From b74b04bd6701f2b4aee5b84174db2297dbdc9636 Mon Sep 17 00:00:00 2001 From: ranapratap55 Date: Tue, 21 Oct 2025 10:50:40 +0530 Subject: [PATCH 1/5] [AMDGPU][Clang] Support for type inferring extended image builtins for AMDGPU --- clang/include/clang/Basic/BuiltinsAMDGPU.def | 41 + clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp | 133 +- clang/lib/Sema/SemaAMDGPU.cpp | 42 +- .../CodeGen/builtins-extended-image-load.c | 1529 +++++++++++++++++ ...iltins-extended-image-param-gfx1100-err.cl | 21 + 5 files changed, 1736 insertions(+), 30 deletions(-) create mode 100644 clang/test/CodeGen/builtins-extended-image-load.c create mode 100644 clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def index 8428fa97fe445..0fe22f583a117 100644 --- a/clang/include/clang/Basic/BuiltinsAMDGPU.def +++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def @@ -958,6 +958,47 @@ TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f32_f32, "V4fifffQtV4ibii", "n TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4hifQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_f32_f32, "fiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_f32_f32, "fiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_f32_f32, "fiffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f32_f32, "V4fiffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f16_f32, "V4hiffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_f32_f32, "fifffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f32_f32, "V4fifffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f16_f32, "V4hifffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f32_f32, "V4fifffffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f16_f32, "V4hifffffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") #undef BUILTIN #undef TARGET_BUILTIN diff --git a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp index 5049a0ab0a395..23bb9ff40a97b 100644 --- a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp @@ -133,8 +133,8 @@ Value *EmitAMDGPUWorkGroupSize(CodeGenFunction &CGF, unsigned Index) { } llvm::MDBuilder MDHelper(CGF.getLLVMContext()); - llvm::MDNode *RNode = MDHelper.createRange(APInt(16, 1), - APInt(16, CGF.getTarget().getMaxOpenCLWorkGroupSize() + 1)); + llvm::MDNode *RNode = MDHelper.createRange( + APInt(16, 1), APInt(16, CGF.getTarget().getMaxOpenCLWorkGroupSize() + 1)); LD->setMetadata(llvm::LLVMContext::MD_range, RNode); LD->setMetadata(llvm::LLVMContext::MD_noundef, llvm::MDNode::get(CGF.getLLVMContext(), {})); @@ -253,8 +253,7 @@ emitAMDGCNImageOverloadedReturnType(clang::CodeGen::CodeGenFunction &CGF, } // Emit an intrinsic that has 1 float or double operand, and 1 integer. -static Value *emitFPIntBuiltin(CodeGenFunction &CGF, - const CallExpr *E, +static Value *emitFPIntBuiltin(CodeGenFunction &CGF, const CallExpr *E, unsigned IntrinsicID) { llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0)); llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1)); @@ -456,8 +455,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Y = EmitScalarExpr(E->getArg(1)); llvm::Value *Z = EmitScalarExpr(E->getArg(2)); - llvm::Function *Callee = CGM.getIntrinsic(Intrinsic::amdgcn_div_scale, - X->getType()); + llvm::Function *Callee = + CGM.getIntrinsic(Intrinsic::amdgcn_div_scale, X->getType()); llvm::Value *Tmp = Builder.CreateCall(Callee, {X, Y, Z}); @@ -477,8 +476,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Src2 = EmitScalarExpr(E->getArg(2)); llvm::Value *Src3 = EmitScalarExpr(E->getArg(3)); - llvm::Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas, - Src0->getType()); + llvm::Function *F = + CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas, Src0->getType()); llvm::Value *Src3ToBool = Builder.CreateIsNotNull(Src3); return Builder.CreateCall(F, {Src0, Src1, Src2, Src3ToBool}); } @@ -619,13 +618,13 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_frexp_expf: { Value *Src0 = EmitScalarExpr(E->getArg(0)); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_frexp_exp, - { Builder.getInt32Ty(), Src0->getType() }); + {Builder.getInt32Ty(), Src0->getType()}); return Builder.CreateCall(F, Src0); } case AMDGPU::BI__builtin_amdgcn_frexp_exph: { Value *Src0 = EmitScalarExpr(E->getArg(0)); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_frexp_exp, - { Builder.getInt16Ty(), Src0->getType() }); + {Builder.getInt16Ty(), Src0->getType()}); return Builder.CreateCall(F, Src0); } case AMDGPU::BI__builtin_amdgcn_fract: @@ -646,8 +645,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_ballot_w64: { llvm::Type *ResultType = ConvertType(E->getType()); llvm::Value *Src = EmitScalarExpr(E->getArg(0)); - Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_ballot, { ResultType }); - return Builder.CreateCall(F, { Src }); + Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_ballot, {ResultType}); + return Builder.CreateCall(F, {Src}); } case AMDGPU::BI__builtin_amdgcn_inverse_ballot_w32: case AMDGPU::BI__builtin_amdgcn_inverse_ballot_w64: { @@ -671,8 +670,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // FIXME-GFX10: How should 32 bit mask be handled? Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_icmp, - { Builder.getInt64Ty(), Src0->getType() }); - return Builder.CreateCall(F, { Src0, Src1, Src2 }); + {Builder.getInt64Ty(), Src0->getType()}); + return Builder.CreateCall(F, {Src0, Src1, Src2}); } case AMDGPU::BI__builtin_amdgcn_fcmp: case AMDGPU::BI__builtin_amdgcn_fcmpf: { @@ -682,8 +681,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // FIXME-GFX10: How should 32 bit mask be handled? Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_fcmp, - { Builder.getInt64Ty(), Src0->getType() }); - return Builder.CreateCall(F, { Src0, Src1, Src2 }); + {Builder.getInt64Ty(), Src0->getType()}); + return Builder.CreateCall(F, {Src0, Src1, Src2}); } case AMDGPU::BI__builtin_amdgcn_class: case AMDGPU::BI__builtin_amdgcn_classf: @@ -695,11 +694,12 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Intrinsic::amdgcn_fmed3); case AMDGPU::BI__builtin_amdgcn_ds_append: case AMDGPU::BI__builtin_amdgcn_ds_consume: { - Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append ? - Intrinsic::amdgcn_ds_append : Intrinsic::amdgcn_ds_consume; + Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append + ? Intrinsic::amdgcn_ds_append + : Intrinsic::amdgcn_ds_consume; Value *Src0 = EmitScalarExpr(E->getArg(0)); - Function *F = CGM.getIntrinsic(Intrin, { Src0->getType() }); - return Builder.CreateCall(F, { Src0, Builder.getFalse() }); + Function *F = CGM.getIntrinsic(Intrin, {Src0->getType()}); + return Builder.CreateCall(F, {Src0, Builder.getFalse()}); } case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_i32: case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_v2i32: @@ -919,12 +919,10 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // The builtins take these arguments as vec4 where the last element is // ignored. The intrinsic takes them as vec3. - RayOrigin = Builder.CreateShuffleVector(RayOrigin, RayOrigin, - {0, 1, 2}); - RayDir = - Builder.CreateShuffleVector(RayDir, RayDir, {0, 1, 2}); - RayInverseDir = Builder.CreateShuffleVector(RayInverseDir, RayInverseDir, - {0, 1, 2}); + RayOrigin = Builder.CreateShuffleVector(RayOrigin, RayOrigin, {0, 1, 2}); + RayDir = Builder.CreateShuffleVector(RayDir, RayDir, {0, 1, 2}); + RayInverseDir = + Builder.CreateShuffleVector(RayInverseDir, RayInverseDir, {0, 1, 2}); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_image_bvh_intersect_ray, {NodePtr->getType(), RayDir->getType()}); @@ -998,8 +996,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Value *Rtn = Builder.CreateExtractValue(Call, 0); Value *A = Builder.CreateExtractValue(Call, 1); llvm::Type *RetTy = ConvertType(E->getType()); - Value *I0 = Builder.CreateInsertElement(PoisonValue::get(RetTy), Rtn, - (uint64_t)0); + Value *I0 = + Builder.CreateInsertElement(PoisonValue::get(RetTy), Rtn, (uint64_t)0); // ds_bvh_stack_push8_pop2_rtn returns {i64, i32} but the builtin returns // <2 x i64>, zext the second value. if (A->getType()->getPrimitiveSizeInBits() < @@ -1138,6 +1136,83 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_image_sample_cube_v4f16_f32: return