AMDGPU/clang: Add builtins for llvm.amdgcn.ballot

[arsenm]

Summary

I wasn't sure what the best strategy was for the wave size
difference. I went for an explicit, enforced builtin for each. The
other option would be to just assume wave64, and IRGen the different
mangling + zext. I didn't see an obvious way to check the wave size
where builtins are emitted, and it might be beneficial to force you to
acknowledge the wave size difference? Or it might be an unnecessary
complication.

The behavior is also slightly odd when directly specifying
-target-feature to cc1 for +/- the size (since we have both positive
and negative forms of both sizes), but this is probably unimportant.

We're also still missing a predefined macro for the wave size, which
we probably need.

[arsenm]

ping

[arsenm]

ping

[sameerds]

The documentation for HIP __ballot seems to indicate that the user does not have to explicitly specify the warp size. How is that achieved with these new builtins? Can this be captured in a lit test here?

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

clang/lib/Basic/Targets/AMDGPU.cpp

348

So the implication here is that wave32 is the preferred choice on newer architectures, and hence the default when available?

[yaxunl]

clang/lib/Basic/Targets/AMDGPU.cpp

353

what's the default wave front size in backend for gfx10* before this change?

clang/test/CodeGenOpenCL/amdgpu-features.cl

7

what happens if both +wavefrontsize32 and +wavefrontsize64 are specified?

[arsenm]

In D82087#2140778, @sameerds wrote:

The documentation for HIP __ballot seems to indicate that the user does not have to explicitly specify the warp size. How is that achieved with these new builtins? Can this be captured in a lit test here?

This seems like a defect in the design to me

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

[arsenm]

clang/lib/Basic/Targets/AMDGPU.cpp

348

Yes, this has always been the case

353

It's always been wave32

[arsenm]

In D82087#2140778, @sameerds wrote:

The documentation for HIP __ballot seems to indicate that the user does not have to explicitly specify the warp size. How is that achieved with these new builtins? Can this be captured in a lit test here?

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

I think if the language interface insists on fixing the wave size, then I think the correct solution is to implement this in the header based on a wave size macro (which we're desperately missing). The library implementation should be responsible for inserting the extension to 64-bit for wave32

[sameerds]

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

I think if the language interface insists on fixing the wave size, then I think the correct solution is to implement this in the header based on a wave size macro (which we're desperately missing). The library implementation should be responsible for inserting the extension to 64-bit for wave32

Not sure if the frontend should try to infer warpsize and the mask size, or even whether it can in all cases. But this can result in wrong behaviour when the program passes 32-bit mask but then gets compiled for a 64-bit mask. It's easy to say that the programmer must not assume a warp-size, but it would be useful if the language can altogether avoid the confusion.

In D82087#2144712, @arsenm wrote:

In D82087#2140778, @sameerds wrote:

The documentation for HIP __ballot seems to indicate that the user does not have to explicitly specify the warp size. How is that achieved with these new builtins? Can this be captured in a lit test here?

This seems like a defect in the design to me

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

I tend to agree. The HIP vote/ballot builtins are also missing a mask parameter, whose type needs to match the wavesize.

[yaxunl]

In D82087#2146170, @sameerds wrote:

https://github.com/ROCm-Developer-Tools/HIP/blob/master/docs/markdown/hip_kernel_language.md#warp-vote-and-ballot-functions

I think if the language interface insists on fixing the wave size, then I think the correct solution is to implement this in the header based on a wave size macro (which we're desperately missing). The library implementation should be responsible for inserting the extension to 64-bit for wave32

Agree that FE should have a predefined macro for wave front size. Since it is per amdgpu target and not per language, it should be named as __amdgpu_wavefront_size__ or something similar, then it could be used by all languages.

Then we need to initialize warpSize in HIP header by this macro
https://github.com/ROCm-Developer-Tools/HIP/blob/386a0e0123d67b95b4c0ebb3ebcf1d1615758146/include/hip/hcc_detail/device_functions.h#L300

I tends to think we should define __ballot in HIP header conditionally by the wavefront size so that the return type is correct for both wave32 and wave64 mode. We should assume normal HIP compilation always have -mcpu specified so that wavefront size is known at compile time. If -mcpu is not specified probably we should not define warpSize or __ballot.

[JonChesterfield]

A macro for wavefront size would make targeting gfx10 for openmp easier.

We currently use an int32_t for nvptx and an int64_t for amdgcn in various runtime function interfaces. I'd like to be able to set the latter based on said macro.

