[LIBCLC][BINDLESS][CUDA] always inline redirection functs #18699
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 functions have effectively zero register overhead, therefore no usage circumstance that brings downside to always inlining. Signed-off-by: JackAKirk <[email protected]>
|
Marking as ready for review: will fix format later. |
frasercrmck
reviewed
May 29, 2025
| @@ -149,7 +149,7 @@ void __nvvm_sust_3d_v4i32_clamp(write_only image3d_t, int, int, int, int, int, | |||
|
|
|||
| int __nvvm_suq_width(long) __asm("llvm.nvvm.suq.width"); | |||
| int __nvvm_suq_height(long) __asm("llvm.nvvm.suq.height"); | |||
| int __nvvm_suq_depth(long arg) { | |||
| __attribute__((always_inline)) int __nvvm_suq_depth(long arg) { | |||
There was a problem hiding this comment.
It might be prudent to define some kind of IMAGE_DEF_ATTRS macro like CLC_DEF which is set to __attribute__((always_inline))? That way more attributes could be added less noisily in the future if needed?
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Outdated suggestions cannot be applied.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.
These functions at most do some casting, and have effectively zero register overhead at default opt level, therefore there should be no usage circumstance that brings a downside to always inlining.
This brings the nvptx libclc image backend in line with the amd one which requires no such changes. amd libclc backend already does the same thing via consistent usage of the _CLC_DECL macro for all functions. Whilst not immediately obvious to the libclc programmer, _CLC_DECL macro calls
__attribute__((always_inline)).There's a few cases that had low register usage that I've added the
inlinehint to also, being probably overly cautious.