[commit: ghc] master: Add preprocessor defines when SSE is enabled (bab8dc7)

[Johan Tibell]

Repository : ssh://darcs.haskell.org//srv/darcs/ghc

On branch  : master

http://hackage.haskell.org/trac/ghc/changeset/bab8dc7925a374ef91df3bca3d9a2aa9755212e1

>---------------------------------------------------------------

commit bab8dc7925a374ef91df3bca3d9a2aa9755212e1
Author: Johan Tibell <johan.tibell at gmail.com>
Date:   Thu Jan 10 15:30:21 2013 -0800

    Add preprocessor defines when SSE is enabled
    
    This will add the following preprocessor defines when Haskell source
    files are compiled:
    
     * __SSE__ - If any version of SSE is enabled
     * __SSE2__ - If SSE2 or greater is enabled
     * __SSE4_2_ - If SSE4.2 is enabled
    
    Note that SSE2 is enabled by default on x86-64.

>---------------------------------------------------------------

 compiler/main/DriverPipeline.hs   |    8 ++++++++
 compiler/main/DynFlags.hs         |   27 +++++++++++++++++++++++++++
 compiler/nativeGen/X86/CodeGen.hs |   12 ++----------
 3 files changed, 37 insertions(+), 10 deletions(-)

diff --git a/compiler/main/DriverPipeline.hs b/compiler/main/DriverPipeline.hs
index 866ae8c..feaac3b 100644
--- a/compiler/main/DriverPipeline.hs
+++ b/compiler/main/DriverPipeline.hs
@@ -1962,12 +1962,20 @@ doCpp dflags raw include_cc_opts input_fn output_fn = do
         -- remember, in code we *compile*, the HOST is the same our TARGET,
         -- and BUILD is the same as our HOST.
 
+    let sse2 = isSse2Enabled dflags
+        sse4_2 = isSse4_2Enabled dflags
+        sse_defs =
+          [ "-D__SSE__=1" | sse2 || sse4_2 ] ++
+          [ "-D__SSE2__=1" | sse2 || sse4_2 ] ++
+          [ "-D__SSE4_2__=1" | sse4_2 ]
+
     cpp_prog       (   map SysTools.Option verbFlags
                     ++ map SysTools.Option include_paths
                     ++ map SysTools.Option hsSourceCppOpts
                     ++ map SysTools.Option target_defs
                     ++ map SysTools.Option hscpp_opts
                     ++ map SysTools.Option cc_opts
+                    ++ map SysTools.Option sse_defs
                     ++ [ SysTools.Option     "-x"
                        , SysTools.Option     "c"
                        , SysTools.Option     input_fn
diff --git a/compiler/main/DynFlags.hs b/compiler/main/DynFlags.hs
index 3538629..216ca10 100644
--- a/compiler/main/DynFlags.hs
+++ b/compiler/main/DynFlags.hs
@@ -118,6 +118,10 @@ module DynFlags (
         tAG_MASK,
         mAX_PTR_TAG,
         tARGET_MIN_INT, tARGET_MAX_INT, tARGET_MAX_WORD,
+
+        -- * SSE
+        isSse2Enabled,
+        isSse4_2Enabled,
   ) where
 
 #include "HsVersions.h"
@@ -2153,6 +2157,11 @@ dynamic_flags = [
   , Flag "monly-4-regs" (NoArg (addWarn "The -monly-4-regs flag does nothing; it will be removed in a future GHC release"))
   , Flag "msse2"        (NoArg (setGeneralFlag Opt_SSE2))
   , Flag "msse4.2"      (NoArg (setGeneralFlag Opt_SSE4_2))
+    -- at some point we should probably have a single SSE flag that
+    -- contains the SSE version, instead of having a different flag
+    -- per version. That would make it easier to e.g. check if SSE2 is
+    -- enabled as you wouldn't have to check if either Opt_SSE2 or
+    -- Opt_SSE4_2 is set (as the latter implies the former).
 
      ------ Warning opts -------------------------------------------------
   , Flag "W"      (NoArg (mapM_ setWarningFlag minusWOpts))
@@ -3371,3 +3380,21 @@ makeDynFlagsConsistent dflags
           arch = platformArch platform
           os   = platformOS   platform
 
+-- -----------------------------------------------------------------------------
+-- SSE
+
+isSse2Enabled :: DynFlags -> Bool
+isSse2Enabled dflags = isSse4_2Enabled dflags || isSse2Enabled'
+  where
+    isSse2Enabled' = case platformArch (targetPlatform dflags) of
+        ArchX86_64 -> -- SSE2 is fixed on for x86_64.  It would be
+                      -- possible to make it optional, but we'd need to
+                      -- fix at least the foreign call code where the
+                      -- calling convention specifies the use of xmm regs,
+                      -- and possibly other places.
+                      True
+        ArchX86    -> gopt Opt_SSE2 dflags
+        _          -> False
+
+isSse4_2Enabled :: DynFlags -> Bool
+isSse4_2Enabled dflags = gopt Opt_SSE4_2 dflags
diff --git a/compiler/nativeGen/X86/CodeGen.hs b/compiler/nativeGen/X86/CodeGen.hs
index 30cf060..d014709 100644
--- a/compiler/nativeGen/X86/CodeGen.hs
+++ b/compiler/nativeGen/X86/CodeGen.hs
@@ -71,20 +71,12 @@ is32BitPlatform = do
 sse2Enabled :: NatM Bool
 sse2Enabled = do
   dflags <- getDynFlags
-  case platformArch (targetPlatform dflags) of
-      ArchX86_64 -> -- SSE2 is fixed on for x86_64.  It would be
-                    -- possible to make it optional, but we'd need to
-                    -- fix at least the foreign call code where the
-                    -- calling convention specifies the use of xmm regs,
-                    -- and possibly other places.
-                    return True
-      ArchX86    -> return (gopt Opt_SSE2 dflags || gopt Opt_SSE4_2 dflags)
-      _          -> panic "sse2Enabled: Not an X86* arch"
+  return (isSse2Enabled dflags)
 
 sse4_2Enabled :: NatM Bool
 sse4_2Enabled = do
   dflags <- getDynFlags
-  return (gopt Opt_SSE4_2 dflags)
+  return (isSse4_2Enabled dflags)
 
 if_sse2 :: NatM a -> NatM a -> NatM a
 if_sse2 sse2 x87 = do