZipVector $ V.zipWith ($) f a
+
+-- no ZeroTestable instance for Vectors, so define here
+instance (ZeroTestable.C a) => ZeroTestable.C (Vector a) where
+  isZero = V.all isZero
diff --git a/src/Crypto/Lol/Types/IrreducibleChar2.hs b/src/Crypto/Lol/Types/IrreducibleChar2.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/IrreducibleChar2.hs
@@ -0,0 +1,60 @@
+{-# LANGUAGE NoImplicitPrelude, RebindableSyntax, ScopedTypeVariables,
+             FlexibleInstances, TypeFamilies, UndecidableInstances, PolyKinds #-}
+
+-- | (Orphan) instance of 'IrreduciblePoly' for characteristic 2 fields.
+
+module Crypto.Lol.Types.IrreducibleChar2 () where
+
+import Crypto.Lol.LatticePrelude hiding ((^))
+import Crypto.Lol.Reflects
+import Crypto.Lol.Types.PrimeField
+
+import Data.Type.Natural (N2)
+
+-- conway
+--generate in Python (or choose any irreducible polynomial)
+-- to generate with Sage, start sage and type:
+--      conway_polynomial(p,e)
+-- then copy and paste
+instance (CharOf a ~ N2, Ring a) => IrreduciblePoly a where
+  irreduciblePoly = do
+    pn <- taggedProxy
+    let n = proxy value pn :: Int
+    return $ case n of
+      1 -> X^1 + 1
+      2 -> X^2 + X^1 + 1
+      3 -> X^3 + X^1 + 1
+      4 -> X^4 + X^1 + 1
+      5 -> X^5 + X^2 + 1
+      6 -> X^6 + X^4 + X^3 + X^1 + 1
+      7 -> X^7 + X^1 + 1
+      8 -> X^8 + X^4 + X^3 + X^2 + 1
+      9 -> X^9 + X^4 + 1
+      10 -> X^10 + X^6 + X^5 + X^3 + X^2 + X^1 + 1
+      11 -> X^11 + X^2 + 1
+      12 -> X^12 + X^7 + X^6 + X^5 + X^3 + X^1 + 1
+      13 -> X^13 + X^4 + X^3 + X^1 + 1
+      14 -> X^14 + X^7 + X^5 + X^3 + 1
+      15 -> X^15 + X^5 + X^4 + X^2 + 1
+      16 -> X^16 + X^5 + X^3 + X^2 + 1
+      17 -> X^17 + X^3 + 1
+      18 -> X^18 + X^12 + X^10 + X^1 + 1
+      19 -> X^19 + X^5 + X^2 + X^1 + 1
+      20 -> X^20 + X^10 + X^9 + X^7 + X^6 + X^5 + X^4 + X^1 + 1
+      21 -> X^21 + X^6 + X^5 + X^2 + 1
+      22 -> X^22 + X^12 + X^11 + X^10 + X^9 + X^8 + X^6 + X^5 + 1
+      23 -> X^23 + X^5 + 1
+      24 -> X^24 + X^16 + X^15 + X^14 + X^13 + X^10 + X^9 + X^7 + X^5 + X^3 + 1
+      25 -> X^25 + X^8 + X^6 + X^2 + 1
+      26 -> X^26 + X^14 + X^10 + X^8 + X^7 + X^6 + X^4 + X^1 + 1
+      27 -> X^27 + X^12 + X^10 + X^9 + X^7 + X^5 + X^3 + X^2 + 1
+      28 -> X^28 + X^13 + X^7 + X^6 + X^5 + X^2 + 1
+      29 -> X^29 + X^2 + 1
+      30 -> X^30 + X^17 + X^16 + X^13 + X^11 + X^7 + X^5 + X^3 + X^2 + X^1 + 1
+      31 -> X^31 + X^3 + 1
+      32 -> X^32 + X^15 + X^9 + X^7 + X^4 + X^3 + 1 
+      otherwise -> 
+        error $ "The ConwayPoly instance for N2 included with the library (and exported by Crypto.Lol) only contains " ++
+                "irreducible polynomials for characteristic-2 fields up to GF(2^32). You need a polynomial " ++ 
+                "for GF(2^" ++ (show n) ++ "). Define your own instance of ConwayPoly and do " ++
+                "not import Crypto.Lol."
diff --git a/src/Crypto/Lol/Types/Numeric.hs b/src/Crypto/Lol/Types/Numeric.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/Numeric.hs
@@ -0,0 +1,213 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleInstances, GADTs,
+             MultiParamTypeClasses, NoImplicitPrelude, RebindableSyntax,
+             ScopedTypeVariables, TypeOperators #-}
+
+-- we have some orphan instances here for instances of
+-- package classes with Prelude data types
+{-# OPTIONS_GHC -fno-warn-orphans #-}
+
+-- | This module imports NumericPrelude and defines constraint
+-- synonyms for NumericPrelude classes to help with code readability,
+-- and defines saner versions of some NumericPrelude functions
+
+module Crypto.Lol.Types.Numeric
+( module Crypto.Lol.Types.Numeric -- everything we define here
+, module NumericPrelude         -- re-export
+, Int64                         -- commonly used
+) where
+
+import Control.DeepSeq
+import Control.Monad.Random
+
+import           Algebra.IntegralDomain (divUp)
+-- NumericPrelude has silly types for these functions
+import           NumericPrelude         hiding (abs, max, min, (^))
+import qualified NumericPrelude.Numeric (abs)
+import qualified Prelude                (max, min)
+
+import qualified Algebra.Absolute             (C)
+import qualified Algebra.Additive             (C)
+import qualified Algebra.Algebraic            (C)
+import qualified Algebra.Field                (C)
+import qualified Algebra.IntegralDomain       (C)
+import qualified Algebra.Module               (C)
+import qualified Algebra.PrincipalIdealDomain (C)
+import qualified Algebra.RealField            (C)
+import qualified Algebra.RealIntegral         (C)
+import qualified Algebra.RealRing             (C)
+import qualified Algebra.RealTranscendental   (C)
+import qualified Algebra.Ring                 (C)
+import qualified Algebra.ToInteger            (C)
+import qualified Algebra.ToRational           (C, realToField)
+import qualified Algebra.Transcendental       (C)
+import qualified Algebra.ZeroTestable         (C)
+import           MathObj.Polynomial
+
+import Data.Int (Int64)
+
+-- | The Prelude definition of 'max'.
+max :: Ord a => a -> a -> a
+max = Prelude.max
+
+-- | The Prelude definition of 'min'.
+min :: Ord a => a -> a -> a
+min = Prelude.min
+
+-- | The sane definition of 'abs' from
+-- 'NumericPrelude.Numeric'
+-- rather than the default from 'NumericPrelude'.
+abs :: Absolute a => a -> a
+abs = NumericPrelude.Numeric.abs
+
+-- | The hidden NP function from 'Algebra.ToRational'.
+realToField :: (Field b, ToRational a) => a -> b
+realToField = Algebra.ToRational.realToField
+
+-- use this if you need:
+{- isZero -}
+-- | Sane synonym for 'Algebra.ZeroTestable.C'.
+type ZeroTestable a = (Algebra.ZeroTestable.C a)
+
+{- - + negate -}
+-- | Sane synonym for 'Algebra.Additive.C'.
+type Additive a = (Algebra.Additive.C a)
+
+{- Additive, plus: * fromIntegral -}
+-- | Sane synonym for 'Algebra.Ring.C'.
+type Ring a = (Algebra.Ring.C a)
+
+{- Ring and Additive, plus: *> -}
+-- | Sane synonym for 'Algebra.Module.C'.
+type Module a v = (Algebra.Module.C a v)
+
+{- Ring, plus: div, mod, divmod -}
+-- | Sane synonym for 'Algebra.IntegralDomain.C'.
+type IntegralDomain a = (Algebra.IntegralDomain.C a)
+
+{- Ring, plus: abs signum toRational' -}
+-- | Sane synonym for 'Algebra.ToRational.C'.
+type ToRational a = (Algebra.ToRational.C a)
+
+{- Ring, plus: / recip fromRational -}
+-- | Sane synonym for 'Algebra.Field.C'.
+type Field a = (Algebra.Field.C a)
+
+{- Ring, plus: abs and rounding functions -}
+-- | Sane synonym for 'Algebra.RealRing.C'.
+type RealRing a = (Algebra.RealRing.C a)
+
+{- Field, plus: abs signum round floor ceiling -}
+-- | Sane synonym for 'Algebra.RealField.C'.
+type RealField a = (Algebra.RealField.C a)
+
+{- Field, plus: sqrt root ^/ -}
+-- | Sane synonym for 'Algebra.Algebraic.C'.
+type Algebraic a = (Algebra.Algebraic.C a)
+
+{- Algebraic, plus: pi exp log sin atan -}
+-- | Sane synonym for 'Algebra.Transcendental.C'.
+type Transcendental a = (Algebra.Transcendental.C a)
+
+{- Transcendental and RealField, plus atan2 -}
+-- | Sane synonym for 'Algebra.RealTranscendental.C'.
+type RealTranscendental a = (Algebra.RealTranscendental.C a)
+
+{- Transcendental, plus: == <= >= < > -}
+-- | Convenient synonym for @(Ord a, Transcendental a)@
+type OrdFloat a = (Ord a, Transcendental a)
+
+{- ToRational and Ring, plus: toInteger div mod divmod quot rem quotrem -}
+-- | Sane synonym for 'Algebra.ToInteger.C'.
+type ToInteger a = (Algebra.ToInteger.C a)
+
+-- | Sane synonym for 'Algebra.Absolute.C'.
+type Absolute a = (Algebra.Absolute.C a)
+
+-- | Sane synonym for 'Algebra.RealIntegral.C'.
+type RealIntegral a = (Algebra.RealIntegral.C a)
+
+-- | Sane synonym for 'Algebra.PrincipalIdealDomain.C'.
+type PID a = (Algebra.PrincipalIdealDomain.C a)
+
+-- | Sane synonym for 'MathObj.Polynomial.T'.
+type Polynomial a = MathObj.Polynomial.T a
+
+-- | IntegralDomain instance for Double
+instance Algebra.IntegralDomain.C Double where
+    _ `div` 0 = error "cannot divide Double by 0\n"
+    a `div` b = a / b
+    _ `mod` _ = 0
+
+-- NFData instance for Polynomial, missing from NP
+instance (NFData r) => NFData (Polynomial r) where
+  rnf = rnf . coeffs
+
+-- | Our custom exponentiation, overriding NP's version that
+-- requires 'Integer' exponent.
+-- Copied from http://hackage.haskell.org/package/base-4.7.0.0/docs/src/GHC-Real.html#%5E
+{-# SPECIALISE [1] (^) ::
+        Integer -> Integer -> Integer,
+        Integer -> Int -> Integer,
+        Int -> Int -> Int,
+        Int64 -> Int64 -> Int64
+  #-}
+(^) :: forall a i . (Ring a, ToInteger i) => a -> i -> a
+x0 ^ y0 | y0 < 0    = error "Negative exponent"
+        | y0 == 0   = 1
+        | otherwise = f x0 y0
+    where -- f : x0 ^ y0 = x ^ y
+          f :: a -> i -> a -- a polymorphic local binding needs a sig
+          f x y | even y    = f (x * x) (y `quot` 2)
+                | y == 1    = x
+                | otherwise = g (x * x) ((y - 1) `quot` 2) x
+          -- g : x0 ^ y0 = (x ^ y) * z
+          g :: a -> i -> a -> a
+          g x y z | even y = g (x * x) (y `quot` 2) z
+                  | y == 1 = x * z
+                  | otherwise = g (x * x) ((y - 1) `quot` 2) (x * z)
+
+-- | Inverse of @a@ modulo @q@, in range @0..q-1@.  (Argument order is
+-- infix-friendly.)