emitAMDGCNImageOverloadedReturnType( *this, E, Intrinsic::amdgcn_image_sample_cube, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_1d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_1d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_1d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_1d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_1d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_1d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_1d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_1d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_d_1d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_2d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_2d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_d_2d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_3d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_3d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_3d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_3d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_3d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_3d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_3d_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_3d_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_d_3d, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_cube_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_cube_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_cube, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_cube_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_cube_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_cube, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_1darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_1darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_1darray_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_1darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_1darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_1darray_v4f16_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_d_1darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_lz_2darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_l_2darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_v4f32_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_v4f16_f32: + case clang::AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_sample_d_2darray, false); + case clang::AMDGPU::BI__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32: + return emitAMDGCNImageOverloadedReturnType( + *this, E, Intrinsic::amdgcn_image_gather4_lz_2d, false); case AMDGPU::BI__builtin_amdgcn_mfma_scale_f32_16x16x128_f8f6f4: case AMDGPU::BI__builtin_amdgcn_mfma_scale_f32_32x32x64_f8f6f4: { llvm::FixedVectorType *VT = FixedVectorType::get(Builder.getInt32Ty(), 8); @@ -1627,7 +1702,7 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Src1 = EmitScalarExpr(E->getArg(1)); llvm::Value *Src2 = EmitScalarExpr(E->getArg(2)); Function *F = CGM.getIntrinsic(Intrinsic::fshr, Src0->getType()); - return Builder.CreateCall(F, { Src0, Src1, Src2 }); + return Builder.CreateCall(F, {Src0, Src1, Src2}); } case AMDGPU::BI__builtin_amdgcn_fence: { ProcessOrderScopeAMDGCN(EmitScalarExpr(E->getArg(0)), diff --git a/clang/lib/Sema/SemaAMDGPU.cpp b/clang/lib/Sema/SemaAMDGPU.cpp index e32f4376a5ebf..18760f31ab298 100644 --- a/clang/lib/Sema/SemaAMDGPU.cpp +++ b/clang/lib/Sema/SemaAMDGPU.cpp @@ -153,7 +153,47 @@ bool SemaAMDGPU::CheckAMDGCNBuiltinFunctionCall(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_image_sample_3d_v4f32_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_3d_v4f16_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_cube_v4f32_f32: - case AMDGPU::BI__builtin_amdgcn_image_sample_cube_v4f16_f32: { + case AMDGPU::BI__builtin_amdgcn_image_sample_cube_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_1d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_1d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_3d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_3d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_cube_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_lz_cube_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_1d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_1d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_1darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_1darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_3d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_3d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_cube_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_cube_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_1d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_1d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_1darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_1darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_2darray_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_3d_v4f32_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_d_3d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32: { StringRef FeatureList( getASTContext().BuiltinInfo.getRequiredFeatures(BuiltinID)); if (!Builtin::evaluateRequiredTargetFeatures(FeatureList, diff --git a/clang/test/CodeGen/builtins-extended-image-load.c b/clang/test/CodeGen/builtins-extended-image-load.c new file mode 100644 index 0000000000000..7841a651dbada --- /dev/null +++ b/clang/test/CodeGen/builtins-extended-image-load.c @@ -0,0 +1,1529 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5 +// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1100 -target-feature +extended-image-insts %s -emit-llvm -o - | FileCheck %s + +typedef int int4 __attribute__((ext_vector_type(4))); +typedef float float4 __attribute__((ext_vector_type(4))); +typedef _Float16 half; +typedef half half4 __attribute__((ext_vector_type(4))); + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_gather4_lz_2d_v4f32_f32_r( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(1, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_gather4_lz_2d_v4f32_f32_g( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32.v8i32.v4i32(i32 2, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_g(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(2, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_gather4_lz_2d_v4f32_f32_b( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32.v8i32.v4i32(i32 4, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_b(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(4, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_gather4_lz_2d_v4f32_f32_a( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32.v8i32.v4i32(i32 8, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_a(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(8, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_1d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.1d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP2]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP3]] +// +float4 test_amdgcn_image_sample_lz_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f32_f32(100, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_1d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.1d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_sample_l_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_d_1d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_d_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_2d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.2d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_sample_lz_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_2d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.2d.v4f32.f32.v8i32.v4i32(i32 10, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_l_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f32_f32(10, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_d_2d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP6]], align 32 +// CHECK-NEXT: [[TMP7:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP7]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP8]] +// +float4 test_amdgcn_image_sample_d_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f32_f32(100, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_3d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.3d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_lz_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_3d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.3d.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP6]] +// +float4 test_amdgcn_image_sample_l_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_d_3d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP9]], align 32 +// CHECK-NEXT: [[TMP10:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP11:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], float [[TMP6]], float [[TMP7]], float [[TMP8]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP10]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP11]] +// +float4 test_amdgcn_image_sample_d_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_cube_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.cube.