[x] CMake
[ ] Logger
[x] Config
[x] CLI
[x] -h/--help Print Grouped Help with Built-in Defaults to stdout and Exit 0 before Loading config.toml or Creating Files (PR-0033)
[x] Unknown Options Rejected with pi-engine: unknown option: <arg> and Try 'pi-engine --help' for usage. on stderr, Exit 2 (PR-0033)
[x] Missing and Malformed CLI Values Are Usage Errors (Exit 2); Configuration, Runtime, and Verification Failures Stay Exit 1 (PR-0033)
[x] CLI Parsing Separated from Override Application (CommandLine::parse before ConfigLoader::load) (PR-0033)
[x] Parser/Help Unit Test (CommandLineTest) and Real-Process CLI Option Regression (CommandLineCliTest) Added (PR-0033)
[x] PR-0033 Verification Gate (complete Debug build, 68/68 CTest, 4/4 changed-path ASan/UBSan tests)
[x] Project directory structure
[x] README Algorithm, Architecture Direction, and Current Status
[x] Initial scheduler module skeleton
[x] CTest
[x] GitHub Actions CI
[x] Automated Build
[x] Automated Test
[x] GCC Build Matrix
[x] Clang Build Matrix
[x] clang-format Check
[x] clang-tidy Static Analysis
[x] AddressSanitizer
[x] UBSan
[x] Artifact Upload
[ ] Benchmark Workflow
[ ] Release Workflow
[x] CPUID
[x] AVX Detect
[x] AVX2 Detect
[x] AVX512 Detect
[ ] Huge Pages
[ ] NUMA
[x] Arena
[x] Pool
[x] Alignment
[x] Scratch Buffer
[x] Task
[x] Worker
[x] Scheduler
[x] ThreadPool
[x] Reference Queue
[x] Lock-Free Queue
[x] Work Stealing
[x] Task Synchronization
[x] Task Handle / Future
[x] Task Join
[x] Shared Task State
[x] Task Execute Lifecycle
[x] Task Exception Capture
[x] TaskHandle Exception Inspection and Rethrow
[x] TaskHandle Validity
[x] TaskHandle Completion Wait
[x] Scheduler / ThreadPool Unified Submit Handle API
[x] Scheduler Synchronization Test Scenarios
[x] Scheduler Lifecycle State
[x] Concurrent Start / Stop Serialization
[x] Idempotent Start / Stop
[x] Scheduler Restart After Stop
[x] Submission Rejection While Stopping
[x] Drain Shutdown Semantics
[x] Accepted Outstanding Task Accounting
[x] Two-Phase Worker Stop and Join
[x] Global / Local Queue Role Contract
[x] Cross-Pool Worker Submission Routing
[x] Global Queue Capacity Rejection
[x] Lock-Free Queue Capacity Validation
[x] Lock-Free Queue Sequence / Contention Correctness
[x] Lock-Free Queue Slot Reuse Coverage
[x] Lock-Free Queue Concurrent Tests
[x] Lock-Free Queue Exactly-Once MPMC Coverage
[x] Concurrent Submit / Stop Acceptance Cutoff Test
[x] Concurrent Accepted / Rejected Execution Accounting
[x] Multi-Caller Repeated Start / Stop Test
[x] Dedicated Scheduler Concurrency Test Target
[x] Work-Stealing Behavior Tests
[x] Worker-Local Child Concentration Test
[x] Multiple Thief Thread Observation
[x] Root / Thief Thread Identity Validation
[x] Global Fallback Exclusion Test Queue
[x] Work-Stealing Load / Behavior Test Separation
[x] Scheduler Correctness Hardening (PR-0017)
[x] GMP
[x] In-Place Sign Negation
[ ] FFT
[x] Node
[x] Merge
[x] Recursive Split
[x] Parallel API Refactoring
[x] Chudnovsky Leaf Integration
[x] Input Range Validation
[x] Leaf and Small-Range P/Q/T Correctness
[x] Small Known-Digits Integration Test
[x] Binary Splitting Boundary and Regression Coverage
[x] Parallel Cutoff Policy
[x] Parallel Execution
[x] Parallel Merge
[x] Staged Parallel DAG
[x] Parallel Submission Fallback
[x] Worker-Context Deadlock Avoidance
[x] Observable Parallel Binary Splitting Test
[x] Formula
[x] Precision Control
[x] Precision and Guard-Digit Policy
[x] Overflow-Safe Digit-to-Term Conversion
[x] Estimated Binary Precision
[x] GMP Power-of-Ten Operation
[x] GMP Floor Integer Square Root
[x] GMP Floor Division
[x] End-to-End Sequential Pi Calculation
[x] End-to-End Parallel Pi Calculation
[x] Guard-Digit Rounding
[x] Normalized Decimal Output
[x] Production Known-Digits Coverage
[x] Guard-Precision Stability at 1,000 Digits
[x] Split / Finalize / Format Timing Metrics