+modinv :: (PID i, Eq i) => i -> i -> Maybe i
+modinv a q = let (d, (_, inv)) = extendedGCD q a
+             in if d == one
+                then Just $ inv `mod` q
+                else Nothing
+
+-- | Decompose an element into a list of "centered" digits with respect
+-- to relative radices.
+decomp :: (IntegralDomain z, Ord z) => [z] -> z -> [z]
+decomp [] v = [v]
+decomp (b:bs) v = let (q,r) = v `divModCent` b
+                  in r : decomp bs q
+
+-- | Yield @ceil (log_b(x))@.
+logCeil :: (ToInteger i) => i -> i -> Int
+logCeil _ 1 = 0
+logCeil b x = 1 + logCeil b (x `divUp` b)
+
+-- | Deterministically round to the nearest multiple of @i@.
+roundMult :: (RealField r, ToInteger i) => i -> r -> i
+roundMult 1 r  = round r
+roundMult i r = let r' = r / fromIntegral i in i * round r'
+
+-- | Randomly round to the nearest larger or smaller multiple of @i@,
+-- where the round-off term has expectation zero.
+roundScalarCentered :: (RealField r, Random r, ToInteger i,
+                        MonadRandom mon)
+                      => i -> r -> mon i
+roundScalarCentered p x =
+  let x' = x / fromIntegral p
+      mod1 = x' - floor x'
+  in do prob <- getRandomR (zero, one)
+        return $ p * if prob < mod1
+                     then ceiling x'
+                     else floor x'
+
+-- | Variant of 'Algebra.IntegralDomain.divMod' in which the remainder
+-- is in the range @[-b\/2,b\/2)@.
+divModCent :: (IntegralDomain i, Ord i) => i -> i -> (i,i)
+divModCent a b = let (q,r) = a `divMod` b
+                 in if 2*r < b -- divMod returns non-neg remainder
+                    then (q,r)
+                    else (q+1,r-b)
diff --git a/src/Crypto/Lol/Types/PrimeField.hs b/src/Crypto/Lol/Types/PrimeField.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/PrimeField.hs
@@ -0,0 +1,36 @@
+{-# LANGUAGE NoImplicitPrelude, RebindableSyntax, PolyKinds, TypeFamilies, 
+             DataKinds, FlexibleContexts, ConstraintKinds #-}
+
+-- | Prime-order fields.
+
+module Crypto.Lol.Types.PrimeField where
+
+import Crypto.Lol.LatticePrelude as LP
+import Crypto.Lol.Reflects
+
+import MathObj.Polynomial
+
+-- | Constraint synonym for prime-order fields.
+type PrimeField fp = (Enumerable fp, Eq fp, ZeroTestable fp, Field fp,
+       IrreduciblePoly fp)
+
+-- | The characteristic of a field, represented as a type.
+type family CharOf (fp :: k) :: Nat
+
+-- | Represents prime-order fields over which we can get irreducible
+-- polynomials of desired degree.  (An instance of this class is
+-- defined in 'Crypto.Lol.Types.IrreducibleChar2' and exported from
+-- 'Crypto.Lol'.)
+class (Ring fp, Prime (CharOf fp)) => IrreduciblePoly fp where
+  irreduciblePoly :: (Reflects deg Int) => Tagged deg (Polynomial fp)
+
+-- | Convenience function for writing 'IrreduciblePoly' instances.
+taggedProxy :: Tagged s (Proxy s)
+taggedProxy = tag Proxy
+
+-- | Convenience data type for writing 'IrreduciblePoly' instances.
+data X = X
+
+-- | Convenience function for writing 'IrreduciblePoly' instances.
+(^) :: (Ring a) => X -> Int -> Polynomial a
+X ^ i | i >= 0 = fromCoeffs $ (replicate i 0) ++ [1]
diff --git a/src/Crypto/Lol/Types/ZPP.hs b/src/Crypto/Lol/Types/ZPP.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/ZPP.hs
@@ -0,0 +1,23 @@
+{-# LANGUAGE FlexibleContexts, TypeFamilies #-}
+
+-- | A class for integers mod a prime power.
+
+module Crypto.Lol.Types.ZPP
+( ZPP(..)
+) where
+
+import Crypto.Lol.LatticePrelude
+import Crypto.Lol.Types.FiniteField
+
+-- | Represents integers modulo a prime power.
+class (PrimeField (ZPOf zq), Ring zq, Ring (ZPOf zq)) => ZPP zq where
+
+  -- | An implementation of the integers modulo the prime base.
+  type ZPOf zq
+
+  -- | The prime and exponent of the modulus.
+  modulusZPP :: Tagged zq PP
+
+  -- | Lift from @Z_p@ to a representative.
+  liftZp :: ZPOf zq -> zq
+
diff --git a/src/Crypto/Lol/Types/ZmStar.hs b/src/Crypto/Lol/Types/ZmStar.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/ZmStar.hs
@@ -0,0 +1,69 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts,
+             NoImplicitPrelude, PolyKinds, RebindableSyntax,
+             ScopedTypeVariables, TypeFamilies, TypeOperators, 
+             UndecidableInstances #-}
+
+-- | A collection of helper functions for working with @Z_m^*@
+
+module Crypto.Lol.Types.ZmStar
+( order, partitionCosets
+) where
+
+import Crypto.Lol.Factored
+import Crypto.Lol.LatticePrelude as LP hiding (null)
+import Crypto.Lol.Reflects
+import Crypto.Lol.Types.ZqBasic
+
+import Data.List as L (foldl', transpose)
+import Data.Map  (Map, elems, empty, insertWith')
+import Data.Set  as S (Set, difference, findMin, fromList, map, null)
+
+
+-- | The multiplicative order of @p@ (the argument) modulo @m@.
+-- Requires @gcd(p,m)=1@.
+order :: forall m . (Reflects m Int) => Int -> Tagged m Int
+order p = tag $
+  let mval = proxy value (Proxy::Proxy m)
+  in if gcd p mval /= 1
+     then error "p and m not coprime"
+     else 1 + (length $ takeWhile (/= one) $
+               tail $ iterate (* (fromIntegral p)) (one :: ZqBasic m Int))
+
+-- given p, returns the cosets of Z_m^* / <p>
+cosets :: forall zm . (Mod zm, ModRep zm ~ Int, Ord zm, Ring zm)
+  => Int -> [Set zm]
+cosets p =
+  let mval = proxy modulus (Proxy::Proxy zm)
+  in if gcd p mval /= 1
+     then error "p and m not coprime"
+     else let zmstar = fromList $ LP.map fromIntegral $ filter ((==) 1 . gcd mval) [1..mval]
+              zp = fromIntegral p
+              -- generates the coset containing x
+              coset x = fromList $ x : takeWhile (/=x) (iterate (*zp) $ zp*x)
+              -- repeatedly removes a (new) coset from the remaining elements
+              genCosets s | null s = []
+              genCosets s = let c = coset (findMin s)
+                            in c : genCosets (difference s c)
+          in genCosets zmstar
+
+-- CJP: could tag this by '(p,m,m') for safety/memoization.
+
+-- | Given @p@, returns a partition of the cosets of @Z_{m\'}^* \/ \<p>@
+-- (specified by representatives), where the cosets in each component
+-- are in bijective correspondence with the cosets of @Z_m^* \/ \<p>@ under
+-- the natural (@mod m@) homomorphism.
+partitionCosets :: forall m m' . (m `Divides` m')
+  => Int -> Tagged '(m, m') [[Int]]
+partitionCosets p =
+  let m'cosets = cosets p
+      -- a map from cosets of Z_m^* / <p> to their preimages under the
+      -- natural homomorphism
+      partition =
+        L.foldl' (\cmap x -> insertWith' (++) (S.map (reduce . lift) x) [x] cmap)
+          (empty :: Map (Set (ZqBasic m Int)) [Set (ZqBasic m' Int)])
+          m'cosets
+      -- transpose the map to get a list of list of sets, where for each
+      -- inner list, there is exactly one m'-(co)set lying above each m-coset
+      part' = transpose $ elems partition
+     -- concat the inner sets to get a list of "CRT cosets" (indexed in Z_m'^*)
+  in return $ LP.map (LP.map (lift . findMin)) part'
diff --git a/src/Crypto/Lol/Types/ZqBasic.hs b/src/Crypto/Lol/Types/ZqBasic.hs
new file mode 100644
--- /dev/null
+++ b/src/Crypto/Lol/Types/ZqBasic.hs
@@ -0,0 +1,256 @@
+{-# LANGUAGE ConstraintKinds, DataKinds, DeriveDataTypeable,
+             FlexibleContexts, FlexibleInstances,
+             GeneralizedNewtypeDeriving, MultiParamTypeClasses,
+             NoImplicitPrelude, PolyKinds, RebindableSyntax,
+             RoleAnnotations, ScopedTypeVariables, 
+             StandaloneDeriving, TypeFamilies, UndecidableInstances #-}
+
+-- | An implementation of modular arithmetic, i.e., the ring Zq.
+
+module Crypto.Lol.Types.ZqBasic
+( ZqBasic -- export the type, but not the constructor (for safety)
+) where
+
+import Crypto.Lol.LatticePrelude as LP
+import Crypto.Lol.Reflects
+import Crypto.Lol.CRTrans
+import Crypto.Lol.Types.FiniteField
+import Crypto.Lol.Types.ZPP
+import Crypto.Lol.Gadget
+
+import Control.Applicative
+import Control.DeepSeq        (NFData)
+import Control.Monad          (liftM)
+import Data.Coerce
+import Data.Maybe
+import Data.Typeable
+import NumericPrelude.Numeric as NP (round)
+import System.Random
+import Test.QuickCheck
+
+-- for the Unbox instances
+import qualified Data.Vector.Generic         as V
+import qualified Data.Vector.Generic.Mutable as M
+import qualified Data.Vector.Unboxed         as U
+
+import Foreign.Storable
+
+-- for the Elt instance
+import qualified Data.Array.Repa.Eval as E
+
+import qualified Algebra.Additive       as Additive (C)
+import qualified Algebra.Field          as Field (C)
+import qualified Algebra.IntegralDomain as IntegralDomain (C)
+import qualified Algebra.Ring           as Ring (C)
+import qualified Algebra.ZeroTestable   as ZeroTestable (C)
+
+-- | The ring @Z_q@ of integers modulo 'q', using underlying integer
+-- type 'z'.
+newtype ZqBasic q z = ZqB z
+                    deriving (Eq, Ord, ZeroTestable.C, E.Elt, Show, NFData, Storable)
+
+-- the q argument, though phantom, matters for safety
+type role ZqBasic nominal representational
+
+--deriving instance (U.Unbox i) => V.Vector U.Vector (ZqBasic q i)
+--deriving instance (U.Unbox i) => M.MVector U.MVector (ZqBasic q i)
+--deriving instance (U.Unbox i) => U.Unbox (ZqBasic q i)
+
+-- convenience synonym for many instances
+type ReflectsTI q z = (Reflects q z, ToInteger z)
+
+reduce' :: forall q z . (ReflectsTI q z) => z -> ZqBasic q z
+reduce' = coerce . (`mod` proxy value (Proxy::Proxy q))
+
+-- puts value in range [-q/2, q/2)
+decode' :: forall q z . (ReflectsTI q z) => ZqBasic q z -> z
+decode' = let qval = proxy value (Proxy::Proxy q)
+          in \(ZqB x) -> if 2 * x < qval
+                         then x
+                         else x - qval
+
+instance (ReflectsTI q z, Enum z) => Enumerable (ZqBasic q z) where
+  values = let qval :: z = proxy value (Proxy::Proxy q)
+           in coerce [0..(qval-1)]
+
+instance (ReflectsTI q z) => Mod (ZqBasic q z) where
+  type ModRep (ZqBasic q z) = z
+
+  modulus = retag (value :: Tagged q z)
+
+type instance CharOf (ZqBasic p z) = p
+
+instance (PPow pp, zq ~ ZqBasic pp z, 
+          PrimeField (ZPOf zq), Ring zq, Ring (ZPOf zq)) 
+         => ZPP (ZqBasic (pp :: PrimePower) z) where
+
+  type ZPOf (ZqBasic pp z) = ZqBasic (PrimePP pp) z
+
+  modulusZPP = retag (ppPPow :: Tagged pp PP)
+
+  liftZp = coerce
+
+instance (ReflectsTI q z) => Reduce z (ZqBasic q z) where
+  reduce = reduce'
+
+instance (Reflects q z, Ring (ZqBasic q z)) => Reduce Integer (ZqBasic q z) where
+  reduce = fromInteger
+
+instance (ReflectsTI q z) => Lift' (ZqBasic q z) where
+  type LiftOf (ZqBasic q z) = z
+  lift = decode'
+
+instance (ReflectsTI q z, ReflectsTI q' z, Ring z)
+         => Rescale (ZqBasic q z) (ZqBasic q' z) where
+
+    rescale = rescaleMod
+
+instance (Reflects p z, ReflectsTI q z,
+          Field (ZqBasic p z), Field (ZqBasic q z))
+         => Encode (ZqBasic p z) (ZqBasic q z) where
+
+    lsdToMSD = let pval :: z = proxy value (Proxy::Proxy p)
+                   negqval :: z = negate $ proxy value (Proxy::Proxy q)
+               in (reduce' negqval, recip $ reduce' pval)
+
+-- instance of CRTrans
+instance (Reflects q z, PID z, r ~ (ZqBasic q z), Mod r, Enumerable r,
+          Show z) -- for DT.trace
+         => CRTrans (ZqBasic q z) where
+
+  crtInfo =
+    --DT.trace ("ZqBasic.crtInfo: q = " ++ 
+    --          show (proxy value (Proxy::Proxy q) :: z)) $
+    let qval :: z = proxy value (Proxy::Proxy q)
+    in \m -> (,) <$> omegaPowMod m <*>
+  -- CJP: using coerce depends on modinv returning in [0..q-1]
+                     (coerce $ fromIntegral (valueHat m) `modinv` qval)
+
+-- instance of CRTEmbed
+instance (ReflectsTI q z, Ring (ZqBasic q z)) => CRTEmbed (ZqBasic q z) where
+  type CRTExt (ZqBasic q z) = Complex Double
+
+  toExt (ZqB x) = fromReal $ fromIntegral x
+  fromExt x = reduce' $ NP.round $ real x
+
+-- instance of Additive
+instance (ReflectsTI q z, Additive z) => Additive.C (ZqBasic q z) where
+  -- CJP: "LHS too complicated to desugar"; might be fixed in 7.10:
+  -- https://ghc.haskell.org/trac/ghc/ticket/8848
+  {-# SPECIALIZE instance ReflectsTI q Int => Additive.C (ZqBasic q Int) #-}
+  {-# SPECIALIZE instance ReflectsTI q Int64 => Additive.C (ZqBasic q Int64) #-}
+  
+  zero = ZqB zero
+  
+  (+) = let qval = proxy value (Proxy::Proxy q)
+        in \ (ZqB x) (ZqB y) ->
+        let z = x + y
+        in ZqB (if z >= qval then z - qval else z)
+
+  negate (ZqB x) = reduce' $ negate x
+
+-- instance of Ring
+instance (ReflectsTI q z, Ring z) => Ring.C (ZqBasic q z) where
+    (ZqB x) * (ZqB y) = reduce' $ x * y
+
+    fromInteger x =
+      let qval = toInteger (proxy value (Proxy::Proxy q) :: z)
+    -- this is safe as long as type z can hold the value of q
+      in ZqB $ fromInteger $ x `mod` qval
+
+-- instance of Field
+instance (ReflectsTI q z, PID z, Show z) => Field.C (ZqBasic q z) where
+
+  recip = let qval = proxy value (Proxy::Proxy q)
+              -- safe because modinv returns in range 0..qval-1
+          in \(ZqB x) -> ZqB $ 
+               fromMaybe (error $ "ZqB.recip fail: " ++ 
+                         show x ++ "\t" ++ show qval) $ modinv x qval
+
+-- (canonical) instance of IntegralDomain, needed for FastCyc
+instance (Field (ZqBasic q z)) => IntegralDomain.C (ZqBasic q z) where
+    divMod a b = (a/b, zero)
+
+-- Gadget-related instances
+instance (ReflectsTI q z, Additive z)
+         => Gadget TrivGad (ZqBasic q z) where
+  
+  gadget = tag [one]
+
+instance (ReflectsTI q z, Ring z) => Decompose TrivGad (ZqBasic q z) where
+  type DecompOf (ZqBasic q z) = z
+  decompose x = tag [lift x]
+
+instance (ReflectsTI q z, Ring z) => Correct TrivGad (ZqBasic q z) where
+  correct a = case untag a of
+    [b] -> b
+    _ -> error "Correct TrivGad: wrong length"
+
+instance (ReflectsTI q z, Additive z, Reflects b z)
+         => Gadget (BaseBGad b) (ZqBasic q z) where
+  
+  gadget = let qval = proxy value (Proxy :: Proxy q)
+               bval = proxy value (Proxy :: Proxy b)
+               k = logCeil bval qval
+           in tag $ map reduce' (take k (iterate (*bval) one))
+
+instance (ReflectsTI q z, Ring z, Reflects b z) => Decompose (BaseBGad b) (ZqBasic q z) where
+  type DecompOf (ZqBasic q z) = z
+  decompose = let qval = proxy value (Proxy :: Proxy q)
+                  bval = proxy value (Proxy :: Proxy b)
+                  k = logCeil bval qval
+                  radices = replicate (k-1) bval
+              in tag . decomp radices . lift
+
+-- TODO: implement Correct for BaseBGad b
+
+-- instance of Random
+instance (ReflectsTI q z, Random z) => Random (ZqBasic q z) where
+  random = let high = proxy value (Proxy::Proxy q) - 1
+           in \g -> let (x,g') = randomR (0,high) g
+                    in (ZqB x, g')
+
+  randomR _ = error "randomR non-sensical for Zq types"
+
+-- instance of Arbitrary
+instance (ReflectsTI q z, Random z) => Arbitrary (ZqBasic q z) where
+  arbitrary =
+    let qval :: z = proxy value (Proxy::Proxy q)
+    in fromIntegral <$> choose (0, qval-1)
+
+  shrink = shrinkNothing
+
+-- CJP: restored manual Unbox instances, until we have a better way
+-- (NewtypeDeriving or TH)
+
+newtype instance U.MVector s (ZqBasic q z) = MV_ZqBasic (U.MVector s z)
+newtype instance U.Vector (ZqBasic q z) = V_ZqBasic (U.Vector z)
+
+-- Unbox, when underlying representation is
+instance (U.Unbox z) => U.Unbox (ZqBasic q z)
+
+{- purloined and tweaked from code in `vector` package that defines
+types as unboxed -}
+instance (U.Unbox z) => M.MVector U.MVector (ZqBasic q z) where
+  basicLength (MV_ZqBasic v) = M.basicLength v
+  basicUnsafeSlice z n (MV_ZqBasic v) = MV_ZqBasic $ M.basicUnsafeSlice z n v
+  basicOverlaps (MV_ZqBasic v1) (MV_ZqBasic v2) = M.basicOverlaps v1 v2
+  basicInitialize (MV_ZqBasic v) = M.basicInitialize v
+  basicUnsafeNew n = MV_ZqBasic `liftM` M.basicUnsafeNew n
+  basicUnsafeReplicate n (ZqB x) = MV_ZqBasic `liftM` M.basicUnsafeReplicate n x
+  basicUnsafeRead (MV_ZqBasic v) z = ZqB `liftM` M.basicUnsafeRead v z
+  basicUnsafeWrite (MV_ZqBasic v) z (ZqB x) = M.basicUnsafeWrite v z x
+  basicClear (MV_ZqBasic v) = M.basicClear v
+  basicSet (MV_ZqBasic v) (ZqB x) = M.basicSet v x
+  basicUnsafeCopy (MV_ZqBasic v1) (MV_ZqBasic v2) = M.basicUnsafeCopy v1 v2
+  basicUnsafeMove (MV_ZqBasic v1) (MV_ZqBasic v2) = M.basicUnsafeMove v1 v2
+  basicUnsafeGrow (MV_ZqBasic v) n = MV_ZqBasic `liftM` M.basicUnsafeGrow v n
+
+instance (U.Unbox z) => V.Vector U.Vector (ZqBasic q z) where
+  basicUnsafeFreeze (MV_ZqBasic v) = V_ZqBasic `liftM` V.basicUnsafeFreeze v
+  basicUnsafeThaw (V_ZqBasic v) = MV_ZqBasic `liftM` V.basicUnsafeThaw v
+  basicLength (V_ZqBasic v) = V.basicLength v
+  basicUnsafeSlice z n (V_ZqBasic v) = V_ZqBasic $ V.basicUnsafeSlice z n v
+  basicUnsafeIndexM (V_ZqBasic v) z = ZqB `liftM` V.basicUnsafeIndexM v z
+  basicUnsafeCopy (MV_ZqBasic mv) (V_ZqBasic v) = V.basicUnsafeCopy mv v
+  elemseq _ = seq
diff --git a/test-suite/CycTests.hs b/test-suite/CycTests.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/CycTests.hs
@@ -0,0 +1,105 @@
+{-# LANGUAGE RankNTypes, ScopedTypeVariables, NoImplicitPrelude, RebindableSyntax,
+             TypeOperators, FlexibleContexts, ConstraintKinds, TypeFamilies,
+             DataKinds #-}
+module CycTests (cycTests) where
+
+import TestTypes
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.Cyclotomic.Cyc
+import Crypto.Lol.LatticePrelude
+import Crypto.Lol.Cyclotomic.Tensor.CTensor
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+import Crypto.Lol.Types.FiniteField
+import Crypto.Lol.Types.IrreducibleChar2
+import Crypto.Lol.Types.ZPP
+
+import Control.Monad (join, liftM2)
+
+import Data.Array.Repa.Eval (Elt)
+import Data.Type.Natural hiding (zero)
+import Data.Vector.Unboxed (Vector, Unbox)
+import Data.Vector.Storable (Storable)
+
+import Test.Framework (testGroup, Test, defaultMain)
+import Test.Framework.Providers.QuickCheck2 (testProperty)
+import Test.QuickCheck (Property, property, Arbitrary)
+
+cycTests = [testGroup "coeffsPow" $ groupC $ wrapCmm'rToBool prop_coeffsBasis,
+            testGroup "crtSet" $ groupC $ wrapProxyCmm'rToBool prop_crtSet_pairs]
+
+
+
+
+
+type BasisCtx t m m' r = 
+  (m `Divides` m', ZPP r, CElt t r, CElt t (ZPOf r))
+
+prop_coeffsBasis :: forall t m m' r . (BasisCtx t m m' r)
+  => Proxy m -> Cyc t m' r -> Bool
+prop_coeffsBasis _ x = 
+  let xs = map embed (coeffsCyc Pow x :: [Cyc t m r])
+      bs = proxy powBasis (Proxy::Proxy m)
+  in (sum $ zipWith (*) xs bs) == x
+
+-- verifies that CRT set elements satisfy c_i * c_j = delta_ij * c_i
+-- necessary (not sufficient?) condition
+prop_crtSet_pairs :: forall t m m' r . (BasisCtx t m m' r)
+  => Proxy m -> Proxy (Cyc t m' r) -> Bool
+prop_crtSet_pairs pm _ = 
+  let crtset = proxy crtSet pm :: [Cyc t m' r]
+      pairs = join (liftM2 (,)) crtset
+  in and $ map (\(a,b) -> if a == b then a*b == a else a*b == zero) pairs
+
+type BasisWrapCtx t m m' r =
+  (BasisCtx t m m' r, Show (Cyc t m' r), Arbitrary (t m' r))
+
+wrapCmm'rToBool :: (BasisWrapCtx t m m' r)
+  => (Proxy m -> Cyc t m' r -> Bool) 
+     -> Proxy (Cyc t) -> Proxy '(m,m',r) -> Property
+wrapCmm'rToBool f _ _ = property $ f Proxy
+
+wrapProxyCmm'rToBool :: (BasisWrapCtx t m m' r)
+  => (Proxy m -> Proxy (Cyc t m' r) -> Bool) 
+     -> Proxy (Cyc t) -> Proxy '(m,m',r) -> Property
+wrapProxyCmm'rToBool f _ _ = property $ f Proxy Proxy
+
+groupC ::
+  (forall t m m' r . 
