NVIDIA / NVIDIA/cutlass

[QST] CuTE - how to emit SASS loads with ".constant" suffix?

Open
#2,301 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

? - Needs Triage inactive-30d inactive-90d question
Dominant language
C++
Stars
10.5k
Forks
2.1k
Avg merge
3d 11h
Merged PRs (30d)
7

Description

I've noticed that some cuBLASlt kernels on Ampere use this instruction for async gmem=>smem loads:

LDGSTS.E.BYPASS.LTC128B.128.CONSTANT

However, my own kernels using CuTE's SM80_CP_ASYNC_CACHEGLOBAL<uint128_t> use this instruction instead:

LDGSTS.E.BYPASS.LTC128B.128

I believe this may have something to do with the propagation of __restrict__ and const in CuTE. All my pointers are __restrict__ (and const where applicable) but it doesn't seem like the compiler is able to take advantage of this.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No source file or test is named. Start by locating CuTE's SM80_CP_ASYNC_CACHEGLOBAL<uint128_t> implementation and compare its generated SASS with the cuBLASLt instruction; done means determining whether const or restrict propagation can produce the .CONSTANT suffix.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
performance
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.