Repository navigation
[xegpu] Remove obsolete fused_attention_schedule - #312
Conversation
|
Part of larger rewrite where the new gpu pipeline will also replace |
| "d_head": args.d_head, | ||
| "wg_tile": [1, args.wg_rows], | ||
| "sg_rows": args.sg_rows, | ||
| "subgroup_size": args.subgroup_size, |
There was a problem hiding this comment.
Note that I'm dropping the subgroup_size option for now. It is essentially hard-coded to 16 in the attention schedule. We can turn it into a parameter when needed.
charithaintc
left a comment
There was a problem hiding this comment.
LGTM. I would keep fused_attention.py is possible. usually its easier to debug perf/layout issues with it. maybe move it to an IR test from e2e test.
| # Operand A shape is (wg_rows, d_head) | ||
| wg_rows, reduction_tile = ir.ShapedType(target_op.results[0].type).shape | ||
| wg_rows, d_head = ir.ShapedType(target_op.operands[0].type).shape | ||
| assert d_head == 64, "annotate_attention expects d_head to be 64" |
There was a problem hiding this comment.
I think d_head = 32, 128 should also work (possibly with spill). if we really want to fix it maybe remove dhead option from fused_attention.py
There was a problem hiding this comment.
Yes, we should add some tile size selection logic based on the problem size. I'll address this in another PR.
fused_attention.pyexample CLI parameters are propagated to the pipeline.fused_attention_schedule.extract_handletransform op supports parameter handles as well.Assisted-by: Claude