[clang] [llvm] [HLSL] Add WaveMultiPrefixCountBits (PR #238132)

LLVM Continuous Integration via cfe-commits cfe-commits at lists.llvm.org
Fri Nov 6 01:27:58 PST 2026


llvm-ci wrote:

LLVM Buildbot has detected a new failure on builder `intel-sycl-gpu` running on `intel-sycl-gpu-01` while building `clang,llvm` at step 6 "test-build-unified-tree-check-all".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/225/builds/24704

<details>
<summary>Here is the relevant piece of the build log for the reference</summary>

```
Step 6 (test-build-unified-tree-check-all) failure: test (failure)
******************** TEST 'Clang :: CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl' FAILED ********************
Exit Code: 2

Command Output (stdout):
--
# RUN: at line 1
/home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/clang -cc1 -internal-isystem /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/lib/clang/24/include -nostdsysteminc -finclude-default-header    -triple dxil-pc-shadermodel6.5-library /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl -emit-llvm    -disable-llvm-passes -o - | /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/FileCheck /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl
# executed command: /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/clang -cc1 -internal-isystem /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/lib/clang/24/include -nostdsysteminc -finclude-default-header -triple dxil-pc-shadermodel6.5-library /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl -emit-llvm -disable-llvm-passes -o -
# executed command: /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/FileCheck /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl
# RUN: at line 4
/home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/clang -cc1 -internal-isystem /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/lib/clang/24/include -nostdsysteminc -finclude-default-header    -triple dxil-pc-shadermodel6.5-library /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl -emit-llvm    -disable-llvm-passes -o - |    /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/opt -S -strip-convergence-intrinsics -dxil-op-lower |    /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/FileCheck /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl --check-prefix=DXIL
# executed command: /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/bin/clang -cc1 -internal-isystem /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/build/lib/clang/24/include -nostdsysteminc -finclude-default-header -triple dxil-pc-shadermodel6.5-library /home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl -emit-llvm -disable-llvm-passes -o -
# .---command stdout------------
# | ; ModuleID = '/home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl'
# | source_filename = "/home/test-user/llvm-buildbot-worker/intel-sycl-gpu/llvm-project/clang/test/CodeGenHLSL/builtins/WaveMultiPrefixCountBits.hlsl"
# | target datalayout = "e-m:e-ve-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
# | target triple = "dxilv1.5-pc-shadermodel6.5-library"
# | 
# | ; Function Attrs: alwaysinline convergent mustprogress norecurse nounwind
# | define hidden noundef i32 @_Z9test_boolbDv4_j(i1 noundef %value, <4 x i32> noundef %mask) #0 {
# | entry:
# |   %0 = call token @llvm.experimental.convergence.entry()
# |   %value.addr = alloca i32, align 4
# |   %mask.addr = alloca <4 x i32>, align 4
# |   %storedv = zext i1 %value to i32
# |   store i32 %storedv, ptr %value.addr, align 4
# |   store <4 x i32> %mask, ptr %mask.addr, align 4
# |   %1 = load i32, ptr %value.addr, align 4
# |   %loadedv = icmp ne i32 %1, 0
# |   %2 = load <4 x i32>, ptr %mask.addr, align 4
# |   %3 = extractelement <4 x i32> %2, i64 0
# |   %4 = extractelement <4 x i32> %2, i64 1
# |   %5 = extractelement <4 x i32> %2, i64 2
# |   %6 = extractelement <4 x i32> %2, i64 3
# |   %hlsl.wave.multi.prefix.bit.count = call i32 @llvm.dx.wave.multi.prefix.bit.count(i1 %loadedv, i32 %3, i32 %4, i32 %5, i32 %6) [ "convergencectrl"(token %0) ]
# |   ret i32 %hlsl.wave.multi.prefix.bit.count
# | }
# | 
# | ; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)
# | declare token @llvm.experimental.convergence.entry() #1
# | 
# | ; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)
# | declare i32 @llvm.dx.wave.multi.prefix.bit.count(i1, i32, i32, i32, i32) #1
# | 
# | ; Function Attrs: alwaysinline convergent mustprogress norecurse nounwind
# | define hidden noundef i32 @_Z9test_truev() #0 {
# | entry:
# |   %0 = call token @llvm.experimental.convergence.entry()
# |   %hlsl.wave.multi.prefix.bit.count = call i32 @llvm.dx.wave.multi.prefix.bit.count(i1 true, i32 1, i32 2, i32 4, i32 8) [ "convergencectrl"(token %0) ]
# |   ret i32 %hlsl.wave.multi.prefix.bit.count
...

```

</details>

https://github.com/llvm/llvm-project/pull/238132


More information about the cfe-commits mailing list