Vectorise even with -O0

[ghc-hetmet.git] / compiler / main / DynFlags.hs

diff --git a/compiler/main/DynFlags.hs b/compiler/main/DynFlags.hs
index 39459d3..9e28d4c 100644 (file)
--- a/compiler/main/DynFlags.hs
+++ b/compiler/main/DynFlags.hs
@@ -60,8 +60,6 @@ module DynFlags (
     compilerInfo,
   ) where
 
--- XXX This define is a bit of a hack, and should be done more nicely
-#define FAST_STRING_NOT_NEEDED 1
 #include "HsVersions.h"
 
 import Module
@@ -239,6 +237,7 @@ data DynFlag
    -- optimisation opts
    | Opt_Strictness
    | Opt_FullLaziness
+   | Opt_StaticArgumentTransformation
    | Opt_CSE
    | Opt_LiberateCase
    | Opt_SpecConstr
@@ -305,6 +304,7 @@ data DynFlags = DynFlags {
   ruleCheck            :: Maybe String,
 
   specConstrThreshold   :: Maybe Int,  -- Threshold for SpecConstr
+  specConstrCount      :: Maybe Int,   -- Max number of specialisations for any one function
   liberateCaseThreshold :: Maybe Int,   -- Threshold for LiberateCase 
 
   stolen_x86_regs      :: Int,         
@@ -496,6 +496,7 @@ defaultDynFlags =
         shouldDumpSimplPhase    = const False,
        ruleCheck               = Nothing,
        specConstrThreshold     = Just 200,
+       specConstrCount         = Just 3,
         liberateCaseThreshold   = Just 200,
        stolen_x86_regs         = 4,
        cmdlineHcIncludes       = [],
@@ -708,6 +709,7 @@ optLevelFlags
 
     , ([2],    Opt_LiberateCase)
     , ([2],    Opt_SpecConstr)
+    , ([2],    Opt_StaticArgumentTransformation)
 
     , ([0,1,2], Opt_DoLambdaEtaExpansion)
                -- This one is important for a tiresome reason:
@@ -827,6 +829,7 @@ getCoreToDo dflags
     liberate_case = dopt Opt_LiberateCase dflags
     rule_check    = ruleCheck dflags
     vectorisation = dopt Opt_Vectorise dflags
+    static_args   = dopt Opt_StaticArgumentTransformation dflags
 
     maybe_rule_check phase = runMaybe rule_check (CoreDoRuleCheck phase)
 
@@ -870,11 +873,18 @@ getCoreToDo dflags
            MaxSimplifierIterations max_iter
        ]
 
-    core_todo = 
+    core_todo =
      if opt_level == 0 then
-       [simpl_phase 0 ["final"] max_iter]
+       [runWhen vectorisation (CoreDoPasses [ simpl_gently, CoreDoVectorisation ]),
+        simpl_phase 0 ["final"] max_iter]
      else {- opt_level >= 1 -} [ 
 
+    -- We want to do the static argument transform before full laziness as it
+    -- may expose extra opportunities to float things outwards. However, to fix
+    -- up the output of the transformation we need at do at least one simplify
+    -- after this before anything else
+           runWhen static_args CoreDoStaticArgs,
+
        -- initial simplify: mk specialiser happy: minimum effort please
         simpl_gently,
 
@@ -1185,6 +1195,10 @@ dynamic_flags = [
                 upd (\dfs -> dfs{ specConstrThreshold = Just n })))
   ,  ( "fno-spec-constr-threshold",   NoArg (
                 upd (\dfs -> dfs{ specConstrThreshold = Nothing })))
+  ,  ( "fspec-constr-count",                 IntSuffix (\n ->
+                upd (\dfs -> dfs{ specConstrCount = Just n })))
+  ,  ( "fno-spec-constr-count",   NoArg (
+                upd (\dfs -> dfs{ specConstrCount = Nothing })))
   ,  ( "fliberate-case-threshold",    IntSuffix (\n ->
                 upd (\dfs -> dfs{ liberateCaseThreshold = Just n })))
   ,  ( "fno-liberate-case-threshold", NoArg (
@@ -1245,6 +1259,7 @@ fFlags = [
   ( "warn-tabs",                        Opt_WarnTabs ),
   ( "print-explicit-foralls",           Opt_PrintExplicitForalls ),
   ( "strictness",                       Opt_Strictness ),
+  ( "static-argument-transformation",   Opt_StaticArgumentTransformation ),
   ( "full-laziness",                    Opt_FullLaziness ),
   ( "liberate-case",                    Opt_LiberateCase ),
   ( "spec-constr",                      Opt_SpecConstr ),