test(lit/vmi): add SimdVF per-token FP8 cast target IR#487
Open
WenboCodes wants to merge 27 commits into
Open
Conversation
Add design-reference .pto capturing the pto.vmi lowering for the
SimdVF per-token cast-to-FP8 kernel, verbatim from
PTO-Gym/docs/simtvf-per-token-cast-to-fp8-pto-vmi-mapping.md.
The file holds the original TileLang SimdVF source in its header
comment and the forward-looking pseudo MLIR (vreduce_max{sub_group},
vbr{to,sub_group}, vcvt{to,rnd,sat}, pset_b32 "PAT_ALL", etc.) which
uses surface ops not yet implemented in the VMI dialect. It is a
no-op (RUN: true) spec, not an executable regression test.
Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Add a design-reference
.ptototest/lit/vmi/capturing the targetpto.vmilowering for the SimdVF per-token cast-to-FP8 kernel.Contents (verbatim from
PTO-Gym/docs/simtvf-per-token-cast-to-fp8-pto-vmi-mapping.md):with T.SimdVF():source (section 1 of the doc).Why this shape
The VMI code in the source doc is explicitly design-level pseudo MLIR. It uses surface ops that are not yet implemented in the PTOAS VMI dialect:
pto.vmi.vlds/pto.vmi.vstspto.vmi.load/pto.vmi.storepto.vmi.pset_b32 "PAT_ALL"pto.vmi.create_maskpto.vmi.vreduce_max {sub_group = 2}group_reduce_addf/group_reduce_addi; there is no group max-reducepto.vmi.vbr {to = 256, sub_group = 2}pto.vmi.group_broadcast {num_groups = N}pto.vmi.vcvt {to = ..., rnd = "R", sat = "SAT"}pto.vmi.truncfpto.vmi.vsts {mode = "ONE_PER_SUB_GROUP"}So it cannot be parsed by
ptoas/pto-test-opt. It is stored as a no-op (RUN: true) forward-looking spec rather than an executable regression test. See the upstream doc's section 8 ("需要 PTO/VMI surface 补齐的点") for the full list of missing surface points.Test plan
lit test/lit/vmi/vmi_simdvf_per_token_cast_to_fp8.pto→ passes as no-op (RUN: true).pto-test-optexecution, by design.🤖 Generated with Claude Code