[jdoerfert]

Can we land this? I'd like to use the new intrinsics as I don't understand the old ones.

[arsenm]

In D82087#3797883, @jdoerfert wrote:

Can we land this? I'd like to use the new intrinsics as I don't understand the old ones.

What do you think about using the two separate builtins, vs. one magic builtin that auto-changes the wavesize?

[JonChesterfield]

In D82087#3803464, @arsenm wrote:

In D82087#3797883, @jdoerfert wrote:

Can we land this? I'd like to use the new intrinsics as I don't understand the old ones.

What do you think about using the two separate builtins, vs. one magic builtin that auto-changes the wavesize?

The magic one would also change its return type, or always be i64 with high bits (zext or maybe sext or maybe copy of low), so less magic seems clearer

[arsenm]

Rebase. Use _w32/_w64 suffixes since some other wave specific builtins seem to have gone with that convention

[sameerds]

clang/test/CodeGenOpenCL/amdgpu-features.cl

7

Shouldn't this be separately an error in itself? Is it tested elsewhere?

[arsenm]

clang/test/CodeGenOpenCL/amdgpu-features.cl

7

It looks like you end up with both features set by clang, and wave64 wins in codegen

[arsenm]

This doesn't work correctly for unspecified wavesize for non-wave32 targets

[arsenm]

Fix unknown target handling, diagnose some more of the errors

[sameerds]

clang/lib/Basic/Targets/AMDGPU.cpp

353

I would have preferred this to be a separate change, just like the FIXME for diagnosing wavefrontsize32 on targets that don't support it. But not feeling strongly enough to block this change!

[sameerds]

And note that the change description is written in a first-person train of thought. Please do rewrite it!

[arsenm]

f4bcd7f598331457cfe74e459b489d4098369511

clang/include/clang/Basic/BuiltinsAMDGPU.def

 BUILTIN(__builtin_amdgcn_cubeid, "ffff", "nc")
 BUILTIN(__builtin_amdgcn_cubesc, "ffff", "nc")
 BUILTIN(__builtin_amdgcn_cubetc, "ffff", "nc")
 BUILTIN(__builtin_amdgcn_cubema, "ffff", "nc")
 BUILTIN(__builtin_amdgcn_s_sleep, "vIi", "n")
 BUILTIN(__builtin_amdgcn_s_incperflevel, "vIi", "n")
 BUILTIN(__builtin_amdgcn_s_decperflevel, "vIi", "n")
 BUILTIN(__builtin_amdgcn_s_setprio, "vIs", "n")
-BUILTIN(__builtin_amdgcn_uicmp, "WUiUiUiIi", "nc")
-BUILTIN(__builtin_amdgcn_uicmpl, "WUiWUiWUiIi", "nc")
-BUILTIN(__builtin_amdgcn_sicmp, "WUiiiIi", "nc")
-BUILTIN(__builtin_amdgcn_sicmpl, "WUiWiWiIi", "nc")
-BUILTIN(__builtin_amdgcn_fcmp, "WUiddIi", "nc")
-BUILTIN(__builtin_amdgcn_fcmpf, "WUiffIi", "nc")
 BUILTIN(__builtin_amdgcn_ds_swizzle, "iiIi", "nc")
 BUILTIN(__builtin_amdgcn_ds_permute, "iii", "nc")
 BUILTIN(__builtin_amdgcn_ds_bpermute, "iii", "nc")
 BUILTIN(__builtin_amdgcn_readfirstlane, "ii", "nc")
 BUILTIN(__builtin_amdgcn_readlane, "iii", "nc")
 BUILTIN(__builtin_amdgcn_fmed3f, "ffff", "nc")
 BUILTIN(__builtin_amdgcn_ds_faddf, "ff*3fIiIiIb", "n")
 BUILTIN(__builtin_amdgcn_ds_fminf, "ff*3fIiIiIb", "n")

 BUILTIN(__builtin_amdgcn_msad_u8, "UiUiUiUi", "nc")
 BUILTIN(__builtin_amdgcn_sad_hi_u8, "UiUiUiUi", "nc")
 BUILTIN(__builtin_amdgcn_sad_u16, "UiUiUiUi", "nc")
 BUILTIN(__builtin_amdgcn_qsad_pk_u16_u8, "WUiWUiUiWUi", "nc")
 BUILTIN(__builtin_amdgcn_mqsad_pk_u16_u8, "WUiWUiUiWUi", "nc")
 BUILTIN(__builtin_amdgcn_mqsad_u32_u8, "V4UiWUiUiV4Ui", "nc")
 
 //===----------------------------------------------------------------------===//
+// Ballot builtins.
+//===----------------------------------------------------------------------===//
+
+TARGET_BUILTIN(__builtin_amdgcn_ballot_w32, "Uib", "nc", "wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_ballot_w64, "LUib", "nc", "wavefrontsize64")
+
+// Deprecated intrinsics in favor of __builtin_amdgn_ballot_{w32|w64}
+BUILTIN(__builtin_amdgcn_uicmp, "WUiUiUiIi", "nc")
+BUILTIN(__builtin_amdgcn_uicmpl, "WUiWUiWUiIi", "nc")
+BUILTIN(__builtin_amdgcn_sicmp, "WUiiiIi", "nc")
+BUILTIN(__builtin_amdgcn_sicmpl, "WUiWiWiIi", "nc")
+BUILTIN(__builtin_amdgcn_fcmp, "WUiddIi", "nc")
+BUILTIN(__builtin_amdgcn_fcmpf, "WUiffIi", "nc")
+
+//===----------------------------------------------------------------------===//
 // CI+ only builtins.
 //===----------------------------------------------------------------------===//
 TARGET_BUILTIN(__builtin_amdgcn_s_dcache_inv_vol, "v", "n", "ci-insts")
 TARGET_BUILTIN(__builtin_amdgcn_buffer_wbinvl1_vol, "v", "n", "ci-insts")
 TARGET_BUILTIN(__builtin_amdgcn_ds_gws_sema_release_all, "vUi", "n", "ci-insts")
 TARGET_BUILTIN(__builtin_amdgcn_is_shared, "bvC*0", "nc", "flat-address-space")
 TARGET_BUILTIN(__builtin_amdgcn_is_private, "bvC*0", "nc", "flat-address-space")
 

clang/lib/Basic/Targets/AMDGPU.cpp

 //===--- AMDGPU.cpp - Implement AMDGPU target feature support -------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements AMDGPU TargetInfo objects.
 //
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPU.h"
 #include "clang/Basic/Builtins.h"
 #include "clang/Basic/CodeGenOptions.h"
+#include "clang/Basic/Diagnostic.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
 #include "clang/Basic/TargetBuiltins.h"
 
 using namespace clang;
 using namespace clang::targets;
 
 namespace clang {

 
 ArrayRef<const char *> AMDGPUTargetInfo::getGCCRegNames() const {
   return llvm::makeArrayRef(GCCRegNames);
 }
 
 bool AMDGPUTargetInfo::initFeatureMap(
     llvm::StringMap<bool> &Features, DiagnosticsEngine &Diags, StringRef CPU,
     const std::vector<std::string> &FeatureVec) const {
+  const bool IsNullCPU = CPU.empty();
+  bool IsWave32Capable = false;
 
   using namespace llvm::AMDGPU;
 
   // XXX - What does the member GPU mean if device name string passed here?
   if (isAMDGCN(getTriple())) {
     switch (llvm::AMDGPU::parseArchAMDGCN(CPU)) {
     case GK_GFX1103:
     case GK_GFX1102:
     case GK_GFX1101:
     case GK_GFX1100:
+      IsWave32Capable = true;
       Features["ci-insts"] = true;
       Features["dot1-insts"] = true;
       Features["dot5-insts"] = true;
       Features["dot6-insts"] = true;
       Features["dot7-insts"] = true;
       Features["dot8-insts"] = true;
       Features["dl-insts"] = true;
       Features["flat-address-space"] = true;
       Features["16-bit-insts"] = true;
       Features["dpp"] = true;
       Features["gfx8-insts"] = true;
       Features["gfx9-insts"] = true;
       Features["gfx10-insts"] = true;
       Features["gfx10-3-insts"] = true;
       Features["gfx11-insts"] = true;
       break;
     case GK_GFX1036:
     case GK_GFX1035:
     case GK_GFX1034:
     case GK_GFX1033:
     case GK_GFX1032:
     case GK_GFX1031:
     case GK_GFX1030:
+      IsWave32Capable = true;
       Features["ci-insts"] = true;
       Features["dot1-insts"] = true;
       Features["dot2-insts"] = true;
       Features["dot5-insts"] = true;
       Features["dot6-insts"] = true;
       Features["dot7-insts"] = true;
       Features["dl-insts"] = true;
       Features["flat-address-space"] = true;
case GK_GFX1011:
       Features["dot1-insts"] = true;
       Features["dot2-insts"] = true;
       Features["dot5-insts"] = true;
       Features["dot6-insts"] = true;
       Features["dot7-insts"] = true;
       [[fallthrough]];
     case GK_GFX1013:
     case GK_GFX1010:
+      IsWave32Capable = true;
       Features["dl-insts"] = true;
       Features["ci-insts"] = true;
       Features["flat-address-space"] = true;
       Features["16-bit-insts"] = true;
       Features["dpp"] = true;
       Features["gfx8-insts"] = true;
       Features["gfx9-insts"] = true;
       Features["gfx10-insts"] = true;
if (isAMDGCN(getTriple())) {
     case GK_R630:
     case GK_R600:
       break;
     default:
       llvm_unreachable("Unhandled GPU!");
     }
   }
 
