9b0d746407
Merge separate g_tls_sll_head[] and g_tls_sll_count[] arrays into unified TinyTLSSLL struct to improve L1D cache locality. Expected performance gain: +12-18% from reducing cache line splits (2 loads → 1 load per operation). Changes: - core/hakmem_tiny.h: Add TinyTLSSLL type (16B aligned, head+count+pad) - core/hakmem_tiny.c: Replace separate arrays with g_tls_sll[8] - core/box/tls_sll_box.h: Update Box API (13 sites) for unified access - Updated 32+ files: All g_tls_sll_head[i] → g_tls_sll[i].head - Updated 32+ files: All g_tls_sll_count[i] → g_tls_sll[i].count - core/hakmem_tiny_integrity.h: Unified canary guards - core/box/integrity_box.c: Simplified canary validation - Makefile: Added core/box/tiny_sizeclass_hist_box.o to link Build: ✅ PASS (10K ops sanity test) Warnings: Only pre-existing LTO type mismatches (unrelated) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
90 lines
2.8 KiB
C
90 lines
2.8 KiB
C
// prewarm_box.c - Box Prewarm Implementation
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include "../hakmem_tiny.h" // MUST BE FIRST: Base types
|
|
#include "../tiny_tls.h" // TinyTLSSlab type definition
|
|
#include "../hakmem_tiny_config.h" // TINY_NUM_CLASSES
|
|
#include "../hakmem_tiny_superslab.h" // SuperSlab
|
|
#include "../hakmem_tiny_integrity.h" // HAK_CHECK_CLASS_IDX
|
|
#include "prewarm_box.h"
|
|
#include "capacity_box.h" // box_cap_init(), box_cap_avail()
|
|
#include "carve_push_box.h" // box_carve_and_push()
|
|
|
|
// External declarations
|
|
extern __thread TinyTLSSlab g_tls_slabs[TINY_NUM_CLASSES];
|
|
extern __thread TinyTLSSLL g_tls_sll[TINY_NUM_CLASSES];
|
|
extern SuperSlab* superslab_refill(int class_idx);
|
|
|
|
// ============================================================================
|
|
// Box Prewarm API Implementation
|
|
// ============================================================================
|
|
|
|
int box_prewarm_tls(int class_idx, int count) {
|
|
// PRIORITY 1: Bounds check
|
|
HAK_CHECK_CLASS_IDX(class_idx, "box_prewarm_tls");
|
|
|
|
if (count <= 0) return 0;
|
|
|
|
// Step 1: Ensure capacity system is initialized
|
|
// This is critical to prevent the double-free bug
|
|
box_cap_init();
|
|
|
|
// Step 2: Check available capacity
|
|
uint32_t avail = box_cap_avail(class_idx);
|
|
if (avail == 0) {
|
|
// TLS SLL already at capacity
|
|
return 0;
|
|
}
|
|
|
|
// Limit count to available capacity
|
|
uint32_t want = (uint32_t)count;
|
|
if (want > avail) {
|
|
want = avail;
|
|
}
|
|
|
|
// Step 3: Ensure SuperSlab is available
|
|
TinyTLSSlab* tls = &g_tls_slabs[class_idx];
|
|
if (!tls->ss) {
|
|
// Try to allocate SuperSlab
|
|
if (superslab_refill(class_idx) == NULL) {
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
fprintf(stderr, "[BOX_PREWARM] Failed to allocate SuperSlab for class %d\n",
|
|
class_idx);
|
|
#endif
|
|
return 0;
|
|
}
|
|
// Reload tls pointer after superslab_refill
|
|
tls = &g_tls_slabs[class_idx];
|
|
}
|
|
|
|
// Step 4: Atomically carve and push blocks
|
|
// This uses Box Carve-Push which guarantees no orphaned blocks
|
|
uint32_t pushed = box_carve_and_push(class_idx, want);
|
|
|
|
#if !HAKMEM_BUILD_RELEASE
|
|
if (pushed < want) {
|
|
fprintf(stderr, "[BOX_PREWARM] Partial prewarm: requested=%u pushed=%u class=%d\n",
|
|
want, pushed, class_idx);
|
|
}
|
|
#endif
|
|
|
|
return (int)pushed;
|
|
}
|
|
|
|
int box_prewarm_needed(int class_idx, int target_count) {
|
|
// PRIORITY 1: Bounds check
|
|
HAK_CHECK_CLASS_IDX(class_idx, "box_prewarm_needed");
|
|
|
|
if (target_count <= 0) return 0;
|
|
|
|
// Check current count
|
|
uint32_t current = g_tls_sll[class_idx].count;
|
|
if (current >= (uint32_t)target_count) {
|
|
// Already at or above target
|
|
return 0;
|
|
}
|
|
|
|
// Return how many more blocks needed
|
|
return (target_count - (int)current);
|
|
}
|