diff --git a/docs/c-api.rst b/docs/c-api.rst
--- a/docs/c-api.rst
+++ b/docs/c-api.rst
@@ -31,7 +31,7 @@
 changes to the configuration must be made *before* calling
 :c:func:`futhark_context_new`.  A configuration object must not be
 freed before any context objects for which it is used.  The same
-configuration may be used for multiple concurrent contexts.
+configuration may *not* be used for multiple concurrent contexts.
 
 .. c:struct:: futhark_context_config
 
@@ -65,6 +65,34 @@
 
    With a nonzero flag, print a running log to standard error of what
    the program is doing.
+
+.. c:function:: int futhark_context_config_set_tuning_param(struct futhark_context_config *cfg, const char *param_name, size_t new_value)
+
+   Set the value of a tuning parameter.  Returns zero on success, and
+   non-zero if the parameter cannot be set.  This is usually because a
+   parameter of the given name does not exist.  See
+   :c:func:`futhark_get_tuning_param_count` and
+   :c:func:`futhark_get_tuning_param_name` for how to query which
+   parameters are available.  Most of the tuning parameters are
+   applied only when the context is created, but some may be changed
+   even after the context is active.  At the moment, only parameters
+   of class "threshold" may change after the context has been created.
+   Use :c:func:`futhark_get_tuning_param_class` to determine the class
+   of a tuning parameter.
+
+.. c:function:: int futhark_get_tuning_param_count(void)
+
+   Return the number of available tuning parameters.  Useful for
+   knowing how to call :c:func:`futhark_get_tuning_param_name` and
+   :c:func:`futhark_get_tuning_param_class`.
+
+.. c:function:: const char* futhark_get_tuning_param_name(int i)
+
+   Return the name of tuning parameter *i*, counting from zero.
+
+.. c:function:: const char* futhark_get_tuning_param_class(int i)
+
+   Return the class of tuning parameter *i*, counting from zero.
 
 Context
 -------
diff --git a/docs/error-index.rst b/docs/error-index.rst
--- a/docs/error-index.rst
+++ b/docs/error-index.rst
@@ -169,3 +169,28 @@
 first, meaning that the size is available by the time ``a`` is bound.
 In many other cases, we can lift out the "size-producing" expressions
 into a separate ``let``-binding preceding the problematic expressions.
+
+Other errors
+------------
+
+.. _nested-entry:
+
+"Entry points may not be declared inside modules."
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+This occurs when the program uses the ``entry`` keyword inside a
+module::
+
+  module m = {
+    entry f x = x + 1
+  }
+
+Entry points can only be declared at the top level of a file.  When we
+wish to make a function from inside a module available as an entry
+point, we must define a wrapper function::
+
+  module m = {
+    let f x = x + 1
+  }
+
+  entry f = m.f
diff --git a/docs/language-reference.rst b/docs/language-reference.rst
--- a/docs/language-reference.rst
+++ b/docs/language-reference.rst
@@ -453,7 +453,7 @@
         : | `id`
    size : "[" `id` "]"
    pat:   `id`
-      : | `literal`
+      : | `pat_literal`
       : |  "_"
       : | "(" ")"
       : | "(" `pat` ")"
@@ -462,6 +462,11 @@
       : | "{" `fieldid` ["=" `pat`] ("," `fieldid` ["=" `pat`])* "}"
       : | `constructor` `pat`*
       : | `pat` ":" `type`
+   pat_literal: [ "-" ] `intnumber`
+              | [ "-" ] `floatnumber`
+              | `charlit`
+              | "true"
+              | "false"
    loopform :   "for" `id` "<" `exp`
             : | "for" `pat` "in" `exp`
             : | "while" `exp`
diff --git a/docs/man/futhark-bench.rst b/docs/man/futhark-bench.rst
--- a/docs/man/futhark-bench.rst
+++ b/docs/man/futhark-bench.rst
@@ -50,7 +50,8 @@
 --exclude-case=TAG
 
   Do not run test cases that contain the given tag.  Cases marked with
-  "nobench" or "disable" are ignored by default.
+  "nobench", "disable", or "no_foo" (where *foo* is the backend used)
+  are ignored by default.
 
 --futhark=program
 
diff --git a/docs/man/futhark-cuda.rst b/docs/man/futhark-cuda.rst
--- a/docs/man/futhark-cuda.rst
+++ b/docs/man/futhark-cuda.rst
@@ -144,15 +144,15 @@
   to the CUDA documentation for which options are supported.  Be
   careful - some options can easily result in invalid results.
 
---print-sizes
-
-  Print all sizes that can be set with ``-size`` or ``--tuning``.
-
---size=ASSIGNMENT
+--param=ASSIGNMENT
 
-  Set a configurable run-time parameter to the given
+  Set a tuning parameter to the given
   value. ``ASSIGNMENT`` must be of the form ``NAME=INT`` Use
   ``--print-sizes`` to see which names are available.
+
+--print-params
+
+  Print all tuning parameters that can be set with ``--param`` or ``--tuning``.
 
 --tuning=FILE
 
diff --git a/docs/man/futhark-opencl.rst b/docs/man/futhark-opencl.rst
--- a/docs/man/futhark-opencl.rst
+++ b/docs/man/futhark-opencl.rst
@@ -170,11 +170,16 @@
   end.  When ``-r`` is used, only the last run will be profiled.
   Implied by ``-D``.
 
---size=ASSIGNMENT
+--param=ASSIGNMENT
 
-  Set a configurable run-time parameter to the given
+  Set a tuning parameter to the given
   value. ``ASSIGNMENT`` must be of the form ``NAME=INT`` Use
   ``--print-sizes`` to see which names are available.
+
+--print-params
+
+  Print all tuning parameters that can be set with ``--param`` or
+  ``--tuning``.
 
 --tuning=FILE
 
diff --git a/docs/man/futhark-test.rst b/docs/man/futhark-test.rst
--- a/docs/man/futhark-test.rst
+++ b/docs/man/futhark-test.rst
@@ -144,7 +144,8 @@
 --exclude=tag
 
   Do not run test cases that contain the given tag.  Cases marked with
-  "disable" are ignored by default.
+  "disable" are ignored by default, as are cases marked "no_foo",
+  where *foo* is the backend used.
 
 -i
   Only interpret - do not run any compilers.
diff --git a/docs/server-protocol.rst b/docs/server-protocol.rst
--- a/docs/server-protocol.rst
+++ b/docs/server-protocol.rst
@@ -133,6 +133,11 @@
 
 Corresponds to :c:func:`futhark_context_report`.
 
+``set_tuning_param``
+....................
+
+Corresponds to :c:func:`futhark_context_config_set_tuning_param`.
+
 Environment Variables
 ---------------------
 
diff --git a/futhark.cabal b/futhark.cabal
--- a/futhark.cabal
+++ b/futhark.cabal
@@ -1,6 +1,6 @@
 cabal-version: 2.4
 name:           futhark
-version:        0.20.3
+version:        0.20.4
 synopsis:       An optimising compiler for a functional, array-oriented language.
 
 description:    Futhark is a small programming language designed to be compiled to
@@ -327,7 +327,7 @@
     , filepath >=1.4.1.1
     , free >=4.12.4
     , futhark-data >= 1.0.2.0
-    , futhark-server >= 1.1.0.0
+    , futhark-server >= 1.1.1.0
     , githash >=0.1.6.1
     , half >= 0.3
     , haskeline
diff --git a/rts/c/server.h b/rts/c/server.h
--- a/rts/c/server.h
+++ b/rts/c/server.h
@@ -2,10 +2,14 @@
 
