feat: implement SM-Constrained GEMM API #744
Open
+162
−16
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
yzh119
reviewed
Jan 21, 2025
There was a problem hiding this comment.
Hi @lanchongyizu , great work! Can you also support this for SM80 template?
For SM80 API, we might support setting a 3D tuple of num_ctas (on m, n, k dimension, correspondingly).
yzh119
reviewed
Jan 21, 2025
| @@ -33,6 +33,7 @@ | |||
| @pytest.mark.parametrize("dtype", DTYPES) | |||
| @pytest.mark.parametrize("device", CUDA_DEVICES) | |||
| @pytest.mark.parametrize("backend", ["auto", "sm90", "sm80"]) | |||
| @pytest.mark.parametrize("num_ctas", [0, 4, 16, 64]) | |||
There was a problem hiding this comment.
What's the expected behavior of num_ctas=0?
As requested in flashinfer-ai#591, this PR implements the `plan` function of GEMM with `num_ctas` as an argument to specify the grid size.
yzh119
force-pushed
the
sm_constrained_gemm
branch
from
841b423 to
eac553b
Compare
yzh119
reviewed
Jan 24, 2025
| @@ -121,8 +121,7 @@ cudaError_t CutlassSegmentGEMMSM90Run(void* float_buffer, size_t float_buffer_si | |||
|
|
|||
| cutlass::KernelHardwareInfo hw_info; | |||
| cudaGetDevice(&hw_info.device_id); | |||
| hw_info.sm_count = | |||
| cutlass::KernelHardwareInfo::query_device_multiprocessor_count(hw_info.device_id); | |||
| hw_info.sm_count = num_ctas; | |||
There was a problem hiding this comment.
I tried nsys profiler and it turns out this value can't control the number of SMs this kernel used.
A more fundamental approach might be using green context.
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Outdated suggestions cannot be applied.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.
As requested in #591, this PR implements the
planfunction of GEMMwith
num_ctasas an argument to specify the grid size.@yzh119