ci: enable sccache pilot (heavy workload — APR-MONO)#894
Merged
Conversation
Phase 3 sccache pilot — heavy workload tier. APR-MONO monorepo has 60+ crates with the largest dep graph in the fleet, so it should produce the strongest hit-rate signal for fleet-wide rollout vs shelve. Replaces paiml-mcp-agent-toolkit as heavy pilot (unworkable due to local bare-repo config + master-only policy). Joins: - copia (light workload, PR #30) - bashrs (medium workload, PR #197) Spec: docs/specifications/build-performance.md §7 Phase 3 Falsifier: examples/falsify_f9_cache_hit_rate.rs (>=20% miss-rate gate) (Refs PMAT-151)
noahgift
added a commit
that referenced
this pull request
Apr 19, 2026
…-TWO-001 teacher Re-dated 2026-04-15 → 2026-04-19 since v0.31.0 tag was never cut; rolls up everything that landed between PR #748 (initial 0.30→0.31 bump) and PR #888 (MCP spec v1.2.0 + parity epic closure). New in [0.31.0] - 2026-04-19 (beyond original 2026-04-15 rc1 content): - MCP Server M1–M3: 9 apr tools over stdio JSON-RPC 2.0, YAML-codegen'd schemas (FALSIFY-MCP-008), notifications/cancelled + progress, JSON Schema Draft 7 meta-validation in CI. - apr code — Claude Code parity epic CLOSED at v5.1 (14/3/4 over 21 rows; PMAT-CODE-PARITY-MATRIX-001 both closure conditions met). 10 tickets closed in one cycle (P0×4, P1×5, P2×2). - Contracts harness: `pv check-parity` SEMANTIC gate + new apr-claude-proxy-v1.yaml DRAFT. - SHIP-TWO-001 teacher shipped: paiml/qwen2.5-coder-7b-apache-q4k-v1 (7.5 GB .apr, Apache-2.0) — first artifact to pass full apr publish contract. `apr validate-manifest` + `--live` + FALSIFY-PM-007 safetensors dtype Poka-Yoke. - Perf: decode hot-path hygiene HP-001/002/003 (184→382 tok/s on 1.5B Q4_K_M, 2.07×); 32-tok bench 442.8→479.9; FlashDecoding gated for small models. - CI: sccache pilot (#894), nextest opt-in (#897). - Flaky perf test fixes: tui_load, F-203, RP-002-prop, citl-neural. Previous [0.31.0] - 2026-04-15 entry was an rc draft; merged into the final 2026-04-19 cut to keep the release history single-authoritative. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
5 tasks
noahgift
added a commit
that referenced
this pull request
Apr 19, 2026
…-TWO-001 teacher (#899) Re-dated 2026-04-15 → 2026-04-19 since v0.31.0 tag was never cut; rolls up everything that landed between PR #748 (initial 0.30→0.31 bump) and PR #888 (MCP spec v1.2.0 + parity epic closure). New in [0.31.0] - 2026-04-19 (beyond original 2026-04-15 rc1 content): - MCP Server M1–M3: 9 apr tools over stdio JSON-RPC 2.0, YAML-codegen'd schemas (FALSIFY-MCP-008), notifications/cancelled + progress, JSON Schema Draft 7 meta-validation in CI. - apr code — Claude Code parity epic CLOSED at v5.1 (14/3/4 over 21 rows; PMAT-CODE-PARITY-MATRIX-001 both closure conditions met). 10 tickets closed in one cycle (P0×4, P1×5, P2×2). - Contracts harness: `pv check-parity` SEMANTIC gate + new apr-claude-proxy-v1.yaml DRAFT. - SHIP-TWO-001 teacher shipped: paiml/qwen2.5-coder-7b-apache-q4k-v1 (7.5 GB .apr, Apache-2.0) — first artifact to pass full apr publish contract. `apr validate-manifest` + `--live` + FALSIFY-PM-007 safetensors dtype Poka-Yoke. - Perf: decode hot-path hygiene HP-001/002/003 (184→382 tok/s on 1.5B Q4_K_M, 2.07×); 32-tok bench 442.8→479.9; FlashDecoding gated for small models. - CI: sccache pilot (#894), nextest opt-in (#897). - Flaky perf test fixes: tui_load, F-203, RP-002-prop, citl-neural. Previous [0.31.0] - 2026-04-15 entry was an rc draft; merged into the final 2026-04-19 cut to keep the release history single-authoritative. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Why
build-performance.md §7 Phase 3 requires diverse workloads to produce a representative hit-rate signal. APR-MONO delivers the upper bound — if sccache wins here, it wins everywhere.
Falsifier
`examples/falsify_f9_cache_hit_rate.rs` — F9 gate in infra repo. >=20 samples, miss-rate delta vs baseline.
Test plan
(Refs PMAT-151)
🤖 Generated with Claude Code