52386401b3
Major Features: - Debug counter infrastructure for Refill Stage tracking - Free Pipeline counters (ss_local, ss_remote, tls_sll) - Diagnostic counters for early return analysis - Unified larson.sh benchmark runner with profiles - Phase 6-3 regression analysis documentation Bug Fixes: - Fix SuperSlab disabled by default (HAKMEM_TINY_USE_SUPERSLAB) - Fix profile variable naming consistency - Add .gitignore patterns for large files Performance: - Phase 6-3: 4.79 M ops/s (has OOM risk) - With SuperSlab: 3.13 M ops/s (+19% improvement) This is a clean repository without large log files. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
84 lines
2.8 KiB
C
84 lines
2.8 KiB
C
/* hakmem_ace_metrics.h - ACE Learning Layer Metrics Collection
|
||
*
|
||
* メトリクス収集システム:
|
||
* - Fast metrics (1Hz): throughput, LLC miss, mutex wait, backlog
|
||
* - Slow metrics (60s): fragmentation ratio, RSS
|
||
* - Inline helpers for hot-path tracking
|
||
*/
|
||
|
||
#ifndef HAKMEM_ACE_METRICS_H
|
||
#define HAKMEM_ACE_METRICS_H
|
||
|
||
#include <stdint.h>
|
||
#include <stdbool.h>
|
||
#include <stdatomic.h>
|
||
#include <time.h>
|
||
|
||
/* メトリクス構造体 */
|
||
struct hkm_ace_metrics {
|
||
/* Fast metrics (updated every 0.5-1s) */
|
||
uint64_t throughput_ops; /* Operations per second */
|
||
double llc_miss_rate; /* LLC miss rate (0.0-1.0) */
|
||
uint64_t mutex_wait_ns; /* Cumulative mutex contention time */
|
||
uint32_t remote_free_backlog[8]; /* Per-class remote free backlog */
|
||
|
||
/* Slow metrics (updated every 60s) */
|
||
double fragmentation_ratio; /* allocated / reserved bytes */
|
||
uint64_t rss_mb; /* Resident set size in MB */
|
||
|
||
/* Timestamp */
|
||
uint64_t timestamp_ms; /* Collection timestamp */
|
||
};
|
||
|
||
/* グローバルカウンタ（hot-path用） */
|
||
extern _Atomic uint64_t g_ace_alloc_count;
|
||
extern _Atomic uint64_t g_ace_free_count;
|
||
extern _Atomic uint64_t g_ace_mutex_wait_ns;
|
||
|
||
/* 初期化・破棄 */
|
||
void hkm_ace_metrics_init(void);
|
||
void hkm_ace_metrics_destroy(void);
|
||
|
||
/* メトリクス収集 */
|
||
void hkm_ace_metrics_collect(struct hkm_ace_metrics *out);
|
||
void hkm_ace_metrics_collect_fast(struct hkm_ace_metrics *out);
|
||
void hkm_ace_metrics_collect_slow(struct hkm_ace_metrics *out);
|
||
|
||
/* Hot-path inline helpers */
|
||
|
||
/* アロケーションカウント（高速インクリメント） */
|
||
static inline void hkm_ace_track_alloc(void) {
|
||
atomic_fetch_add_explicit(&g_ace_alloc_count, 1, memory_order_relaxed);
|
||
}
|
||
|
||
/* フリーカウント（高速インクリメント） */
|
||
static inline void hkm_ace_track_free(void) {
|
||
atomic_fetch_add_explicit(&g_ace_free_count, 1, memory_order_relaxed);
|
||
}
|
||
|
||
/* ミューテックス待機時間トラッキング
|
||
* 使い方:
|
||
* uint64_t t0 = hkm_ace_mutex_timer_start();
|
||
* pthread_mutex_lock(&lock);
|
||
* hkm_ace_mutex_timer_end(t0);
|
||
*/
|
||
static inline uint64_t hkm_ace_mutex_timer_start(void) {
|
||
struct timespec ts;
|
||
clock_gettime(CLOCK_MONOTONIC, &ts);
|
||
return (uint64_t)ts.tv_sec * 1000000000ULL + (uint64_t)ts.tv_nsec;
|
||
}
|
||
|
||
static inline void hkm_ace_mutex_timer_end(uint64_t start_ns) {
|
||
struct timespec ts;
|
||
clock_gettime(CLOCK_MONOTONIC, &ts);
|
||
uint64_t end_ns = (uint64_t)ts.tv_sec * 1000000000ULL + (uint64_t)ts.tv_nsec;
|
||
uint64_t delta = (end_ns > start_ns) ? (end_ns - start_ns) : 0;
|
||
atomic_fetch_add_explicit(&g_ace_mutex_wait_ns, delta, memory_order_relaxed);
|
||
}
|
||
|
||
/* LLC miss monitoring (rdpmc wrapper) */
|
||
bool hkm_ace_llc_available(void);
|
||
void hkm_ace_llc_read(uint64_t *misses, uint64_t *references);
|
||
|
||
#endif /* HAKMEM_ACE_METRICS_H */
|