[OpenCL] Add decls for cl_intel_bfloat16_conversions #150393
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
These map to SPIR-V instructions, which are long supported by the llvm SPIR-V target [1] and the llvm-spirv translator [2]. Intel's offline compiler (ocloc) and OpenCL implementation trivially supports these, by having these same declarations [3] and relying on llvm-spirv to map calls to them to their corresponding SPIR-V instructions. [1]: https://github.com/llvm/llvm-project/blob/531cf8298b08eacdf670bac8c28db97a5dc8cb01/llvm/lib/Target/SPIRV/SPIRVBuiltins.td#L1546C11-L1546C27 [2]: https://github.com/KhronosGroup/SPIRV-LLVM-Translator/blob/10c7569b3c4cb456fbfdcc86c3de45d46c7f5fa8/lib/SPIRV/OCLUtil.h#L327 [3]: https://github.com/intel/intel-graphics-compiler/blob/342c4fb729ff6a20a41e19adc8329ad18ba05660/IGC/BiFModule/Languages/OpenCL/opencl_cth_released.h#L6899
llvmbot
added
clang
Member
|
@llvm/pr-subscribers-backend-x86 @llvm/pr-subscribers-clang Author: Mészáros Gergely (Maetveis) ChangesThese map to SPIR-V instructions, which are long supported by the llvm SPIR-V target 1 and the llvm-spirv translator 2. Intel's offline compiler (ocloc) and OpenCL implementation trivially supports these, by having these same declarations 3 and relying on llvm-spirv to map calls to them to their corresponding SPIR-V instructions. Full diff: https://github.com/llvm/llvm-project/pull/150393.diff 1 Files Affected:
diff --git a/clang/lib/Headers/opencl-c.h b/clang/lib/Headers/opencl-c.h
index e1e0fdeadb0aa..f65b4b314cffd 100644
--- a/clang/lib/Headers/opencl-c.h
+++ b/clang/lib/Headers/opencl-c.h
@@ -18410,6 +18410,22 @@ intel_sub_group_avc_mce_convert_to_sic_result(
#pragma OPENCL EXTENSION cl_intel_device_side_avc_motion_estimation : end
#endif // cl_intel_device_side_avc_motion_estimation
+#if defined(cl_intel_bfloat16_conversions)
+ushort __ovld intel_convert_bfloat16_as_ushort(float source);
+ushort2 __ovld intel_convert_bfloat162_as_ushort2(float2 source);
+ushort3 __ovld intel_convert_bfloat163_as_ushort3(float3 source);
+ushort4 __ovld intel_convert_bfloat164_as_ushort4(float4 source);
+ushort8 __ovld intel_convert_bfloat168_as_ushort8(float8 source);
+ushort16 __ovld intel_convert_bfloat1616_as_ushort16(float16 source);
+
+float __ovld intel_convert_as_bfloat16_float(ushort source);
+float2 __ovld intel_convert_as_bfloat162_float2(ushort2 source);
+float3 __ovld intel_convert_as_bfloat163_float3(ushort3 source);
+float4 __ovld intel_convert_as_bfloat164_float4(ushort4 source);
+float8 __ovld intel_convert_as_bfloat168_float8(ushort8 source);
+float16 __ovld intel_convert_as_bfloat1616_float16(ushort16 source);
+#endif // cl_intel_bfloat16_conversions
+
#ifdef cl_amd_media_ops
uint __ovld amd_bitalign(uint, uint, uint);
uint2 __ovld amd_bitalign(uint2, uint2, uint2);
|
Contributor
Author
|
FYI: See the llvm SPIR-V backend doing the right thing if given a declaration on compiler explorer: https://godbolt.org/z/ac1ThYePG |
Contributor
Author
|
ping @michalpaszkowski |
michalpaszkowski
approved these changes
Jul 29, 2025
Member
|
LGTM! Thanks for the patch! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
These map to SPIR-V instructions, which are long supported by the llvm SPIR-V target 1 and the llvm-spirv translator 2.
Intel's offline compiler (ocloc) and OpenCL implementation trivially supports these, by having these same declarations 3 and relying on llvm-spirv to map calls to them to their corresponding SPIR-V instructions.