+       (BasisWrapCtx t m m' r) 
+       => Proxy (Cyc t) 
+          -> Proxy '(m,m',r) 
+          -> Property)
+  -> [Test]
+-- since we don't have any Tensor-level tests for coeffs/basis functions,
+-- we need to test all Tensors here.
+groupC f =
+  [testGroup "FC CT" $ groupMM'R (f (Proxy::Proxy (Cyc CT))),
+   testGroup "FC RT" $ groupMM'R (f (Proxy::Proxy (Cyc RT)))]
+
+type BasisWrapCCtx m m' r =
+  (BasisWrapCtx RT m m' r,
+   BasisWrapCtx CT m m' r)
+
+groupMM'R :: 
+  (forall m m' r . (BasisWrapCCtx m m' r) => Proxy '(m, m', r) -> Property) 
+  -> [Test]
+groupMM'R f = [testProperty "F1/F7/PP8" $ f (Proxy::Proxy '(F1, F7, Zq (PP2 N3))), 
+               testProperty "F1/F7/PP2" $ f (Proxy::Proxy '(F1, F7, Zq (PP2 N1)))] -- add some more test cases
+
+
+
+
+-- for crtSet, take all pairwise products 
+-- if elts are equal, id
+-- if not, zero
+
+-- also do a cardinality check
+
+-- checks cardinality of the CRT set
+{-
+prop_crtSet_card pm _
+  let inferLen = length $ (proxy crtSetDec pm :: [t m' r])
+      expectLen = 
+  in  
+-}
diff --git a/test-suite/Main.hs b/test-suite/Main.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/Main.hs
@@ -0,0 +1,19 @@
+
+
+--module Tests where
+
+import SHETests
+import TensorTests
+import CycTests
+import ZqTests
+
+import Test.Framework
+
+main :: IO ()
+main = do
+  flip defaultMainWithArgs ["--threads=1","--maximum-generated-tests=1000"]
+    [  testGroup "Tensor Tests" tensorTests
+     , testGroup "Cyc Tests" cycTests
+     , testGroup "SHE Tests" sheTests
+     , testGroup "Zq Tests" zqTests
+    ]
diff --git a/test-suite/SHETests.hs b/test-suite/SHETests.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/SHETests.hs
@@ -0,0 +1,442 @@
+{-# LANGUAGE ScopedTypeVariables, NoImplicitPrelude, RebindableSyntax, 
+             DataKinds, TypeOperators, NoMonomorphismRestriction, NoMonoLocalBinds,
+             ConstraintKinds, TypeFamilies, FlexibleContexts, PartialTypeSignatures, 
+             RankNTypes, MultiParamTypeClasses, FlexibleInstances, UndecidableInstances, 
+             RebindableSyntax, GADTs, PolyKinds, KindSignatures #-}
+
+module SHETests (sheTests) where
+
+import TestTypes
+
+import Control.Applicative hiding ((<$$>))
+import Control.Monad
+import Control.Monad.Random
+
+import Crypto.Lol.LatticePrelude hiding (lift)
+import Crypto.Lol.Cyclotomic.Cyc
+import Crypto.Lol.Applications.SymmSHE
+import Crypto.Lol.CRTrans
+import Crypto.Lol.Gadget
+import Crypto.Lol.Cyclotomic.Linear
+
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+import qualified Crypto.Lol.Cyclotomic.Tensor.CTensor as CT
+import Crypto.Lol.Cyclotomic.Tensor.CTensor hiding (CT)
+import Crypto.Lol.Types.ZqBasic
+
+import Data.Array.Repa.Eval (Elt)
+import Data.Type.Natural hiding (zero,one)
+import Data.Typeable
+import Data.Vector.Unboxed (Unbox)
+import Data.Vector.Storable (Storable)
+
+import Test.Framework
+import Test.Framework.Providers.QuickCheck2
+import Test.QuickCheck hiding (generate,output)
+import Test.QuickCheck.Monadic (monadicIO, assert)
+
+v = 1 :: Double
+
+sheTests = 
+  [testGroup "Tunnel" $ tunnelTests,
+   testGroup "Dec . Enc (Unrestricted)" $ groupCEnc $ wrapEnc prop_encDec,
+   testGroup "Dec . Enc (MSD)" $ groupCEnc $ wrapEnc prop_encDec_MSD,
+   testGroup "AddPub" $ groupCEnc $ wrapEnc prop_addPub,
+   testGroup "MulPub" $ groupCEnc $ wrapEnc prop_mulPub,
+   testGroup "ScalarPub" $ groupCEnc $ wrapScalar prop_addScalar,
+   testGroup "CTAdd" $ groupCEnc $ wrapMath prop_ctadd,
+   testGroup "CTMul" $ groupCEnc $ wrapMath prop_ctmul,
+   testGroup "CT zero" $ groupCEnc $ wrapConst prop_ctzero,
+   testGroup "CT one" $ groupCEnc $ wrapConst prop_ctone,
+   testGroup "ModSwPT" modSwPTTests,
+   testGroup "KSLin" $ groupCKS $ wrapKSLin prop_ksLin,
+   testGroup "KSQuad" $ groupCKS $ wrapKSQuad prop_ksQuad,
+   testGroup "Embed" $ groupCTwEm $ wrapEm prop_ctembed,
+   testGroup "Twace" $ groupCTwEm $ wrapTw prop_cttwace
+  ]
+
+type EncDecCtx c m m' zp zq =
+  (GenSKCtx c m (LiftOf zp) Double,
+   EncryptCtx c m m' (LiftOf zp) zp zq,
+   -- constraints from decryptUnrestricted
+   ToSDCtx c m' zp zq, Lift' zq, Reduce (LiftOf zq) zp)
+
+prop_encDec :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq) 
+  => Proxy '(m', zq) -> Cyc c m zp -> Property
+prop_encDec _ x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let x' = decryptUnrestricted sk $ y
+  assert $ x == x'
+
+prop_encDec_MSD :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq) 
+  => Proxy '(m', zq) -> Cyc c m zp -> Property
+prop_encDec_MSD _ x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let x' = decryptUnrestricted sk $ toMSD y
+  assert $ x == x'
+
+prop_addPub :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq)
+  => Proxy '(m', zq) -> Cyc c m zp -> Property
+prop_addPub _ x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let y' = addPublic x y
+      x' = decryptUnrestricted sk y'
+  assert $ x' == (x+x)
+
+prop_mulPub :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq)
+  => Proxy '(m', zq) -> Cyc c m zp -> Property
+prop_mulPub _ x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let y' = mulPublic x y
+      x' = decryptUnrestricted sk y'
+  assert $ x' == (x*x)
+
+prop_addScalar :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq)
+  => Proxy '(m', zq) -> zp -> Cyc c m zp -> Property
+prop_addScalar _ s x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let y' = addScalar s y
+      x' = decryptUnrestricted sk y'
+  assert $ x' == ((scalarCyc s)+x)
+
+prop_ctadd :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq) 
+  => Proxy '(m', zq) -> Cyc c m zp -> Cyc c m zp -> Property
+prop_ctadd _ x1 x2 = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y1 :: CT m zp (Cyc c m' zq) <- encrypt sk x1
+  y2 :: CT m zp (Cyc c m' zq) <- encrypt sk x2
+  let y' = y1+y2
+      x' = decryptUnrestricted sk y'
+  assert $ x1+x2 == x'
+
+prop_ctmul :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq)
+  => Proxy '(m', zq) -> Cyc c m zp -> Cyc c m zp -> Property
+prop_ctmul _ x1 x2 = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y1 :: CT m zp (Cyc c m' zq) <- encrypt sk x1
+  y2 :: CT m zp (Cyc c m' zq) <- encrypt sk x2
+  let y' = y1*y2
+      x' = decryptUnrestricted sk y'
+  assert $ x1*x2 == x'
+
+prop_ctzero :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq) 
+  => Proxy '(m', zq) -> Proxy (Cyc c m zp) -> Property
+prop_ctzero _ _ = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  let z = decryptUnrestricted sk (zero :: CT m zp (Cyc c m' zq))
+  assert $ zero == z
+
+prop_ctone :: forall m zp c m' zq . 
