Skip to content

Add an optional GPU simulator CI job (simulated H100) - #1

Open
saqibkh wants to merge 1 commit into
ourarash:mainfrom
saqibkh:add-gpu-sim-ci
Open

saqibkh wants to merge 1 commit into
ourarash:mainfrom
saqibkh:add-gpu-sim-ci

Conversation

@saqibkh

@saqibkh saqibkh commented Nov 16, 2026

Copy link
Copy Markdown

The existing CI only compiles the examples because the hosted runners have no GPU. This adds a separate gpu-sim.yml that builds them for a simulated H100 (PTX only, shared cudart) and runs each one with --quick. Every example has to exit 0 and end with PASS; a SKIP, a missing device or a wrong result fails the job. It runs on workflow_dispatch and on pull requests that touch .cu/.cuh files or CMake.

It runs 89 examples (basics, memory, warp, reduction, scan, histogram, compaction, matmul 00-07, transpose, sort, convolution, stencil, ml, sparse, libraries, applications). Left out: matmul 08-11 (Tensor Core WMMA, mma.sync, Hopper wgmma), the compute-sanitizer tests and the PyTorch extension. None fail.

Fork run: https://github.com/saqibkh/cuda-algorithm-lab/actions/runs/43776769664 (green). Cold setup was about 20 minutes, then 4 minutes to build and 1.5 minutes to run.

I am the author of PantheonSim, so please weigh this accordingly.

This branch has not been deployed

No deployments
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