refactor apply_w8a8_block_fp8_linear in fp

[ChangyiYang]

See also: #4353

Motivation

refactor apply_w8a8_block_fp8_linear to make the logic more clear and more meaningful

Modifications

only refactor apply_w8a8_block_fp8_linear in fb8_utils

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[ChangyiYang]

Hi! I have create a new commit. Here are some bullet points

• Refactored dispatch logic to return only the selected matmul implementation function at initialization time to avoid filtering overhead.
• For implementations requiring runtime validation (e.g., shape or dtype constraints), added fallback to Triton when conditions are not met.
• Moved runtime condition checks ahead of computation to avoid redundant operations and unnecessary overhead.
• Extracted w8a8_block_fp8_matmul_deepgemm from w8a8_block_fp8_matmul to enable direct usage of DeepGEMM kernels when applicable.
• Retained the function w8a8_block_fp8_matmul as the unified entry point for matrix multiplication, to avoid requiring users to manually check whether DeepGEMM is available each time they use it.

Some choices I make, which may potentially have some improve space

• In w8a8_block_fp8_matmul_deepgemm, use the same assertion check as w8a8_block_fp8_matmul, which may have some unnecessary check

Please check if any futhur modification is needed!

[ChangyiYang]

Hi ! Here are some modification for this commit

• split out w8a8_block_fp8_matmul_triton
• retain w8a8_block_fp8_matmul only for testing purpose
• for benchmarks that use deepgemm, directly call w8a8_block_fp8_matmul_deepgemm

Feel free to tell me if there anything needs further modification!

[ChangyiYang]

Hi! Typo fixed. Feel free to tell me if there is any more issue!

[zhyncs]

cc @HaiShaw

[ChangyiYang]

@Alcanderian hi, I click update branch to merge from main and seems needs to be approved again. Can you kindly approve again?

[Alcanderian]

@ChangyiYang We have to fix the CI
https://github.com/sgl-project/sglang/actions/runs/15267675566/job/42936151803?pr=6545#step:4:325
https://github.com/sgl-project/sglang/actions/runs/15267675566/job/42936151795?pr=6545#step:4:339
https://github.com/sgl-project/sglang/actions/runs/15267675566/job/42936151794?pr=6545#step:4:3685
https://github.com/sgl-project/sglang/actions/runs/15267675566/job/42936151793?pr=6545#step:4:1352

[ChangyiYang]

@Alcanderian thanks for pointing that out! fix that now. Can you run CI again?

[ChangyiYang]

Hi! I think I have fixed 2 minor bugs and the CI seems be correct ( the failing one is not actually failing and 9 pipelines are blocking). Feel free to tell me if any more adjustment needed before merge! Thanks you all guys for the help :)