+  (EncDecCtx c m m' zp zq)
+  => Proxy '(m', zq) -> Proxy (Cyc c m zp) -> Property
+prop_ctone _ _ = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  let z = decryptUnrestricted sk (one :: CT m zp (Cyc c m' zq))
+  assert $ one == z
+
+type EncDecWrapCtx c m m' zp zq =
+  (EncDecCtx c m m' zp zq, Show (Cyc c m zp), Arbitrary (c m zp), Show zp, Arbitrary zp)
+
+wrapEnc :: (EncDecWrapCtx c m m' zp zq)
+  => (Proxy '(m', zq) -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy '(m, m', zp, zq) -> Property
+wrapEnc f _ _ = property $ f Proxy
+
+wrapScalar :: (EncDecWrapCtx c m m' zp zq)
+  => (Proxy '(m', zq) -> zp -> Cyc c m zp -> Property)
+     -> Proxy (Cyc c) -> Proxy '(m, m', zp, zq) -> Property
+wrapScalar f _ _ = property $ f Proxy
+
+wrapMath :: (EncDecWrapCtx c m m' zp zq)
+  => (Proxy '(m', zq) -> Cyc c m zp -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy '(m, m', zp, zq) -> Property
+wrapMath f _ _ = property $ f Proxy
+
+wrapConst :: (EncDecWrapCtx c m m' zp zq)
+  => (Proxy '(m', zq) -> Proxy (Cyc c m zp) -> Property) 
+     -> Proxy (Cyc c) -> Proxy '(m, m', zp, zq) -> Property
+wrapConst f _ _ = property $ f Proxy Proxy
+
+groupCEnc :: 
+  (forall c m m' zp zq . (EncDecWrapCtx c m m' zp zq)
+     => Proxy (Cyc c)
+     -> Proxy '(m, m', zp, zq)
+     -> Property) 
+  -> [Test]
+groupCEnc f =
+  [testGroup "CT" $ groupTypesEnc (f (Proxy::Proxy (Cyc CT.CT))),
+   testGroup "RT" $ groupTypesEnc (f (Proxy::Proxy (Cyc RT)))]
+
+type EncDecWrapCCtx m m' zp zq =
+  (EncDecWrapCtx RT m m' zp zq,
+   EncDecWrapCtx CT.CT m m' zp zq)
+
+groupTypesEnc :: 
+  (forall m m' zp zq . (EncDecWrapCCtx m m' zp zq)
+    => Proxy '(m, m', zp, zq)
+    -> Property)
+  -> [Test]
+groupTypesEnc f = [testProperty "F7/F7 /ZP2/ZQ2" $ f (Proxy::Proxy '(F7, F7,  ZP2, ZQ2)),
+                   testProperty "F7/F21/ZP2/ZQ2" $ f (Proxy::Proxy '(F7, F21, ZP2, ZQ2)),
+                   testProperty "F2/F8 /ZP2/Q536871001" $ f (Proxy::Proxy '(F2,F8,ZP2,Zq Q536871001)),
+                   testProperty "F1/F8 /ZP2/Q536871001" $ f (Proxy::Proxy '(F1,F8,ZP2,Zq Q536871001)),
+                   testProperty "F4/F12/ZP2/SmoothZQ1" $ f (Proxy::Proxy '(F4,F12,ZP2,SmoothZQ1)),
+                   testProperty "F4/F8/ZP3/SmoothQ1" $ f (Proxy::Proxy '(F4,F8,ZP3, Zq SmoothQ1)),
+                   testProperty "F7/F7 /ZP4/ZQ2" $ f (Proxy::Proxy '(F7, F7,  ZP4, ZQ2)),
+                   testProperty "F7/F21/ZP4/ZQ2" $ f (Proxy::Proxy '(F7, F21, ZP4, ZQ2)),
+                   testProperty "F1/F4/ZP4/ZQ1" $ f (Proxy::Proxy '(F1,F4,ZP4,ZQ1)),
+                   testProperty "F4/F4/ZP4/ZQ1" $ f (Proxy::Proxy '(F4,F4,ZP4,ZQ1)),
+                   testProperty "F14/F14/ZP4/ZQ1" $ f (Proxy::Proxy '(F14,F14,ZP4,ZQ1)),
+                   testProperty "F28/F28/ZP4/ZQ1" $ f (Proxy::Proxy '(F28,F28,ZP4,ZQ1)),
+                   testProperty "F28/F28/ZP4/Q80221" $ f (Proxy::Proxy '(F28,F28,ZP4,Zq Q80221)),
+                   testProperty "F1/F8 /ZP4/Q536871001" $ f (Proxy::Proxy '(F1,F8,ZP4,Zq Q536871001)),
+                   testProperty "F2/F8 /ZP4/Q536871001" $ f (Proxy::Proxy '(F2,F8,ZP4,Zq Q536871001)),
+                   testProperty "F4/F12/ZP8/SmoothZQ1" $ f (Proxy::Proxy '(F4,F12,ZP8,SmoothZQ1))
+                  ]
+
+-- one-off tests, no wrapper
+
+prop_modSwPT :: forall m zp c m' zq z v zp' .
+  (EncryptCtx c m m' z zp zq,
+   GenSKCtx c m z v,
+   DecryptCtx c m m' z zp' zq,
+   ModSwitchPTCtx c m' zp zp' zq,
+   RescaleCyc (Cyc c) zp zp', Mod zp',
+   CElt c Int64, CElt c zq,
+   z ~ LiftOf zp', v ~ Double, ModRep zp' ~ ModRep zp) 
+  => Proxy '(m', zq, zp') -> Cyc c m zp -> Property
+prop_modSwPT _ x = monadicIO $ do
+  let p = proxy modulus (Proxy::Proxy zp)
+      p' = proxy modulus (Proxy::Proxy zp')
+      x' = (fromIntegral $ p `div` p') * x
+  sk :: SK (Cyc c m' z) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x'
+  let y' = modSwitchPT y :: CT m zp' (Cyc c m' zq)
+      x'' = decrypt sk y'
+  assert $ x'' == rescaleCyc Dec x'
+
+modSwPTTests = 
+  [testProperty "RT/F7/F21/ZQ1/ZP4/ZP8" (prop_modSwPT (Proxy::Proxy '(F21, ZQ1, ZP4)) :: Cyc RT F7 ZP8 -> Property),
+   testProperty "RT/F7/F42/ZQ1/ZP2/ZP4" (prop_modSwPT (Proxy::Proxy '(F42, ZQ1, ZP2)) :: Cyc RT F7 ZP4 -> Property),
+   testProperty "CT/F7/F21/ZQ1/ZP4/ZP8" (prop_modSwPT (Proxy::Proxy '(F21, ZQ1, ZP4)) :: Cyc CT.CT F7 ZP8 -> Property),
+   testProperty "CT/F7/F42/ZQ1/ZP2/ZP4" (prop_modSwPT (Proxy::Proxy '(F42, ZQ1, ZP2)) :: Cyc CT.CT F7 ZP4 -> Property)]
+
+
+tunnelTests = 
+  [testProperty "RT/F7/F21/ZQ1/ZP4/ZP8" 
+    (prop_ringTunnel (Proxy::Proxy '(F40,ZQ1,F20,F60,TrivGad,ZQ2)) :: Cyc RT F8 ZP4 -> Property)]
+
+prop_ringTunnel :: forall c e r s e' r' s' z zp zq zq' gad . 
+  (TunnelCtx c e r s e' r' s' z zp zq zq' gad, 
+   EncryptCtx c r r' z zp zq,
+   GenSKCtx c r' z Double,
+   GenSKCtx c s' z Double,
+   DecryptCtx c s s' z zp zq,
+   Random (Cyc c s zp),
+   e ~ FGCD r s, Fact e) 
+  => Proxy '(r', zq, s, s', gad, zq') -> Cyc c r zp -> Property
+prop_ringTunnel _ x = monadicIO $ do
+  let totr = proxy totientFact (Proxy::Proxy r)
+      tote = proxy totientFact (Proxy::Proxy e)
+      basisSize = totr `div` tote
+  -- choose a random linear function of the appropriate size
+  bs :: [Cyc c s zp] <- replicateM basisSize getRandom
+  let f = (linearDec bs) \\ (gcdDivides (Proxy::Proxy r) (Proxy::Proxy s)) :: Linear c zp e r s 
+      expected = evalLin f x \\ (gcdDivides (Proxy::Proxy r) (Proxy::Proxy s))
+  skin :: SK (Cyc c r' (LiftOf zp)) <- genSK v
+  skout :: SK (Cyc c s' (LiftOf zp)) <- genSK v
+  y :: CT r zp (Cyc c r' zq) <- encrypt skin x
+  tunn <- proxyT (tunnelCT f skout skin) (Proxy::Proxy (gad,zq'))
+  let y' = tunn y
+      actual = decrypt skout y' :: Cyc c s zp
+  assert $ expected == actual
+
+
+
+
+
+
+
+
+
+
+
+
+type KsCtx m zp z c m' zq gad zq' deczq = 
+  (GenSKCtx c m' z Double,
+   z ~ LiftOf zp, 
+   EncryptCtx c m m' z zp zq,
+   KeySwitchCtx gad c m' zp zq zq', 
+   KSHintCtx gad c m' z zq',
+   RescaleCyc (Cyc c) zq deczq,
+   DecryptCtx c m m' z zp deczq)
+
+prop_ksLin :: forall m zp z c m' zq gad zq' deczq . (KsCtx m zp z c m' zq gad zq' deczq) 
+  => Proxy '(m', zq, gad, zq', deczq) -> Cyc c m zp -> Property
+prop_ksLin (_ :: Proxy '(m', zq, gad, zq', deczq)) x = monadicIO $ do
+  sk1 :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  sk2 :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk1 x
+  ks <- proxyT (keySwitchLinear sk2 sk1) (Proxy::Proxy (gad,zq'))
+  let y' :: CT m zp (Cyc c m' zq) = ks y
+      x' = decrypt sk2 (rescaleLinearCT y' :: CT m zp (Cyc c m' deczq))
+  assert $ x == x'
+
+prop_ksQuad :: forall m zp z c m' zq gad zq' deczq . (KsCtx m zp z c m' zq gad zq' deczq) 
+  => Proxy '(m', zq, gad, zq', deczq) -> Cyc c m zp -> Cyc c m zp -> Property
+prop_ksQuad (_ :: Proxy '(m', zq, gad, zq', deczq)) x1 x2 = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y1 :: CT m zp (Cyc c m' zq) <- encrypt sk x1
+  y2 :: CT m zp (Cyc c m' zq) <- encrypt sk x2
+  ks <- proxyT (keySwitchQuadCirc sk) (Proxy::Proxy (gad,zq'))
+  let y' = ks (y1*y2)
+      x' = decrypt sk (rescaleLinearCT y' :: CT m zp (Cyc c m' deczq))
+  assert $ x1*x2 == x'
+
+type KsWrapCtx m zp z c m' zq gad zq' deczq = 
+  (KsCtx m zp z c m' zq gad zq' deczq, Show (Cyc c m zp), Arbitrary (c m zp))
+
+wrapKSLin :: forall m zp z c m' zq gad zq' deczq . (KsWrapCtx m zp z c m' zq gad zq' deczq)
+  => (Proxy '(m', zq, gad, zq', deczq) -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy gad -> Proxy '(m, m', zp, zq, zq', deczq) -> Property
+wrapKSLin f _ _ _ = property $ f Proxy
+
+wrapKSQuad :: forall m zp z c m' zq gad zq' deczq . (KsWrapCtx m zp z c m' zq gad zq' deczq)
+  => (Proxy '(m', zq, gad, zq',deczq) -> Cyc c m zp -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy gad -> Proxy '(m, m', zp, zq, zq', deczq) -> Property
+wrapKSQuad f _ _ _ = property $ f Proxy
+
+groupCKS :: 
+  (forall c m m' zp z zq zq' gad deczq . (KsWrapCtx m zp z c m' zq gad zq' deczq)
+     => Proxy (Cyc c)
+     -> Proxy gad
+     -> Proxy '(m, m', zp, zq, zq', deczq)
+     -> Property)
+  -> [Test]
+groupCKS f =
+  [testGroup "CT" $ groupGadKS $ f (Proxy::Proxy (Cyc CT.CT)),
+   testGroup "RT" $ groupGadKS $ f (Proxy::Proxy (Cyc RT))]
+
+type KsWrapCCtx m zp z m' zq gad zq' deczq = 
+  (KsWrapCtx m zp z RT m' zq gad zq' deczq,
+   KsWrapCtx m zp z CT.CT m' zq gad zq' deczq)
+
+groupGadKS :: 
+  (forall m m' zp z zq zq' gad deczq . (KsWrapCCtx m zp z m' zq gad zq' deczq)
+     => Proxy gad
+     -> Proxy '(m, m', zp, zq, zq', deczq)
+     -> Property) 
+  -> [Test]
+groupGadKS f =
+  [testGroup "TrivGad" $ groupTypesKS (f (Proxy::Proxy TrivGad))]
+   --testGroup "Base16" $ groupTypesKS (f (Proxy::Proxy (BaseBGad N16)))]
+
+type KsWrapCGadCtx m zp z m' zq zq' deczq = 
+  (KsWrapCCtx m zp z m' zq TrivGad zq' deczq)
+
+groupTypesKS :: 
+  (forall m m' zp z zq zq' deczq . (KsWrapCGadCtx m zp z m' zq zq' deczq)
+    => Proxy '(m, m', zp, zq, zq', deczq) 
+    -> Property) 
+  -> [Test]
+groupTypesKS f = 
+  [testProperty "F1/F7/ZP2/ZQ1/ZQ2" $ f (Proxy::Proxy '(F1, F7, ZP2, ZQ1, ZQ2, ZQ1)),
+   testProperty "F2/F4/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F2, F4, ZP8, SmoothZQ1, SmoothZQ2, SmoothZQ1)),
+   testProperty "F4/F12/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F4, F12, ZP2, SmoothZQ1, SmoothZQ2, SmoothZQ1)),
+   testProperty "F8/F64/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F8, F64, ZP2, SmoothZQ1, SmoothZQ2, SmoothZQ1)),
+   testProperty "F3/F27/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F3, F27, ZP2, SmoothZQ1, SmoothZQ2, SmoothZQ1)),
+   testProperty "F2/F4/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F2, F4, ZP8, SmoothZQ2, SmoothZQ3, SmoothZQ1)),
+   testProperty "F4/F12/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F4, F12, ZP2, SmoothZQ2, SmoothZQ3, SmoothZQ1)),
+   testProperty "F8/F64/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F8, F64, ZP2, SmoothZQ2, SmoothZQ3, SmoothZQ1)),
+   testProperty "F3/F27/ZP2/SmoothZQ1/SmoothZQ2" $ f (Proxy::Proxy '(F3, F27, ZP2, SmoothZQ2, SmoothZQ3, SmoothZQ1))]
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+type TwEmCtx c m m' t t' zp zq =
+  (EncryptCtx c m m' (LiftOf zp) zp zq,
+   GenSKCtx c m (LiftOf zp) Double, 
+   DecryptCtx c m m' (LiftOf zp) zp zq, 
+   t `Divides` t', m `Divides` t, m' `Divides` t', m ~ FGCD m' t)
+
+prop_ctembed :: forall c m m' t t' zp zq . 
