Skip to content

x86_64 CPU parity: AVX2+FMA quantized matmul kernels and kernel-machine routing #89

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
gabewillen merged 10 commits into from
Jul 2, 2026
Merged

x86_64 CPU parity: AVX2+FMA quantized matmul kernels and kernel-machine routing #89

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
a6b60b3
feat: ship v1.27 ryzen avx2 fma support
gabewillen Jun 25, 2026
07c7a3a
feat: view-sliced parallel matmul cutover with cross-engine compariso…
gabewillen Jul 2, 2026
e3f3c38
feat: x86_64 AVX2+FMA quantized matmul kernels (Q4_K, Q4_0/Q4_1/Q5_0/…
gabewillen Jul 2, 2026
fa8cf37
feat: route domain matmuls through kernel machine; add f32 matrix-vec…
gabewillen Jul 2, 2026
36145f0
test: cover all whisper linear-variant instantiations and embeddings …
gabewillen Jul 2, 2026
4912bab
chore: refresh quality gates timing data
gabewillen Jul 2, 2026
fd3c26b
fix: gate host AVX flags on build-host CPU probe; run rejected wavefr…
gabewillen Jul 2, 2026
adbe964
docs: document sign_epi8 -128 precondition on dot_i8_pairs_i32x8_avx2
gabewillen Jul 2, 2026
82dbe47
fix: bind reserved computes to the stored reservation's tensor lifecycle
gabewillen Jul 2, 2026
4850c04
fix: route duplicate graph actors through serial wavefront dispatch
gabewillen Jul 2, 2026
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
62 changes: 31 additions & 31 deletions .codex/config.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,124 +6,124 @@ codex_hooks = true

[agents.gsd-advisor-researcher]
description = "Researches a single gray area decision and returns a structured comparison table with rationale. Spawned by discuss-phase advisor mode."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-advisor-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-advisor-researcher.toml"

[agents.gsd-ai-researcher]
description = "Researches a chosen AI framework's official docs to produce implementation-ready guidance — best practices, syntax, core patterns, and pitfalls distilled for the specific use case. Writes the Framework Quick Reference and Implementation Guidance sections of AI-SPEC.md. Spawned by /gsd-ai-integration-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-ai-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-ai-researcher.toml"

[agents.gsd-assumptions-analyzer]
description = "Deeply analyzes codebase for a phase and returns structured assumptions with evidence. Spawned by discuss-phase assumptions mode."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-assumptions-analyzer.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-assumptions-analyzer.toml"

[agents.gsd-code-fixer]
description = "Applies fixes to code review findings from REVIEW.md. Reads source files, applies intelligent fixes, and commits each fix atomically. Spawned by /gsd-code-review-fix."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-code-fixer.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-code-fixer.toml"

[agents.gsd-code-reviewer]
description = "Reviews source files for bugs, security issues, and code quality problems. Produces structured REVIEW.md with severity-classified findings. Spawned by /gsd-code-review."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-code-reviewer.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-code-reviewer.toml"

[agents.gsd-codebase-mapper]
description = "Explores codebase and writes structured analysis documents. Spawned by map-codebase with a focus area (tech, arch, quality, concerns). Writes documents directly to reduce orchestrator context load."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-codebase-mapper.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-codebase-mapper.toml"

[agents.gsd-debug-session-manager]
description = "Manages multi-cycle /gsd-debug checkpoint and continuation loop in isolated context. Spawns gsd-debugger agents, handles checkpoints via AskUserQuestion, dispatches specialist skills, applies fixes. Returns compact summary to main context. Spawned by /gsd-debug command."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-debug-session-manager.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-debug-session-manager.toml"

[agents.gsd-debugger]
description = "Investigates bugs using scientific method, manages debug sessions, handles checkpoints. Spawned by /gsd-debug orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-debugger.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-debugger.toml"

[agents.gsd-doc-verifier]
description = "Verifies factual claims in generated docs against the live codebase. Returns structured JSON per doc."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-doc-verifier.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-doc-verifier.toml"

[agents.gsd-doc-writer]
description = "Writes and updates project documentation. Spawned with a doc_assignment block specifying doc type, mode (create/update/supplement), and project context."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-doc-writer.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-doc-writer.toml"