 // Forward declarations of things that we technically don't know until
 // the application header file is included, but which we need.
+struct futhark_context_config;
 struct futhark_context;
 char *futhark_context_get_error(struct futhark_context *ctx);
 int futhark_context_sync(struct futhark_context *ctx);
 int futhark_context_clear_caches(struct futhark_context *ctx);
+int futhark_context_config_set_tuning_param(struct futhark_context_config *cfg,
+                                            const char *param_name,
+                                            size_t new_value);
 
 typedef int (*restore_fn)(const void*, FILE *, struct futhark_context*, void*);
 typedef void (*store_fn)(const void*, FILE *, struct futhark_context*, void*);
@@ -166,6 +170,7 @@
 
 struct server_state {
   struct futhark_prog prog;
+  struct futhark_context *cfg;
   struct futhark_context *ctx;
   int variables_capacity;
   struct variable *variables;
@@ -273,7 +278,9 @@
   if (err != 0) {
     failure();
     char *error = futhark_context_get_error(s->ctx);
-    puts(error);
+    if (error != NULL) {
+      puts(error);
+    }
     free(error);
   }
 }
@@ -524,6 +531,19 @@
   free(report);
 }
 
+void cmd_set_tuning_param(struct server_state *s, const char *args[]) {
+  const char *param = get_arg(args, 0);
+  const char *val_s = get_arg(args, 1);
+  size_t val = atol(val_s);
+  int err = futhark_context_config_set_tuning_param(s->cfg, param, val);
+
+  error_check(s, err);
+
+  if (err != 0) {
+    printf("Failed to set tuning parameter %s to %ld\n", param, (long)val);
+  }
+}
+
 char *next_word(char **line) {
   char *p = *line;
 
@@ -608,17 +628,22 @@
     cmd_unpause_profiling(s, tokens+1);
   } else if (strcmp(command, "report") == 0) {
     cmd_report(s, tokens+1);
+  } else if (strcmp(command, "set_tuning_param") == 0) {
+    cmd_set_tuning_param(s, tokens+1);
   } else {
     futhark_panic(1, "Unknown command: %s\n", command);
   }
 }
 
-void run_server(struct futhark_prog *prog, struct futhark_context *ctx) {
+void run_server(struct futhark_prog *prog,
+                struct futhark_context_config *cfg,
+                struct futhark_context *ctx) {
   char *line = NULL;
   size_t buflen = 0;
   ssize_t linelen;
 
   struct server_state s = {
+    .cfg = cfg,
     .ctx = ctx,
     .variables_capacity = 100,
     .prog = *prog
diff --git a/rts/c/tuning.h b/rts/c/tuning.h
--- a/rts/c/tuning.h
+++ b/rts/c/tuning.h
@@ -2,7 +2,7 @@
 
 static char* load_tuning_file(const char *fname,
                               void *cfg,
-                              int (*set_size)(void*, const char*, size_t)) {
+                              int (*set_tuning_param)(void*, const char*, size_t)) {
   const int max_line_len = 1024;
   char* line = (char*) malloc(max_line_len);
 
@@ -20,7 +20,7 @@
     if (eql) {
       *eql = 0;
       int value = atoi(eql+1);
-      if (set_size(cfg, line, (size_t)value) != 0) {
+      if (set_tuning_param(cfg, line, (size_t)value) != 0) {
         char* err = (char*) malloc(max_line_len + 50);
         snprintf(err, max_line_len + 50, "Unknown name '%s' on line %d.", line, lineno);
         free(line);
diff --git a/rts/python/opencl.py b/rts/python/opencl.py
--- a/rts/python/opencl.py
+++ b/rts/python/opencl.py
@@ -55,7 +55,7 @@
                 device_matches += 1
     raise Exception('No OpenCL platform and device matching constraints found.')
 
-def size_assignment(s):
+def param_assignment(s):
     name, value = s.split('=')
     return (name, int(value))
 
diff --git a/src/Futhark/CLI/Autotune.hs b/src/Futhark/CLI/Autotune.hs
--- a/src/Futhark/CLI/Autotune.hs
+++ b/src/Futhark/CLI/Autotune.hs
@@ -75,19 +75,29 @@
       val' <- readMaybe val
       return (thresh, val')
 
-type RunDataset = Int -> Path -> IO (Either String ([(String, Int)], Int))
+type RunDataset = Server -> Int -> Path -> IO (Either String ([(String, Int)], Int))
 
 type DatasetName = String
 
-serverOptions :: Path -> AutotuneOptions -> [String]
-serverOptions path opts =
+serverOptions :: AutotuneOptions -> [String]
+serverOptions opts =
   "--default-threshold" :
   show (optDefaultThreshold opts) :
   "-L" :
-  map opt path
-    ++ optExtraOptions opts
+  optExtraOptions opts
+
+setTuningParam :: Server -> String -> Int -> IO ()
+setTuningParam server name val =
+  either (error . T.unpack . T.unlines . failureMsg) (const $ pure ())
+    =<< cmdSetTuningParam server (T.pack name) (T.pack (show val))
+
+setTuningParams :: Server -> Path -> IO ()
+setTuningParams server = mapM_ (uncurry $ setTuningParam server)
+
+restoreTuningParams :: AutotuneOptions -> Server -> Path -> IO ()
+restoreTuningParams opts server = mapM_ opt
   where
-    opt (name, val) = "--size=" ++ name ++ "=" ++ show val
+    opt (name, _) = setTuningParam server name (optDefaultThreshold opts)
 
 prepare :: AutotuneOptions -> FutharkExe -> FilePath -> IO [(DatasetName, RunDataset, T.Text)]
 prepare opts futhark prog = do
@@ -116,7 +126,10 @@
         case runExpectedResult trun of
           Succeeds expected
             | null (runTags trun `intersect` ["notune", "disable"]) ->
-              Just (runDescription trun, run entry_point trun expected)
+              Just
+                ( runDescription trun,
+                  \server -> run server entry_point trun expected
+                )
           _ -> Nothing
 
   fmap concat . forM truns $ \ios -> do
@@ -125,7 +138,7 @@
     forM cases $ \(dataset, do_run) ->
       return (dataset, do_run, iosEntryPoint ios)
   where
-    run entry_point trun expected timeout path = do
+    run server entry_point trun expected timeout path = do
       let bestRuntime :: ([RunResult], T.Text) -> ([(String, Int)], Int)
           bestRuntime (runres, errout) =
             ( comparisons (T.unpack errout),
@@ -135,24 +148,26 @@
           ropts = runOptions timeout opts
 
       when (optVerbose opts > 1) $
-        putStrLn $ "Running with options: " ++ unwords (serverOptions path opts)
+        putStrLn $ "Trying path: " ++ show path
 
-      -- XXX: it is really inefficient to start a new server for every
-      -- run, but unfortunately we can only set threshold parameters
-      -- on startup.
-      let progbin = "." </> dropExtension prog
-      withServer (futharkServerCfg progbin (serverOptions path opts)) $ \server ->
-        either (Left . T.unpack) (Right . bestRuntime)
-          <$> benchmarkDataset
-            server
-            ropts
-            futhark
-            prog
-            entry_point
-            (runInput trun)
-            expected
-            (testRunReferenceOutput prog entry_point trun)
+      -- Setting the tuning parameters is a stateful action, so we
+      -- must be careful to restore the defaults below.  This is
+      -- because we rely on parameters not in 'path' to have their
+      -- default value.
+      setTuningParams server path
 