+  (TwEmCtx c m m' t t' zp zq)
+  => Proxy '(m', zq, t, t') -> Cyc c m zp -> Property
+prop_ctembed _ x = monadicIO $ do
+  sk :: SK (Cyc c m' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt sk x
+  let y' = embedCT y :: CT t zp (Cyc c t' zq)
+      x' = decrypt (embedSK sk) y'
+  assert $ (embed x :: Cyc c t zp) == x'
+
+prop_cttwace :: forall c m m' t t' zp zq . 
+  (TwEmCtx c t t' m m' zp zq)
+  => Proxy '(m', zq, t, t') -> Cyc c m zp -> Property
+prop_cttwace _ x = monadicIO $ do
+  sk :: SK (Cyc c t' (LiftOf zp)) <- genSK v
+  y :: CT m zp (Cyc c m' zq) <- encrypt (embedSK sk) x
+  let y' = twaceCT y :: CT t zp (Cyc c t' zq)
+      x' = decrypt sk y'
+  assert $ (twace x :: Cyc c t zp) == x'
+
+type TwEmWrapCtx c m m' t t' zp zq = 
+  (TwEmCtx c m m' t t' zp zq, Show (Cyc c m zp), Show (Cyc c t zp), Arbitrary (c m zp), Arbitrary (c t zp))
+
+wrapEm :: (TwEmWrapCtx c m m' t t' zp zq)
+  => (Proxy '(m', zq, t, t') -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy '(m, m', t, t', zp, zq) -> Property
+wrapEm f _ _ = property $ f Proxy
+
+wrapTw :: (TwEmWrapCtx c t t' m m' zp zq)
+  => (Proxy '(m', zq, t, t') -> Cyc c m zp -> Property) 
+     -> Proxy (Cyc c) -> Proxy '(t, t', m, m', zp, zq) -> Property
+wrapTw f _ _ = property $ f Proxy
+
+groupCTwEm :: 
+  (forall c m m' t t' zp zq . (TwEmWrapCtx c m m' t t' zp zq)
+     => Proxy (Cyc c)
+     -> Proxy '(m, m', t, t', zp, zq)
+     -> Property) 
+  -> [Test]
+groupCTwEm f =
+  [testGroup "CT" $ groupTypesTwEm (f (Proxy::Proxy (Cyc CT.CT))),
+   testGroup "RT" $ groupTypesTwEm (f (Proxy::Proxy (Cyc RT)))]
+
+type TwEmWrapCCtx m m' t t' zp zq =
+  (TwEmWrapCtx RT m m' t t' zp zq,
+   TwEmWrapCtx CT.CT m m' t t' zp zq)
+
+groupTypesTwEm :: 
+  (forall m m' t t' zp zq . (TwEmWrapCCtx m m' t t' zp zq)
+    => Proxy '(m, m', t, t', zp, zq) 
+    -> Property) 
+  -> [Test]
+groupTypesTwEm f = 
+  [testProperty "F1/F7/F3/F21/ZP2/ZQ1" $ f (Proxy::Proxy '(F1, F7, F3, F21, ZP2, ZQ1))]
diff --git a/test-suite/TensorTests.hs b/test-suite/TensorTests.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/TensorTests.hs
@@ -0,0 +1,364 @@
+{-# LANGUAGE NoImplicitPrelude, RebindableSyntax, ScopedTypeVariables, 
+             DataKinds, TypeOperators, KindSignatures, RankNTypes, GADTs,
+             MultiParamTypeClasses, ConstraintKinds, FlexibleInstances, RebindableSyntax,
+             FlexibleContexts, UndecidableInstances, TypeFamilies, DeriveDataTypeable #-}
+
+module TensorTests (tensorTests) where
+
+
+import TestTypes
+
+import Crypto.Lol.CRTrans
+import Crypto.Lol.LatticePrelude as LP hiding (round)
+import Crypto.Lol.Cyclotomic.Tensor
+import Crypto.Lol.Cyclotomic.Tensor.CTensor
+import Crypto.Lol.Cyclotomic.Tensor.RepaTensor
+
+import Control.Applicative
+import Control.Monad.Random
+
+import Data.Array.Repa.Eval (Elt)
+import Data.Constraint
+import Data.Maybe
+import Data.Vector.Unboxed as U
+import Data.Vector.Storable (Storable)
+
+import Test.Framework
+import Test.Framework.Providers.QuickCheck2
+import Test.QuickCheck hiding (generate,output)
+
+tensorTests = 
+  [testGroup "fmap comparison" $ groupTMR $ wrapTmrToBool prop_fmap,
+   testGroup "fmap comparison 2" $ groupTMR $ wrapTmrToBool prop_fmap2,
+   testGroup "Extension Mult" $ groupExtTests $ wrap2TmrToBool prop_mul_ext,
+
+   -- inverse property
+   tremTests, 
+   gInvGTests,
+   testGroup "CRTInv.CRT" $ groupTMR $ wrapTmrToBool prop_crt_inv,
+   testGroup "LInv.L" $ groupTMR $ wrapTmrToBool prop_l_inv,
+
+   -- commutative property
+   gCommuteTests,
+   embedCommuteTests,
+   twaceCommuteTests,
+   testGroup "Scalar" $ groupTMR $ wrapRToBool prop_scalar_crt,
+   twaceInvarTests
+   ]
+
+
+type TMRCtx t m r = (Tensor t, Fact m, m `Divides` m, CRTrans r, TElt t r, CRTEmbed r, TElt t (CRTExt r))
+
+prop_fmap :: (TMRCtx t m r) => t m r -> Bool
+prop_fmap x = fmapT id x == x \\ witness entailFullT x \\ witness entailIndexT x
+
+prop_fmap2 :: (TMRCtx t m r) => t m r -> Bool
+prop_fmap2 x = (fmapT id x) == (fmap id x) \\ witness entailFullT x \\ witness entailIndexT x
+
+-- tests that multiplication in the extension ring matches CRT multiplication
+prop_mul_ext :: forall t m r . (TMRCtx t m r)
+  => t m r -> t m r -> Bool
+prop_mul_ext x y = 
+  let m = proxy valueFact (Proxy::Proxy m)
+  in case (crtInfo m :: Maybe (CRTInfo r)) of
+       Nothing -> error "mul have a CRT to call prop_mul_ext"
+       Just _ -> (let z = x * y
+                      z' = fmapT fromExt $ (fmapT toExt x) * (fmapT toExt y)
+                  in z == z') \\ witness entailFullT x \\ witness entailFullT (fmap toExt x) \\ witness entailIndexT x
+
+gInvGTests = testGroup "GInv.G == id" [
+  testGroup "Pow basis" $ groupTMR $ wrapTmrToBool prop_ginv_pow,
+  testGroup "Dec basis" $ groupTMR $ wrapTmrToBool prop_ginv_dec,
+  testGroup "CRT basis" $ groupTMR $ wrapTmrToBool prop_ginv_crt]
+
+-- divG . mulG == id in Pow basis
+prop_ginv_pow :: (TMRCtx t m r) => t m r -> Bool
+prop_ginv_pow x = (fromMaybe (error "could not divide by G in prop_ginv_pow") $ 
+  divGPow $ mulGPow x) == x \\ witness entailFullT x
+
+-- divG . mulG == id in Dec basis
+prop_ginv_dec :: (TMRCtx t m r) => t m r -> Bool
+prop_ginv_dec x = (fromMaybe (error "could not divide by G in prop_ginv_dec") $ 
+  divGDec $ mulGDec x) == x \\ witness entailFullT x
+
+-- divG . mulG == id in CRT basis
+prop_ginv_crt :: (TMRCtx t m r) => t m r -> Bool
+prop_ginv_crt x = fromMaybe (error "no CRT in prop_ginv_crt") $ do
+  divGCRT' <- divGCRT
+  mulGCRT' <- mulGCRT
+  return $ (divGCRT' $ mulGCRT' x) == x \\ witness entailFullT x
+
+-- crtInv . crt == id
+prop_crt_inv :: (TMRCtx t m r) => t m r -> Bool
+prop_crt_inv x = fromMaybe (error "no CRT in prop_crt_inv") $ do
+  crt' <- crt
+  crtInv' <- crtInv
+  return $ (crtInv' $ crt' x) == x \\ witness entailFullT x
+
+-- lInv . l == id
+prop_l_inv :: (TMRCtx t m r) => t m r -> Bool
+prop_l_inv x = (lInv $ l x) == x \\ witness entailFullT x
+
+-- scalarCRT = crt . scalarPow
+prop_scalar_crt :: forall t m r . (TMRCtx t m r)
+                   => Proxy (t m r) -> r -> Bool
+prop_scalar_crt _ r = fromMaybe (error "no CRT in prop_scalar_crt") $ do
+  scalarCRT' <- scalarCRT
+  crt' <- crt
+  return $ (scalarCRT' r :: t m r) == (crt' $ scalarPow r)
+  \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+gCommuteTests = testGroup "G commutes with L" [
+  testGroup "Dec basis" $ groupTMR $ wrapTmrToBool prop_g_dec,
+  testGroup "CRT basis" $ groupTMR $ wrapTmrToBool prop_g_crt]
+
+-- mulGDec == lInv. mulGPow . l
+prop_g_dec :: (TMRCtx t m r) => t m r -> Bool
+prop_g_dec x = (mulGDec x) == (lInv $ mulGPow $ l x) \\ witness entailFullT x
+
+prop_g_crt :: (TMRCtx t m r) => t m r -> Bool
+prop_g_crt x = fromMaybe (error "no CRT in prop_g_crt") $ do
+  mulGCRT' <- mulGCRT
+  crt' <- crt
+  crtInv' <- crtInv
+  return $ (mulGCRT' x) == (crt' $ mulGPow $ crtInv' x) \\ witness entailFullT x
+
+type TMRWrapCtx t m r = (TMRCtx t m r, Show (t m r), Arbitrary (t m r), Show r, Arbitrary r)
+
+wrap2TmrToBool :: (TMRWrapCtx t m r) => (t m r -> t m r -> Bool) 
+     -> Proxy t -> Proxy '(m,r) -> Property
+wrap2TmrToBool f _ _ = property f
+
+wrapTmrToBool :: (TMRWrapCtx t m r) => (t m r -> Bool) 
+     -> Proxy t -> Proxy '(m,r) -> Property
+wrapTmrToBool f _ _ = property f
+
+wrapRToBool :: (TMRWrapCtx t m r) => (Proxy (t m r) -> r -> Bool)
+     -> Proxy t -> Proxy '(m,r) -> Property
+wrapRToBool f _ _ = property $ f Proxy 
+
+groupTMR :: (forall t m r . (TMRWrapCtx t m r)
+             => Proxy t
+             -> Proxy '(m,r) 
+             -> Property) -> [Test]
+groupTMR f =
+  [testGroup "CT" $ groupMR (f (Proxy::Proxy CT)),
+   testGroup "RT" $ groupMR (f (Proxy::Proxy RT))]
+
+groupExtTests :: (forall t m r . (TMRWrapCtx t m r)
+             => Proxy t
+             -> Proxy '(m,r) 
+             -> Property) -> [Test]
+groupExtTests f =
+  [testGroup "CT" $ groupMRExt (f (Proxy::Proxy CT)),
+   testGroup "RT" $ groupMRExt (f (Proxy::Proxy RT))]
+
+type MRWrapCtx m r = (TMRWrapCtx CT m r, TMRWrapCtx RT m r)
+
+groupMR :: (forall m r . (MRWrapCtx m r) => Proxy '(m, r) -> Property) 
+            -> [Test]
+groupMR f = [testProperty "F7/Q29" $ f (Proxy::Proxy '(F7, Zq Q29)),