-  return TargetInfo::initFeatureMap(Features, Diags, CPU, FeatureVec);
+  if (!TargetInfo::initFeatureMap(Features, Diags, CPU, FeatureVec))
+    return false;
+
+  // FIXME: Not diagnosing wavefrontsize32 on wave64 only targets.
+  const bool HaveWave32 =
+      (IsWave32Capable || IsNullCPU) && Features.count("wavefrontsize32");
+  const bool HaveWave64 = Features.count("wavefrontsize64");
+
+  // TODO: Should move this logic into TargetParser
+  if (HaveWave32 && HaveWave64) {
+    Diags.Report(diag::err_invalid_feature_combination)
+        << "'wavefrontsize32' and 'wavefrontsize64' are mutually exclusive";
+    return false;
+  }
+
+  // Don't assume any wavesize with an unknown subtarget.
+  if (!IsNullCPU) {
+    // Default to wave32 if available, or wave64 if not
+    if (!HaveWave32 && !HaveWave64) {
+      StringRef DefaultWaveSizeFeature =
+          IsWave32Capable ? "wavefrontsize32" : "wavefrontsize64";
+      Features.insert(std::make_pair(DefaultWaveSizeFeature, true));
+    }
+  }
+
+  return true;
 }
 
 void AMDGPUTargetInfo::fillValidCPUList(
     SmallVectorImpl<StringRef> &Values) const {
   if (isAMDGCN(getTriple()))
     llvm::AMDGPU::fillValidArchListAMDGCN(Values);
   else
     llvm::AMDGPU::fillValidArchListR600(Values);

clang/lib/CodeGen/CGBuiltin.cpp

   case AMDGPU::BI__builtin_amdgcn_fracth:
     return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_fract);
   case AMDGPU::BI__builtin_amdgcn_lerp:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_lerp);
   case AMDGPU::BI__builtin_amdgcn_ubfe:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_ubfe);
   case AMDGPU::BI__builtin_amdgcn_sbfe:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_sbfe);
+  case AMDGPU::BI__builtin_amdgcn_ballot_w32:
+  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 });
+  }
   case AMDGPU::BI__builtin_amdgcn_uicmp:
   case AMDGPU::BI__builtin_amdgcn_uicmpl:
   case AMDGPU::BI__builtin_amdgcn_sicmp:
   case AMDGPU::BI__builtin_amdgcn_sicmpl: {
     llvm::Value *Src0 = EmitScalarExpr(E->getArg(0));
     llvm::Value *Src1 = EmitScalarExpr(E->getArg(1));
     llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
 

clang/test/CodeGenOpenCL/amdgpu-features-illegal.cl

+// RUN: not %clang_cc1 -triple amdgcn -target-feature +wavefrontsize32 -target-feature +wavefrontsize64 -o /dev/null %s 2>&1 | FileCheck %s
+// RUN: not %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize32 -target-feature +wavefrontsize64 -o /dev/null %s 2>&1 | FileCheck %s
+
+// CHECK: error: invalid feature combination: 'wavefrontsize32' and 'wavefrontsize64' are mutually exclusive
+
+kernel void test() {}

clang/test/CodeGenOpenCL/amdgpu-features.cl

 // REQUIRES: amdgpu-registered-target
 
 // Check that appropriate features are defined for every supported AMDGPU
 // "-target" and "-mcpu" options.
 