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_lz_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_cube_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.cube.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP6]] +// +float4 test_amdgcn_image_sample_l_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_1darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.1darray.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_amdgcn_image_sample_lz_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f32_f32(1, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_1darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.1darray.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_l_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_d_1darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.d.1darray.v4f32.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP6]] +// +float4 test_amdgcn_image_sample_d_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_lz_2darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.lz.2darray.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_amdgcn_image_sample_lz_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_l_2darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.l.2darray.v4f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP6]] +// +float4 test_amdgcn_image_sample_l_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_sample_d_2darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP7]], align 32 +// CHECK-NEXT: [[TMP8:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP9:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.d.2darray.v4f32.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], float [[TMP6]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP8]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP9]] +// +float4 test_amdgcn_image_sample_d_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_1d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.1d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP2]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP3]] +// +half4 test_amdgcn_image_sample_lz_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f16_f32(100, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_1d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.1d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_amdgcn_image_sample_l_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_d_1d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.d.1d.v4f16.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_d_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_2d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.2d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_amdgcn_image_sample_lz_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_2d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.2d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_l_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_d_2d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP6]], align 32 +// CHECK-NEXT: [[TMP7:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP8:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.d.2d.v4f16.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP7]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP8]] +// +half4 test_amdgcn_image_sample_d_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_3d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.3d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_lz_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_3d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.3d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP6]] +// +half4 test_amdgcn_image_sample_l_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_d_3d_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP9]], align 32 +// CHECK-NEXT: [[TMP10:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP11:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.d.3d.v4f16.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], float [[TMP6]], float [[TMP7]], float [[TMP8]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP10]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP11]] +// +half4 test_amdgcn_image_sample_d_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_cube_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.cube.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_lz_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_cube_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.cube.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP6]] +// +half4 test_amdgcn_image_sample_l_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_1darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.1darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_amdgcn_image_sample_lz_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_1darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.1darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_l_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_d_1darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.d.1darray.v4f16.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP6]] +// +half4 test_amdgcn_image_sample_d_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_lz_2darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.lz.2darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_amdgcn_image_sample_lz_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_l_2darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.l.2darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP6]] +// +half4 test_amdgcn_image_sample_l_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_amdgcn_image_sample_d_2darray_v4f16_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP7]], align 32 +// CHECK-NEXT: [[TMP8:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP9:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.d.2darray.v4f16.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], float [[TMP6]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP8]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP9]] +// +half4 test_amdgcn_image_sample_d_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_lz_2d_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.amdgcn.image.sample.lz.2d.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP4]] +// +float test_amdgcn_image_sample_lz_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_f32_f32(1, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_l_2d_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call float @llvm.amdgcn.image.sample.l.2d.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP5]] +// +float test_amdgcn_image_sample_l_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_d_2d_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP6]], align 32 +// CHECK-NEXT: [[TMP7:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP8:%.*]] = call float @llvm.amdgcn.image.sample.d.2d.f32.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP7]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP8]] +// +float test_amdgcn_image_sample_d_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_f32_f32(1, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_lz_2darray_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call float @llvm.amdgcn.image.sample.lz.2darray.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP5]] +// +float test_amdgcn_image_sample_lz_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_l_2darray_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP6:%.*]] = call float @llvm.amdgcn.image.sample.l.2darray.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP5]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP6]] +// +float test_amdgcn_image_sample_l_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_amdgcn_image_sample_d_2darray_f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP7]], align 32 +// CHECK-NEXT: [[TMP8:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP9:%.