2a7ca101d7
Sparse-checkout from NVIDIA/cccl main branch to complete cccl_upstream: Added: - python/cuda_cccl/ (226 files) — Python bindings for device-level algorithms Critical for muh toolchain: cuda.compute.reduce_into, scan, radix_sort, etc. Includes 204 .py files with full test coverage for all 27 algorithms - ci/ (163 files) — Build/test infrastructure build_cub.sh, test_cub.sh, build_and_test_targets.sh, matrix.yaml Directly maps to our [INFRA-CI] and [INFRA-BUILD] items - .agent/skills/ (7 files) — NVIDIA's own agent skills for CCCL cccl-style/SKILL.md, cccl-test/SKILL.md, sass-diff/SKILL.md - docs/ (491 files) — Official CCCL documentation CI references, CMake guides, Python compute docs, libcudacxx PTX docs - test/ (12 files) — Top-level integration tests (cuda_smoke, stdpar) - Root configs: .clang-format, .clang-tidy, CONTRIBUTING.md, pyproject.toml - CLAUDE.md symlink → AGENTS.md (NVIDIA's standard) cccl_upstream now mirrors full NVIDIA/cccl structure: Before: 42M (cub + thrust + libcudacxx + cudax + c + examples + benchmarks) After: 53M (+python +ci +docs +.agent +test +configs) This completes the CCCL base needed for: - [muh-bench] items: ci/util/build_and_test_targets.sh for targeted builds - [CCCL-verify] items: python/cuda_cccl/tests/ as reference implementations - [CCCL-test] items: ci/test_cub.sh, ci/test_thrust.sh - Agent workflow: .agent/skills/ for consistent style and test patterns
71 lines
2.1 KiB
YAML
71 lines
2.1 KiB
YAML
# # CCCL PR benchmark request config.
|
|
#
|
|
# ## Overview:
|
|
#
|
|
# This file is used to request benchmark comparisons in PR CI.
|
|
#
|
|
# This file must match ci/bench.template.yaml to merge.
|
|
# CI branch protections will fail if they differ. Reset before merging.
|
|
#
|
|
# To update the defaults (e.g. new GPU pools), modify both this file and
|
|
# ci/bench.template.yaml together in the same PR.
|
|
#
|
|
# !! Strongly consider appending the following to your **commit messages** while benchmarking. !!
|
|
# This prevents wasteful non-benchmark CI jobs if they are not needed.
|
|
#
|
|
# [bench-only]
|
|
#
|
|
# To skip compile-time benchmark telemetry on unrelated changes, use:
|
|
#
|
|
# [skip-compile-time-bench]
|
|
#
|
|
# ## Quick start:
|
|
#
|
|
# 1. Add one or more benchmark regexes under benchmarks.filters.cub and/or
|
|
# benchmarks.filters.python.
|
|
# 2. Enable at least one GPU by uncommenting or adding entries in benchmarks.gpus.
|
|
# 3. Push and inspect the dispatched benchmark jobs/artifacts.
|
|
# 4. Remove/reset benchmark-request edits before final merge.
|
|
|
|
benchmarks:
|
|
|
|
# Benchmark filters grouped by project.
|
|
filters:
|
|
# CUB C++ benchmark filters (regex matched against ninja target names).
|
|
cub:
|
|
# Examples:
|
|
# - '^cub\.bench\.for_each\.base'
|
|
# - '^cub\.bench\.reduce\.(sum|min)\.'
|
|
|
|
# Python benchmark filters (regex matched against paths under benchmarks/).
|
|
python:
|
|
# Examples:
|
|
# - 'compute/reduce/sum\.py'
|
|
# - 'compute/transform/.*\.py'
|
|
|
|
# Select GPUs. These are limited and shared, be intentional and conservative.
|
|
gpus:
|
|
# - "t4" # sm_75, 16 GB
|
|
# - "rtx2080" # sm_75, 8 GB
|
|
# - "rtxa6000" # sm_86, 48 GB
|
|
# - "l4" # sm_89, 24 GB
|
|
# - "rtx4090" # sm_89, 24 GB
|
|
# - "h100" # sm_90, 80 GB
|
|
# - "rtxpro6000" # sm_120
|
|
|
|
# Extra .devcontainer/launch.sh -d args
|
|
# launch_args: "--cuda 13.3 --host gcc14"
|
|
launch_args: "" # Latest nvcc + gcc
|
|
|
|
# Advanced:
|
|
base_ref: "origin/main"
|
|
test_ref: "HEAD"
|
|
arch: "native"
|
|
nvbench_args: >-
|
|
--timeout 30
|
|
--skip-time 15e-6
|
|
--stopping-criterion entropy
|
|
--throttle-threshold 90
|
|
--throttle-recovery-delay 0.15
|
|
nvbench_compare_args: ""
|