+// RUN: %clang_cc1 -triple amdgcn -S -emit-llvm -o - %s | FileCheck --check-prefix=NOCPU %s
+// RUN: %clang_cc1 -triple amdgcn -target-feature +wavefrontsize32 -S -emit-llvm -o - %s | FileCheck --check-prefix=NOCPU-WAVE32 %s
+// RUN: %clang_cc1 -triple amdgcn -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck --check-prefix=NOCPU-WAVE64 %s
+
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx600 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX600 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx601 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX601 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx602 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX602 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx700 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX700 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx701 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX701 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx702 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX702 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx703 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX703 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx704 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX704 %s

 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1034 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1034 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1035 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1035 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1036 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1036 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1100 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1100 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1101 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1101 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1102 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1102 %s
 // RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103 %s
 
-// GFX600: "target-features"="+s-memtime-inst"
-// GFX601: "target-features"="+s-memtime-inst"
-// GFX602: "target-features"="+s-memtime-inst"
-// GFX700: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX701: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX702: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX703: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX704: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX705: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst"
-// GFX801: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst"
-// GFX802: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst"
-// GFX803: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst"
-// GFX805: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst"
-// GFX810: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst"
-// GFX900: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX902: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX904: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX906: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX908: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst"
-// GFX909: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX90A: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst"
-// GFX90C: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX940: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1013: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1030: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1031: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1032: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1033: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1034: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1035: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1036: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst"
-// GFX1100: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts"
-// GFX1101: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts"
-// GFX1102: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts"
-// GFX1103: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts"
+// RUN: %clang_cc1 -triple amdgcn -target-cpu gfx1103 -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck --check-prefix=GFX1103-W64 %s
+
+// NOCPU-NOT: "target-features"
+// NOCPU-WAVE32: "target-features"="+wavefrontsize32"
+// NOCPU-WAVE64: "target-features"="+wavefrontsize64"
+
+// GFX600: "target-features"="+s-memtime-inst,+wavefrontsize64"
+// GFX601: "target-features"="+s-memtime-inst,+wavefrontsize64"
+// GFX602: "target-features"="+s-memtime-inst,+wavefrontsize64"
+// GFX700: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX701: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX702: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX703: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX704: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX705: "target-features"="+ci-insts,+flat-address-space,+s-memtime-inst,+wavefrontsize64"
+// GFX801: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX802: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX803: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX805: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX810: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX900: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX902: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX904: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX906: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX908: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX909: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX90A: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX90C: "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX940: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1013: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+flat-address-space,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1030: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1031: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1032: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1033: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1034: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1035: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1036: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
+// GFX1100: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
+// GFX1101: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
+// GFX1102: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
+// GFX1103: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
+// GFX1103-W64: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dot8-insts,+dpp,+flat-address-space,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize64"
 
 kernel void test() {}

clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl

 }
 
 // CHECK-LABEL: @test_groupstaticsize
 // CHECK: call i32 @llvm.amdgcn.groupstaticsize()
 void test_groupstaticsize(global uint* out)
 {
   *out = __builtin_amdgcn_groupstaticsize();
 }
+
+// CHECK-LABEL: @test_ballot_wave32(
+// CHECK: call i32 @llvm.amdgcn.ballot.i32(i1 %{{.+}})
+void test_ballot_wave32(global uint* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ballot_w32(a == b);
+}

clang/test/CodeGenOpenCL/builtins-amdgcn-wave32.cl

+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -D__AMDGCN_WAVEFRONT_SIZE=32 -target-feature +wavefrontsize32 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx1010 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx1010 -target-feature +wavefrontsize32 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx1100 -target-feature +wavefrontsize32 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+
+typedef unsigned int uint;
+
+
+// CHECK-LABEL: @test_ballot_wave32(
+// CHECK: call i32 @llvm.amdgcn.ballot.i32(i1 %{{.+}})
+void test_ballot_wave32(global uint* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ballot_w32(a == b);
+}
+
+// CHECK-LABEL: @test_ballot_wave32_target_attr(
+// CHECK: call i32 @llvm.amdgcn.ballot.i32(i1 %{{.+}})
+__attribute__((target("wavefrontsize32")))
+void test_ballot_wave32_target_attr(global uint* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ballot_w32(a == b);
+}
+
+#if __AMDGCN_WAVEFRONT_SIZE != 32
+#error Wrong wavesize detected
+#endif

