generic sycl: refactor kernel mainloops #2070
Merged
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
github-actions
bot
added
the
platform:gpu-generic
t4c1
force-pushed
the
reorg_mainloop
branch
2 times, most recently
from
fe9cb1d
to
6b6d294
Compare
mgouicem
reviewed
Sep 3, 2024
|
Thanks for the PR. Would you have some performance data you can share? |
|
No, we did not run any benchmarks. For now this can be taken as improvement in how code is laid out, although it should improve performance as well. |
ShanoToni
approved these changes
Sep 4, 2024
|
make test |
densamoilov
reviewed
Sep 6, 2024
src/gpu/generic/sycl/sycl_utils.hpp
Outdated
| @@ -35,6 +37,18 @@ inline bool md_dims_in_range(const dnnl::impl::memory_desc_t *desc) { | |||
| return true; | |||
| } | |||
|
|
|||
| inline ::sycl::nd_range<1> get_range(const exec_ctx_t &ctx, int work_amount) { | |||
| auto eng = dnnl::engine(ctx.stream()->engine(), true); | |||
| auto device = dnnl::sycl_interop::get_device(eng); | |||
There was a problem hiding this comment.
Please don't use public API inside the library, use the internal API.
const auto *sycl_engine_impl = utils::downcast<const xpu::sycl::engine_impl_t *>(eng->impl());
auto device = sycl_engine_impl->device());
t4c1
force-pushed
the
reorg_mainloop
branch
from
5cee3f2
to
f1be8f4
Compare
kala855
approved these changes
Sep 18, 2024
densamoilov
approved these changes
Sep 19, 2024
|
@t4c1 can you please address the conflict that appeared after I merged the cublaslt PR. |
t4c1
force-pushed
the
reorg_mainloop
branch
from
f1be8f4
to
aafa2d8
Compare
t4c1
force-pushed
the
reorg_mainloop
branch
from
aafa2d8
to
194e8a2
Compare
|
done |
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.
Refactors mainloops in many sycl kernels so that consecutive workitems work on consecutive calculations. This should improve performace on Nvidia and AMD GPUs, as coalesced memory transfers can be used in more cases. The amount of duplicated code is also reduced.
This PR should be easier to review with hiding of whitespace changes, as in many places there is difference only in indentation levels.