+             testProperty "F12/SmoothZQ1" $ f (Proxy::Proxy '(F12, SmoothZQ1)),
+             testProperty "F1/Q17" $ f (Proxy::Proxy '(F1, Zq Q17)),
+             testProperty "F2/Q17" $ f (Proxy::Proxy '(F2, Zq Q17)),
+             testProperty "F4/Q17" $ f (Proxy::Proxy '(F4, Zq Q17)),
+             testProperty "F8/Q17" $ f (Proxy::Proxy '(F8, Zq Q17)),
+             testProperty "F21/Q8191" $ f (Proxy::Proxy '(F21, Zq Q8191)),
+             testProperty "F42/Q8191" $ f (Proxy::Proxy '(F42, Zq Q8191)),
+             testProperty "F42/ZQ1" $ f (Proxy::Proxy '(F42, ZQ1)),
+             testProperty "F42/ZQ2" $ f (Proxy::Proxy '(F42, ZQ2))]
+
+-- we can't include a large modulus here because there is not enough
+-- precision in Doubles to handle the error
+groupMRExt :: (forall m r . (MRWrapCtx m r) => Proxy '(m, r) -> Property) 
+            -> [Test]
+groupMRExt f = [testProperty "F7/Q29" $ f (Proxy::Proxy '(F7, Zq Q29)),
+             testProperty "F1/Q17" $ f (Proxy::Proxy '(F1, Zq Q17)),
+             testProperty "F2/Q17" $ f (Proxy::Proxy '(F2, Zq Q17)),
+             testProperty "F4/Q17" $ f (Proxy::Proxy '(F4, Zq Q17)),
+             testProperty "F8/Q17" $ f (Proxy::Proxy '(F8, Zq Q17)),
+             testProperty "F21/Q8191" $ f (Proxy::Proxy '(F21, Zq Q8191)),
+             testProperty "F42/Q8191" $ f (Proxy::Proxy '(F42, Zq Q8191)),
+             testProperty "F42/ZQ1" $ f (Proxy::Proxy '(F42, ZQ1)),
+             testProperty "F42/ZQ2" $ f (Proxy::Proxy '(F42, ZQ2))]
+
+
+
+
+
+
+
+
+
+
+
+type TMM'RCtx t m m' r = (Tensor t, m `Divides` m', TElt t r, Ring r, CRTrans r)
+
+-- groups related tests
+tremTests = testGroup "Tr.Em == id" [
+  testGroup "Pow basis" $ groupTMM'R $ wrapTmm'rToBool prop_trem_pow,
+  testGroup "Dec basis" $ groupTMM'R $ wrapTmm'rToBool prop_trem_dec,
+  testGroup "CRT basis" $ groupTMM'R $ wrapTmm'rToBool prop_trem_crt]
+
+-- tests that twace . embed == id in the Pow basis
+prop_trem_pow :: forall t m m' r . (TMM'RCtx t m m' r)
+  => Proxy m' -> t m r -> Bool
+prop_trem_pow _ x = (twacePowDec $ (embedPow x :: t m' r)) == x \\ witness entailFullT x
+
+-- tests that twace . embed == id in the Dec basis
+prop_trem_dec :: forall t m m' r . (TMM'RCtx t m m' r)
+  => Proxy m' -> t m r -> Bool
+prop_trem_dec _ x = (twacePowDec $ (embedDec x :: t m' r)) == x \\ witness entailFullT x
+
+-- tests that twace . embed == id in the CRT basis
+prop_trem_crt :: forall t m m' r . (TMM'RCtx t m m' r)
+  => Proxy m' -> t m r -> Bool
+prop_trem_crt _ x = fromMaybe (error "no CRT in prop_trem_crt") $
+  (x==) <$> (twaceCRT <*> (embedCRT <*> pure x :: Maybe (t m' r))) \\ witness entailFullT x
+
+embedCommuteTests = testGroup "Em commutes with L" [
+  testGroup "Dec basis" $ groupTMM'R $ wrapTmm'rToBool prop_embed_dec,
+  testGroup "CRT basis" $ groupTMM'R $ wrapTmm'rToBool prop_embed_crt]
+
+-- embedDec == lInv . embedPow . l
+prop_embed_dec :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> t m r -> Bool
+prop_embed_dec _ x = (embedDec x :: t m' r) == (lInv $ embedPow $ l x) 
+  \\ proxy entailFullT (Proxy::Proxy (t m' r))
+
+-- embedCRT = crt . embedPow . crtInv
+prop_embed_crt :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> t m r -> Bool
+prop_embed_crt _ x = fromMaybe (error "no CRT in prop_embed_crt") $ do
+  crt' <- crt
+  crtInv' <- crtInv
+  embedCRT' <- embedCRT
+  return $ (embedCRT' x :: t m' r) == (crt' $ embedPow $ crtInv' x) 
+    \\ proxy entailFullT (Proxy::Proxy (t m' r))
+
+twaceCommuteTests = testGroup "Tw commutes with L" [
+  testGroup "Dec basis" $ groupTMM'R $ wrapTm'mrToBool prop_twace_dec,
+  testGroup "CRT basis" $ groupTMM'R $ wrapTm'mrToBool prop_twace_crt]
+
+-- twacePowDec = lInv . twacePowDec . l
+prop_twace_dec :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m -> t m' r -> Bool
+prop_twace_dec _ x = (twacePowDec x :: t m r) == (lInv $ twacePowDec $ l x)
+  \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+-- twaceCRT = crt . twacePowDec . crtInv
+prop_twace_crt :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m -> t m' r -> Bool
+prop_twace_crt _ x = fromMaybe (error "no CRT in prop_trace_crt") $ do
+  twaceCRT' <- twaceCRT
+  crt' <- crt
+  crtInv' <- crtInv
+  return $ (twaceCRT' x :: t m r) == (crt' $ twacePowDec $ crtInv' x)
+    \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+twaceInvarTests = testGroup "Twace invariants" [
+  testGroup "Tw and Em ID for equal indices" $ groupTMR $ wrapTmrToBool $ prop_twEmID,
+  testGroup "Invar1 Pow basis" $ groupTMM'R $ wrapProxyTmm'rToBool prop_twace_invar1_pow,
+  testGroup "Invar1 Dec basis" $ groupTMM'R $ wrapProxyTmm'rToBool prop_twace_invar1_dec,
+  testGroup "Invar1 CRT basis" $ groupTMM'R $ wrapProxyTmm'rToBool prop_twace_invar1_crt,
+  testGroup "Invar2 Pow/Dec basis" $ groupTMM'R $ wrapProxyTmm'rToBool prop_twace_invar2_powdec,
+  testGroup "Invar2 CRT basis" $ groupTMM'R $ wrapProxyTmm'rToBool prop_twace_invar2_crt
+  ]
+
+prop_twEmID :: forall t m r . (Tensor t, TElt t r, CRTrans r, Fact m, m `Divides` m) => t m r -> Bool
+prop_twEmID x = ((twacePowDec x) == x) &&
+                  (((fromMaybe (error "twemid_crt") twaceCRT) x) == x) &&
+                  ((embedPow x) == x) &&
+                  ((embedDec x) == x) &&
+                  (((fromMaybe (error "twemid_crt") embedCRT) x) == x) \\ witness entailFullT x
+
+-- twace mhat'/g' = mhat*totm'/totm/g (Pow basis)
+prop_twace_invar1_pow :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> Proxy (t m r) -> Bool
+prop_twace_invar1_pow _ _ = fromMaybe (error "could not divide by G in prop_twace_invar1_pow") $ do
+  let mhat = proxy valueHatFact (Proxy::Proxy m)
+      mhat' = proxy valueHatFact (Proxy::Proxy m')
+      totm = proxy totientFact (Proxy::Proxy m)
+      totm' = proxy totientFact (Proxy::Proxy m')
+  output :: t m r <- divGPow $ scalarPow $ fromIntegral $ mhat * totm' `div` totm
+  input :: t m' r <- divGPow $ scalarPow $ fromIntegral mhat'
+  return $ (twacePowDec input) == output \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+-- twace mhat'/g' = mhat*totm'/totm/g (Dec basis)
+prop_twace_invar1_dec :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> Proxy (t m r) -> Bool
+prop_twace_invar1_dec _ _ = fromMaybe (error "could not divide by G in prop_twace_invar1_dec") $ do
+  let mhat = proxy valueHatFact (Proxy::Proxy m)
+      mhat' = proxy valueHatFact (Proxy::Proxy m')
+      totm = proxy totientFact (Proxy::Proxy m)
+      totm' = proxy totientFact (Proxy::Proxy m')
+  output :: t m r <- divGDec $ lInv $ scalarPow $ fromIntegral $ mhat * totm' `div` totm
+  input :: t m' r <- divGDec $ lInv $ scalarPow $ fromIntegral mhat'
+  return $ (twacePowDec input) == output \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+-- twace mhat'/g' = mhat*totm'/totm/g (CRT basis)
+prop_twace_invar1_crt :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> Proxy (t m r) -> Bool
+prop_twace_invar1_crt _ _ = fromMaybe (error "no CRT in prop_twace_invar1_crt") $ do
+  let mhat = proxy valueHatFact (Proxy::Proxy m)
+      mhat' = proxy valueHatFact (Proxy::Proxy m')
+      totm = proxy totientFact (Proxy::Proxy m)
+      totm' = proxy totientFact (Proxy::Proxy m')
+  scalarCRT1 <- scalarCRT
+  scalarCRT2 <- scalarCRT
+  divGCRT1 <- divGCRT
+  divGCRT2 <- divGCRT
+  twaceCRT' <- twaceCRT
+  let output :: t m r = divGCRT1 $ scalarCRT1 $ fromIntegral $ mhat * totm' `div` totm
+      input :: t m' r = divGCRT2 $ scalarCRT2 $ fromIntegral mhat'
+  return $ (twaceCRT' input) == output \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+-- twace preserves scalars in Pow/Dec basis
+prop_twace_invar2_powdec :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> Proxy (t m r) -> Bool
+prop_twace_invar2_powdec _ _ = 
+  let output = scalarPow $ one :: t m r
+      input = scalarPow $ one :: t m' r
+  in (twacePowDec input) == output \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+-- twace preserves scalars in Pow/Dec basis
+prop_twace_invar2_crt :: forall t m m' r . (TMM'RCtx t m m' r) => Proxy m' -> Proxy (t m r) -> Bool
+prop_twace_invar2_crt _ _ = fromMaybe (error "no CRT in prop_twace_invar2_crt") $ do
