From: Peter Collingbourne <peter@pcc.me.uk>
Date: Sat, 4 Dec 2010 01:51:33 +0000 (+0000)
Subject: Implement -cl-mad-enable
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=4400cb8c548a90c472444b57690117563f4be195;p=clang

Implement -cl-mad-enable

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@120881 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/include/clang/Driver/CC1Options.td b/include/clang/Driver/CC1Options.td
index 13919aafc0..f34a4841f8 100644
--- a/include/clang/Driver/CC1Options.td
+++ b/include/clang/Driver/CC1Options.td
@@ -607,3 +607,5 @@ def cl_unsafe_math_optimizations : Flag<"-cl-unsafe-math-optimizations">,
   HelpText<"OpenCL only. Allow unsafe floating-point optimizations.  Also implies -cl-no-signed-zeros and -cl-mad-enable">;
 def cl_fast_relaxed_math : Flag<"-cl-fast-relaxed-math">,
   HelpText<"OpenCL only. Sets -cl-finite-math-only and -cl-unsafe-math-optimizations, and defines __FAST_RELAXED_MATH__">;
+def cl_mad_enable : Flag<"-cl-mad-enable">,
+  HelpText<"OpenCL only. Enable less precise MAD instructions to be generated.">;
diff --git a/include/clang/Frontend/CodeGenOptions.h b/include/clang/Frontend/CodeGenOptions.h
index 84f500383d..52fb144541 100644
--- a/include/clang/Frontend/CodeGenOptions.h
+++ b/include/clang/Frontend/CodeGenOptions.h
@@ -56,6 +56,8 @@ public:
   unsigned HiddenWeakVTables : 1; /// Emit weak vtables, RTTI, and thunks with
                                   /// hidden visibility
   unsigned InstrumentFunctions : 1; /// Set when -finstrument-functions is enabled
+  unsigned LessPreciseFPMAD  : 1; /// Enable less precise MAD instructions to be
+                                  /// generated.
   unsigned MergeAllConstants : 1; /// Merge identical constants.
   unsigned NoCommon          : 1; /// Set when -fno-common or C++ is enabled.
   unsigned NoImplicitFloat   : 1; /// Set when -mno-implicit-float is enabled.
@@ -123,6 +125,7 @@ public:
     HiddenWeakTemplateVTables = 0;
     HiddenWeakVTables = 0;
     InstrumentFunctions = 0;
+    LessPreciseFPMAD = 0;
     MergeAllConstants = 1;
     NoCommon = 0;
     NoImplicitFloat = 0;
diff --git a/lib/CodeGen/BackendUtil.cpp b/lib/CodeGen/BackendUtil.cpp
index 2808b4a350..f12715ac27 100644
--- a/lib/CodeGen/BackendUtil.cpp
+++ b/lib/CodeGen/BackendUtil.cpp
@@ -183,6 +183,7 @@ bool EmitAssemblyHelper::AddEmitPasses(BackendAction Action,
     llvm::FloatABIType = llvm::FloatABI::Default;
   }
 
+  llvm::LessPreciseFPMADOption = CodeGenOpts.LessPreciseFPMAD;
   llvm::NoInfsFPMath = CodeGenOpts.NoInfsFPMath;
   llvm::NoNaNsFPMath = CodeGenOpts.NoNaNsFPMath;
   NoZerosInBSS = CodeGenOpts.NoZeroInitializedInBSS;
diff --git a/lib/Frontend/CompilerInvocation.cpp b/lib/Frontend/CompilerInvocation.cpp
index c30d7d3f6d..f7656425cc 100644
--- a/lib/Frontend/CompilerInvocation.cpp
+++ b/lib/Frontend/CompilerInvocation.cpp
@@ -897,6 +897,7 @@ static void ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args, InputKind IK,
   Opts.DisableFPElim = Args.hasArg(OPT_mdisable_fp_elim);
   Opts.FloatABI = Args.getLastArgValue(OPT_mfloat_abi);
   Opts.HiddenWeakVTables = Args.hasArg(OPT_fhidden_weak_vtables);
+  Opts.LessPreciseFPMAD = Args.hasArg(OPT_cl_mad_enable);
   Opts.LimitFloatPrecision = Args.getLastArgValue(OPT_mlimit_float_precision);
   Opts.NoInfsFPMath = Opts.NoNaNsFPMath = Args.hasArg(OPT_cl_finite_math_only)||
                                           Args.hasArg(OPT_cl_fast_relaxed_math);