*]] = call float @llvm.amdgcn.image.sample.d.2darray.f32.f32.f32.v8i32.v4i32(i32 1, float [[TMP0]], float [[TMP1]], float [[TMP2]], float [[TMP3]], float [[TMP4]], float [[TMP5]], float [[TMP6]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP8]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP9]] +// +float test_amdgcn_image_sample_d_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} diff --git a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl new file mode 100644 index 0000000000000..90ba5c23d9649 --- /dev/null +++ b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl @@ -0,0 +1,21 @@ +// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1100 -target-feature +extended-image-insts -S -verify=expected -o - %s +// REQUIRES: amdgpu-registered-target + +typedef int int4 __attribute__((ext_vector_type(4))); +typedef float float4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(1, f32, f32, tex, vec4i32, 0, f32, 110); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_3d_v4f16_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_d_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_f32_f32(f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2d_f32_f32' must be a constant integer}} +} From 901f395cadd7eef9afdb418bb97dfa64854a11c8 Mon Sep 17 00:00:00 2001 From: ranapratap55 Date: Fri, 24 Oct 2025 17:15:21 +0530 Subject: [PATCH 2/5] Modified half 'h' to 'x' in def and adds target check tests --- clang/include/clang/Basic/BuiltinsAMDGPU.def | 94 ++++---- clang/lib/Sema/SemaAMDGPU.cpp | 1 + ...image-load.c => builtins-extended-image.c} | 3 +- clang/test/CodeGen/builtins-image-load.c | 3 +- clang/test/CodeGen/builtins-image-store.c | 3 +- ...iltins-extended-image-param-gfx1100-err.cl | 212 ++++++++++++++++- ...uiltins-extended-image-param-gfx942-err.cl | 225 ++++++++++++++++++ .../builtins-image-load-param-gfx1100-err.cl | 2 +- .../builtins-image-load-param-gfx942-err.cl | 2 +- .../builtins-image-store-param-gfx1100-err.cl | 2 +- .../builtins-image-store-param-gfx942-err.cl | 2 +- 11 files changed, 488 insertions(+), 61 deletions(-) rename clang/test/CodeGen/{builtins-extended-image-load.c => builtins-extended-image.c} (99%) create mode 100644 clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def index 0fe22f583a117..6947162a3b962 100644 --- a/clang/include/clang/Basic/BuiltinsAMDGPU.def +++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def @@ -889,115 +889,115 @@ TARGET_BUILTIN(__builtin_amdgcn_cooperative_atomic_store_8x16B, "vV4i*V4iIicC*", // Image builtins //===----------------------------------------------------------------------===// TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f32_i32, "V4fiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f16_i32, "V4hiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f16_i32, "V4xiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_f32_i32, "fiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_f32_i32, "fiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_f32_i32, "fiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_f32_i32, "fiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f32_i32, "vV4fiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f16_i32, "vV4hiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f16_i32, "vV4xiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_f32_i32, "vfiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_f32_i32, "vfiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_f32_i32, "vfiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_f32_i32, "vfiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f16_f32, "V4hifQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f16_f32, "V4xifQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f16_f32, "V4hiffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f16_f32, "V4xiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_f32_f32, "fiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f16_f32, "V4hiffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f16_f32, "V4xiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_f32_f32, "fifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4hifQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4xifQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_f32_f32, "fiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f16_f32, "V4hiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_f32_f32, "fiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f16_f32, "V4hifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f16_f32, "V4hiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_f32_f32, "fiffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f32_f32, "V4fiffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f16_f32, "V4hiffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f16_f32, "V4xiffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_f32_f32, "fifffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f32_f32, "V4fifffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f16_f32, "V4hifffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f16_f32, "V4xifffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f32_f32, "V4fifffffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f16_f32, "V4hifffffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f16_f32, "V4xifffffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") #undef BUILTIN diff --git a/clang/lib/Sema/SemaAMDGPU.cpp b/clang/lib/Sema/SemaAMDGPU.cpp index 18760f31ab298..139c4abc040df 100644 --- a/clang/lib/Sema/SemaAMDGPU.cpp +++ b/clang/lib/Sema/SemaAMDGPU.cpp @@ -174,6 +174,7 @@ bool SemaAMDGPU::CheckAMDGCNBuiltinFunctionCall(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_image_sample_l_1darray_v4f16_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_f32_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_v4f16_f32: + case AMDGPU::BI__builtin_amdgcn_image_sample_l_2d_v4f32_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_f32_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f32_f32: case AMDGPU::BI__builtin_amdgcn_image_sample_l_2darray_v4f16_f32: diff --git a/clang/test/CodeGen/builtins-extended-image-load.c b/clang/test/CodeGen/builtins-extended-image.c similarity index 99% rename from clang/test/CodeGen/builtins-extended-image-load.c rename to clang/test/CodeGen/builtins-extended-image.c index 7841a651dbada..0dbf81dabd77b 100644 --- a/clang/test/CodeGen/builtins-extended-image-load.c +++ b/clang/test/CodeGen/builtins-extended-image.c @@ -3,8 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half; -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); // CHECK-LABEL: define dso_local <4 x float> @test_amdgcn_image_gather4_lz_2d_v4f32_f32_r( // CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0:[0-9]+]] { diff --git a/clang/test/CodeGen/builtins-image-load.c b/clang/test/CodeGen/builtins-image-load.c index 8442124416338..e455e8e0a8de4 100644 --- a/clang/test/CodeGen/builtins-image-load.c +++ b/clang/test/CodeGen/builtins-image-load.c @@ -3,8 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half; -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); // CHECK-LABEL: define dso_local float @test_builtin_image_load_2d( // CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0:[0-9]+]] { diff --git a/clang/test/CodeGen/builtins-image-store.c b/clang/test/CodeGen/builtins-image-store.c index 5309a16df7033..f930d3e786d51 100644 --- a/clang/test/CodeGen/builtins-image-store.c +++ b/clang/test/CodeGen/builtins-image-store.c @@ -2,8 +2,7 @@ // RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 %s -emit-llvm -o - | FileCheck %s typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half; -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); // CHECK-LABEL: define dso_local void @test_builtin_image_store_2d( // CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0:[0-9]+]] { diff --git a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl index 90ba5c23d9649..0f87f9850b49d 100644 --- a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl @@ -3,19 +3,223 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { - return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(1, f32, f32, tex, vec4i32, 0, f32, 110); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(1, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_g(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(2, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_b(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(4, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_a(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(8, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_lz_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f32_f32(i32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_1d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_d_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_1d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_lz_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, 103); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_d_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f32_f32(i32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2d_v4f32_f32' must be a constant integer}} +} +float4 test_amdgcn_image_sample_lz_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f32_f32(i32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_3d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_3d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_d_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_3d_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_lz_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_cube_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_cube_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_lz_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f32_f32(1, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_1darray_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_d_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_1darray_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_lz_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_l_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2darray_v4f32_f32' must be a constant integer}} +} + +float4 test_amdgcn_image_sample_d_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2darray_v4f32_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f16_f32(i32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f16_f32(i32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_1d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_d_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f16_f32(i32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_1d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f16_f32(100, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_d_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_3d_v4f16_f32' must be a constant integer}} } half4 test_amdgcn_image_sample_l_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { - return __builtin_amdgcn_image_sample_l_3d_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 120, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_3d_v4f16_f32' must be a constant integer}} + return __builtin_amdgcn_image_sample_l_3d_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_3d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_d_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_3d_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_cube_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f16_f32(i32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_cube_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f16_f32(i32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f16_f32(i32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_1darray_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_d_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_1darray_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_lz_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_l_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2darray_v4f16_f32' must be a constant integer}} +} + +half4 test_amdgcn_image_sample_d_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2darray_v4f16_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_lz_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_f32_f32(1, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2d_f32_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_l_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2d_f32_f32' must be a constant integer}} } float test_amdgcn_image_sample_d_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { - return __builtin_amdgcn_image_sample_d_2d_f32_f32(f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 120, 110); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2d_f32_f32' must be a constant integer}} + return __builtin_amdgcn_image_sample_d_2d_f32_f32(1, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2d_f32_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_lz_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_2darray_f32_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_l_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_l_2darray_f32_f32' must be a constant integer}} +} + +float test_amdgcn_image_sample_d_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_d_2darray_f32_f32' must be a constant integer}} } diff --git a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl new file mode 100644 index 0000000000000..80e2130831580 --- /dev/null +++ b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl @@ -0,0 +1,225 @@ +// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx942 -verify=GFX94 -S -o - %s +// REQUIRES: amdgpu-registered-target + +typedef int int4 __attribute__((ext_vector_type(4))); +typedef float float4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(1, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_gather4_lz_2d_v4f32_f32_r' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_g(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(2, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_gather4_lz_2d_v4f32_f32_g' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_b(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(4, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_gather4_lz_2d_v4f32_f32_b' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_a(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_gather4_lz_2d_v4f32_f32(8, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_gather4_lz_2d_v4f32_f32_a' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_lz_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f32_f32(105, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_1d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_1d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_d_1d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_1d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_lz_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f32_f32(10, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_d_2d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f32_f32(105, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2d_v4f32_f32' needs target feature extended-image-insts}} +} +float4 test_amdgcn_image_sample_lz_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f32_f32(105, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_3d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_3d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_d_3d_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_3d_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_lz_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_cube_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_cube_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_cube_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_lz_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f32_f32(1, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_1darray_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_1darray_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_d_1darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_1darray_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_lz_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2darray_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_l_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2darray_v4f32_f32' needs target feature extended-image-insts}} +} + +float4 