+      either (Left . T.unpack) (Right . bestRuntime)
+        <$> benchmarkDataset
+          server
+          ropts
+          futhark
+          prog
+          entry_point
+          (runInput trun)
+          expected
+          (testRunReferenceOutput prog entry_point trun)
+        <* restoreTuningParams opts server path
+
 --- Benchmarking a program
 
 data DatasetResult = DatasetResult [(String, Int)] Double
@@ -187,7 +202,7 @@
 thresholdForest prog = do
   thresholds <-
     getThresholds
-      <$> readProcess ("." </> dropExtension prog) ["--print-sizes"] ""
+      <$> readProcess ("." </> dropExtension prog) ["--print-params"] ""
   let root (v, _) = ((v, False), [])
   return $
     unfoldForest (unfold thresholds) $
@@ -226,11 +241,12 @@
 
 tuneThreshold ::
   AutotuneOptions ->
+  Server ->
   [(DatasetName, RunDataset, T.Text)] ->
   Path ->
   (String, Path) ->
   IO Path
-tuneThreshold opts datasets already_tuned (v, _v_path) = do
+tuneThreshold opts server datasets already_tuned (v, _v_path) = do
   tune_result <-
     foldM tuneDataset Nothing datasets
   case tune_result of
@@ -259,6 +275,7 @@
 
           sample_run <-
             run
+              server
               (optTimeout opts)
               ((v, maybe thresholdMax snd thresholds) : already_tuned)
 
@@ -281,7 +298,7 @@
 
                   runner :: Int -> Int -> IO (Maybe Int)
                   runner timeout' threshold = do
-                    res <- run timeout' ((v, threshold) : already_tuned)
+                    res <- run server timeout' ((v, threshold) : already_tuned)
                     case res of
                       Right (_, runTime) ->
                         return $ Just runTime
@@ -365,8 +382,12 @@
   when (optVerbose opts > 0) $
     putStrLn $ ("Threshold forest:\n" ++) $ drawForest $ map (fmap show) forest
 
-  foldM (tuneThreshold opts datasets) [] $ tuningPaths forest
+  let progbin = "." </> dropExtension prog
+  putStrLn $ "Running with options: " ++ unwords (serverOptions opts)
 
+  withServer (futharkServerCfg progbin (serverOptions opts)) $ \server ->
+    foldM (tuneThreshold opts server datasets) [] $ tuningPaths forest
+
 runAutotuner :: AutotuneOptions -> FilePath -> IO ()
 runAutotuner opts prog = do
   best <- tune opts prog
@@ -460,11 +481,7 @@
 
 -- | Run @futhark autotune@
 main :: String -> [String] -> IO ()
-main = mainWithOptions
-  initialAutotuneOptions
-  commandLineOptions
-  "options... program"
-  $ \progs config ->
-    case progs of
-      [prog] -> Just $ runAutotuner config prog
-      _ -> Nothing
+main = mainWithOptions initialAutotuneOptions commandLineOptions "options... program" $ \progs config ->
+  case progs of
+    [prog] -> Just $ runAutotuner config prog
+    _ -> Nothing
diff --git a/src/Futhark/CLI/Bench.hs b/src/Futhark/CLI/Bench.hs
--- a/src/Futhark/CLI/Bench.hs
+++ b/src/Futhark/CLI/Bench.hs
@@ -506,12 +506,16 @@
     max_timeout :: Int
     max_timeout = maxBound `div` 1000000
 
+excludeBackend :: BenchOptions -> BenchOptions
+excludeBackend config =
+  config {optExcludeCase = "no_" <> optBackend config : optExcludeCase config}
+
 -- | Run @futhark bench@.
 main :: String -> [String] -> IO ()
 main = mainWithOptions initialBenchOptions commandLineOptions "options... programs..." $ \progs config ->
   case progs of
     [] -> Nothing
-    _ -> Just $ runBenchmarks config progs
+    _ -> Just $ runBenchmarks (excludeBackend config) progs
 
 --- The following extracted from hstats package by Marshall Beddoe:
 --- https://hackage.haskell.org/package/hstats-0.3
diff --git a/src/Futhark/CLI/Test.hs b/src/Futhark/CLI/Test.hs
--- a/src/Futhark/CLI/Test.hs
+++ b/src/Futhark/CLI/Test.hs
@@ -739,9 +739,17 @@
       "Number of tests to run concurrently."
   ]
 
+excludeBackend :: TestConfig -> TestConfig
+excludeBackend config =
+  config
+    { configExclude =
+        "no_" <> T.pack (configBackend (configPrograms config)) :
+        configExclude config
+    }
+
 -- | Run @futhark test@.
 main :: String -> [String] -> IO ()
 main = mainWithOptions defaultConfig commandLineOptions "options... programs..." $ \progs config ->
   case progs of
     [] -> Nothing
-    _ -> Just $ runTests config progs
+    _ -> Just $ runTests (excludeBackend config) progs
diff --git a/src/Futhark/CodeGen/Backends/CCUDA.hs b/src/Futhark/CodeGen/Backends/CCUDA.hs
--- a/src/Futhark/CodeGen/Backends/CCUDA.hs
+++ b/src/Futhark/CodeGen/Backends/CCUDA.hs
@@ -248,10 +248,10 @@
 
 callKernel :: GC.OpCompiler OpenCL ()
 callKernel (GetSize v key) =
-  GC.stm [C.cstm|$id:v = ctx->sizes.$id:key;|]
+  GC.stm [C.cstm|$id:v = *ctx->tuning_params.$id:key;|]
 callKernel (CmpSizeLe v key x) = do
   x' <- GC.compileExp x
-  GC.stm [C.cstm|$id:v = ctx->sizes.$id:key <= $exp:x';|]
+  GC.stm [C.cstm|$id:v = *ctx->tuning_params.$id:key <= $exp:x';|]
   sizeLoggingCode v key x'
 callKernel (GetSizeMax v size_class) =
   let field = "max_" ++ cudaSizeClass size_class
diff --git a/src/Futhark/CodeGen/Backends/CCUDA/Boilerplate.hs b/src/Futhark/CodeGen/Backends/CCUDA/Boilerplate.hs
--- a/src/Futhark/CodeGen/Backends/CCUDA/Boilerplate.hs
+++ b/src/Futhark/CodeGen/Backends/CCUDA/Boilerplate.hs
@@ -92,27 +92,27 @@
       size_var_inits = map (\k -> [C.cinit|$string:(zEncodeString (pretty k))|]) $ M.keys sizes
       size_class_inits = map (\c -> [C.cinit|$string:(pretty c)|]) $ M.elems sizes
 
-  GC.earlyDecl [C.cedecl|static const char *size_names[] = { $inits:size_name_inits };|]
-  GC.earlyDecl [C.cedecl|static const char *size_vars[] = { $inits:size_var_inits };|]
-  GC.earlyDecl [C.cedecl|static const char *size_classes[] = { $inits:size_class_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_names[] = { $inits:size_name_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_vars[] = { $inits:size_var_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_classes[] = { $inits:size_class_inits };|]
 
 generateConfigFuns :: M.Map Name SizeClass -> GC.CompilerM OpenCL () String
 generateConfigFuns sizes = do
-  let size_decls = map (\k -> [C.csdecl|typename int64_t $id:k;|]) $ M.keys sizes
+  let size_decls = map (\k -> [C.csdecl|typename int64_t *$id:k;|]) $ M.keys sizes
       num_sizes = M.size sizes
-  GC.earlyDecl [C.cedecl|struct sizes { $sdecls:size_decls };|]
+  GC.earlyDecl [C.cedecl|struct tuning_params { $sdecls:size_decls };|]
   cfg <- GC.publicDef "context_config" GC.InitDecl $ \s ->
     ( [C.cedecl|struct $id:s;|],
       [C.cedecl|struct $id:s { struct cuda_config cu_cfg;
                               int profiling;
-                              typename int64_t sizes[$int:num_sizes];
+                              typename int64_t tuning_params[$int:num_sizes];
                               int num_nvrtc_opts;
                               const char **nvrtc_opts;
                             };|]
     )
 
   let size_value_inits = zipWith sizeInit [0 .. M.size sizes -1] (M.elems sizes)
-      sizeInit i size = [C.cstm|cfg->sizes[$int:i] = $int:val;|]
+      sizeInit i size = [C.cstm|cfg->tuning_params[$int:i] = $int:val;|]
         where
           val = fromMaybe 0 $ sizeDefault size
   GC.publicDef_ "context_config_new" GC.InitDecl $ \s ->
@@ -129,8 +129,8 @@
                          cfg->nvrtc_opts[0] = NULL;
                          $stms:size_value_inits
                          cuda_config_init(&cfg->cu_cfg, $int:num_sizes,
-                                          size_names, size_vars,
-                                          cfg->sizes, size_classes);
+                                          tuning_param_names, tuning_param_vars,
+                                          cfg->tuning_params, tuning_param_classes);
                          return cfg;
                        }|]
     )
@@ -247,39 +247,39 @@
                        }|]
     )
 
