Sycltla BF16 GEMM

[xinyu-intel]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS ABOVE HAVE BEEN CONSIDERED.

Purpose

measured on B60

M	oneDNN us	oneDNN TFLOPS	TLA us	TLA TFLOPS	TLA tile	Speedup
1	81.7	0.377	75.3	0.433	8x256	1.08x
2	79.1	0.743	75.5	0.888	8x256	1.05x
4	92.8	1.672	76.3	1.762	8x256	1.22x
8	83.2	2.918	76.5	3.502	8x256	1.09x
16	88.0	6.260	77.5	6.945	16x256	1.14x
32	91.4	12.648	77.9	13.811	32x256	1.17x
64	90.4	23.864	80.7	26.569	64x256	1.12x
128	89.7	47.696	85.4	50.147	128x256	1.05x
256	128.0	68.233	118.4	72.252	256x256	1.08x
384	177.6	72.471	177.0	72.770	384x128	1.00x
512	236.7	73.384	234.6	73.338	128x128	1.01x
640	285.3	75.223	263.2	81.720	128x128	1.08x
768	342.0	76.050	295.0	87.200	256x128	1.16x
896	343.7	87.460	362.4	82.887	128x128	0.95x
1024	395.7	86.732	410.5	83.668	256x128	0.96x
4096	1456.2	94.490	1500.9	91.505	256x256	0.97x

Test Plan

Test Result

(Optional) Documentation Update

BEFORE SUBMITTING, PLEASE READ https://docs.vllm.ai/en/latest/contributing (anything written below this line will be removed by GitHub Actions)