Skip to content

profiling.md

Latest commit

 

History

History
96 lines (68 loc) · 2.4 KB

profiling.md

File metadata and controls

96 lines (68 loc) · 2.4 KB

QR v2 Nsight Compute Profiling

This profiles the GPU Mode QR v2 problem from reference-kernels and downloads Nsight Compute details that AI agents can read directly. The full .ncu-rep GUI report is still included for local inspection.

1. Install and Register

curl -fsSL https://raw.githubusercontent.com/gpu-mode/popcorn-cli/main/install.sh | bash
popcorn register discord

Restart your terminal if popcorn is not found after installation.

2. Get the QR v2 Starter Submission

mkdir -p qr-v2-profile
cd qr-v2-profile
curl -O https://raw.githubusercontent.com/gpu-mode/reference-kernels/main/problems/linalg/qr_v2/submission.py

The profiler uses the hosted GPU Mode NCU service:

export POPCORN_BREV_PROFILER_URL=https://http--brev-profiler-proxy--dxfjds728w5v.code.run

3. Profile One QR v2 Shape

This profiles benchmarks[0] from reference-kernels/problems/linalg/qr_v2/task.yml:

popcorn submit submission.py \
  --leaderboard qr_v2 \
  --profile-brev \
  --benchmark-index 0 \
  --no-tui

The first QR v2 benchmark shape is:

batch: 20; n: 32; cond: 1; seed: 43214

4. Read the Details

After the run finishes, the CLI downloads and extracts files like:

profile.0-batch-20-n-32-cond-1-seed-43214.zip
profile.0-batch-20-n-32-cond-1-seed-43214/ncu-details.txt
profile.0-batch-20-n-32-cond-1-seed-43214/ncu-details.csv
profile.0-batch-20-n-32-cond-1-seed-43214/profile.ncu-rep   # optional GUI report

Use ncu-details.txt or ncu-details.csv as the default artifact for AI analysis. The CLI prints clickable links for these detail files.

The last line printed by the CLI opens the optional GUI report on macOS:

open -a "NVIDIA Nsight Compute" 'profile.0-batch-20-n-32-cond-1-seed-43214/profile.ncu-rep'

Profile All QR v2 Benchmark Shapes

Omit --benchmark-index:

popcorn submit submission.py \
  --leaderboard qr_v2 \
  --profile-brev \
  --no-tui

This profiles every entry in the benchmarks: list in QR v2 task.yml, not the tests: list. It will produce one zip plus extracted details and optional .ncu-rep files per benchmark shape.

Normal Submit Commands

For correctness testing:

popcorn submit submission.py --leaderboard qr_v2 --gpu B200 --mode test --no-tui

For leaderboard submission:

popcorn submit submission.py --leaderboard qr_v2 --gpu B200 --mode leaderboard --no-tui