-  GC.publicDef_ "context_config_set_size" GC.InitDecl $ \s ->
-    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value);|],
-      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value) {
+  GC.publicDef_ "context_config_set_tuning_param" GC.InitDecl $ \s ->
+    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t new_value);|],
+      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t new_value) {
 
                          for (int i = 0; i < $int:num_sizes; i++) {
-                           if (strcmp(size_name, size_names[i]) == 0) {
-                             cfg->sizes[i] = size_value;
+                           if (strcmp(param_name, tuning_param_names[i]) == 0) {
+                             cfg->tuning_params[i] = new_value;
                              return 0;
                            }
                          }
 
-                         if (strcmp(size_name, "default_group_size") == 0) {
-                           cfg->cu_cfg.default_block_size = size_value;
+                         if (strcmp(param_name, "default_group_size") == 0) {
+                           cfg->cu_cfg.default_block_size = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_num_groups") == 0) {
-                           cfg->cu_cfg.default_grid_size = size_value;
+                         if (strcmp(param_name, "default_num_groups") == 0) {
+                           cfg->cu_cfg.default_grid_size = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_threshold") == 0) {
-                           cfg->cu_cfg.default_threshold = size_value;
+                         if (strcmp(param_name, "default_threshold") == 0) {
+                           cfg->cu_cfg.default_threshold = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_tile_size") == 0) {
-                           cfg->cu_cfg.default_tile_size = size_value;
+                         if (strcmp(param_name, "default_tile_size") == 0) {
+                           cfg->cu_cfg.default_tile_size = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_reg_tile_size") == 0) {
-                           cfg->cu_cfg.default_reg_tile_size = size_value;
+                         if (strcmp(param_name, "default_reg_tile_size") == 0) {
+                           cfg->cu_cfg.default_reg_tile_size = new_value;
                            return 0;
                          }
 
@@ -337,7 +337,7 @@
                          typename CUdeviceptr global_failure;
                          typename CUdeviceptr global_failure_args;
                          struct cuda_context cuda;
-                         struct sizes sizes;
+                         struct tuning_params tuning_params;
                          // True if a potentially failing kernel has been enqueued.
                          typename int32_t failure_is_an_option;
 
@@ -346,9 +346,9 @@
                        };|]
     )
 
-  let set_sizes =
+  let set_tuning_params =
         zipWith
-          (\i k -> [C.cstm|ctx->sizes.$id:k = cfg->sizes[$int:i];|])
+          (\i k -> [C.cstm|ctx->tuning_params.$id:k = &cfg->tuning_params[$int:i];|])
           [(0 :: Int) ..]
           $ M.keys sizes
       max_failure_args =
@@ -396,7 +396,7 @@
                  $stms:init_kernel_fields
 
                  $stms:final_inits
-                 $stms:set_sizes
+                 $stms:set_tuning_params
 
                  init_constants(ctx);
                  // Clear the free list of any deallocations that occurred while initialising constants.
diff --git a/src/Futhark/CodeGen/Backends/COpenCL.hs b/src/Futhark/CodeGen/Backends/COpenCL.hs
--- a/src/Futhark/CodeGen/Backends/COpenCL.hs
+++ b/src/Futhark/CodeGen/Backends/COpenCL.hs
@@ -315,10 +315,10 @@
 
 callKernel :: GC.OpCompiler OpenCL ()
 callKernel (GetSize v key) =
-  GC.stm [C.cstm|$id:v = ctx->sizes.$id:key;|]
+  GC.stm [C.cstm|$id:v = *ctx->tuning_params.$id:key;|]
 callKernel (CmpSizeLe v key x) = do
   x' <- GC.compileExp x
-  GC.stm [C.cstm|$id:v = ctx->sizes.$id:key <= $exp:x';|]
+  GC.stm [C.cstm|$id:v = *ctx->tuning_params.$id:key <= $exp:x';|]
   sizeLoggingCode v key x'
 callKernel (GetSizeMax v size_class) =
   let field = "max_" ++ pretty size_class
diff --git a/src/Futhark/CodeGen/Backends/COpenCL/Boilerplate.hs b/src/Futhark/CodeGen/Backends/COpenCL/Boilerplate.hs
--- a/src/Futhark/CodeGen/Backends/COpenCL/Boilerplate.hs
+++ b/src/Futhark/CodeGen/Backends/COpenCL/Boilerplate.hs
@@ -92,23 +92,23 @@
       size_class_inits = map (\c -> [C.cinit|$string:(pretty c)|]) $ M.elems sizes
       num_sizes = M.size sizes
 
-  GC.earlyDecl [C.cedecl|static const char *size_names[] = { $inits:size_name_inits };|]
-  GC.earlyDecl [C.cedecl|static const char *size_vars[] = { $inits:size_var_inits };|]
-  GC.earlyDecl [C.cedecl|static const char *size_classes[] = { $inits:size_class_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_names[] = { $inits:size_name_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_vars[] = { $inits:size_var_inits };|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_classes[] = { $inits:size_class_inits };|]
 
-  let size_decls = map (\k -> [C.csdecl|typename int64_t $id:k;|]) $ M.keys sizes
-  GC.earlyDecl [C.cedecl|struct sizes { $sdecls:size_decls };|]
+  let size_decls = map (\k -> [C.csdecl|typename int64_t *$id:k;|]) $ M.keys sizes
+  GC.earlyDecl [C.cedecl|struct tuning_params { $sdecls:size_decls };|]
   cfg <- GC.publicDef "context_config" GC.InitDecl $ \s ->
     ( [C.cedecl|struct $id:s;|],
       [C.cedecl|struct $id:s { struct opencl_config opencl;
-                              typename int64_t sizes[$int:num_sizes];
+                              typename int64_t tuning_params[$int:num_sizes];
                               int num_build_opts;
                               const char **build_opts;
                             };|]
     )
 
   let size_value_inits = zipWith sizeInit [0 .. M.size sizes -1] (M.elems sizes)
-      sizeInit i size = [C.cstm|cfg->sizes[$int:i] = $int:val;|]
+      sizeInit i size = [C.cstm|cfg->tuning_params[$int:i] = $int:val;|]
         where
           val = fromMaybe 0 $ sizeDefault size
   GC.publicDef_ "context_config_new" GC.InitDecl $ \s ->
@@ -124,8 +124,8 @@
                          cfg->build_opts[0] = NULL;
                          $stms:size_value_inits
                          opencl_config_init(&cfg->opencl, $int:num_sizes,
-                                            size_names, size_vars,
-                                            cfg->sizes, size_classes);
+                                            tuning_param_names, tuning_param_vars,
+                                            cfg->tuning_params, tuning_param_classes);
                          return cfg;
                        }|]
     )
@@ -263,39 +263,39 @@
                        }|]
     )
 
