Files
hakmem/core/box/tiny_route_box.h

65 lines
2.4 KiB
C
Raw Normal View History

P-Tier + Tiny Route Policy: Aggressive Superslab Management + Safe Routing ## Phase 1: Utilization-Aware Superslab Tiering (案B実装済) - Add ss_tier_box.h: Classify SuperSlabs into HOT/DRAINING/FREE based on utilization - HOT (>25%): Accept new allocations - DRAINING (≤25%): Drain only, no new allocs - FREE (0%): Ready for eager munmap - Enhanced shared_pool_release_slab(): - Check tier transition after each slab release - If tier→FREE: Force remaining slots to EMPTY and call superslab_free() immediately - Bypasses LRU cache to prevent registry bloat from accumulating DRAINING SuperSlabs - Test results (bench_random_mixed_hakmem): - 1M iterations: ✅ ~1.03M ops/s (previously passed) - 10M iterations: ✅ ~1.15M ops/s (previously: registry full error) - 50M iterations: ✅ ~1.08M ops/s (stress test) ## Phase 2: Tiny Front Routing Policy (新規Box) - Add tiny_route_box.h/c: Single 8-byte table for class→routing decisions - ROUTE_TINY_ONLY: Tiny front exclusive (no fallback) - ROUTE_TINY_FIRST: Try Tiny, fallback to Pool if fails - ROUTE_POOL_ONLY: Skip Tiny entirely - Profiles via HAKMEM_TINY_PROFILE ENV: - "hot": C0-C3=TINY_ONLY, C4-C6=TINY_FIRST, C7=POOL_ONLY - "conservative" (default): All TINY_FIRST - "off": All POOL_ONLY (disable Tiny) - "full": All TINY_ONLY (microbench mode) - A/B test results (ws=256, 100k ops random_mixed): - Default (conservative): ~2.90M ops/s - hot: ~2.65M ops/s (more conservative) - off: ~2.86M ops/s - full: ~2.98M ops/s (slightly best) ## Design Rationale ### Registry Pressure Fix (案B) - Problem: DRAINING tier SS occupied registry indefinitely - Solution: When total_active_blocks→0, immediately free to clear registry slot - Result: No more "registry full" errors under stress ### Routing Policy Box (新) - Problem: Tiny front optimization scattered across ENV/branches - Solution: Centralize routing in single table, select profiles via ENV - Benefit: Safe A/B testing without touching hot path code - Future: Integrate with RSS budget/learning layers for dynamic profile switching ## Next Steps (性能最適化) - Profile Tiny front internals (TLS SLL, FastCache, Superslab backend latency) - Identify bottleneck between current ~2.9M ops/s and mimalloc ~100M ops/s - Consider: - Reduce shared pool lock contention - Optimize unified cache hit rate - Streamline Superslab carving logic 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 18:01:25 +09:00
// tiny_route_box.h - Box: Tiny Front Routing Policy
//
// Purpose:
// Decide, per Tiny class, whether allocation should go through Tiny front
// or directly to the Pool/backend. This keeps routing policy in a single,
// cheap table lookup, without getenv() or complex logic in the hot path.
//
// Box Theory:
// - Single Responsibility:
// Only decides "Tiny vs Pool vs Tiny+Fallback" per class.
// - Clear Boundary:
// Front Gate / Alloc Gatekeeper calls tiny_route_get(class_idx) once.
// Tiny Fast Path and Pool backend remain unchanged.
// - Reversible / A/B:
// Profiles are selected via HAKMEM_TINY_PROFILE ENV at init time.
// Hot path is stable; routing can be tuned without touching fast code.
#ifndef TINY_ROUTE_BOX_H
#define TINY_ROUTE_BOX_H
#include <stdint.h>
#include <stdio.h>
P-Tier + Tiny Route Policy: Aggressive Superslab Management + Safe Routing ## Phase 1: Utilization-Aware Superslab Tiering (案B実装済) - Add ss_tier_box.h: Classify SuperSlabs into HOT/DRAINING/FREE based on utilization - HOT (>25%): Accept new allocations - DRAINING (≤25%): Drain only, no new allocs - FREE (0%): Ready for eager munmap - Enhanced shared_pool_release_slab(): - Check tier transition after each slab release - If tier→FREE: Force remaining slots to EMPTY and call superslab_free() immediately - Bypasses LRU cache to prevent registry bloat from accumulating DRAINING SuperSlabs - Test results (bench_random_mixed_hakmem): - 1M iterations: ✅ ~1.03M ops/s (previously passed) - 10M iterations: ✅ ~1.15M ops/s (previously: registry full error) - 50M iterations: ✅ ~1.08M ops/s (stress test) ## Phase 2: Tiny Front Routing Policy (新規Box) - Add tiny_route_box.h/c: Single 8-byte table for class→routing decisions - ROUTE_TINY_ONLY: Tiny front exclusive (no fallback) - ROUTE_TINY_FIRST: Try Tiny, fallback to Pool if fails - ROUTE_POOL_ONLY: Skip Tiny entirely - Profiles via HAKMEM_TINY_PROFILE ENV: - "hot": C0-C3=TINY_ONLY, C4-C6=TINY_FIRST, C7=POOL_ONLY - "conservative" (default): All TINY_FIRST - "off": All POOL_ONLY (disable Tiny) - "full": All TINY_ONLY (microbench mode) - A/B test results (ws=256, 100k ops random_mixed): - Default (conservative): ~2.90M ops/s - hot: ~2.65M ops/s (more conservative) - off: ~2.86M ops/s - full: ~2.98M ops/s (slightly best) ## Design Rationale ### Registry Pressure Fix (案B) - Problem: DRAINING tier SS occupied registry indefinitely - Solution: When total_active_blocks→0, immediately free to clear registry slot - Result: No more "registry full" errors under stress ### Routing Policy Box (新) - Problem: Tiny front optimization scattered across ENV/branches - Solution: Centralize routing in single table, select profiles via ENV - Benefit: Safe A/B testing without touching hot path code - Future: Integrate with RSS budget/learning layers for dynamic profile switching ## Next Steps (性能最適化) - Profile Tiny front internals (TLS SLL, FastCache, Superslab backend latency) - Identify bottleneck between current ~2.9M ops/s and mimalloc ~100M ops/s - Consider: - Reduce shared pool lock contention - Optimize unified cache hit rate - Streamline Superslab carving logic 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 18:01:25 +09:00
// Routing policy per Tiny class.
