[SYCL] Fix using some of math built-ins when ESIMD is included #14793
Conversation
This file contains 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
AlexeySachkov
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
AlexeySachkov
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
This reverts commit 1f1be9c.
… and support new functions (intel#13383)" This is a partial revert of bcca7a8. Notable changes: - new tests for `popcount`, `clz` and `ctz` built-ins were preserved - public definitions of those ESIMD APIs were preserved - the implementation of the latter was changed, though: - drop template args around `__spirv_ocl_*` intrinsics to use ones that are auto-declared by the compiler - added `#ifdef __SYCL_DEVICE_ONLY__`, because the compiler only declares `__spirv_ocl_*` intrinsics for device compilation
AlexeySachkov
force-pushed
the
private/asachkov/revert-13383
branch
from
1b00cda to
97abb43
Compare
AlexeySachkov
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
AlexeySachkov
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
AlexeySachkov
changed the title
sarnex
reviewed
Jul 29, 2024
| @@ -72,23 +72,6 @@ template <typename T, int N> | |||
| __ESIMD_INTRIN __ESIMD_raw_vec_t(T, N) | |||
| __spirv_ocl_native_powr(__ESIMD_raw_vec_t(T, N), __ESIMD_raw_vec_t(T, N)); | |||
|
|
|||
| template <typename T, int N> | |||
| __ESIMD_INTRIN __ESIMD_raw_vec_t(T, N) | |||
There was a problem hiding this comment.
instead what if we SFINAE these to only work when N==1, so we use the compiler generated ones by default?
There was a problem hiding this comment.
Whatever, I'll deal with this later. Will approve/merge for now.
sarnex
approved these changes
Jul 29, 2024
sarnex
added a commit
to sarnex/llvm
that referenced
this pull request
Jul 29, 2024
intel#14793)" This reverts commit 0228c23.
sarnex
added a commit
to sarnex/llvm
that referenced
this pull request
Aug 5, 2024
intel#14793)" This reverts commit 0228c23.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
ESIMD headers declare some of
__spirv_ocl_*built-ins as template functions, but those built-ins are also automatically declared by the compiler implicitly when used.On Windows, redeclarations in headers cause compilation issues, because by some reason they take priority, but template arguments for them couldn't be inferred.
This commit effectively introduces new tests to cover affected scenarios and reverts a couple of ESIMD commits to fix the issue:
__spirv_ocl_*built-insI suppose that both PRs were made in attempt to move away from custom ESIMD intrinsic to standard SPIR-V ones, but that should be done without manually declaring the latter. A bigger refactoring might be needed to use auto-declared SPIR-V built-ins in ESIMD because of presence and usage of single-element vectors in ESIMD (which do not exist in SPIR-V).