[clang] [libclc] [AMDGPU][libclc] Enable cl_khr_subgroup_non_uniform_arithmetic and use in libclc (PR #200747)

[via cfe-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: Wenju He (wenju-he)

<details>
<summary>Changes</summary>

Per OpenCL spec, sub_group_non_uniform_reduce_* functions requires support for the cl_khr_subgroup_non_uniform_arithmetic extension.

---
Full diff: https://github.com/llvm/llvm-project/pull/200747.diff


3 Files Affected:

- (modified) clang/lib/Basic/Targets/AMDGPU.h (+1) 
- (modified) clang/test/Misc/amdgcn.languageOptsOpenCL.cl (+4) 
- (modified) libclc/opencl/lib/generic/subgroup/sub_group_non_uniform_reduce.cl (+4) 


``````````diff
diff --git a/clang/lib/Basic/Targets/AMDGPU.h b/clang/lib/Basic/Targets/AMDGPU.h
index f1bfca4c42417..29327067dbbff 100644
--- a/clang/lib/Basic/Targets/AMDGPU.h
+++ b/clang/lib/Basic/Targets/AMDGPU.h
@@ -324,6 +324,7 @@ class LLVM_LIBRARY_VISIBILITY AMDGPUTargetInfo final : public TargetInfo {
       Opts["cl_khr_mipmap_image"] = true;
       Opts["cl_khr_mipmap_image_writes"] = true;
       Opts["cl_khr_subgroups"] = true;
+      Opts["cl_khr_subgroup_non_uniform_arithmetic"] = true;
       Opts["cl_amd_media_ops"] = true;
       Opts["cl_amd_media_ops2"] = true;
 
diff --git a/clang/test/Misc/amdgcn.languageOptsOpenCL.cl b/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
index 5e611329b518d..3971bb99b113d 100644
--- a/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
+++ b/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
@@ -158,6 +158,10 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_subgroups: enable
 
+#ifndef cl_khr_subgroup_non_uniform_arithmetic
+#error "Missing cl_khr_subgroup_non_uniform_arithmetic define"
+#endif
+
 #ifndef cl_amd_media_ops
 #error "Missing cl_amd_media_ops define"
 #endif
diff --git a/libclc/opencl/lib/generic/subgroup/sub_group_non_uniform_reduce.cl b/libclc/opencl/lib/generic/subgroup/sub_group_non_uniform_reduce.cl
index e00717b979ea1..d46afb88afa59 100644
--- a/libclc/opencl/lib/generic/subgroup/sub_group_non_uniform_reduce.cl
+++ b/libclc/opencl/lib/generic/subgroup/sub_group_non_uniform_reduce.cl
@@ -8,6 +8,8 @@
 
 #include "clc/subgroup/clc_sub_group_non_uniform_reduce.h"
 
+#ifdef cl_khr_subgroup_non_uniform_arithmetic
+
 #define __CLC_BODY "sub_group_non_uniform_reduce.inc"
 #include "clc/integer/gentype.inc"
 
@@ -28,3 +30,5 @@ _CLC_DEF _CLC_OVERLOAD int
 sub_group_non_uniform_reduce_logical_xor(int predicate) {
   return __clc_sub_group_non_uniform_reduce_logical_xor(predicate);
 }
+
+#endif // cl_khr_subgroup_non_uniform_arithmetic

``````````

</details>


https://github.com/llvm/llvm-project/pull/200747