-  GC.publicDef_ "context_config_set_size" GC.InitDecl $ \s ->
-    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value);|],
-      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value) {
+  GC.publicDef_ "context_config_set_tuning_param" GC.InitDecl $ \s ->
+    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t new_value);|],
+      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t new_value) {
 
                          for (int i = 0; i < $int:num_sizes; i++) {
-                           if (strcmp(size_name, size_names[i]) == 0) {
-                             cfg->sizes[i] = size_value;
+                           if (strcmp(param_name, tuning_param_names[i]) == 0) {
+                             cfg->tuning_params[i] = new_value;
                              return 0;
                            }
                          }
 
-                         if (strcmp(size_name, "default_group_size") == 0) {
-                           cfg->opencl.default_group_size = size_value;
+                         if (strcmp(param_name, "default_group_size") == 0) {
+                           cfg->opencl.default_group_size = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_num_groups") == 0) {
-                           cfg->opencl.default_num_groups = size_value;
+                         if (strcmp(param_name, "default_num_groups") == 0) {
+                           cfg->opencl.default_num_groups = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_threshold") == 0) {
-                           cfg->opencl.default_threshold = size_value;
+                         if (strcmp(param_name, "default_threshold") == 0) {
+                           cfg->opencl.default_threshold = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_tile_size") == 0) {
-                           cfg->opencl.default_tile_size = size_value;
+                         if (strcmp(param_name, "default_tile_size") == 0) {
+                           cfg->opencl.default_tile_size = new_value;
                            return 0;
                          }
 
-                         if (strcmp(size_name, "default_reg_tile_size") == 0) {
-                           cfg->opencl.default_reg_tile_size = size_value;
+                         if (strcmp(param_name, "default_reg_tile_size") == 0) {
+                           cfg->opencl.default_reg_tile_size = new_value;
                            return 0;
                          }
 
@@ -320,7 +320,7 @@
                          typename cl_mem global_failure;
                          typename cl_mem global_failure_args;
                          struct opencl_context opencl;
-                         struct sizes sizes;
+                         struct tuning_params tuning_params;
                          // True if a potentially failing kernel has been enqueued.
                          typename cl_int failure_is_an_option;
                        };|]
@@ -350,9 +350,9 @@
                      $stms:ctx_opencl_inits
   }|]
 
-  let set_sizes =
+  let set_tuning_params =
         zipWith
-          (\i k -> [C.cstm|ctx->sizes.$id:k = cfg->sizes[$int:i];|])
+          (\i k -> [C.cstm|ctx->tuning_params.$id:k = &cfg->tuning_params[$int:i];|])
           [(0 :: Int) ..]
           $ M.keys sizes
       max_failure_args =
@@ -380,7 +380,7 @@
                      $stms:(map loadKernel (M.toList kernels))
 
                      $stms:final_inits
-                     $stms:set_sizes
+                     $stms:set_tuning_params
 
                      init_constants(ctx);
                      // Clear the free list of any deallocations that occurred while initialising constants.
diff --git a/src/Futhark/CodeGen/Backends/GenericC.hs b/src/Futhark/CodeGen/Backends/GenericC.hs
--- a/src/Futhark/CodeGen/Backends/GenericC.hs
+++ b/src/Futhark/CodeGen/Backends/GenericC.hs
@@ -1639,24 +1639,24 @@
   ops <- asks envOperations
   profilereport <- gets $ DL.toList . compProfileItems
 
-  publicDef_ "get_num_sizes" InitDecl $ \s ->
+  publicDef_ "get_tuning_param_count" InitDecl $ \s ->
     ( [C.cedecl|int $id:s(void);|],
       [C.cedecl|int $id:s(void) {
-                return sizeof(size_names)/sizeof(size_names[0]);
+                return sizeof(tuning_param_names)/sizeof(tuning_param_names[0]);
               }|]
     )
 
-  publicDef_ "get_size_name" InitDecl $ \s ->
+  publicDef_ "get_tuning_param_name" InitDecl $ \s ->
     ( [C.cedecl|const char* $id:s(int);|],
       [C.cedecl|const char* $id:s(int i) {
-                return size_names[i];
+                return tuning_param_names[i];
               }|]
     )
 
-  publicDef_ "get_size_class" InitDecl $ \s ->
+  publicDef_ "get_tuning_param_class" InitDecl $ \s ->
     ( [C.cedecl|const char* $id:s(int);|],
       [C.cedecl|const char* $id:s(int i) {
-                return size_classes[i];
+                return tuning_param_classes[i];
               }|]
     )
 
diff --git a/src/Futhark/CodeGen/Backends/GenericC/CLI.hs b/src/Futhark/CodeGen/Backends/GenericC/CLI.hs
--- a/src/Futhark/CodeGen/Backends/GenericC/CLI.hs
+++ b/src/Futhark/CodeGen/Backends/GenericC/CLI.hs
@@ -93,25 +93,25 @@
                   }|]
       },
     Option
-      { optionLongName = "print-sizes",
+      { optionLongName = "print-params",
         optionShortName = Nothing,
         optionArgument = NoArgument,
-        optionDescription = "Print all sizes that can be set with --size or --tuning.",
+        optionDescription = "Print all tuning parameters that can be set with --param or --tuning.",
         optionAction =
           [C.cstm|{
-                int n = futhark_get_num_sizes();
+                int n = futhark_get_tuning_param_count();
                 for (int i = 0; i < n; i++) {
-                  printf("%s (%s)\n", futhark_get_size_name(i),
-                                      futhark_get_size_class(i));
+                  printf("%s (%s)\n", futhark_get_tuning_param_name(i),
+                                      futhark_get_tuning_param_class(i));
                 }
                 exit(0);
               }|]
       },
     Option
-      { optionLongName = "size",
+      { optionLongName = "param",
         optionShortName = Nothing,
         optionArgument = RequiredArgument "ASSIGNMENT",
-        optionDescription = "Set a configurable run-time parameter to the given value.",
+        optionDescription = "Set a tuning parameter to the given value.",
         optionAction =
           [C.cstm|{
                 char *name = optarg;
@@ -120,7 +120,7 @@
                 int value = atoi(value_str);
                 if (equals != NULL) {
                   *equals = 0;
-                  if (futhark_context_config_set_size(cfg, name, (size_t)value) != 0) {
+                  if (futhark_context_config_set_tuning_param(cfg, name, (size_t)value) != 0) {
                     futhark_panic(1, "Unknown size: %s\n", name);
                   }
                 } else {
@@ -135,7 +135,7 @@
         optionAction =
           [C.cstm|{
                 char *ret = load_tuning_file(optarg, cfg, (int(*)(void*, const char*, size_t))
-                                                          futhark_context_config_set_size);
+                                                          futhark_context_config_set_tuning_param);
                 if (ret != NULL) {
                   futhark_panic(1, "When loading tuning from '%s': %s\n", optarg, ret);
                 }}|]
diff --git a/src/Futhark/CodeGen/Backends/GenericC/Server.hs b/src/Futhark/CodeGen/Backends/GenericC/Server.hs
--- a/src/Futhark/CodeGen/Backends/GenericC/Server.hs
+++ b/src/Futhark/CodeGen/Backends/GenericC/Server.hs
@@ -53,25 +53,25 @@
                   }|]
       },
     Option
-      { optionLongName = "print-sizes",
+      { optionLongName = "print-params",
         optionShortName = Nothing,
         optionArgument = NoArgument,
-        optionDescription = "Print all sizes that can be set with --size or --tuning.",
+        optionDescription = "Print all tuning parameters that can be set with --param or --tuning.",
         optionAction =
           [C.cstm|{
-                int n = futhark_get_num_sizes();
+                int n = futhark_get_tuning_param_count();
                 for (int i = 0; i < n; i++) {
-                  printf("%s (%s)\n", futhark_get_size_name(i),
-                                      futhark_get_size_class(i));
+                  printf("%s (%s)\n", futhark_get_tuning_param_name(i),
+                                      futhark_get_tuning_param_class(i));
                 }
                 exit(0);
               }|]
       },
     Option
-      { optionLongName = "size",
+      { optionLongName = "param",
         optionShortName = Nothing,
         optionArgument = RequiredArgument "ASSIGNMENT",
-        optionDescription = "Set a configurable run-time parameter to the given value.",
+        optionDescription = "Set a tuning parameter to the given value.",
         optionAction =
           [C.cstm|{
                 char *name = optarg;
@@ -80,7 +80,7 @@
                 int value = atoi(value_str);
                 if (equals != NULL) {
                   *equals = 0;
-                  if (futhark_context_config_set_size(cfg, name, value) != 0) {
+                  if (futhark_context_config_set_tuning_param(cfg, name, value) != 0) {
                     futhark_panic(1, "Unknown size: %s\n", name);
                   }
                 } else {
@@ -95,7 +95,7 @@
         optionAction =
           [C.cstm|{
                 char *ret = load_tuning_file(optarg, cfg, (int(*)(void*, const char*, size_t))
-                                                          futhark_context_config_set_size);
+                                                          futhark_context_config_set_tuning_param);
                 if (ret != NULL) {
                   futhark_panic(1, "When loading tuning from '%s': %s\n", optarg, ret);
                 }}|]
@@ -326,7 +326,7 @@
   }
 