[agents.gsd-domain-researcher]
description = "Researches the business domain and real-world application context of the AI system being built. Surfaces domain expert evaluation criteria, industry-specific failure modes, regulatory context, and what \"good\" looks like for practitioners in this field — before the eval-planner turns it into measurable rubrics. Spawned by /gsd-ai-integration-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-domain-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-domain-researcher.toml"

[agents.gsd-eval-auditor]
description = "Retroactive audit of an implemented AI phase's evaluation coverage. Checks implementation against the AI-SPEC.md evaluation plan. Scores each eval dimension as COVERED/PARTIAL/MISSING. Produces a scored EVAL-REVIEW.md with findings, gaps, and remediation guidance. Spawned by /gsd-eval-review orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-eval-auditor.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-eval-auditor.toml"

[agents.gsd-eval-planner]
description = "Designs a structured evaluation strategy for an AI phase. Identifies critical failure modes, selects eval dimensions with rubrics, recommends tooling, and specifies the reference dataset. Writes the Evaluation Strategy, Guardrails, and Production Monitoring sections of AI-SPEC.md. Spawned by /gsd-ai-integration-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-eval-planner.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-eval-planner.toml"

[agents.gsd-executor]
description = "Executes GSD plans with atomic commits, deviation handling, checkpoint protocols, and state management. Spawned by execute-phase orchestrator or execute-plan command."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-executor.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-executor.toml"

[agents.gsd-framework-selector]
description = "Presents an interactive decision matrix to surface the right AI/LLM framework for the user's specific use case. Produces a scored recommendation with rationale. Spawned by /gsd-ai-integration-phase and /gsd-select-framework orchestrators."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-framework-selector.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-framework-selector.toml"

[agents.gsd-integration-checker]
description = "Verifies cross-phase integration and E2E flows. Checks that phases connect properly and user workflows complete end-to-end."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-integration-checker.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-integration-checker.toml"

[agents.gsd-intel-updater]
description = "Analyzes codebase and writes structured intel files to .planning/intel/."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-intel-updater.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-intel-updater.toml"

[agents.gsd-nyquist-auditor]
description = "Fills Nyquist validation gaps by generating tests and verifying coverage for phase requirements"
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-nyquist-auditor.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-nyquist-auditor.toml"

[agents.gsd-pattern-mapper]
description = "Analyzes codebase for existing patterns and produces PATTERNS.md mapping new files to closest analogs. Read-only codebase analysis spawned by /gsd-plan-phase orchestrator before planning."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-pattern-mapper.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-pattern-mapper.toml"

[agents.gsd-phase-researcher]
description = "Researches how to implement a phase before planning. Produces RESEARCH.md consumed by gsd-planner. Spawned by /gsd-plan-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-phase-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-phase-researcher.toml"

[agents.gsd-plan-checker]
description = "Verifies plans will achieve phase goal before execution. Goal-backward analysis of plan quality. Spawned by /gsd-plan-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-plan-checker.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-plan-checker.toml"

[agents.gsd-planner]
description = "Creates executable phase plans with task breakdown, dependency analysis, and goal-backward verification. Spawned by /gsd-plan-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-planner.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-planner.toml"

[agents.gsd-project-researcher]
description = "Researches domain ecosystem before roadmap creation. Produces files in .planning/research/ consumed during roadmap creation. Spawned by /gsd-new-project or /gsd-new-milestone orchestrators."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-project-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-project-researcher.toml"

[agents.gsd-research-synthesizer]
description = "Synthesizes research outputs from parallel researcher agents into SUMMARY.md. Spawned by /gsd-new-project after 4 researcher agents complete."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-research-synthesizer.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-research-synthesizer.toml"

[agents.gsd-roadmapper]
description = "Creates project roadmaps with phase breakdown, requirement mapping, success criteria derivation, and coverage validation. Spawned by /gsd-new-project orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-roadmapper.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-roadmapper.toml"

[agents.gsd-security-auditor]
description = "Verifies threat mitigations from PLAN.md threat model exist in implemented code. Produces SECURITY.md. Spawned by /gsd-secure-phase."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-security-auditor.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-security-auditor.toml"

[agents.gsd-ui-auditor]
description = "Retroactive 6-pillar visual audit of implemented frontend code. Produces scored UI-REVIEW.md. Spawned by /gsd-ui-review orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-ui-auditor.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-ui-auditor.toml"

