[THLO] fold sizes more aggressively #2505
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
![copybara-service[bot]](https://avatars.githubusercontent.com/in/44061?s=60&v=4){kind=small}
copybara-service
bot
force-pushed
the
test_524010738
branch
from
2f7c09f
to
0f6766b
Compare
Canonicalization will get these, but you can end up in a situation where not folding creates dynamic shapes for following ops that require static shapes. That IR wouldn't verify anymore. I'm having trouble creating a test case for this. PiperOrigin-RevId: 524234633
copybara-service
bot
force-pushed
the
test_524010738
branch
from
0f6766b
to
22f394d
Compare
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 24, 2023
#MIGRATION_3P_TRITON__GIT_TO_THIRD_PARTY # Commits integrated - 726bdb984f2bcb48adfaa341ee7b0263be227b98 [FRONTEND][BACKEND] Fix constexpr assignment ; revert #24... by Zahi Moudallal <128723247+zahimoud@users.noreply.github.com> - 87a223d76fe32a28ca563c94215a95f505794c6d bump triton_shared (#2501) by Maksim Levental <maksim.levental@gmail.com> - 721897fcc4f942aa97d2e9ba3787a5e213758177 upgrade llvm to `b1115f8c` (NFC) (#2403) by Mehdi Amini <mamini@nvidia.com> - 05dc28be0e72dd496300a31b99a21a5a5118f8e9 [CI] refactor workflows (#2504) by Philippe Tillet <phil@openai.com> - 376acb610b5888263ee61713ff0a71e1d5908d69 [BUILD] Fix macos x86 build (#2505) by Thomas Raoux <thomas.raoux@openai.com> - 768fc1fcd98ecfc0892f8982b0bb009dd7bb11ea [FRONTEND] change hash to not require ptxas (#2476) by ian Bearman <ianb@microsoft.com> - e36d1665ca2f816212fc80ee2633caa66a0066bf [BACKEND] Fix unsupported view op created during optimiza... by Thomas Raoux <thomas.raoux@openai.com> - a980ec50f1ed3176e2603c25f73f0ddc031cf1d8 [BACKEND] Fixing f8e5m2 to bf16 conversion on A100 (#2508) by Zahi Moudallal <128723247+zahimoud@users.noreply.github.com> - a4f373938c9a4ba67105c5394c168945af4c990e [RUNTIME] Filter out paths that don't exist in json group... by Horace He <chilli@meta.com> - be1de890e1f9bdf0910521b5a536c332a1c1aa2f [BACKEND] Replace assert(0) with llvm::report_fatal_error... by Keren Zhou <kerenzhou@openai.com> - 0d57820be9ca360cf62cc3a7dc21aecc45a1c53a update triton-shared ref (#2506) by ian Bearman <ianb@microsoft.com> - bdf464e4a8f80ad6bd6a7b470cb3d36efd61c8a2 Make kernel_static_print test work when called twice. (#2... by Justin Lebar <justin.lebar@gmail.com> - 30186f401ec52d9addac79a60f418792875f7d11 Fix segfault in assertion test. (#2520) by Justin Lebar <justin.lebar@gmail.com> - dc9e3063d73d2410e1855e1ff258aa90a6158548 [HOPPER] Move to tl.make_block_ptr in flash_attention bac... by runseny <145632023+runseny@users.noreply.github.com> - b0c166b9e3f2f58c0906fa41f261787ebf3fef0d [BACKEND] Fixing bug in elementwise conversion (#2517) by Zahi Moudallal <128723247+zahimoud@users.noreply.github.com> - 4f4c07e7d586aae3daa802ce86a9aa935f8cda17 [CI] add text file containing LLVM commit hash by Ashay Rane <ashay@users.noreply.github.com> - 7af27fadee0fce2218a1353feea2f76ea25ad005 update hash to 76ce4736721a by Phil Tillet <phil@openai.com> - f192611ff3bdacb8d1d1cad084dfe4cd277a0ec9 Bump LLVM version to https://github.com/llvm/llvm-project... by Goran Flegar <gflegar@google.com> PiperOrigin-RevId: 576212898
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.
[THLO] fold sizes more aggressively
Canonicalization will get these, but you can end up in a situation where not
folding creates dynamic shapes for following ops that require static shapes.
That IR wouldn't verify anymore.
I'm having trouble creating a test case for this.