   if (entry_point != NULL) {
-    run_server(&prog, ctx);
+    run_server(&prog, cfg, ctx);
   }
 
   futhark_context_free(ctx);
diff --git a/src/Futhark/CodeGen/Backends/MulticoreC.hs b/src/Futhark/CodeGen/Backends/MulticoreC.hs
--- a/src/Futhark/CodeGen/Backends/MulticoreC.hs
+++ b/src/Futhark/CodeGen/Backends/MulticoreC.hs
@@ -195,14 +195,14 @@
                            }|]
     )
 
-  GC.earlyDecl [C.cedecl|static const char *size_names[0];|]
-  GC.earlyDecl [C.cedecl|static const char *size_vars[0];|]
-  GC.earlyDecl [C.cedecl|static const char *size_classes[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_names[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_vars[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_classes[0];|]
 
-  GC.publicDef_ "context_config_set_size" GC.InitDecl $ \s ->
-    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value);|],
-      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value) {
-                     (void)cfg; (void)size_name; (void)size_value;
+  GC.publicDef_ "context_config_set_tuning_param" GC.InitDecl $ \s ->
+    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t param_value);|],
+      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t param_value) {
+                     (void)cfg; (void)param_name; (void)param_value;
                      return 1;
                    }|]
     )
diff --git a/src/Futhark/CodeGen/Backends/PyOpenCL.hs b/src/Futhark/CodeGen/Backends/PyOpenCL.hs
--- a/src/Futhark/CodeGen/Backends/PyOpenCL.hs
+++ b/src/Futhark/CodeGen/Backends/PyOpenCL.hs
@@ -140,13 +140,13 @@
                 [Assign (Var "default_reg_tile_size") $ Var "optarg"]
             },
           Option
-            { optionLongName = "size",
+            { optionLongName = "param",
               optionShortName = Nothing,
-              optionArgument = RequiredArgument "size_assignment",
+              optionArgument = RequiredArgument "param_assignment",
               optionAction =
                 [ Assign
                     ( Index
-                        (Var "sizes")
+                        (Var "params")
                         ( IdxExp
                             ( Index
                                 (Var "optarg")
diff --git a/src/Futhark/CodeGen/Backends/SequentialC/Boilerplate.hs b/src/Futhark/CodeGen/Backends/SequentialC/Boilerplate.hs
--- a/src/Futhark/CodeGen/Backends/SequentialC/Boilerplate.hs
+++ b/src/Futhark/CodeGen/Backends/SequentialC/Boilerplate.hs
@@ -38,6 +38,13 @@
                         }|]
     )
 
+  GC.publicDef_ "context_config_set_profiling" GC.InitDecl $ \s ->
+    ( [C.cedecl|void $id:s(struct $id:cfg* cfg, int flag);|],
+      [C.cedecl|void $id:s(struct $id:cfg* cfg, int flag) {
+                         (void)cfg; (void)flag;
+                       }|]
+    )
+
   GC.publicDef_ "context_config_set_logging" GC.InitDecl $ \s ->
     ( [C.cedecl|void $id:s(struct $id:cfg* cfg, int flag);|],
       [C.cedecl|void $id:s(struct $id:cfg* cfg, int detail) {
@@ -100,14 +107,14 @@
                                }|]
     )
 
-  GC.earlyDecl [C.cedecl|static const char *size_names[0];|]
-  GC.earlyDecl [C.cedecl|static const char *size_vars[0];|]
-  GC.earlyDecl [C.cedecl|static const char *size_classes[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_names[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_vars[0];|]
+  GC.earlyDecl [C.cedecl|static const char *tuning_param_classes[0];|]
 
-  GC.publicDef_ "context_config_set_size" GC.InitDecl $ \s ->
-    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value);|],
-      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *size_name, size_t size_value) {
-                         (void)cfg; (void)size_name; (void)size_value;
+  GC.publicDef_ "context_config_set_tuning_param" GC.InitDecl $ \s ->
+    ( [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t param_value);|],
+      [C.cedecl|int $id:s(struct $id:cfg* cfg, const char *param_name, size_t param_value) {
+                         (void)cfg; (void)param_name; (void)param_value;
                          return 1;
                        }|]
     )
diff --git a/src/Futhark/Compiler/Program.hs b/src/Futhark/Compiler/Program.hs
--- a/src/Futhark/Compiler/Program.hs
+++ b/src/Futhark/Compiler/Program.hs
@@ -204,7 +204,7 @@
   [FilePath] ->
   m (E.Warnings, Imports, VNameSource)
 readLibrary extra_eps fps =
-  typeCheckProgram emptyBasis . setEntryPoints extra_eps fps
+  typeCheckProgram emptyBasis . setEntryPoints (E.defaultEntryPoint : extra_eps) fps
     =<< readUntypedLibrary fps
 
 -- | Read and type-check Futhark imports (no @.fut@ extension; may
diff --git a/src/Futhark/IR/Pretty.hs b/src/Futhark/IR/Pretty.hs
--- a/src/Futhark/IR/Pretty.hs
+++ b/src/Futhark/IR/Pretty.hs
@@ -10,6 +10,7 @@
 -- the interface from 'Pretty'.
 module Futhark.IR.Pretty
   ( prettyTuple,
+    prettyTupleLines,
     pretty,
     PrettyRep (..),
     ppTuple',
@@ -305,7 +306,7 @@
   ppr (Lambda [] (Body _ stms []) []) | stms == mempty = text "nilFn"
   ppr (Lambda params body rettype) =
     text "\\" <+> ppTuple' params
-      <+/> colon <+> ppTuple' rettype <+> text "->"
+      <+/> colon <+> ppTupleLines' rettype <+> text "->"
       </> indent 2 (ppr body)
 
 instance Pretty EntryPointType where
@@ -351,3 +352,7 @@
 -- | Like 'prettyTuple', but produces a 'Doc'.
 ppTuple' :: Pretty a => [a] -> Doc
 ppTuple' ets = braces $ commasep $ map (align . ppr) ets
+
+-- | Like 'prettyTupleLines', but produces a 'Doc'.
+ppTupleLines' :: Pretty a => [a] -> Doc
+ppTupleLines' ets = braces $ stack $ punctuate comma $ map (align . ppr) ets
diff --git a/src/Futhark/IR/Primitive.hs b/src/Futhark/IR/Primitive.hs
--- a/src/Futhark/IR/Primitive.hs
+++ b/src/Futhark/IR/Primitive.hs
@@ -178,7 +178,7 @@
   = IntType IntType
   | FloatType FloatType
   | Bool
-  | -- | An informationless type - still takes up space!
+  | -- | An informationless type - An array of this type takes up no space.
     Unit
   deriving (Eq, Ord, Show)
 
diff --git a/src/Futhark/Optimise/InPlaceLowering/LowerIntoStm.hs b/src/Futhark/Optimise/InPlaceLowering/LowerIntoStm.hs
--- a/src/Futhark/Optimise/InPlaceLowering/LowerIntoStm.hs
+++ b/src/Futhark/Optimise/InPlaceLowering/LowerIntoStm.hs
@@ -150,7 +150,7 @@
               traverse (toSubExp "index") $
                 fixSlice (fmap pe64 slice) $ map (pe64 . Var) gtids
 
