Z-Data 0.8.3.0 → 0.8.4.0

raw patch · 5 files changed

+25/−16 lines, 5 filesPVP: major bump suggested

API removals or changes: PVP suggests a major version bump

API changes (from Hackage documentation)

- Z.Data.Vector: arrVec :: Vec v a => IArray v a -> v a

+ Z.Data.Vector: arrVec :: (Vec v a, Vec u a, IArray v ~ IArray u) => v a -> u a

- Z.Data.Vector.Base: arrVec :: Vec v a => IArray v a -> v a

+ Z.Data.Vector.Base: arrVec :: (Vec v a, Vec u a, IArray v ~ IArray u) => v a -> u a

Files

ChangeLog.md +5/−0
Z-Data.cabal +2/−1
Z/Data/Vector/Base.hs +3/−3
cbits/bytes.c +2/−2
third_party/Turbo-Base64/turbob64sse.c +13/−10

ChangeLog.md view

@@ -1,5 +1,10 @@ # Revision history for Z-Data +## 0.8.4.0  -- 2021-06-03++* `arrVec` from `Z.Data.Vector` now could convert between vector(array slice) and array back and forth.+* Fix a CPP bug in `no-avx` build flag.+ ## 0.8.3.0  -- 2021-06-02  * Fix numeric literals' `Print` instances when precedence > 6 and  value < 0.

Z-Data.cabal view

@@ -1,6 +1,6 @@ cabal-version:      2.4 name:               Z-Data-version:            0.8.3.0+version:            0.8.4.0 synopsis:           Array, vector and text description:        This package provides array, slice and text operations license:            BSD-3-Clause@@ -242,6 +242,7 @@     third_party/Turbo-Base64/turbob64c.c     third_party/Turbo-Base64/turbob64d.c     third_party/Turbo-Base64/turbob64sse.c+    -- third_party/Turbo-Base64/turbob64avx2.c, this file is conditionally included from bytes.c     third_party/utf8rewind/source/internal/casemapping.c     third_party/utf8rewind/source/internal/codepoint.c     third_party/utf8rewind/source/internal/composition.c

Z/Data/Vector/Base.hs view

@@ -135,10 +135,10 @@     -- | Create a vector by slicing an array(with offset and length).     fromArr :: IArray v a -> Int -> Int -> v a --- | Construct a 'Vec' by slicing a whole array.-arrVec :: Vec v a => IArray v a -> v a+-- | Change vector types based on same array type, e.g. construct an array from a slice, or vice-versa.+arrVec :: (Vec v a, Vec u a, IArray v ~ IArray u) => v a -> u a {-# INLINE arrVec #-}-arrVec arr = fromArr arr 0 (sizeofArr arr)+arrVec bs = let (arr, s, l) = toArr bs in fromArr arr s l  instance Vec Array a where     type IArray Array = Array

cbits/bytes.c view

@@ -259,7 +259,7 @@     tb64avx2enc(input+off, (size_t)len, output+output_off); #elif defined(__AVX__) && !defined(NO_AVX)     tb64avxenc(input+off, (size_t)len, output+output_off);-#elif defined(__SSE3__)+#elif defined(__SSSE3__)     tb64sseenc(input+off, (size_t)len, output+output_off); #else     tb64xenc(input+off, (size_t)len, output+output_off);@@ -271,7 +271,7 @@     return (HsInt)tb64avx2dec(input+off, (size_t)len, output); #elif defined(__AVX__) && !defined(NO_AVX)     return (HsInt)tb64avxdec(input+off, (size_t)len, output);-#elif defined(__SSE3__)+#elif defined(__SSSE3__)     return (HsInt)tb64ssedec(input+off, (size_t)len, output); #else     return (HsInt)tb64xdec(input+off, (size_t)len, output);

third_party/Turbo-Base64/turbob64sse.c view

@@ -34,12 +34,14 @@ 
 #include <string.h>
 
-  #if defined(__AVX__)
+  #if defined(__AVX__) && !defined(NO_AVX)
 #include <immintrin.h>
-#define FUNPREF tb64avx
+#define FUNPREF_ENC tb64avxenc
+#define FUNPREF_DEC tb64avxdec
   #elif defined(__SSE4_1__)
 #include <smmintrin.h>
-#define FUNPREF tb64sse
+#define FUNPREF_ENC tb64sseenc
+#define FUNPREF_DEC tb64ssedec
   #elif defined(__SSSE3__)
     #ifdef __powerpc64__
 #define __SSE__   1
@@ -47,7 +49,8 @@ #define __SSE3__  1
 #define NO_WARN_X86_INTRINSICS 1
     #endif
-#define FUNPREF tb64sse
+#define FUNPREF_ENC tb64sseenc
+#define FUNPREF_DEC tb64ssedec
 #include <tmmintrin.h>
   #elif defined(__SSE2__)
 #include <emmintrin.h>
@@ -195,11 +198,11 @@ #elif defined(__SSSE3__) //----------------- SSSE3 / SSE4.1 / AVX (derived from the AVX2 functions ) -----------------------------------------------------------------
 
 #define OVD 4
-size_t TEMPLATE2(FUNPREF, dec)(const unsigned char *in, size_t inlen, unsigned char *out) {
+size_t FUNPREF_DEC(const unsigned char *in, size_t inlen, unsigned char *out) {
   if(inlen >= 16+OVD) {
     const unsigned char *ip;
           unsigned char *op; 
-      #ifdef __AVX__
+      #if defined(__AVX__) && !defined(NO_AVX)
     #define ND 64
       #else
     #define ND 32
@@ -241,7 +244,7 @@       CHECK1(MM_B64CHK(iv3, shifted3, check_asso, check_values, vx));
         #endif
     }
-      #ifdef __AVX__
+      #if defined(__AVX__) && !defined(NO_AVX)
     for(; ip < (in+inlen)-(16+OVD); ip += 16, op += (16/4)*3) {
       #else
     if(ip < (in+inlen)-(16+OVD)) {
@@ -250,7 +253,7 @@       __m128i ov0, shifted0; MM_MAP8TO6(iv0, shifted0,delta_asso, delta_values, ov0); MM_PACK8TO6(ov0, cpv);
       _mm_storeu_si128((__m128i*) op, ov0);                                                  
       CHECK1(MM_B64CHK(iv0, shifted0, check_asso, check_values, vx));
-        #ifndef __AVX__
+        #if defined(__AVX__) && !defined(NO_AVX)
       ip += 16; op += (16/4)*3;
         #endif
     }
@@ -262,11 +265,11 @@ }
 
 #define OVE 8
-size_t TEMPLATE2(FUNPREF, enc)(const unsigned char* in, size_t inlen, unsigned char *out) { 
+size_t FUNPREF_ENC(const unsigned char* in, size_t inlen, unsigned char *out) { 
   const unsigned char *ip = in; 
         unsigned char *op = out;
   size_t   outlen = TB64ENCLEN(inlen); 
-    #ifdef __AVX__
+    #if defined(__AVX__) && !defined(NO_AVX)
   #define NE 64
     #else
   #define NE 32