clang/test/CodeGenOpenCL/builtins-amdgcn-wave64.cl

+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx900 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx900 -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx1010 -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+// RUN: %clang_cc1 -cl-std=CL2.0 -triple amdgcn-unknown-unknown -target-cpu gfx1100 -target-feature +wavefrontsize64 -S -emit-llvm -o - %s | FileCheck -enable-var-scope %s
+
+typedef unsigned long ulong;
+
+// CHECK-LABEL: @test_ballot_wave64(
+// CHECK: call i64 @llvm.amdgcn.ballot.i64(i1 %{{.+}})
+void test_ballot_wave64(global ulong* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ballot_w64(a == b);
+}
+
+// CHECK-LABEL: @test_ballot_wave64_target_attr(
+// CHECK: call i64 @llvm.amdgcn.ballot.i64(i1 %{{.+}})
+__attribute__((target("wavefrontsize64")))
+void test_ballot_wave64_target_attr(global ulong* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ballot_w64(a == b);
+}
+
+#if __AMDGCN_WAVEFRONT_SIZE != 64
+#error Wrong wavesize detected
+#endif

clang/test/OpenMP/amdgcn-attributes.cpp

 }
 
 int callable(int x) {
   // ALL-LABEL: @_Z8callablei(i32 noundef %x) #1
   return x + 1;
 }
 
 // DEFAULT: attributes #0 = { convergent noinline norecurse nounwind optnone "frame-pointer"="none" "kernel" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="true" }
-// CPU: attributes #0 = { convergent noinline norecurse nounwind optnone "frame-pointer"="none" "kernel" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst" "uniform-work-group-size"="true" }
+// CPU: attributes #0 = { convergent noinline norecurse nounwind optnone "frame-pointer"="none" "kernel" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="true" }
 // NOIEEE: attributes #0 = { convergent noinline norecurse nounwind optnone "amdgpu-ieee"="false" "frame-pointer"="none" "kernel" "no-nans-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="true" }
 // UNSAFEATOMIC: attributes #0 = { convergent noinline norecurse nounwind optnone "amdgpu-unsafe-fp-atomics"="true" "frame-pointer"="none" "kernel" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="true" }
 
 // DEFAULT: attributes #1 = { convergent mustprogress noinline nounwind optnone "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
-// CPU: attributes #1 = { convergent mustprogress noinline nounwind optnone "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst" }
+// CPU: attributes #1 = { convergent mustprogress noinline nounwind optnone "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" }
 // NOIEEE: attributes #1 = { convergent mustprogress noinline nounwind optnone "amdgpu-ieee"="false" "frame-pointer"="none" "no-nans-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // UNSAFEATOMIC: attributes #1 = { convergent mustprogress noinline nounwind optnone "amdgpu-unsafe-fp-atomics"="true" "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }

clang/test/SemaOpenCL/builtins-amdgcn-error-wave32.cl

+// RUN: %clang_cc1 -triple amdgcn-- -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx900 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx900 -target-feature +wavefrontsize64 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 -target-feature +wavefrontsize64 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 -target-feature -wavefrontsize32 -verify -S -o - %s
+
+typedef unsigned int uint;
+
+void test_ballot_wave32(global uint* out, int a, int b) {
+  *out = __builtin_amdgcn_ballot_w32(a == b);  // expected-error {{'__builtin_amdgcn_ballot_w32' needs target feature wavefrontsize32}}
+}
+
+// FIXME: Should error for subtargets that don't support wave32
+__attribute__((target("wavefrontsize32")))
+void test_ballot_wave32_target_attr(global uint* out, int a, int b) {
+  *out = __builtin_amdgcn_ballot_w32(a == b);
+}

clang/test/SemaOpenCL/builtins-amdgcn-error-wave64.cl

+// RUN: %clang_cc1 -triple amdgcn-- -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-feature +wavefrontsize32 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 -target-feature +wavefrontsize32 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 -target-feature -wavefrontsize64 -verify -S -o - %s
+// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1010 -verify -S -o - %s
+
+typedef unsigned long ulong;
+
+void test_ballot_wave64(global ulong* out, int a, int b) {
+  *out = __builtin_amdgcn_ballot_w64(a == b);  // expected-error {{'__builtin_amdgcn_ballot_w64' needs target feature wavefrontsize64}}
+}
+
+__attribute__((target("wavefrontsize64")))
+void test_ballot_wave64_target_attr(global ulong* out, int a, int b) {
+  *out = __builtin_amdgcn_ballot_w64(a == b);
+}