Files
hakmem/core/front/tiny_ring_cache.c
Moe Charm (CI) 03ba62df4d Phase 23 Unified Cache + PageFaultTelemetry generalization: Mid/VM page-fault bottleneck identified
Summary:
- Phase 23 Unified Cache: +30% improvement (Random Mixed 256B: 18.18M → 23.68M ops/s)
- PageFaultTelemetry: Extended to generic buckets (C0-C7, MID, L25, SSM)
- Measurement-driven decision: Mid/VM page-faults (80-100K) >> Tiny (6K) → prioritize Mid/VM optimization

Phase 23 Changes:
1. Unified Cache implementation (core/front/tiny_unified_cache.{c,h})
   - Direct SuperSlab carve (TLS SLL bypass)
   - Self-contained pop-or-refill pattern
   - ENV: HAKMEM_TINY_UNIFIED_CACHE=1, HAKMEM_TINY_UNIFIED_C{0-7}=128

2. Fast path pruning (tiny_alloc_fast.inc.h, tiny_free_fast_v2.inc.h)
   - Unified ON → direct cache access (skip all intermediate layers)
   - Alloc: unified_cache_pop_or_refill() → immediate fail to slow
   - Free: unified_cache_push() → fallback to SLL only if full

PageFaultTelemetry Changes:
3. Generic bucket architecture (core/box/pagefault_telemetry_box.{c,h})
   - PF_BUCKET_{C0-C7, MID, L25, SSM} for domain-specific measurement
   - Integration: hak_pool_try_alloc(), l25_alloc_new_run(), shared_pool_allocate_superslab_unlocked()

4. Measurement results (Random Mixed 500K / 256B):
   - Tiny C2-C7: 2-33 pages, high reuse (64-3.8 touches/page)
   - SSM: 512 pages (initialization footprint)
   - MID/L25: 0 (unused in this workload)
   - Mid/Large VM benchmarks: 80-100K page-faults (13-16x higher than Tiny)

Ring Cache Enhancements:
5. Hot Ring Cache (core/front/tiny_ring_cache.{c,h})
   - ENV: HAKMEM_TINY_HOT_RING_ENABLE=1, HAKMEM_TINY_HOT_RING_C{0-7}=size
   - Conditional compilation cleanup

Documentation:
6. Analysis reports
   - RANDOM_MIXED_BOTTLENECK_ANALYSIS.md: Page-fault breakdown
   - RANDOM_MIXED_SUMMARY.md: Phase 23 summary
   - RING_CACHE_ACTIVATION_GUIDE.md: Ring cache usage
   - CURRENT_TASK.md: Updated with Phase 23 results and Phase 24 plan

Next Steps (Phase 24):
- Target: Mid/VM PageArena/HotSpanBox (page-fault reduction 80-100K → 30-40K)
- Tiny SSM optimization deferred (low ROI, ~6K page-faults already optimal)
- Expected improvement: +30-50% for Mid/Large workloads

Generated with Claude Code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-17 02:47:58 +09:00

