Skip to content

perf(solve): optimize exact and factorized solve kernels #152

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
acgetchell merged 1 commit into from
Jun 9, 2026
Merged

perf(solve): optimize exact and factorized solve kernels #152

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions docs/RELEASING.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,7 @@ git remote -v
Ensure your local `main` is up to date before beginning:

```bash
git checkout main
git switch main
git pull --ff-only
```

Expand All @@ -50,7 +50,7 @@ but keep them minimal and release-critical.
1. Create the release branch

```bash
git checkout -b "release/$TAG"
git switch -c "release/$TAG"
```

2. Bump versions
Expand Down Expand Up @@ -200,7 +200,7 @@ If you discover issues after generating the changelog:
1. Sync your local `main` to the merge commit

```bash
git checkout main
git switch main
git pull --ff-only
```

Expand Down
130 changes: 65 additions & 65 deletions justfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -165,19 +165,6 @@ action-lint: _ensure-actionlint
bench:
cargo bench --features bench

# Compile benchmarks without running them, treating warnings as errors.
# This catches bench/release-profile-only warnings that won't show up in normal debug-profile runs.
bench-compile:
RUSTFLAGS='-D warnings' cargo bench --no-run --features bench
RUSTFLAGS='-D warnings' cargo bench --no-run --features bench,exact --bench exact

# Run the cheaper latest measurements used for latest-vs-last reports.
bench-latest: bench-vs-linalg-la-stack bench-exact

# Run latest measurements and render the latest-vs-last performance report.
bench-latest-vs-last baseline="last": bench-latest python-sync
uv run bench-compare {{baseline}}

# Compare latest measurements against a saved baseline.
# Defaults to the `last` full-release baseline.
bench-compare baseline="last" suite="all" scope="release-signal": python-sync
Expand All @@ -186,53 +173,22 @@ bench-compare baseline="last" suite="all" scope="release-signal": python-sync
baseline="{{baseline}}"
uv run bench-compare "$baseline" --suite "{{suite}}" --scope "{{scope}}"

# Backward-compatible alias for the GitHub Actions release-asset comparison.
performance-archive-published current_tag="" baseline_tag="":
just performance-github-assets "{{current_tag}}" "{{baseline_tag}}"

# Compare stored GitHub Actions release benchmark assets without local cargo runs.
performance-github-assets current_tag="" baseline_tag="": python-sync
#!/usr/bin/env bash
set -euo pipefail
current_tag="{{current_tag}}"
baseline_tag="{{baseline_tag}}"
if [[ -n "$current_tag" || -n "$baseline_tag" ]]; then
if [[ -z "$current_tag" || -z "$baseline_tag" ]]; then
echo "current_tag and baseline_tag must be provided together" >&2
exit 2
fi
uv run archive-performance "$current_tag" "$baseline_tag" --github-assets --generate-in-temp-worktree --worktree-ref "$current_tag" --output-only --output target/bench-reports/github-assets-performance.md
else
uv run archive-performance --published-latest --github-assets --generate-in-temp-worktree --output-only --output target/bench-reports/github-assets-performance.md
fi

# Compare the current tree against the latest published release locally.
performance-local: python-sync
uv run archive-performance --current-vs-latest --generate-in-temp-worktree --output-only --output target/bench-reports/performance.md

# Generate local release-signal measurements in a temp worktree, then promote/archive docs.
performance-release current_tag="" baseline_tag="": python-sync
#!/usr/bin/env bash
set -euo pipefail
current_tag="{{current_tag}}"
baseline_tag="{{baseline_tag}}"
if [[ -n "$current_tag" || -n "$baseline_tag" ]]; then
if [[ -z "$current_tag" || -z "$baseline_tag" ]]; then
echo "current_tag and baseline_tag must be provided together" >&2
exit 2
fi
uv run archive-performance "$current_tag" "$baseline_tag" --generate-in-temp-worktree --worktree-ref HEAD
else
uv run archive-performance --infer-release --generate-in-temp-worktree --worktree-ref HEAD
fi
# Compile benchmarks without running them, treating warnings as errors.
# This catches bench/release-profile-only warnings that won't show up in normal debug-profile runs.
bench-compile:
RUSTFLAGS='-D warnings' cargo bench --no-run --features bench
RUSTFLAGS='-D warnings' cargo bench --no-run --features bench,exact --bench exact

# Run the exact-arithmetic benchmark suite.
bench-exact:
cargo bench --features bench,exact --bench exact

# Save a full Criterion baseline for the previous release signal.
bench-save-last:
just bench-save-baseline last
# Run the cheaper latest measurements used for latest-vs-last reports.
bench-latest: bench-vs-linalg-la-stack bench-exact

# Run latest measurements and render the latest-vs-last performance report.
bench-latest-vs-last baseline="last": bench-latest python-sync
uv run bench-compare {{baseline}}

# Save a Criterion baseline. Defaults to all release-signal benchmark suites.
bench-save-baseline tag suite="all":
Expand All @@ -256,6 +212,10 @@ bench-save-baseline tag suite="all":
;;
esac

# Save a full Criterion baseline for the previous release signal.
bench-save-last:
just bench-save-baseline last

# Bench the la-stack vs nalgebra/faer comparison suite.
bench-vs-linalg filter="":
#!/usr/bin/env bash
Expand Down Expand Up @@ -340,6 +300,10 @@ check-fast:
ci: check bench-compile test-all examples
@echo "🎯 CI checks complete!"