-          let res_dims = arrayDims $ snd bindee_dec
+          let res_dims = take (length slice') $ arrayDims $ snd bindee_dec
               ret' = WriteReturns cs (Shape res_dims) src [(Slice $ map DimFix slice', se)]
 
           v_aliased <- newName v
diff --git a/src/Futhark/Pass/ExtractKernels/Interchange.hs b/src/Futhark/Pass/ExtractKernels/Interchange.hs
--- a/src/Futhark/Pass/ExtractKernels/Interchange.hs
+++ b/src/Futhark/Pass/ExtractKernels/Interchange.hs
@@ -205,8 +205,7 @@
       auxing map_aux . fmap subExpsRes . letTupExp' "withacc_inter" $
         Op $ Screma w (iota_w : map paramName acc_params ++ arrs) (mapSOAC maplam)
     let pat = Pat $ rearrangeShape perm $ patElems map_pat
-        perm' = [0 .. patSize pat -1]
-    pure $ WithAccStm perm' pat inputs' acc_lam'
+    pure $ WithAccStm perm pat inputs' acc_lam'
     where
       newAccLamParams ps = do
         let (cert_ps, acc_ps) = splitAt (length ps `div` 2) ps
diff --git a/src/Futhark/TypeCheck.hs b/src/Futhark/TypeCheck.hs
--- a/src/Futhark/TypeCheck.hs
+++ b/src/Futhark/TypeCheck.hs
@@ -122,7 +122,7 @@
   show (InvalidPatError pat t desc) =
     "Pat\n" ++ pretty pat
       ++ "\ncannot match value of type\n"
-      ++ prettyTuple t
+      ++ prettyTupleLines t
       ++ end
     where
       end = case desc of
diff --git a/src/Futhark/Util/Pretty.hs b/src/Futhark/Util/Pretty.hs
--- a/src/Futhark/Util/Pretty.hs
+++ b/src/Futhark/Util/Pretty.hs
@@ -7,6 +7,7 @@
     pretty,
     prettyDoc,
     prettyTuple,
+    prettyTupleLines,
     prettyText,
     prettyTextOneLine,
     prettyOneLine,
@@ -53,6 +54,12 @@
 -- | Prettyprint a list enclosed in curly braces.
 prettyTuple :: Pretty a => [a] -> String
 prettyTuple = PP.pretty 80 . ppTuple'
+
+-- | Like 'prettyTuple', but put a linebreak after every element.
+prettyTupleLines :: Pretty a => [a] -> String
+prettyTupleLines = PP.pretty 80 . ppTupleLines'
+  where
+    ppTupleLines' ets = braces $ stack $ punctuate comma $ map (align . ppr) ets
 
 -- | The document @'apply' ds@ separates @ds@ with commas and encloses them with
 -- parentheses.
diff --git a/src/Language/Futhark/Parser/Parser.y b/src/Language/Futhark/Parser/Parser.y
--- a/src/Language/Futhark/Parser/Parser.y
+++ b/src/Language/Futhark/Parser/Parser.y
@@ -389,7 +389,7 @@
 Val    :: { ValBindBase NoInfo Name }
 Val     : let BindingId TypeParams FunParams maybeAscription(TypeExpDecl) '=' Exp
           { let (name, _) = $2
-            in ValBind (if name==defaultEntryPoint then Just NoInfo else Nothing) name (fmap declaredType $5) NoInfo
+            in ValBind Nothing name (fmap declaredType $5) NoInfo
                $3 $4 $7 Nothing mempty (srcspan $1 $>)
           }
 
@@ -661,11 +661,8 @@
        { IndexSection $4 NoInfo (srcspan $1 $>) }
 
 
-PrimLit :: { (PrimValue, SrcLoc) }
-        : true   { (BoolValue True, $1) }
-        | false  { (BoolValue False, $1) }
-
-        | i8lit   { let L loc (I8LIT num)  = $1 in (SignedValue $ Int8Value num, loc) }
+NumLit :: { (PrimValue, SrcLoc) }
+        : i8lit   { let L loc (I8LIT num)  = $1 in (SignedValue $ Int8Value num, loc) }
         | i16lit  { let L loc (I16LIT num) = $1 in (SignedValue $ Int16Value num, loc) }
         | i32lit  { let L loc (I32LIT num) = $1 in (SignedValue $ Int32Value num, loc) }
         | i64lit  { let L loc (I64LIT num) = $1 in (SignedValue $ Int64Value num, loc) }
@@ -679,6 +676,12 @@
         | f32lit { let L loc (F32LIT num) = $1 in (FloatValue $ Float32Value num, loc) }
         | f64lit { let L loc (F64LIT num) = $1 in (FloatValue $ Float64Value num, loc) }
 
+
+PrimLit :: { (PrimValue, SrcLoc) }
+        : true   { (BoolValue True, $1) }
+        | false  { (BoolValue False, $1) }
+        | NumLit { $1 }
+
 Exps1 :: { (UncheckedExp, [UncheckedExp]) }
        : Exps1_ { case reverse (snd $1 : fst $1) of
                     []   -> (snd $1, [])
@@ -789,11 +792,14 @@
                  | CFieldPat                    { [$1] }
 
 CaseLiteral :: { (PatLit, SrcLoc) }
-             : PrimLit  { (PatLitPrim (fst $1), snd $1) }
-             | charlit  { let L loc (CHARLIT x) = $1
+             : charlit  { let L loc (CHARLIT x) = $1
                           in (PatLitInt (toInteger (ord x)), loc) }
+             | PrimLit  { (PatLitPrim (fst $1), snd $1) }
              | intlit   { let L loc (INTLIT x) = $1 in (PatLitInt x, loc) }
              | floatlit { let L loc (FLOATLIT x) = $1 in (PatLitFloat x, loc) }
+             | '-' NumLit  { (PatLitPrim (primNegate (fst $2)), snd $2) }
+             | '-' intlit   { let L loc (INTLIT x) = $2 in (PatLitInt (negate x), loc) }
+             | '-' floatlit { let L loc (FLOATLIT x) = $2 in (PatLitFloat (negate x), loc) }
 
