[DOC] IKET quick start omits the GEMM GPU requirement and references a missing example

Open Beginner friendly
  • Is something wrong?
#3,761 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 4 days

@0z5a is already working on this.

Since Oct 20, 2026.

  • #3758 by @0z5a — open

Assessment

Difficulty
1/5
Estimated time
Under an hour
Newbie friendliness
82/100
Issue type
Documentation
Clarity
Clearly specified
Activity status
Active
Tech stack
python
Domain
documentation

Research direction

Read the IKET profiling guide at media/docs/pythonDSL/guides/iket_profiling.rst, then check whether examples/python/CuTeDSL/dsl_tutorials/iket_compile_options.py exists at the referenced revision. Update the guide to state the quick-start GEMM's Blackwell requirement, direct Hopper users to an SM90-compatible kernel, and remove the missing example reference; done when those instructions match the reported behavior.

Written by the indexing model from the issue text.

Description

Report incorrect documentation

Location: IKET profiling guide, at the v4.8.1 source revision.

Problems: The introduction states that IKET supports Hopper and newer GPUs, but the quick-start command runs a GEMM that uses Blackwell tcgen05 instructions without stating that additional requirement. The guide also refers to examples/python/CuTeDSL/dsl_tutorials/iket_compile_options.py, which is absent at this revision.

Reproduction: On GH200 with CuTe DSL 4.8.1, the documented command

run-iket profile --postprocess perfetto -- \
  python examples/python/CuTeDSL/dsl_tutorials/fp16_gemm_4_iket.py \
  --mnk 512,1024,64

fails while constructing tcgen05.MmaF16BF16Op: the architecture diagnostic lists supported Blackwell targets and reports got sm_90a.

An independent SM90 add-one kernel using the same IKET marker/range APIs runs successfully under run-iket, produces the expected numeric output, and emits a Perfetto trace. This distinguishes the example's architecture requirement from IKET's own Hopper support; the Blackwell GEMM itself was not validated on GH200.

Suggested fix: State that this particular quick-start GEMM requires Blackwell, direct Hopper users to instrument an SM90-compatible kernel, and remove the nonexistent compile-options example reference.

Dominant language
C++
Stars
10.7k
Forks
2.2k
Avg merge
3d 20h
Merged PRs (30d)
4

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

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.

More from NVIDIA/cutlass

All issues in NVIDIA/cutlass

Similar issues

More C++ issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.