test_amdgcn_image_sample_d_2darray_v4f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2darray_v4f32_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1d_v4f16_f32(105, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_1d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1d_v4f16_f32(105, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_1d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_d_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1d_v4f16_f32(105, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_1d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_d_2d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_3d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_3d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_3d_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_3d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_d_3d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_3d_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_3d_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_cube_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_cube_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_cube_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_cube_v4f16_f32(105, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_cube_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_1darray_v4f16_f32(105, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_1darray_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_1darray_v4f16_f32(105, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_1darray_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_d_1darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_1darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_1darray_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_lz_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2darray_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_l_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_v4f16_f32(100, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2darray_v4f16_f32' needs target feature extended-image-insts}} +} + +half4 test_amdgcn_image_sample_d_2darray_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_v4f16_f32(100, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2darray_v4f16_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_lz_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2d_f32_f32(1, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2d_f32_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_l_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2d_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2d_f32_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_d_2d_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2d_f32_f32(1, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2d_f32_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_lz_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_lz_2darray_f32_f32(1, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_lz_2darray_f32_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_l_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_l_2darray_f32_f32(1, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_l_2darray_f32_f32' needs target feature extended-image-insts}} +} + +float test_amdgcn_image_sample_d_2darray_f32_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { + + return __builtin_amdgcn_image_sample_d_2darray_f32_f32(1, f32, f32, f32, f32, f32, f32, f32, tex, vec4i32, 0, 101, 121); //GFX94-error{{'test_amdgcn_image_sample_d_2darray_f32_f32' needs target feature extended-image-insts}} +} diff --git a/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl index 8f609dcbd34f2..c0c462b018300 100644 --- a/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl @@ -3,7 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); float test_builtin_image_load_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl index b8780024f1076..b779c247ca003 100644 --- a/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl @@ -3,7 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); float test_builtin_image_load_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl index 4f6347e1c5286..0c2b821e7bacd 100644 --- a/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl @@ -2,7 +2,7 @@ // REQUIRES: amdgpu-registered-target typedef float float4 __attribute__((ext_vector_type(4))); -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); void test_builtin_image_store_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl index d0085e5403b5f..b41bc25f98b2c 100644 --- a/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl @@ -2,7 +2,7 @@ // REQUIRES: amdgpu-registered-target typedef float float4 __attribute__((ext_vector_type(4))); -typedef half half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half4 __attribute__((ext_vector_type(4))); void test_builtin_image_store_2d(float f32, int i32, __amdgpu_texture_t tex) { From d37ea0257c3ad7108e4099b60e32243616e31097 Mon Sep 17 00:00:00 2001 From: ranapratap55 Date: Sat, 25 Oct 2025 11:16:04 +0530 Subject: [PATCH 3/5] Minor modification to format changes --- clang/include/clang/Basic/BuiltinsAMDGPU.def | 60 +++++++++---------- clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp | 52 ++++++++-------- clang/test/CodeGen/builtins-image-load.c | 3 +- clang/test/CodeGen/builtins-image-store.c | 3 +- .../builtins-image-load-param-gfx1100-err.cl | 2 +- .../builtins-image-load-param-gfx942-err.cl | 2 +- .../builtins-image-store-param-gfx1100-err.cl | 2 +- .../builtins-image-store-param-gfx942-err.cl | 2 +- 8 files changed, 65 insertions(+), 61 deletions(-) diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def index 6947162a3b962..0bad56ead8aaf 100644 --- a/clang/include/clang/Basic/BuiltinsAMDGPU.def +++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def @@ -889,75 +889,75 @@ TARGET_BUILTIN(__builtin_amdgcn_cooperative_atomic_store_8x16B, "vV4i*V4iIicC*", // Image builtins //===----------------------------------------------------------------------===// TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f32_i32, "V4fiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f16_i32, "V4xiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_1d_v4f16_i32, "V4hiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_1darray_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_f32_i32, "fiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_2d_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_f32_i32, "fiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_2darray_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_3d_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_cube_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f32_i32, "V4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f16_i32, "V4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1d_v4f16_i32, "V4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_1darray_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_f32_i32, "fiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f32_i32, "V4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f16_i32, "V4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2d_v4f16_i32, "V4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_f32_i32, "fiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_2darray_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_3d_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f32_i32, "V4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f16_i32, "V4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_load_mip_cube_v4f16_i32, "V4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f32_i32, "vV4fiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f16_i32, "vV4xiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_1d_v4f16_i32, "vV4hiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_1darray_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_f32_i32, "vfiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_2d_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_f32_i32, "vfiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_2darray_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_3d_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_cube_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f32_i32, "vV4fiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f16_i32, "vV4xiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1d_v4f16_i32, "vV4hiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_1darray_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_f32_i32, "vfiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f32_i32, "vV4fiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f16_i32, "vV4xiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2d_v4f16_i32, "vV4hiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_f32_i32, "vfiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_2darray_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_3d_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f32_i32, "vV4fiiiiiQtii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f16_i32, "vV4xiiiiiQtii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_store_mip_cube_v4f16_i32, "vV4hiiiiiQtii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f16_f32, "V4xifQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_1d_v4f16_f32, "V4hifQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f16_f32, "V4xiffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_1darray_v4f16_f32, "V4hiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_f32_f32, "fiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f16_f32, "V4xiffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_2d_v4f16_f32, "V4hiffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_f32_f32, "fifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_2darray_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4xifffQtV4ibii", "nc", "image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4xifQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") diff --git a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp index 23bb9ff40a97b..1bef56120df1f 100644 --- a/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp +++ b/clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp @@ -133,8 +133,8 @@ Value *EmitAMDGPUWorkGroupSize(CodeGenFunction &CGF, unsigned Index) { } llvm::MDBuilder MDHelper(CGF.getLLVMContext()); - llvm::MDNode *RNode = MDHelper.createRange( - APInt(16, 1), APInt(16, CGF.getTarget().getMaxOpenCLWorkGroupSize() + 1)); + llvm::MDNode *RNode = MDHelper.createRange(APInt(16, 1), + APInt(16, CGF.getTarget().getMaxOpenCLWorkGroupSize() + 1)); LD->setMetadata(llvm::LLVMContext::MD_range, RNode); LD->setMetadata(llvm::LLVMContext::MD_noundef, llvm::MDNode::get(CGF.getLLVMContext(), {})); @@ -253,7 +253,8 @@ emitAMDGCNImageOverloadedReturnType(clang::CodeGen::CodeGenFunction &CGF, } // Emit an intrinsic that has 1 float or double operand, and 1 integer. -static Value *emitFPIntBuiltin(CodeGenFunction &CGF, const CallExpr *E, +static Value *emitFPIntBuiltin(CodeGenFunction &CGF, + const CallExpr *E, unsigned IntrinsicID) { llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0)); llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1)); @@ -455,8 +456,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Y = EmitScalarExpr(E->getArg(1)); llvm::Value *Z = EmitScalarExpr(E->getArg(2)); - llvm::Function *Callee = - CGM.getIntrinsic(Intrinsic::amdgcn_div_scale, X->getType()); + llvm::Function *Callee = CGM.getIntrinsic(Intrinsic::amdgcn_div_scale, + X->getType()); llvm::Value *Tmp = Builder.CreateCall(Callee, {X, Y, Z}); @@ -476,8 +477,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Src2 = EmitScalarExpr(E->getArg(2)); llvm::Value *Src3 = EmitScalarExpr(E->getArg(3)); - llvm::Function *F = - CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas, Src0->getType()); + llvm::Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas, + Src0->getType()); llvm::Value *Src3ToBool = Builder.CreateIsNotNull(Src3); return Builder.CreateCall(F, {Src0, Src1, Src2, Src3ToBool}); } @@ -618,13 +619,13 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, case AMDGPU::BI__builtin_amdgcn_frexp_expf: { Value *Src0 = EmitScalarExpr(E->getArg(0)); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_frexp_exp, - {Builder.getInt32Ty(), Src0->getType()}); + { Builder.getInt32Ty(), Src0->getType() }); return Builder.CreateCall(F, Src0); } case AMDGPU::BI__builtin_amdgcn_frexp_exph: { Value *Src0 = EmitScalarExpr(E->getArg(0)); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_frexp_exp, - {Builder.getInt16Ty(), Src0->getType()}); + { Builder.getInt16Ty(), Src0->getType() }); return Builder.CreateCall(F, Src0); } case AMDGPU::BI__builtin_amdgcn_fract: @@ -670,8 +671,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // FIXME-GFX10: How should 32 bit mask be handled? Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_icmp, - {Builder.getInt64Ty(), Src0->getType()}); - return Builder.CreateCall(F, {Src0, Src1, Src2}); + { Builder.getInt64Ty(), Src0->getType() }); + return Builder.CreateCall(F, { Src0, Src1, Src2 }); } case AMDGPU::BI__builtin_amdgcn_fcmp: case AMDGPU::BI__builtin_amdgcn_fcmpf: { @@ -681,8 +682,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // FIXME-GFX10: How should 32 bit mask be handled? Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_fcmp, - {Builder.getInt64Ty(), Src0->getType()}); - return Builder.CreateCall(F, {Src0, Src1, Src2}); + { Builder.getInt64Ty(), Src0->getType() }); + return Builder.CreateCall(F, { Src0, Src1, Src2 }); } case AMDGPU::BI__builtin_amdgcn_class: case AMDGPU::BI__builtin_amdgcn_classf: @@ -694,12 +695,11 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Intrinsic::amdgcn_fmed3); case AMDGPU::BI__builtin_amdgcn_ds_append: case AMDGPU::BI__builtin_amdgcn_ds_consume: { - Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append - ? Intrinsic::amdgcn_ds_append - : Intrinsic::amdgcn_ds_consume; + Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append ? + Intrinsic::amdgcn_ds_append : Intrinsic::amdgcn_ds_consume; Value *Src0 = EmitScalarExpr(E->getArg(0)); - Function *F = CGM.getIntrinsic(Intrin, {Src0->getType()}); - return Builder.CreateCall(F, {Src0, Builder.getFalse()}); + Function *F = CGM.getIntrinsic(Intrin, { Src0->getType() }); + return Builder.CreateCall(F, { Src0, Builder.getFalse() }); } case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_i32: case AMDGPU::BI__builtin_amdgcn_global_load_tr_b64_v2i32: @@ -919,10 +919,12 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, // The builtins take these arguments as vec4 where the last element is // ignored. The intrinsic takes them as vec3. - RayOrigin = Builder.CreateShuffleVector(RayOrigin, RayOrigin, {0, 1, 2}); - RayDir = Builder.CreateShuffleVector(RayDir, RayDir, {0, 1, 2}); - RayInverseDir = - Builder.CreateShuffleVector(RayInverseDir, RayInverseDir, {0, 1, 2}); + RayOrigin = Builder.CreateShuffleVector(RayOrigin, RayOrigin, + {0, 1, 2}); + RayDir = + Builder.CreateShuffleVector(RayDir, RayDir, {0, 1, 2}); + RayInverseDir = Builder.CreateShuffleVector(RayInverseDir, RayInverseDir, + {0, 1, 2}); Function *F = CGM.getIntrinsic(Intrinsic::amdgcn_image_bvh_intersect_ray, {NodePtr->getType(), RayDir->getType()}); @@ -996,8 +998,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, Value *Rtn = Builder.CreateExtractValue(Call, 0); Value *A = Builder.CreateExtractValue(Call, 1); llvm::Type *RetTy = ConvertType(E->getType()); - Value *I0 = - Builder.CreateInsertElement(PoisonValue::get(RetTy), Rtn, (uint64_t)0); + Value *I0 = Builder.CreateInsertElement(PoisonValue::get(RetTy), Rtn, + (uint64_t)0); // ds_bvh_stack_push8_pop2_rtn returns {i64, i32} but the builtin returns // <2 x i64>, zext the second value. if (A->getType()->getPrimitiveSizeInBits() < @@ -1702,7 +1704,7 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID, llvm::Value *Src1 = EmitScalarExpr(E->getArg(1)); llvm::Value *Src2 = EmitScalarExpr(E->getArg(2)); Function *F = CGM.getIntrinsic(Intrinsic::fshr, Src0->getType()); - return Builder.CreateCall(F, {Src0, Src1, Src2}); + return Builder.CreateCall(F, { Src0, Src1, Src2 }); } case AMDGPU::BI__builtin_amdgcn_fence: { ProcessOrderScopeAMDGCN(EmitScalarExpr(E->getArg(0)), diff --git a/clang/test/CodeGen/builtins-image-load.c b/clang/test/CodeGen/builtins-image-load.c index e455e8e0a8de4..8442124416338 100644 --- a/clang/test/CodeGen/builtins-image-load.c +++ b/clang/test/CodeGen/builtins-image-load.c @@ -3,7 +3,8 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half; +typedef half half4 __attribute__((ext_vector_type(4))); // CHECK-LABEL: define dso_local float @test_builtin_image_load_2d( // CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0:[0-9]+]] { diff --git a/clang/test/CodeGen/builtins-image-store.c b/clang/test/CodeGen/builtins-image-store.c index f930d3e786d51..5309a16df7033 100644 --- a/clang/test/CodeGen/builtins-image-store.c +++ b/clang/test/CodeGen/builtins-image-store.c @@ -2,7 +2,8 @@ // RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 %s -emit-llvm -o - | FileCheck %s typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef _Float16 half; +typedef half half4 __attribute__((ext_vector_type(4))); // CHECK-LABEL: define dso_local void @test_builtin_image_store_2d( // CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0:[0-9]+]] { diff --git a/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl index c0c462b018300..8f609dcbd34f2 100644 --- a/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-load-param-gfx1100-err.cl @@ -3,7 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); float test_builtin_image_load_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl index b779c247ca003..b8780024f1076 100644 --- a/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-load-param-gfx942-err.cl @@ -3,7 +3,7 @@ typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); float test_builtin_image_load_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl index 0c2b821e7bacd..4f6347e1c5286 100644 --- a/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-store-param-gfx1100-err.cl @@ -2,7 +2,7 @@ // REQUIRES: amdgpu-registered-target typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); void test_builtin_image_store_2d(float f32, int i32, __amdgpu_texture_t tex) { diff --git a/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl index b41bc25f98b2c..d0085e5403b5f 100644 --- a/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl +++ b/clang/test/SemaOpenCL/builtins-image-store-param-gfx942-err.cl @@ -2,7 +2,7 @@ // REQUIRES: amdgpu-registered-target typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); void test_builtin_image_store_2d(float f32, int i32, __amdgpu_texture_t tex) { From 60d2b3d9e81665a51e53f59acb648e6c605ebf08 Mon Sep 17 00:00:00 2001 From: ranapratap55 Date: Wed, 29 Oct 2025 22:28:14 +0530 Subject: [PATCH 4/5] Adds language sensitive mapping for float16 builtin for HIP and OpenCL --- clang/lib/AST/ASTContext.cpp | 2 +- .../SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl | 6 ++++-- .../SemaOpenCL/builtins-extended-image-param-gfx942-err.cl | 4 +++- 3 files changed, 8 insertions(+), 4 deletions(-) diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp index 32c8f6209a693..00af2f29afab3 100644 --- a/clang/lib/AST/ASTContext.cpp +++ b/clang/lib/AST/ASTContext.cpp @@ -12404,7 +12404,7 @@ static QualType DecodeTypeFromStr(const char *&Str, const ASTContext &Context, case 'x': assert(HowLong == 0 && !Signed && !Unsigned && "Bad modifiers used with 'x'!"); - Type = Context.Float16Ty; + Type = Context.getLangOpts().OpenCL ? Context.HalfTy : Context.Float16Ty; break; case 'y': assert(HowLong == 0 && !Signed && !Unsigned && diff --git a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl index 0f87f9850b49d..47dbdd4e51782 100644 --- a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl +++ b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx1100-err.cl @@ -1,9 +1,11 @@ // RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1100 -target-feature +extended-image-insts -S -verify=expected -o - %s // REQUIRES: amdgpu-registered-target +#pragma OPENCL EXTENSION cl_khr_fp16 : enable + typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { @@ -111,7 +113,7 @@ float4 test_amdgcn_image_sample_d_2darray_v4f32_f32(float4 v4f32, float f32, int half4 test_amdgcn_image_sample_lz_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { - return __builtin_amdgcn_image_sample_lz_1d_v4f16_f32(i32, f32, tex, vec4i32, 0, f32, i32); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1d_v4f16_f32' must be a constant integer}} + return __builtin_amdgcn_image_sample_lz_1d_v4f16_f32(23, f32, tex, vec4i32, 0, i32, 11); //expected-error{{argument to '__builtin_amdgcn_image_sample_lz_1d_v4f16_f32' must be a constant integer}} } half4 test_amdgcn_image_sample_l_1d_v4f16_f32(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { diff --git a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl index 80e2130831580..e60f8c70dc7c4 100644 --- a/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl +++ b/clang/test/SemaOpenCL/builtins-extended-image-param-gfx942-err.cl @@ -1,9 +1,11 @@ // RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx942 -verify=GFX94 -S -o - %s // REQUIRES: amdgpu-registered-target +#pragma OPENCL EXTENSION cl_khr_fp16 : enable + typedef int int4 __attribute__((ext_vector_type(4))); typedef float float4 __attribute__((ext_vector_type(4))); -typedef _Float16 half4 __attribute__((ext_vector_type(4))); +typedef half half4 __attribute__((ext_vector_type(4))); float4 test_amdgcn_image_gather4_lz_2d_v4f32_f32_r(float4 v4f32, float f32, int i32, __amdgpu_texture_t tex, int4 vec4i32) { From 310846d93f91df08d1588c76f5ce18800c21b820 Mon Sep 17 00:00:00 2001 From: ranapratap55 Date: Thu, 30 Oct 2025 15:03:37 +0530 Subject: [PATCH 5/5] Rename the mapping from 'x' to 'e' --- clang/include/clang/Basic/Builtins.def | 1 + clang/include/clang/Basic/BuiltinsAMDGPU.def | 34 ++++++++++---------- clang/lib/AST/ASTContext.cpp | 7 +++- 3 files changed, 24 insertions(+), 18 deletions(-) diff --git a/clang/include/clang/Basic/Builtins.def b/clang/include/clang/Basic/Builtins.def index b856ad145824d..3a5b72e20afab 100644 --- a/clang/include/clang/Basic/Builtins.def +++ b/clang/include/clang/Basic/Builtins.def @@ -43,6 +43,7 @@ // SJ -> sigjmp_buf // K -> ucontext_t // p -> pid_t +// e -> _Float16 for HIP/C++ and __fp16 for OpenCL // . -> "...". This may only occur at the end of the function list. // // Types may be prefixed with the following modifiers: diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def index 0bad56ead8aaf..4e58f5055fb56 100644 --- a/clang/include/clang/Basic/BuiltinsAMDGPU.def +++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def @@ -959,45 +959,45 @@ TARGET_BUILTIN(__builtin_amdgcn_image_sample_3d_v4f16_f32, "V4hifffQtV4ibii", "n TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_cube_v4f16_f32, "V4hifffQtV4ibii", "nc", "image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f32_f32, "V4fifQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4xifQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1d_v4f16_f32, "V4eifQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_1darray_v4f16_f32, "V4eiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_f32_f32, "fiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2d_v4f16_f32, "V4eiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_2darray_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_3d_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_lz_cube_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f16_f32, "V4xiffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1d_v4f16_f32, "V4eiffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_1darray_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_f32_f32, "fifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2d_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_f32_f32, "fiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_2darray_v4f16_f32, "V4eiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_3d_v4f16_f32, "V4eiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_l_cube_v4f16_f32, "V4eiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f32_f32, "V4fifffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f16_f32, "V4xifffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1d_v4f16_f32, "V4eifffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f32_f32, "V4fiffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f16_f32, "V4xiffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_1darray_v4f16_f32, "V4eiffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_f32_f32, "fiffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f32_f32, "V4fiffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f16_f32, "V4xiffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2d_v4f16_f32, "V4eiffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_f32_f32, "fifffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f32_f32, "V4fifffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f16_f32, "V4xifffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_2darray_v4f16_f32, "V4eifffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f32_f32, "V4fifffffffffQtV4ibii", "nc", "extended-image-insts") -TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f16_f32, "V4xifffffffffQtV4ibii", "nc", "extended-image-insts") +TARGET_BUILTIN(__builtin_amdgcn_image_sample_d_3d_v4f16_f32, "V4eifffffffffQtV4ibii", "nc", "extended-image-insts") TARGET_BUILTIN(__builtin_amdgcn_image_gather4_lz_2d_v4f32_f32, "V4fiffQtV4ibii", "nc", "extended-image-insts") #undef BUILTIN diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp index 00af2f29afab3..4beabea1274ec 100644 --- a/clang/lib/AST/ASTContext.cpp +++ b/clang/lib/AST/ASTContext.cpp @@ -12401,10 +12401,15 @@ static QualType DecodeTypeFromStr(const char *&Str, const ASTContext &Context, // Read the base type. switch (*Str++) { default: llvm_unreachable("Unknown builtin type letter!"); + case 'e': + assert(HowLong == 0 && !Signed && !Unsigned && + "Bad modifiers used with 'e'!"); + Type = Context.getLangOpts().OpenCL ? Context.HalfTy : Context.Float16Ty; + break; case 'x': assert(HowLong == 0 && !Signed && !Unsigned && "Bad modifiers used with 'x'!"); - Type = Context.getLangOpts().OpenCL ? Context.HalfTy : Context.Float16Ty; + Type = Context.Float16Ty; break; case 'y': assert(HowLong == 0 && !Signed && !Unsigned &&