+  scalarCRT1 <- scalarCRT
+  scalarCRT2 <- scalarCRT
+  let input = scalarCRT1 one :: t m' r
+      output = scalarCRT2 one :: t m r
+  return $ (twacePowDec input) == output \\ proxy entailFullT (Proxy::Proxy (t m r))
+
+type TMM'RWrapCtx t m m' r = (TMM'RCtx t m m' r, Show (t m' r), Show (t m r), Arbitrary (t m' r), Arbitrary (t m r))
+
+wrapProxyTmm'rToBool :: (TMM'RWrapCtx t m m' r)
+                        => (Proxy m' -> Proxy (t m r) -> Bool) 
+                        -> Proxy t -> Proxy '(m,m',r) -> Property
+wrapProxyTmm'rToBool f _ _ = property $ f Proxy Proxy
+
+wrapTmm'rToBool :: (TMM'RWrapCtx t m m' r) => (Proxy m' -> t m r -> Bool) 
+                   -> Proxy t -> Proxy '(m,m',r) -> Property
+wrapTmm'rToBool f _ _ = property $ f Proxy
+
+wrapTm'mrToBool :: (TMM'RWrapCtx t m m' r) => (Proxy m -> t m' r -> Bool) 
+                   -> Proxy t -> Proxy '(m,m',r) -> Property
+wrapTm'mrToBool f _ _ = property $ f Proxy
+
+groupTMM'R :: (forall t m m' r . TMM'RWrapCtx t m m' r
+               => Proxy t
+               -> Proxy '(m,m',r) 
+               -> Property) -> [Test]
+groupTMM'R f =
+  [testGroup "CT" $ groupMM'R (f (Proxy::Proxy CT)),
+   testGroup "RT" $ groupMM'R (f (Proxy::Proxy RT))]
+
+type MM'RWrapCtx m m' r = (TMM'RWrapCtx CT m m' r, TMM'RWrapCtx RT m m' r)
+
+groupMM'R :: (forall m m' r . (MM'RWrapCtx m m' r)
+              => Proxy '(m, m', r) -> Property) -> [Test]
+groupMM'R f = [testProperty "F1/F7/Q29" $ f (Proxy::Proxy '(F1, F7, Zq Q29)),
+               testProperty "F4/F12/Q536871001" $ f (Proxy::Proxy '(F4, F12, Zq Q536871001)),
+               testProperty "F4/F12/SmoothZQ1" $ f (Proxy::Proxy '(F4, F12, SmoothZQ1)),
+               testProperty "F2/F8/Q17" $ f (Proxy::Proxy '(F2, F8, Zq Q17)),
+               testProperty "F8/F8/Q17" $ f (Proxy::Proxy '(F8, F8, Zq Q17)),
+               testProperty "F12/F12/SmoothZQ1" $ f (Proxy::Proxy '(F2, F8, SmoothZQ1)),
+               testProperty "F4/F8/Q17" $ f (Proxy::Proxy '(F4, F8, Zq Q17)),
+               testProperty "F3/F21/Q8191" $ f (Proxy::Proxy '(F3, F21, Zq Q8191)),
+               testProperty "F7/F21/Q8191" $ f (Proxy::Proxy '(F7, F21, Zq Q8191)),
+               testProperty "F3/F42/Q8191" $ f (Proxy::Proxy '(F3, F42, Zq Q8191)),
+               testProperty "F3/F21/ZQ1" $ f (Proxy::Proxy '(F3, F21, ZQ1)),
+               testProperty "F7/F21/ZQ2" $ f (Proxy::Proxy '(F7, F21, ZQ2)),
+               testProperty "F3/F42/ZQ3" $ f (Proxy::Proxy '(F3, F42, ZQ3))]
diff --git a/test-suite/TestTypes.hs b/test-suite/TestTypes.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/TestTypes.hs
@@ -0,0 +1,90 @@
+{-# LANGUAGE KindSignatures, PolyKinds, DataKinds, FlexibleInstances, RankNTypes,
+             TypeOperators, ConstraintKinds, FlexibleContexts, ScopedTypeVariables,
+             MultiParamTypeClasses, TypeFamilies, NoImplicitPrelude, RebindableSyntax #-}
+
+module TestTypes (
+  ZP2, ZP3, ZP4, ZP8
+, PP2
+, SmoothZQ1, SmoothZQ2, SmoothZQ3
+, SmoothQ1
+, Zq, ZQ1, ZQ2, ZQ3
+, Q17, Q29, Q8191, Q80221, Q536871001) where
+
+import Control.Monad
+import Control.Monad.Random
+
+import Crypto.Lol
+import Crypto.Lol.Reflects
+import Crypto.Lol.Factored
+
+import Data.Type.Natural
+
+import Test.QuickCheck.Monadic
+
+instance (MonadRandom m) => MonadRandom (PropertyM m) where
+  getRandom = run getRandom
+  getRandoms = run getRandoms
+  getRandomR r = run $ getRandomR r
+  getRandomRs r = run $ getRandomRs r
+
+-- three 24-bit moduli, enough to handle rounding for p=32 (depth-4 circuit at ~17 bits per mul)
+data Q18869761
+instance (ToInteger i) => Reflects Q18869761 i where value = return 18869761
+type ZQ1 = Zq Q18869761
+
+data Q19393921
+instance (ToInteger i) => Reflects Q19393921 i where value = return 19393921
+type ZQ2 = (Zq Q19393921, ZQ1)
+
+data Q19918081
+instance (ToInteger i) => Reflects Q19918081 i where value = return 19918081
+type ZQ3 = (Zq Q19918081, ZQ2)
+
+-- a 31-bit modulus, for rounding off after the last four hops
+data Q2149056001
+instance (ToInteger i) => Reflects Q2149056001 i where value = return 2149056001
+type ZQ4 = (Zq Q2149056001, ZQ3)
+
+-- for rounding off after the first hop
+data Q3144961
+instance (ToInteger i) => Reflects Q3144961 i where value = return 3144961
+type ZQ5 = (Zq Q3144961, ZQ4)
+
+data Q7338241
+instance (ToInteger i) => Reflects Q7338241 i where value = return 7338241
+type ZQ6 = (Zq Q7338241, ZQ5)
+
+-- concrete types useful for building tests or real applications
+data Q17
+data Q29
+data Q8191  -- 1028th prime, = 1 mod 21
+data Q80221  -- good for 28
+data Q536871001 
+-- the next three moduli are "good" for any index dividing 128*27*25*7
+data SmoothQ1 
+data SmoothQ2 
+data SmoothQ3 
+
+instance (ToInteger i) => Reflects Q17 i where value = return 17
+instance (ToInteger i) => Reflects Q29 i where value = return 29
+instance (ToInteger i) => Reflects Q8191 i where value = return 8191
+instance (ToInteger i) => Reflects Q536871001 i where value = return 536871001
+instance (ToInteger i) => Reflects Q80221 i where value = return 80221
+instance (ToInteger i) => Reflects SmoothQ1 i where value = return 2148249601
+instance (ToInteger i) => Reflects SmoothQ2 i where value = return 2148854401
+instance (ToInteger i) => Reflects SmoothQ3 i where value = return 2150668801
+
+type Zq (q :: k) = ZqBasic q Z
+type Z = Int64
+
+type PP2 e = ToPP N2 e
+type PP3 e = ToPP N3 e
+
+type ZP2 = Zq (PP2 N1)
+type ZP3 = Zq (PP3 N1)
+type ZP4 = Zq (PP2 N2)
+type ZP8 = Zq (PP2 N3)
+
+type SmoothZQ1 = Zq SmoothQ1
+type SmoothZQ2 = (Zq SmoothQ2, SmoothZQ1)
+type SmoothZQ3 = (Zq SmoothQ3, SmoothZQ2)
diff --git a/test-suite/ZqTests.hs b/test-suite/ZqTests.hs
new file mode 100644
--- /dev/null
+++ b/test-suite/ZqTests.hs
@@ -0,0 +1,56 @@
+{-# LANGUAGE NoImplicitPrelude, RebindableSyntax, ScopedTypeVariables, DataKinds, TypeOperators, PolyKinds, FlexibleContexts, RankNTypes, KindSignatures, MultiParamTypeClasses #-}
+
+module ZqTests (zqTests) where
+
+import Crypto.Lol.Types.ZqBasic
+import Crypto.Lol.LatticePrelude hiding (Nat)
+import Crypto.Lol.Reflects
+
+import Control.Monad
+
+import GHC.TypeLits
+
+import Test.Framework
+import Test.Framework.Providers.QuickCheck2
+import Test.QuickCheck
+
+
+prop_add :: forall (q :: Nat) . (Reflects q Int, KnownNat q) => Proxy q -> Int -> Int -> Bool
+prop_add _ x y = (fromIntegral $ x + y) == ((fromIntegral x) + (fromIntegral y :: ZqBasic q Int))
+
+prop_mul :: forall (q :: Nat) . (Reflects q Int, KnownNat q) => Proxy q -> Int -> Int -> Bool
+prop_mul _ x y = (fromIntegral $ x * y) == ((fromIntegral x) * (fromIntegral y :: ZqBasic q Int))
+
+prop_recip :: forall (q :: Nat) . (Reflects q Int, KnownNat q) => Proxy q -> Int -> Bool
+prop_recip _ x = let qval = proxy value (Proxy::Proxy q)
+                     y = fromIntegral x :: ZqBasic q Int
+                 in if (x `mod` qval) == 0
+                    then True
+                    else (fromIntegral (1::Int)) == (y * (recip y))
+
+type ZqModuli = '[7, 13, 17, 11, 13, 29]
+
+class CallZqProp xs where
+  callProp :: Proxy xs -> Gen Int -> (forall (q :: Nat) . (Reflects q Int, KnownNat q) => Proxy q -> Int -> Int -> Bool) -> [Test]
+
+  callProp2 :: Proxy xs 
+                -> Gen Int 
+                -> (forall (q :: Nat) . (Reflects q Int, KnownNat q) => Proxy q -> Int -> Bool)
+                -> [Test]
+
+instance CallZqProp '[] where
+  callProp _ _ _ = []
+  callProp2 _ _ _ = []
+
+instance (CallZqProp qs, KnownNat q) => CallZqProp (q ': qs) where
+  callProp _ gen f = (testProperty ("q = " ++ (show $ (proxy value (Proxy::Proxy q) :: Int))) $ property $ liftM2 (f (Proxy::Proxy q)) gen gen) : (callProp (Proxy::Proxy qs) gen f)
+  callProp2 _ gen f = (testProperty ("q = " ++ (show $ (proxy value (Proxy::Proxy q) :: Int))) $ property $ liftM (f (Proxy::Proxy q)) gen) : (callProp2 (Proxy::Proxy qs) gen f)
+
+zqModuli :: Proxy ZqModuli
+zqModuli = Proxy
+
+zqTests :: [Test]
+zqTests = 
+  [testGroup "ZqBasic +" $ callProp zqModuli (choose (-100,100)) prop_add,
+   testGroup "ZqBasic *" $ callProp zqModuli (choose (-100,100)) prop_mul,
+   testGroup "ZqBasic recip" $ callProp2 zqModuli (choose (-100,100)) prop_recip]