[agents.gsd-ui-checker]
description = "Validates UI-SPEC.md design contracts against 6 quality dimensions. Produces BLOCK/FLAG/PASS verdicts. Spawned by /gsd-ui-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-ui-checker.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-ui-checker.toml"

[agents.gsd-ui-researcher]
description = "Produces UI-SPEC.md design contract for frontend phases. Reads upstream artifacts, detects design system state, asks only unanswered questions. Spawned by /gsd-ui-phase orchestrator."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-ui-researcher.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-ui-researcher.toml"

[agents.gsd-user-profiler]
description = "Analyzes extracted session messages across 8 behavioral dimensions to produce a scored developer profile with confidence levels and evidence. Spawned by profile orchestration workflows."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-user-profiler.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-user-profiler.toml"

[agents.gsd-verifier]
description = "Verifies phase goal achievement through goal-backward analysis. Checks codebase delivers what phase promised, not just that tasks completed. Creates VERIFICATION.md report."
config_file = "/Users/gabrielwillen/VSCode/stateforward/emel/emel.cpp/.codex/agents/gsd-verifier.toml"
config_file = "/shared/stateforward/emel.cpp/.codex/agents/gsd-verifier.toml"
Comment thread
gabewillen marked this conversation as resolved.
67 changes: 67 additions & 0 deletions .planning/MILESTONES.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,72 @@
# Project Milestones: EMEL

## v1.27 Ryzen AVX2/FMA Kernel Support (Shipped: 2026-06-25)

**Phases completed:** 6 phases, 6 plans, 0 tasks

**Key accomplishments:**

- Added a source-backed x86_64 host feature contract for AVX2, FMA, and F16C
while explicitly no-claiming AVX-512, AVX-VNNI, AMX, BF16, native FP16, GPU,
and broader x86 feature families.

- Added EMEL-owned AVX2/FMA/F16C flash-attention execution for supported
x86_64 requests with deterministic shared fallback/no-claim behavior.

- Added EMEL-owned AVX2/FMA q2_K, q3_K, and q6_K x q8_K hot-path kernels with
block-native operand flow, no whole-tensor f32 dequant substitution, and
allocation-free supported dispatch proof.

- Integrated the optimized x86_64 path through the maintained generator ->
graph -> processor -> kernel chain and paritychecker attribution for `1`,
`10`, `100`, and `1000` token generation.

- Published truthful `kernel_x86_64` benchmark evidence with counter-checked
optimized flash and q2/q3/q6 rows separated from shared/scalar/reference
lanes.

**Audit:** Final source-backed audit passed with 13/13 requirements satisfied.
The initial `XBN-01` benchmark attribution gap was closed before archive, and
the x86_64 unary SML rule debt was removed from the milestone path.

---

## v1.26 I/O Staged Read Loading Strategy (Shipped: 2026-05-08)

**Phases completed:** 12 phases, 12 plans, 0 tasks

**Key accomplishments:**

- Established the canonical `src/emel/io/staged_read` Stateforward.SML strategy
actor under the existing `emel/io` boundary.

- Modeled source span, target window, stage sizing, platform/resource, and
validation behavior through explicit guards/transitions before accepting
staged copy work.

- Implemented deterministic source-span staged copy semantics with monotonic
progress, explicit terminal success, and named deterministic error categories.

- Integrated staged loading through public tensor-to-I/O dispatch while keeping
`model/tensor` as the sole tensor load/bind/evict/residency owner.

- Kept maintained loader, benchmark, paritychecker, and embedded probe lanes on
public runtime contracts with no actor-internal reach-through.

- Closed the source-backed audit by repairing direct tensor staged-load
nonzero-offset source-window behavior and reconciling closeout artifact plus
embedded-probe reporting truth.

**Audit:** Final source-backed audit passed with all active requirements
satisfied after Phase 237 and Phase 238 gap closure. `ESG-02B` remains
deferred/future because real file open/seek/read and per-stage short-read
taxonomy requires a separately approved file-backed staged-read source path.

**Known deferred items at close:** `ESG-02B` plus the carried-forward deferred
items listed in STATE.md.

---

## v1.25 I/O Read Loading Strategy (Shipped: 2026-05-06)

**Phases completed:** 16 phases, 21 plans, 12 tasks
Expand Down
Loading