[x] Standalone End-to-End Benchmark Target
[x] Sequential / Parallel Benchmark Result Validation
[x] Storage Lifecycle and Ownership Contract (PR-0025)
[x] Chunk Identity and Metadata (PR-0025)
[x] Typed Storage Policy Model (PR-0025)
[x] Versioned Chunk Format (PR-0025)
[x] Compression Codec Interface and None Codec (PR-0025)
[x] Compression (LZ4 backend PR-0028; none/LZ4 selection and acceptance PR-0029)
[x] Chunk Index Contract and Durable Codec (PR-0025)
[x] ChunkCodec-backed Local ChunkStore and Reload (PR-0025)
[x] Synchronous StorageManager Boundary (PR-0025)
[x] Deterministic Memory Budget and Eviction Planner (PR-0025)
[x] Storage Configuration and Progress Telemetry Integration (PR-0025)
[x] Large Synthetic P/Q/T Round-trip and Throughput Validation (PR-0025)
[x] In-memory versus Forced Out-of-Core Merge Benchmark (PR-0026)
[x] Synchronous Out-of-Core Merge Stabilization Boundary (PR-0026)
[x] Production StorageMergeCoordinator Wiring (PR-0027)
[x] Asynchronous Storage Writer Contract (PR-0027)
[x] Bounded Asynchronous I/O Queue and Writer Runner (PR-0027)
[x] Asynchronous NodeLifecycle Completion Transition (PR-0027)
[x] Merge Wait and Backpressure (PR-0027)
[x] Asynchronous Reload Reader and Merge Integration (PR-0027)
[x] Asynchronous Storage I/O Progress Telemetry (PR-0027)
[x] Asynchronous I/O Failure and Shutdown Tests (PR-0027)
[x] Async Spill/Reload Performance Baseline (PR-0027)
[x] Concurrent I/O Execution (PR-0028)
[x] PR-0028 Functional Storage Pipeline Verification
[x] Platform Applicability Audit (PR-0028)
[x] Compression Backend Optimization and Selection (PR-0029; LZ4 space/RSS acceptance, latency trade-off documented)
[x] LZ4 Compression Backend and Durable Round-trip (PR-0028)
[ ] NUMA, Huge Pages, and Affinity Optimization (deferred; separate PR requiring suitable hardware)
[x] Separate file-read, CRC32C, and GMP decode timings (PR-0029)
[x] Cold-cache versus warm-cache reload benchmark
[x] Multi-chunk and concurrent I/O throughput benchmark (256 cells × 5 repetitions, 1,280/1,280 successful; cold/warm accepted)
[x] None versus LZ4 size/throughput comparison (5x12-cell acceptance; no latency-speedup claim)
[x] Single-chunk peak RSS instrumentation and release baseline for 100 MiB, 512 MiB, and 1 GiB workloads
[x] Single-chunk peak RSS five-sample distribution and p95 acceptance (12 cells, 60/60 successful; p95 below the recorded PR-0028 release baseline)
[x] Single-chunk five-sample p50/p95 matrix completion
[x] Large-data ASan/UBSan round-trip validation (512 MiB and 1 GiB; LeakSanitizer environment-limited)
[x] PR-0028 High-Performance Measurement Contract
[x] Independent Reader/Writer Concurrency Benchmark (64 cells × 5 repetitions, 320/320 successful; RSS is not emitted by the current benchmark)
[x] I/O, CRC32C, Codec, and GMP Decode Timing Breakdown (PR-0029)
[x] PR-0029 Final Performance Decision and Stabilization (64/64 CTest; Release and sanitizer smoke)
[x] Verified Reload Duplicate Read/Decode Elimination (PR-0029)
[x] Bounded Buffer/Data-Movement Optimization Validation (round-trip and Release baseline)
[x] Bounded Buffer/Data-Movement Repeated Speedup Acceptance (12-cell end-to-end before/after; individual stage attribution limited)
[x] CRC32C SSE4.2/fallback, pre-sized read, and positive GMP export optimization validation
[x] CRC/read/GMP Repeated Speedup Acceptance (combined end-to-end before/after; baseline stage telemetry unavailable)
[x] PR-0029 Measurement Environment and Acceptance Contract
[x] PR-0029 Workload Matrix Baseline (multi-chunk)
[x] P/Q/T Block Format
[x] Block Magic and Format Version
[x] Range and Tree-Level Metadata
[x] P/Q/T Payload Length Metadata
[x] Checksum Type and Value Metadata
[x] Atomic Temporary File Commit
[x] Manifest
[x] Save
[x] Structural Block Validation
[x] Payload Checksum Validation
[x] Manifest Consistency Validation
[x] Resume from Verified Blocks
[x] Corrupt Block Quarantine
[x] Corrupt Block Recalculation
[x] Production Checkpoint Emission (PR-0030)
[x] Production Verified Resume Integration (PR-0030)
[x] Live Checkpoint Telemetry Publication after Manifest Replacement (PR-0031)
[x] Automated Interrupted CLI Resume Regression (PR-0031)
[x] PR-0031 Verification Gate (complete Debug build, 66/66 CTest, 8/8 changed-path ASan/UBSan tests)
[x] Default No-Config Run Leaves Manifest and Root Checkpoint Block under checkpoints/ (PR-0032)
[x] Resume-Only Mode Plans No Recalculated Checkpoint Work; Empty Accepted Plan Reports checkpoints=0/0 and Writes No Files (PR-0032)
[x] CRC
[x] Progress Phase
[x] Thread-Safe Progress Tracker
[x] Immutable Progress Snapshot
[x] Total and Completed Terms
[x] Total and Completed Blocks
[x] Current Merge Level
[x] Active and Queued Tasks
[x] ETA
[x] Speed
[x] Memory
[x] Chunk (storage resident/stored bytes, indexed chunk count, and asynchronous read/write counters; PR-0025/PR-0027)
[x] Checkpoint Bytes (exact manifest-referenced block bytes since PR-0031)
[x] Last Verified Checkpoint (resume-validated input only)
[x] Monotonic Checkpoint Work Completion (PR-0031)
[x] Durable Checkpoint Block Gauge (PR-0031)
[x] Last Committed Checkpoint (PR-0031)
[x] JSON Progress Schema Version 4 (PR-0031)
[x] CLI Text Reporter
[x] CLI JSON Reporter
[x] TTY and Non-TTY Output Modes
[x] Progress Enable / Disable Option
[x] Progress Interval Option
[x] Progress Format Option
[x] Post-CLI Configuration Validation
[x] Resume Progress Reconstruction
[x] Future Reporter Extension Point
[x] Default Text Terminal Completion Shows phase=completed, terms=N/N, Non-Zero Complete Checkpoint Work, Durable Blocks, and Bytes (PR-0032)
[x] Reporter and Output-Stream Failures Propagated as CLI Failure after Runner Join (PR-0032)
[x] Compact TTY Text Record; Full Non-TTY Text Log and JSON Schema v4 Unchanged (PR-0032)
[x] PR-0032 Verification Gate (7/7 changed-path ASan/UBSan tests, complete Debug build, 66/66 CTest)
Run from an empty working directory with no config.toml, using an executable
rebuilt from the current source (cmake --build build --parallel, then
/path/to/repo/build/pi-engine). Relative paths resolve against this working
directory.
[x] No-argument run exits with status 0 and writes pi.txt and parentless relative pi.txt.verify
[x] Final stderr text record shows phase=completed and terms=N/N with equal values
[x] Final record shows checkpoints=C/C with C > 0, durable=D with D > 0, and non-zero checkpoint_bytes
[x] checkpoints/ contains a non-empty manifest and the root checkpoint block it references; checkpoint_bytes equals the referenced block file sizes
[x] On a TTY the progress line is compact (no storage=/io= diagnostics) and the final line ends with a newline
[x] With stderr redirected to a file, every record is a timestamped line containing storage= and io= diagnostics
[x] --progress-format json emits JSONL records with schema_version 4
[x] --no-checkpoint --resume in a fresh directory reports checkpoints=0/0 and creates no checkpoint files
[x] A progress reporter or output-stream failure (verified with stderr redirected to /dev/full) makes the CLI exit non-zero instead of reporting success
[x] Independent Modular P/Q/T Verification
[x] Canonical Output Contract
[x] Streaming SHA-256
[x] Versioned Known Digits
[x] BBP Hexadecimal Digit Calculation
[x] GMP Decimal-to-Hex Digit Extraction
[x] Deterministic BBP Sample Policy
[x] Final Verification Orchestrator
[x] Versioned Verification Manifest
[x] Manifest Hash Revalidation
[x] CLI Verification Controls and Diagnostics
[x] Verifying Output Progress Phase
[x] 128 MiB Bounded-Memory Streaming Verification
[x] Final Verification Release Benchmark
[ ] AVX2
[ ] AVX512
[ ] CPUID
[ ] memcpy
[ ] Prefetch
[ ] rdtsc
[ ] Single Thread
[ ] Multi Thread
[ ] Memory
[ ] Storage