[PATCH] Add R600 div_scale builtin

[Tom Stellard]

On Mon, Jun 23, 2014 at 12:10:34AM +0000, Matt Arsenault wrote:
> http://reviews.llvm.org/D4248
> 

LGTM.

> Files:
>   include/clang/Basic/TargetBuiltins.h
>   include/clang/module.modulemap
>   lib/Basic/Targets.cpp
>   lib/CodeGen/CGBuiltin.cpp
>   lib/CodeGen/CodeGenFunction.h
>   test/CodeGenOpenCL/builtins-r600.cl

> Index: include/clang/Basic/TargetBuiltins.h
> ===================================================================
> --- include/clang/Basic/TargetBuiltins.h
> +++ include/clang/Basic/TargetBuiltins.h
> @@ -72,6 +72,15 @@
>      };
>    }
>  
> +  /// \brief R600 builtins
> +  namespace R600 {
> +  enum {
> +    LastTIBuiltin = clang::Builtin::FirstTSBuiltin - 1,
> +  #define BUILTIN(ID, TYPE, ATTRS) BI##ID,
> +  #include "clang/Basic/BuiltinsR600.def"
> +    LastTSBuiltin
> +  };
> +  }
>  
>    /// \brief X86 builtins
>    namespace X86 {
> Index: include/clang/module.modulemap
> ===================================================================
> --- include/clang/module.modulemap
> +++ include/clang/module.modulemap
> @@ -36,6 +36,7 @@
>    exclude header "Basic/BuiltinsNEON.def"
>    exclude header "Basic/BuiltinsNVPTX.def"
>    exclude header "Basic/BuiltinsPPC.def"
> +  exclude header "Basic/BuiltinsR600.def"
>    exclude header "Basic/BuiltinsX86.def"
>    exclude header "Basic/BuiltinsXCore.def"
>    exclude header "Basic/DiagnosticOptions.def"
> Index: lib/Basic/Targets.cpp
> ===================================================================
> --- lib/Basic/Targets.cpp
> +++ lib/Basic/Targets.cpp
> @@ -1458,6 +1458,8 @@
>    "-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64";
>  
>  class R600TargetInfo : public TargetInfo {
> +  static const Builtin::Info BuiltinInfo[];
> +
>    /// \brief The GPU profiles supported by the R600 target.
>    enum GPUKind {
>      GK_NONE,
> @@ -1504,11 +1506,10 @@
>  
>    void getTargetBuiltins(const Builtin::Info *&Records,
>                           unsigned &NumRecords) const override {
> -    Records = nullptr;
> -    NumRecords = 0;
> +    Records = BuiltinInfo;
> +    NumRecords = clang::R600::LastTSBuiltin - Builtin::FirstTSBuiltin;
>    }
>  
> -
>    void getTargetDefines(const LangOptions &Opts,
>                          MacroBuilder &Builder) const override {
>      Builder.defineMacro("__R600__");
> @@ -1584,6 +1585,12 @@
>    }
>  };
>  
> +const Builtin::Info R600TargetInfo::BuiltinInfo[] = {
> +#define BUILTIN(ID, TYPE, ATTRS)                \
> +  { #ID, TYPE, ATTRS, 0, ALL_LANGUAGES },
> +#include "clang/Basic/BuiltinsR600.def"
> +};
> +
>  } // end anonymous namespace
>  
>  namespace {
> Index: lib/CodeGen/CGBuiltin.cpp
> ===================================================================
> --- lib/CodeGen/CGBuiltin.cpp
> +++ lib/CodeGen/CGBuiltin.cpp
> @@ -1689,6 +1689,8 @@
>    case llvm::Triple::ppc64:
>    case llvm::Triple::ppc64le:
>      return EmitPPCBuiltinExpr(BuiltinID, E);
> +  case llvm::Triple::r600:
> +    return EmitR600BuiltinExpr(BuiltinID, E);
>    default:
>      return nullptr;
>    }
> @@ -5952,3 +5954,38 @@
>    }
>    }
>  }
> +
> +Value *CodeGenFunction::EmitR600BuiltinExpr(unsigned BuiltinID,
> +                                            const CallExpr *E) {
> +  switch (BuiltinID) {
> +  case R600::BI__builtin_amdgpu_div_scale:
> +  case R600::BI__builtin_amdgpu_div_scalef: {
> +    // Translate from the intrinsics's struct return to the builtin's out
> +    // argument.
> +
> +    std::pair<llvm::Value *, unsigned> FlagOutPtr
> +      = EmitPointerWithAlignment(E->getArg(3));
> +
> +    llvm::Value *X = EmitScalarExpr(E->getArg(0));
> +    llvm::Value *Y = EmitScalarExpr(E->getArg(1));
> +    llvm::Value *Z = EmitScalarExpr(E->getArg(2));
> +
> +    llvm::Value *Callee = CGM.getIntrinsic(Intrinsic::AMDGPU_div_scale,
> +                                           X->getType());
> +
> +    llvm::Value *Tmp = Builder.CreateCall3(Callee, X, Y, Z);
> +
> +    llvm::Value *Result = Builder.CreateExtractValue(Tmp, 0);
> +    llvm::Value *Flag = Builder.CreateExtractValue(Tmp, 1);
> +
> +    llvm::Type *RealFlagType
> +      = FlagOutPtr.first->getType()->getPointerElementType();
> +
> +    llvm::Value *FlagExt = Builder.CreateZExt(Flag, RealFlagType);
> +    llvm::StoreInst *FlagStore = Builder.CreateStore(FlagExt, FlagOutPtr.first);
> +    FlagStore->setAlignment(FlagOutPtr.second);
> +    return Result;
> +  } default:
> +    return nullptr;
> +  }
> +}
> Index: lib/CodeGen/CodeGenFunction.h
> ===================================================================
> --- lib/CodeGen/CodeGenFunction.h
> +++ lib/CodeGen/CodeGenFunction.h
> @@ -2252,6 +2252,7 @@
>    llvm::Value *BuildVector(ArrayRef<llvm::Value*> Ops);
>    llvm::Value *EmitX86BuiltinExpr(unsigned BuiltinID, const CallExpr *E);
>    llvm::Value *EmitPPCBuiltinExpr(unsigned BuiltinID, const CallExpr *E);
> +  llvm::Value *EmitR600BuiltinExpr(unsigned BuiltinID, const CallExpr *E);
>  
>    llvm::Value *EmitObjCProtocolExpr(const ObjCProtocolExpr *E);
>    llvm::Value *EmitObjCStringLiteral(const ObjCStringLiteral *E);
> Index: test/CodeGenOpenCL/builtins-r600.cl
> ===================================================================
> --- /dev/null
> +++ test/CodeGenOpenCL/builtins-r600.cl
> @@ -0,0 +1,30 @@
> +// REQUIRES: r600-registered-target
> +// RUN: %clang_cc1 -triple r600-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
> +
> +#pragma OPENCL EXTENSION cl_khr_fp64 : enable
> +
> +// CHECK-LABEL: @test_div_scale_f64
> +// CHECK: call { double, i1 } @llvm.AMDGPU.div.scale.f64(double %a, double %b, i1 true)
> +// CHECK-DAG: [[FLAG:%.+]] = extractvalue { double, i1 } %{{.+}}, 1
> +// CHECK-DAG: [[VAL:%.+]] = extractvalue { double, i1 } %{{.+}}, 0
> +// CHECK: [[FLAGEXT:%.+]] = zext i1 [[FLAG]] to i32
> +// CHECK: store i32 [[FLAGEXT]]
> +void test_div_scale_f64(global double* out, global int* flagout, double a, double b)
> +{
> +  bool flag;
> +  *out = __builtin_amdgpu_div_scale(a, b, true, &flag);
> +  *flagout = flag;
> +}
> +
> +// CHECK-LABEL: @test_div_scale_f32
> +// CHECK: call { float, i1 } @llvm.AMDGPU.div.scale.f32(float %a, float %b, i1 true)
> +// CHECK-DAG: [[FLAG:%.+]] = extractvalue { float, i1 } %{{.+}}, 1
> +// CHECK-DAG: [[VAL:%.+]] = extractvalue { float, i1 } %{{.+}}, 0
> +// CHECK: [[FLAGEXT:%.+]] = zext i1 [[FLAG]] to i32
> +// CHECK: store i32 [[FLAGEXT]]
> +void test_div_scale_f32(global float* out, global int* flagout, float a, float b)
> +{
> +  bool flag;
> +  *out = __builtin_amdgpu_div_scalef(a, b, true, &flag);
> +  *flagout = flag;
> +}

> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits