fffbac9aac
feat: MIR TypeOp/WeakRef/Barrier PoC implementation
...
- Add TypeOpKind, WeakRefOp, BarrierOp enums for unified instructions
- Implement TypeOp instruction combining TypeCheck/Cast
- Implement WeakRef instruction combining WeakNew/WeakLoad
- Implement Barrier instruction combining BarrierRead/BarrierWrite
- Update VM to handle new unified instructions
- Update MIR printer for new instruction formats
- Add feature flags mir_typeop_poc and mir_refbarrier_unify_poc
- Maintain backward compatibility with legacy instructions
This is Phase 8.5 MIR instruction diet PoC implementation.
🤖 Generated with [Claude Code](https://claude.ai/code )
Co-Authored-By: Claude <noreply@anthropic.com >
2025-08-23 19:27:02 +09:00
cc2a820af7
feat(plugin): Fix plugin BoxRef return and Box argument support
...
- Fixed deadlock in FileBox plugin copyFrom implementation (single lock)
- Added TLV Handle (tag=8) parsing in calls.rs for returned BoxRefs
- Improved plugin loader with config path consistency and detailed logging
- Fixed loader routing for proper Handle type_id/fini_method_id resolution
- Added detailed logging for TLV encoding/decoding in plugin_loader_v2
Test docs/examples/plugin_boxref_return.nyash now works correctly:
- cloneSelf() returns FileBox Handle properly
- copyFrom(Box) accepts plugin Box arguments
- Both FileBox instances close and fini correctly
🤖 Generated with [Claude Code](https://claude.ai/code )
Co-Authored-By: Claude <noreply@anthropic.com >
2025-08-21 00:41:26 +09:00
890b58a19e
docs: 実用優先戦略でPhase 9-10完全再設計
...
Phase 8.6+8.7完了受けて、実用価値最大化戦略に基づく設計変更：
📋 Phase 9-10再設計:
- Phase 9: JIT planning → AOT WASM実装（最優先・2-3週間）
- Phase 9.5: HTTPサーバー実用テスト追加（AOT検証・2週間）
- Phase 10: AOT exploration → LLVM Direct AOT（最高性能・4-6ヶ月）
- Cranelift JIT: Phase 12以降の将来オプションに変更
🎯 実用優先戦略の根拠:
- WASM既に動作済み（13.5倍高速化実証済み）
- AOT実装で即座配布価値提供
- Cranelift JITは重複投資（Rust開発環境改善効果限定）
- 時間効率：2-3ヶ月節約でLLVM集中投資
🚀 期待効果:
- `nyash --compile-native app.nyash -o app.exe` 実現
- 配布可能HTTPサーバーデモ（Phase 9.5）
- Everything is Box哲学のネイティブ最適化
- 1000倍高速化目標（Phase 10）
🤖 Generated with [Claude Code](https://claude.ai/code )
Co-Authored-By: Claude <noreply@anthropic.com >
2025-08-14 13:00:44 +09:00
8ec80a35c3
feat(benchmark): add comprehensive performance benchmarking system
...
🚀 Phase 8.2 PoC2 Achievement: 280x WASM performance boost proven\!
## New Features:
- Complete benchmark framework (src/benchmarks.rs)
- CLI integration: --benchmark --iterations options
- 3-backend comparison: Interpreter/VM/WASM
- Automated performance measurement & reporting
## Benchmark Results (100 iterations average):
- WASM: 0.17ms (280x faster than Interpreter\!)
- VM: 16.97ms (2.9x faster than Interpreter)
- Interpreter: 48.59ms (baseline)
## Added Files:
- benchmarks/bench_{light,medium,heavy}.nyash - Test cases
- benchmark_summary_20250814.md - Clean results
- wasm_demo/ - Browser execution environment
## Documentation Updates:
- docs/execution-backends.md - Added benchmark usage
- docs/CURRENT_TASK.md - Phase 8.3 Copilot coordination
- CLAUDE.md - Quick benchmark access
## Copilot Integration Ready:
- Phase 8.3 merge conflict avoidance strategy documented
- Benchmark framework ready for Box operation performance validation
- CLI integration preserved for future enhancements
🤖 Generated with [Claude Code](https://claude.ai/code )
Co-Authored-By: Claude <noreply@anthropic.com >
2025-08-14 07:19:23 +09:00
697223c8cf
docs: move execution-backends.md to docs root and add quick access
...
- Move docs/説明書/execution-backends.md → docs/execution-backends.md
- Add execution backend guide link to CLAUDE.md quick start section
- Update relative path in docs/説明書/README.md
- Improve developer access to VM/WASM execution options
🤖 Generated with [Claude Code](https://claude.ai/code )
Co-Authored-By: Claude <noreply@anthropic.com >
2025-08-14 06:46:57 +09:00