213 lines
7.5 KiB
C

// tiny_ring_cache.c - Phase 21-1: Ring cache implementation
#include "tiny_ring_cache.h"
#include "../box/tls_sll_box.h" // For tls_sll_pop/push (Phase 21-1-C refill)
#include <stdlib.h>
#include <string.h>
// ============================================================================
// TLS Variables (defined here, extern in header)
// ============================================================================
__thread TinyRingCache g_ring_cache_c2 = {NULL, 0, 0, 0, 0};
__thread TinyRingCache g_ring_cache_c3 = {NULL, 0, 0, 0, 0};
__thread TinyRingCache g_ring_cache_c5 = {NULL, 0, 0, 0, 0};
// ============================================================================
// Metrics (Phase 21-1-E, optional for Phase 21-1-C)
// ============================================================================
#if !HAKMEM_BUILD_RELEASE
__thread uint64_t g_ring_cache_hit[8] = {0};
__thread uint64_t g_ring_cache_miss[8] = {0};
__thread uint64_t g_ring_cache_push[8] = {0};
__thread uint64_t g_ring_cache_full[8] = {0};
__thread uint64_t g_ring_cache_refill[8] = {0};
#endif
// ============================================================================
// Init (called at thread start, from hakmem_tiny.c)
// ============================================================================
void ring_cache_init(void) {
if (!ring_cache_enabled()) return;
// C2 init
size_t cap_c2 = ring_capacity_c2();
g_ring_cache_c2.slots = (void**)calloc(cap_c2, sizeof(void*));
if (!g_ring_cache_c2.slots) {
#if !HAKMEM_BUILD_RELEASE
fprintf(stderr, "[Ring-INIT] Failed to allocate C2 ring (%zu slots)\n", cap_c2);
fflush(stderr);
#endif
return;
}
g_ring_cache_c2.capacity = (uint16_t)cap_c2;
g_ring_cache_c2.mask = (uint16_t)(cap_c2 - 1);
g_ring_cache_c2.head = 0;
g_ring_cache_c2.tail = 0;
// C3 init
size_t cap_c3 = ring_capacity_c3();
g_ring_cache_c3.slots = (void**)calloc(cap_c3, sizeof(void*));
if (!g_ring_cache_c3.slots) {
#if !HAKMEM_BUILD_RELEASE
fprintf(stderr, "[Ring-INIT] Failed to allocate C3 ring (%zu slots)\n", cap_c3);
fflush(stderr);
#endif
// Free C2 if C3 failed
free(g_ring_cache_c2.slots);
g_ring_cache_c2.slots = NULL;
return;
}
g_ring_cache_c3.capacity = (uint16_t)cap_c3;
g_ring_cache_c3.mask = (uint16_t)(cap_c3 - 1);
g_ring_cache_c3.head = 0;
g_ring_cache_c3.tail = 0;
// C5 init
size_t cap_c5 = ring_capacity_c5();
g_ring_cache_c5.slots = (void**)calloc(cap_c5, sizeof(void*));
if (!g_ring_cache_c5.slots) {
#if !HAKMEM_BUILD_RELEASE
fprintf(stderr, "[Ring-INIT] Failed to allocate C5 ring (%zu slots)\n", cap_c5);
fflush(stderr);
#endif
// Free C2 and C3 if C5 failed
free(g_ring_cache_c2.slots);
g_ring_cache_c2.slots = NULL;
free(g_ring_cache_c3.slots);
g_ring_cache_c3.slots = NULL;
return;
}
g_ring_cache_c5.capacity = (uint16_t)cap_c5;
g_ring_cache_c5.mask = (uint16_t)(cap_c5 - 1);
g_ring_cache_c5.head = 0;
g_ring_cache_c5.tail = 0;
#if !HAKMEM_BUILD_RELEASE
fprintf(stderr, "[Ring-INIT] C2=%zu slots (%zu bytes), C3=%zu slots (%zu bytes), C5=%zu slots (%zu bytes)\n",
cap_c2, cap_c2 * sizeof(void*),
cap_c3, cap_c3 * sizeof(void*),
cap_c5, cap_c5 * sizeof(void*));
fflush(stderr);
#endif
}
// ============================================================================
// Shutdown (called at thread exit, optional)
// ============================================================================
void ring_cache_shutdown(void) {
if (!ring_cache_enabled()) return;
// Drain rings to TLS SLL before shutdown (prevent leak)
// TODO: Implement drain logic in Phase 21-1-C
// Free ring buffers
if (g_ring_cache_c2.slots) {
free(g_ring_cache_c2.slots);
g_ring_cache_c2.slots = NULL;
}
if (g_ring_cache_c3.slots) {
free(g_ring_cache_c3.slots);
g_ring_cache_c3.slots = NULL;
}
if (g_ring_cache_c5.slots) {
free(g_ring_cache_c5.slots);
g_ring_cache_c5.slots = NULL;
}
#if !HAKMEM_BUILD_RELEASE
fprintf(stderr, "[Ring-SHUTDOWN] C2/C3/C5 rings freed\n");
fflush(stderr);
#endif
}
// ============================================================================
// Refill from TLS SLL (cascade, Phase 21-1-C)
// ============================================================================
// Refill ring from TLS SLL (one-way cascade: SLL → Ring)
// Returns: number of blocks transferred
int ring_refill_from_sll(int class_idx, int target_count) {
if (!ring_cascade_enabled()) return 0;
if (class_idx != 2 && class_idx != 3) return 0;
int transferred = 0;
while (transferred < target_count) {
void* ptr = NULL;
// Pop from TLS SLL
if (!tls_sll_pop(class_idx, &ptr)) {
break; // SLL empty
}
// Push to Ring
if (!ring_cache_push(class_idx, ptr)) {
// Ring full, push back to SLL
tls_sll_push(class_idx, ptr, (uint32_t)-1); // Unlimited capacity
break;
}
transferred++;
}
#if !HAKMEM_BUILD_RELEASE
if (transferred > 0) {
g_ring_cache_refill[class_idx]++; // Count refill operations
fprintf(stderr, "[Ring-REFILL] C%d: %d blocks transferred from SLL to Ring\n",
class_idx, transferred);
fflush(stderr);
}
#endif
return transferred;
}
// ============================================================================
// Stats (Phase 21-1-C/E metrics)
// ============================================================================
void ring_cache_print_stats(void) {
if (!ring_cache_enabled()) return;
#if !HAKMEM_BUILD_RELEASE
// Current occupancy
uint16_t c2_count = (g_ring_cache_c2.tail >= g_ring_cache_c2.head)
? (g_ring_cache_c2.tail - g_ring_cache_c2.head)
: (g_ring_cache_c2.capacity - g_ring_cache_c2.head + g_ring_cache_c2.tail);
uint16_t c3_count = (g_ring_cache_c3.tail >= g_ring_cache_c3.head)
? (g_ring_cache_c3.tail - g_ring_cache_c3.head)
: (g_ring_cache_c3.capacity - g_ring_cache_c3.head + g_ring_cache_c3.tail);
fprintf(stderr, "\n[Ring-STATS] Ring Cache Metrics (C2/C3):\n");
fprintf(stderr, " C2: %u/%u slots occupied\n", c2_count, g_ring_cache_c2.capacity);
fprintf(stderr, " C3: %u/%u slots occupied\n", c3_count, g_ring_cache_c3.capacity);
// Metrics summary (C2/C3 only)
for (int c = 2; c <= 3; c++) {
uint64_t total_allocs = g_ring_cache_hit[c] + g_ring_cache_miss[c];
uint64_t total_frees = g_ring_cache_push[c] + g_ring_cache_full[c];
double hit_rate = (total_allocs > 0) ? (100.0 * g_ring_cache_hit[c] / total_allocs) : 0.0;
double full_rate = (total_frees > 0) ? (100.0 * g_ring_cache_full[c] / total_frees) : 0.0;
if (total_allocs > 0 || total_frees > 0) {
fprintf(stderr, " C%d: hit=%llu miss=%llu (%.1f%% hit), push=%llu full=%llu (%.1f%% full), refill=%llu\n",
c,
(unsigned long long)g_ring_cache_hit[c],
(unsigned long long)g_ring_cache_miss[c],
hit_rate,
(unsigned long long)g_ring_cache_push[c],
(unsigned long long)g_ring_cache_full[c],
full_rate,
(unsigned long long)g_ring_cache_refill[c]);
}
}
fflush(stderr);
#endif
}