typedef enum {
ROUTE_TINY_ONLY = 0, // Tiny front only (no fallback; failure bubbles up)
ROUTE_TINY_FIRST = 1, // Try Tiny front, then fallback to Pool backend
ROUTE_POOL_ONLY = 2, // Skip Tiny entirely, use Pool/backend only
} TinyRoutePolicy;
// Global routing table for Tiny classes (0..7).
// Initialized once from ENV: HAKMEM_TINY_PROFILE.
extern uint8_t g_tiny_route[8];
// Initialize routing table from ENV profile.
// Profiles:
// "hot" C0-C3=TINY_ONLY, C4-C6=TINY_FIRST, C7=POOL_ONLY
// "conservative" 全クラス TINY_FIRSTデフォルト
// "off" 全クラス POOL_ONLYTiny 無効)
// "full" 全クラス TINY_ONLYmicrobench 用)
void tiny_route_init(void);
// Hot path helper: return routing policy for a given class.
// Uses simple array lookup; class_idx is masked to [0,7] defensively.
static inline TinyRoutePolicy tiny_route_get(int class_idx)
{
TinyRoutePolicy p = (TinyRoutePolicy)g_tiny_route[class_idx & 7];
#if HAKMEM_BUILD_RELEASE
if ((class_idx & 7) == 7) {
static int rel_route_logged = 0;
if (!rel_route_logged) {
const char* mode =
(p == ROUTE_TINY_ONLY) ? "TINY_ONLY" :
(p == ROUTE_TINY_FIRST) ? "TINY_FIRST" :
(p == ROUTE_POOL_ONLY) ? "POOL_ONLY" : "UNKNOWN";
fprintf(stderr, "[REL_C7_ROUTE] via tiny_route_get route=%s\n", mode);
rel_route_logged = 1;
}
}
#endif
return p;
P-Tier + Tiny Route Policy: Aggressive Superslab Management + Safe Routing ## Phase 1: Utilization-Aware Superslab Tiering (案B実装済) - Add ss_tier_box.h: Classify SuperSlabs into HOT/DRAINING/FREE based on utilization - HOT (>25%): Accept new allocations - DRAINING (≤25%): Drain only, no new allocs - FREE (0%): Ready for eager munmap - Enhanced shared_pool_release_slab(): - Check tier transition after each slab release - If tier→FREE: Force remaining slots to EMPTY and call superslab_free() immediately - Bypasses LRU cache to prevent registry bloat from accumulating DRAINING SuperSlabs - Test results (bench_random_mixed_hakmem): - 1M iterations: ✅ ~1.03M ops/s (previously passed) - 10M iterations: ✅ ~1.15M ops/s (previously: registry full error) - 50M iterations: ✅ ~1.08M ops/s (stress test) ## Phase 2: Tiny Front Routing Policy (新規Box) - Add tiny_route_box.h/c: Single 8-byte table for class→routing decisions - ROUTE_TINY_ONLY: Tiny front exclusive (no fallback) - ROUTE_TINY_FIRST: Try Tiny, fallback to Pool if fails - ROUTE_POOL_ONLY: Skip Tiny entirely - Profiles via HAKMEM_TINY_PROFILE ENV: - "hot": C0-C3=TINY_ONLY, C4-C6=TINY_FIRST, C7=POOL_ONLY - "conservative" (default): All TINY_FIRST - "off": All POOL_ONLY (disable Tiny) - "full": All TINY_ONLY (microbench mode) - A/B test results (ws=256, 100k ops random_mixed): - Default (conservative): ~2.90M ops/s - hot: ~2.65M ops/s (more conservative) - off: ~2.86M ops/s - full: ~2.98M ops/s (slightly best) ## Design Rationale ### Registry Pressure Fix (案B) - Problem: DRAINING tier SS occupied registry indefinitely - Solution: When total_active_blocks→0, immediately free to clear registry slot - Result: No more "registry full" errors under stress ### Routing Policy Box (新) - Problem: Tiny front optimization scattered across ENV/branches - Solution: Centralize routing in single table, select profiles via ENV - Benefit: Safe A/B testing without touching hot path code - Future: Integrate with RSS budget/learning layers for dynamic profile switching ## Next Steps (性能最適化) - Profile Tiny front internals (TLS SLL, FastCache, Superslab backend latency) - Identify bottleneck between current ~2.9M ops/s and mimalloc ~100M ops/s - Consider: - Reduce shared pool lock contention - Optimize unified cache hit rate - Streamline Superslab carving logic 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 18:01:25 +09:00
}
#endif // TINY_ROUTE_BOX_H