Hi, The below commit is, as it enables half arguments and returns by default for OpenCL, while this is actually an extension that might not be supported on all platforms. I think that this part of the commit should be reverted:
> --- cfe/trunk/lib/Frontend/CompilerInvocation.cpp (original) > +++ cfe/trunk/lib/Frontend/CompilerInvocation.cpp Fri Mar 18 11:58:36 2016 > @@ -1434,6 +1434,7 @@ void CompilerInvocation::setLangDefaults > Opts.LaxVectorConversions = 0; > Opts.DefaultFPContract = 1; > Opts.NativeHalfType = 1; > + Opts.NativeHalfArgsAndReturns = 1; > } Best, Jeroen > On 18 Mar 2016, at 16:58, Pirama Arumuga Nainar via cfe-commits <cfe-commits > at lists.llvm.org> wrote: > > Author: pirama > Date: Fri Mar 18 11:58:36 2016 > New Revision: 263795 > > URL: http://llvm.org/viewvc/llvm-project?rev=263795&view=rev > Log: > Add -fnative-half-arguments-and-returns > > Summary: > r246764 handled __fp16 arguments and returns for AAPCS, but skipped this > handling for OpenCL. Simlar to OpenCL, RenderScript also handles __fp16 > type natively. > > This patch adds the -fnative-half-arguments-and-returns command line > flag to allow such languages to skip this coercion of __fp16. > > Reviewers: srhines, olista01 > > Subscribers: cfe-commits > > Differential Revision: http://reviews.llvm.org/D18138 > > Modified: > cfe/trunk/include/clang/Basic/LangOptions.def > cfe/trunk/include/clang/Driver/CC1Options.td > cfe/trunk/lib/CodeGen/TargetInfo.cpp > cfe/trunk/lib/Frontend/CompilerInvocation.cpp > cfe/trunk/test/CodeGen/arm-fp16-arguments.c > > Modified: cfe/trunk/include/clang/Basic/LangOptions.def > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/LangOptions.def?rev=263795&r1=263794&r2=263795&view=diff > ============================================================================== > --- cfe/trunk/include/clang/Basic/LangOptions.def (original) > +++ cfe/trunk/include/clang/Basic/LangOptions.def Fri Mar 18 11:58:36 2016 > @@ -163,6 +163,7 @@ LANGOPT(ShortEnums , 1, 0, "short > LANGOPT(OpenCL , 1, 0, "OpenCL") > LANGOPT(OpenCLVersion , 32, 0, "OpenCL version") > LANGOPT(NativeHalfType , 1, 0, "Native half type support") > +LANGOPT(NativeHalfArgsAndReturns, 1, 0, "Native half args and returns") > LANGOPT(HalfArgsAndReturns, 1, 0, "half args and returns") > LANGOPT(CUDA , 1, 0, "CUDA") > LANGOPT(OpenMP , 1, 0, "OpenMP support") > > Modified: cfe/trunk/include/clang/Driver/CC1Options.td > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Driver/CC1Options.td?rev=263795&r1=263794&r2=263795&view=diff > ============================================================================== > --- cfe/trunk/include/clang/Driver/CC1Options.td (original) > +++ cfe/trunk/include/clang/Driver/CC1Options.td Fri Mar 18 11:58:36 2016 > @@ -603,6 +603,8 @@ def fno_rtti_data : Flag<["-"], "fno-rtt > HelpText<"Control emission of RTTI data">; > def fnative_half_type: Flag<["-"], "fnative-half-type">, > HelpText<"Use the native half type for __fp16 instead of promoting to > float">; > +def fnative_half_arguments_and_returns : Flag<["-"], > "fnative-half-arguments-and-returns">, > + HelpText<"Use the native __fp16 type for arguments and returns (and skip > ABI-specific lowering)">; > def fallow_half_arguments_and_returns : Flag<["-"], > "fallow-half-arguments-and-returns">, > HelpText<"Allow function arguments and returns of type half">; > > > Modified: cfe/trunk/lib/CodeGen/TargetInfo.cpp > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/TargetInfo.cpp?rev=263795&r1=263794&r2=263795&view=diff > ============================================================================== > --- cfe/trunk/lib/CodeGen/TargetInfo.cpp (original) > +++ cfe/trunk/lib/CodeGen/TargetInfo.cpp Fri Mar 18 11:58:36 2016 > @@ -5106,7 +5106,7 @@ ABIArgInfo ARMABIInfo::classifyArgumentT > // __fp16 gets passed as if it were an int or float, but with the top 16 > bits > // unspecified. This is not done for OpenCL as it handles the half type > // natively, and does not need to interwork with AAPCS code. > - if (Ty->isHalfType() && !getContext().getLangOpts().OpenCL) { > + if (Ty->isHalfType() && > !getContext().getLangOpts().NativeHalfArgsAndReturns) { > llvm::Type *ResType = IsEffectivelyAAPCS_VFP ? > llvm::Type::getFloatTy(getVMContext()) : > llvm::Type::getInt32Ty(getVMContext()); > @@ -5298,7 +5298,7 @@ ABIArgInfo ARMABIInfo::classifyReturnTyp > // __fp16 gets returned as if it were an int or float, but with the top 16 > // bits unspecified. This is not done for OpenCL as it handles the half type > // natively, and does not need to interwork with AAPCS code. > - if (RetTy->isHalfType() && !getContext().getLangOpts().OpenCL) { > + if (RetTy->isHalfType() && > !getContext().getLangOpts().NativeHalfArgsAndReturns) { > llvm::Type *ResType = IsEffectivelyAAPCS_VFP ? > llvm::Type::getFloatTy(getVMContext()) : > llvm::Type::getInt32Ty(getVMContext()); > > Modified: cfe/trunk/lib/Frontend/CompilerInvocation.cpp > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Frontend/CompilerInvocation.cpp?rev=263795&r1=263794&r2=263795&view=diff > ============================================================================== > --- cfe/trunk/lib/Frontend/CompilerInvocation.cpp (original) > +++ cfe/trunk/lib/Frontend/CompilerInvocation.cpp Fri Mar 18 11:58:36 2016 > @@ -1434,6 +1434,7 @@ void CompilerInvocation::setLangDefaults > Opts.LaxVectorConversions = 0; > Opts.DefaultFPContract = 1; > Opts.NativeHalfType = 1; > + Opts.NativeHalfArgsAndReturns = 1; > } > > Opts.CUDA = IK == IK_CUDA || IK == IK_PreprocessedCuda || > @@ -1795,7 +1796,11 @@ static void ParseLangArgs(LangOptions &O > Opts.ModuleFeatures = Args.getAllArgValues(OPT_fmodule_feature); > std::sort(Opts.ModuleFeatures.begin(), Opts.ModuleFeatures.end()); > Opts.NativeHalfType |= Args.hasArg(OPT_fnative_half_type); > - Opts.HalfArgsAndReturns = > Args.hasArg(OPT_fallow_half_arguments_and_returns); > + Opts.NativeHalfArgsAndReturns |= > Args.hasArg(OPT_fnative_half_arguments_and_returns); > + // Enable HalfArgsAndReturns if present in Args or if > NativeHalfArgsAndReturns > + // is enabled. > + Opts.HalfArgsAndReturns = > Args.hasArg(OPT_fallow_half_arguments_and_returns) > + | Opts.NativeHalfArgsAndReturns; > Opts.GNUAsm = !Args.hasArg(OPT_fno_gnu_inline_asm); > > // __declspec is enabled by default for the PS4 by the driver, and also > > Modified: cfe/trunk/test/CodeGen/arm-fp16-arguments.c > URL: > http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/arm-fp16-arguments.c?rev=263795&r1=263794&r2=263795&view=diff > ============================================================================== > --- cfe/trunk/test/CodeGen/arm-fp16-arguments.c (original) > +++ cfe/trunk/test/CodeGen/arm-fp16-arguments.c Fri Mar 18 11:58:36 2016 > @@ -1,5 +1,6 @@ > // RUN: %clang_cc1 -triple armv7a--none-eabi -target-abi aapcs -mfloat-abi > soft -fallow-half-arguments-and-returns -emit-llvm -o - -O1 %s | FileCheck %s > --check-prefix=CHECK --check-prefix=SOFT > // RUN: %clang_cc1 -triple armv7a--none-eabi -target-abi aapcs -mfloat-abi > hard -fallow-half-arguments-and-returns -emit-llvm -o - -O1 %s | FileCheck %s > --check-prefix=CHECK --check-prefix=HARD > +// RUN: %clang_cc1 -triple armv7a--none-eabi -target-abi aapcs -mfloat-abi > soft -fnative-half-arguments-and-returns -emit-llvm -o - -O1 %s | FileCheck > %s --check-prefix=NATIVE > > __fp16 g; > > @@ -10,12 +11,17 @@ void t1(__fp16 a) { g = a; } > // HARD: [[BITCAST:%.*]] = bitcast float [[PARAM]] to i32 > // HARD: [[TRUNC:%.*]] = trunc i32 [[BITCAST]] to i16 > // CHECK: store i16 [[TRUNC]], i16* bitcast (half* @g to i16*) > +// NATIVE: define void @t1(half [[PARAM:%.*]]) > +// NATIVE: store half [[PARAM]], half* @g > > __fp16 t2() { return g; } > // SOFT: define i32 @t2() > // HARD: define arm_aapcs_vfpcc float @t2() > +// NATIVE: define half @t2() > // CHECK: [[LOAD:%.*]] = load i16, i16* bitcast (half* @g to i16*) > // CHECK: [[ZEXT:%.*]] = zext i16 [[LOAD]] to i32 > // SOFT: ret i32 [[ZEXT]] > // HARD: [[BITCAST:%.*]] = bitcast i32 [[ZEXT]] to float > // HARD: ret float [[BITCAST]] > +// NATIVE: [[LOAD:%.*]] = load half, half* @g > +// NATIVE: ret half [[LOAD]] _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
