[Codegen][CPU] Add design doc for the LLVMCPU C-ukernel framework.#24573
Open
bjacob wants to merge 1 commit into
Open
[Codegen][CPU] Add design doc for the LLVMCPU C-ukernel framework.#24573bjacob wants to merge 1 commit into
bjacob wants to merge 1 commit into
Conversation
8af2850 to
cb00012
Compare
678d998 to
e911197
Compare
cb00012 to
6cd0524
Compare
e911197 to
309ffdf
Compare
6cd0524 to
fc522fd
Compare
309ffdf to
8af5f8e
Compare
fc522fd to
5fb2905
Compare
0b48eed to
be4ec36
Compare
5fb2905 to
6ec426f
Compare
e840eb0 to
c47ea03
Compare
6602e64 to
18678c3
Compare
c47ea03 to
c7aa269
Compare
18678c3 to
a9f88d9
Compare
a9f88d9 to
b60f897
Compare
f896f19 to
d467c2c
Compare
b60f897 to
6c0ae40
Compare
d467c2c to
38ca327
Compare
Adds README.md for compiler/plugins/target/LLVMCPU/builtins/ukernel/ as a
single final documentation commit, rather than iterating the doc across
the earlier framework commits (less churn for review).
Covers: scope (the framework is op-agnostic in principle, with
`inner_tiled` as the initial and primary case getting the detailed
treatment — inner-K-loop-only contract, and how specialization splits
between symbol selection on the `MMAIntrinsic` and post-inlining loop
unrolling on the `intrinsics_{m,n,k}` factors); the two structural
pillars (lives in compiler/, bitcode-only; self-contained IR via
`hal.executable_object`); the end-to-end lowering flow; the
`--iree-llvmcpu-enable-llvm-ukernels` user flag; bring-your-own-ukernel;
an authoring guide; testing at both the lit (compiler IR) and
`tests/e2e/matmul` (numerical) levels; and the three seed examples.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Signed-off-by: Benoit Jacob <jacob.benoit.1@gmail.com>
6c0ae40 to
62ad0cd
Compare
38ca327 to
54665f8
Compare
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
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.
Adds README.md for compiler/plugins/target/LLVMCPU/builtins/ukernel/ as a single final documentation commit, rather than iterating the doc across the earlier framework commits (less churn for review).
Covers: scope (the framework is op-agnostic in principle, with
inner_tiledas the initial and primary case getting the detailed treatment — inner-K-loop-only contract, and how specialization splits between symbol selection on theMMAIntrinsicand post-inlining loop unrolling on theintrinsics_{m,n,k}factors); the two structural pillars (lives in compiler/, bitcode-only; self-contained IR viahal.executable_object); the end-to-end lowering flow; the--iree-llvmcpu-enable-llvm-ukernelsuser flag; bring-your-own-ukernel; an authoring guide; testing at both the lit (compiler IR) andtests/e2e/matmul(numerical) levels; and the three seed examples.Progress towards #24574.