 LoopForm :: { LoopFormBase NoInfo Name }
 LoopForm : for VarId '<' Exp
@@ -1130,6 +1136,12 @@
 floatNegate (Float16Value v) = Float16Value (-v)
 floatNegate (Float32Value v) = Float32Value (-v)
 floatNegate (Float64Value v) = Float64Value (-v)
+
+primNegate :: PrimValue -> PrimValue
+primNegate (FloatValue v) = FloatValue $ floatNegate v
+primNegate (SignedValue v) = SignedValue $ intNegate v
+primNegate (UnsignedValue v) = UnsignedValue $ intNegate v
+primNegate (BoolValue v) = BoolValue $ not v
 
 readLine :: ParserMonad (Maybe T.Text)
 readLine = lift $ lift $ lift readLineFromMonad
diff --git a/src/Language/Futhark/TypeChecker.hs b/src/Language/Futhark/TypeChecker.hs
--- a/src/Language/Futhark/TypeChecker.hs
+++ b/src/Language/Futhark/TypeChecker.hs
@@ -460,7 +460,7 @@
       ModExp,
       MTy
     )
-checkModBody maybe_fsig_e body_e loc = do
+checkModBody maybe_fsig_e body_e loc = enteringModule $ do
   (body_e_abs, body_mty, body_e') <- checkOneModExp body_e
   case maybe_fsig_e of
     Nothing ->
@@ -627,6 +627,11 @@
 
 checkValBind :: ValBindBase NoInfo Name -> TypeM (Env, ValBind)
 checkValBind (ValBind entry fname maybe_tdecl NoInfo tparams params body doc attrs loc) = do
+  top_level <- atTopLevel
+  when (not top_level && isJust entry) $
+    typeError loc mempty $
+      withIndexLink "nested-entry" "Entry points may not be declared inside modules."
+
   (fname', tparams', params', maybe_tdecl', rettype, retext, body') <-
     checkFunDef (fname, maybe_tdecl, tparams, params, body, loc)
 
diff --git a/src/Language/Futhark/TypeChecker/Monad.hs b/src/Language/Futhark/TypeChecker/Monad.hs
--- a/src/Language/Futhark/TypeChecker/Monad.hs
+++ b/src/Language/Futhark/TypeChecker/Monad.hs
@@ -13,12 +13,15 @@
     runTypeM,
     askEnv,
     askImportName,
+    atTopLevel,
+    enteringModule,
     bindSpaced,
     qualifyTypeVars,
     lookupMTy,
     lookupImport,
     localEnv,
     TypeError (..),
+    withIndexLink,
     unappliedFunctor,
     unknownVariable,
     unknownType,
@@ -58,6 +61,7 @@
 import qualified Data.Map.Strict as M
 import Data.Maybe
 import qualified Data.Set as S
+import qualified Data.Version as Version
 import Futhark.FreshNames hiding (newName)
 import qualified Futhark.FreshNames
 import Futhark.Util.Console
@@ -65,6 +69,7 @@
 import Language.Futhark
 import Language.Futhark.Semantic
 import Language.Futhark.Warnings
+import qualified Paths_futhark
 import Prelude hiding (mapM, mod)
 
 -- | A note with extra information regarding a type error.
@@ -92,6 +97,24 @@
     text (inRed $ "Error at " <> locStr loc <> ":")
       </> msg <> ppr notes
 
+errorIndexUrl :: Doc
+errorIndexUrl = version_url <> "error-index.html"
+  where
+    version = Paths_futhark.version
+    base_url = "https://futhark.readthedocs.io/en/"
+    version_url
+      | last (Version.versionBranch version) == 0 = base_url <> "latest/"
+      | otherwise = base_url <> "v" <> text (Version.showVersion version) <> "/"
+
+-- | Attach a reference to documentation explaining the error in more detail.
+withIndexLink :: Doc -> Doc -> Doc
+withIndexLink href msg =
+  stack
+    [ msg,
+      "\nFor more information, see:",
+      indent 2 (ppr errorIndexUrl <> "#" <> href)
+    ]
+
 -- | An unexpected functor appeared!
 unappliedFunctor :: MonadTypeChecker m => SrcLoc -> m a
 unappliedFunctor loc =
@@ -131,7 +154,10 @@
 data Context = Context
   { contextEnv :: Env,
     contextImportTable :: ImportTable,
-    contextImportName :: ImportName
+    contextImportName :: ImportName,
+    -- | Currently type-checking at the top level?  If false, we are
+    -- inside a module.
+    contextAtTopLevel :: Bool
   }
 
 data TypeState = TypeState
@@ -176,7 +202,7 @@
   TypeM a ->
   (Warnings, Either TypeError (a, VNameSource))
 runTypeM env imports fpath src (TypeM m) = do
-  let ctx = Context env imports fpath
+  let ctx = Context env imports fpath True
       s = TypeState src mempty
   case runExcept $ runStateT (runReaderT m ctx) s of
     Left (ws, e) -> (ws, Left e)
@@ -190,6 +216,15 @@
 askImportName :: TypeM ImportName
 askImportName = asks contextImportName
 
+-- | Are we type-checking at the top level, or are we inside a nested
+-- module?
+atTopLevel :: TypeM Bool
+atTopLevel = asks contextAtTopLevel
+
+-- | We are now going to type-check the body of a module.
+enteringModule :: TypeM a -> TypeM a
+enteringModule = local $ \ctx -> ctx {contextAtTopLevel = False}
+
 -- | Look up a module type.
 lookupMTy :: SrcLoc -> QualName Name -> TypeM (QualName VName, MTy)
 lookupMTy loc qn = do
@@ -463,11 +498,11 @@
 
 -- | The names that are available in the initial environment.
 topLevelNameMap :: NameMap
-topLevelNameMap = M.filterWithKey (\k _ -> atTopLevel k) intrinsicsNameMap
+topLevelNameMap = M.filterWithKey (\k _ -> available k) intrinsicsNameMap
   where
-    atTopLevel :: (Namespace, Name) -> Bool
-    atTopLevel (Type, _) = True
-    atTopLevel (Term, v) = v `S.member` (type_names <> binop_names <> fun_names)
+    available :: (Namespace, Name) -> Bool
+    available (Type, _) = True
+    available (Term, v) = v `S.member` (type_names <> binop_names <> fun_names)
       where
         type_names = S.fromList $ map (nameFromString . pretty) anyPrimType
         binop_names =
@@ -476,4 +511,4 @@
               (nameFromString . pretty)
               [minBound .. (maxBound :: BinOp)]
         fun_names = S.fromList $ map nameFromString ["shape"]
-    atTopLevel _ = False
+    available _ = False
diff --git a/src/Language/Futhark/TypeChecker/Terms.hs b/src/Language/Futhark/TypeChecker/Terms.hs
--- a/src/Language/Futhark/TypeChecker/Terms.hs
+++ b/src/Language/Futhark/TypeChecker/Terms.hs
@@ -32,7 +32,6 @@
 import qualified Data.Map.Strict as M
 import Data.Maybe
 import qualified Data.Set as S
-import qualified Data.Version as Version
 import Futhark.IR.Primitive (intByteSize)
 import Futhark.Util (nubOrd)
 import Futhark.Util.Pretty hiding (bool, group, space)
@@ -45,7 +44,6 @@
 import Language.Futhark.TypeChecker.Types hiding (checkTypeDecl)
 import qualified Language.Futhark.TypeChecker.Types as Types
 import Language.Futhark.TypeChecker.Unify hiding (Usage)
-import qualified Paths_futhark
 import Prelude hiding (mod)
 
 --- Uniqueness
@@ -693,23 +691,6 @@
     )
 
 --- Errors
-
-errorIndexUrl :: Doc
-errorIndexUrl = version_url <> "error-index.html"
-  where
-    version = Paths_futhark.version
-    base_url = "https://futhark.readthedocs.io/en/"
-    version_url
-      | last (Version.versionBranch version) == 0 = base_url <> "latest/"
-      | otherwise = base_url <> "v" <> text (Version.showVersion version) <> "/"
-
-withIndexLink :: Doc -> Doc -> Doc
-withIndexLink href msg =
-  stack
-    [ msg,
-      "\nFor more information, see:",
-      indent 2 (ppr errorIndexUrl <> "#" <> href)
-    ]
 
 useAfterConsume :: VName -> SrcLoc -> SrcLoc -> TermTypeM a
 useAfterConsume name rloc wloc = do