# Validate CITATION.cff against the Citation File Format schema.
citation-check: _ensure-uv
uvx --from cffconvert==2.0.0 cffconvert --validate -i CITATION.cff

# Clean build artifacts
clean:
cargo clean
Expand Down Expand Up @@ -502,6 +466,46 @@ markdown-fix: _ensure-rumdl

markdown-lint: markdown-check

# Backward-compatible alias for the GitHub Actions release-asset comparison.
performance-archive-published current_tag="" baseline_tag="":
just performance-github-assets "{{current_tag}}" "{{baseline_tag}}"

# Compare stored GitHub Actions release benchmark assets without local cargo runs.
performance-github-assets current_tag="" baseline_tag="": python-sync
#!/usr/bin/env bash
set -euo pipefail
current_tag="{{current_tag}}"
baseline_tag="{{baseline_tag}}"
if [[ -n "$current_tag" || -n "$baseline_tag" ]]; then
if [[ -z "$current_tag" || -z "$baseline_tag" ]]; then
echo "current_tag and baseline_tag must be provided together" >&2
exit 2
fi
uv run archive-performance "$current_tag" "$baseline_tag" --github-assets --generate-in-temp-worktree --worktree-ref "$current_tag" --output-only --output target/bench-reports/github-assets-performance.md
else
uv run archive-performance --published-latest --github-assets --generate-in-temp-worktree --output-only --output target/bench-reports/github-assets-performance.md
fi

# Compare the current tree against the latest published release locally.
performance-local: python-sync
uv run archive-performance --current-vs-latest --generate-in-temp-worktree --output-only --output target/bench-reports/performance.md

# Generate local release-signal measurements in a temp worktree, then promote/archive docs.
performance-release current_tag="" baseline_tag="": python-sync
#!/usr/bin/env bash
set -euo pipefail
current_tag="{{current_tag}}"
baseline_tag="{{baseline_tag}}"
if [[ -n "$current_tag" || -n "$baseline_tag" ]]; then
if [[ -z "$current_tag" || -z "$baseline_tag" ]]; then
echo "current_tag and baseline_tag must be provided together" >&2
exit 2
fi
uv run archive-performance "$current_tag" "$baseline_tag" --generate-in-temp-worktree --worktree-ref HEAD
else
uv run archive-performance --infer-release --generate-in-temp-worktree --worktree-ref HEAD
fi

# Plot: generate a single time-vs-dimension SVG from Criterion results.
plot-vs-linalg metric="lu_solve" stat="median" sample="new" log_y="false": python-sync
#!/usr/bin/env bash
Expand Down Expand Up @@ -690,6 +694,8 @@ shell-check:
echo "No shell files found to check."
fi

shell-fix: shell-fmt

shell-fmt: _ensure-shfmt
#!/usr/bin/env bash
set -euo pipefail
Expand All @@ -706,8 +712,6 @@ shell-fmt: _ensure-shfmt

shell-lint: shell-check

shell-fix: shell-fmt

# Spell check (typos)
spell-check: _ensure-typos
#!/usr/bin/env bash
Expand Down Expand Up @@ -777,6 +781,10 @@ test-python: python-sync
uv run pytest -q

# TOML
toml-check: toml-fmt-check toml-lint

toml-fix: toml-fmt

toml-fmt: _ensure-taplo
#!/usr/bin/env bash
set -euo pipefail
Expand Down Expand Up @@ -816,10 +824,6 @@ toml-lint: _ensure-taplo
echo "No TOML files found to lint."
fi

toml-check: toml-fmt-check toml-lint

toml-fix: toml-fmt

# File validation
validate-json: _ensure-jq
#!/usr/bin/env bash
Expand Down Expand Up @@ -863,10 +867,6 @@ yaml-fix: _ensure-dprint

yaml-lint: yaml-check

# Validate CITATION.cff against the Citation File Format schema.
citation-check: _ensure-uv
uvx --from cffconvert==2.0.0 cffconvert --validate -i CITATION.cff

# GitHub Actions security analysis
zizmor: _ensure-zizmor
zizmor .github
Loading
Loading