Skip to content

Kittens-like on Aster 16/n - Manual rescheduling to avoid VALU / MEM switch bubbles#384

Open
nicolasvasilache wants to merge 2 commits intomainfrom
kit-perf-002-gemm-optimizations
Open

Kittens-like on Aster 16/n - Manual rescheduling to avoid VALU / MEM switch bubbles#384
nicolasvasilache wants to merge 2 commits intomainfrom
kit-perf-002-gemm-optimizations

Conversation

@nicolasvasilache
Copy link
Copy Markdown
Collaborator

No description provided.

@nicolasvasilache
Copy link
Copy Markdown
Collaborator Author

nicolasvasilache commented Mar 9, 2026

Screenshot 2026-03-09 202627
./contrib/kittens/rocprof-kittens-bench.sh bench/bench_perf_sweep_002_gemm_fp16_weak_scaled.py        --m-wg 38 --n-wg 32 --m-waves 2 --n-waves 2        --m-tiles-wg 4 --n-tiles-wg 4 --k-tiles 1 --stages 2 --k-scaling-factor 256

…stalls by switching too quickly

Signed-off-by: Nicolas Vasilache <nicolasvasilache@users.noreply.github.com>
…e at a coarser grain

Signed-off-by: Nicolas Vasilache <nicolasvasilache@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant