Phase v6-1: C6-only route stub (v1/pool fallback) Phase v6-2: Segment v6 + ColdIface v6 + Core v6 HotPath implementation - 2MiB segment / 64KiB page allocation - O(1) ptr→page_meta lookup with segment masking - C6-heavy A/B: SEGV-free but -44% performance (15.3M ops/s) Phase v6-3: Thin-layer optimization (TLS ownership check + batch header + refill batching) - TLS ownership fast-path skip page_meta for 90%+ of frees - Batch header writes during refill (32 allocs = 1 header write) - TLS batch refill (1/32 refill frequency) - C6-heavy A/B: v6-2 15.3M → v6-3 27.1M ops/s (±0% vs baseline) ✅ Phase v6-4: Mixed hang fix (segment metadata lookup correction) - Root cause: metadata lookup was reading mmap region instead of TLS slot - Fix: use TLS slot descriptor with in_use validation - Mixed health: 5M iterations SEGV-free, 35.8M ops/s ✅ Phase v6-refactor: Code quality improvements (macro unification + inline + docs) - Add SMALL_V6_* prefix macros (header, pointer conversion, page index) - Extract inline validation functions (small_page_v6_valid, small_ptr_in_segment_v6) - Doxygen-style comments for all public functions - Result: 0 compiler warnings, maintained +1.2% performance Files: - core/box/smallobject_core_v6_box.h (new, type & API definitions) - core/box/smallobject_cold_iface_v6.h (new, cold iface API) - core/box/smallsegment_v6_box.h (new, segment type definitions) - core/smallobject_core_v6.c (new, C6 alloc/free implementation) - core/smallobject_cold_iface_v6.c (new, refill/retire logic) - core/smallsegment_v6.c (new, segment allocator) - docs/analysis/SMALLOBJECT_CORE_V6_DESIGN.md (new, design document) - core/box/tiny_route_env_box.h (modified, v6 route added) - core/front/malloc_tiny_fast.h (modified, v6 case in route switch) - Makefile (modified, v6 objects added) - CURRENT_TASK.md (modified, v6 status added) Status: - C6-heavy: v6 OFF 27.1M → v6-3 ON 27.1M ops/s (±0%) ✅ - Mixed: v6 ON 35.8M ops/s (C6-only, other classes via v1) ✅ - Build: 0 warnings, fully documented ✅ 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
161 lines
4.7 KiB
C
161 lines
4.7 KiB
C
// smallsegment_v6.c - SmallSegment v6 実装(Phase v6-2)
|
||
|
||
#include <stdlib.h>
|
||
#include <string.h>
|
||
#include <sys/mman.h>
|
||
#include <stdint.h>
|
||
#include <unistd.h>
|
||
#include "box/smallsegment_v6_box.h"
|
||
|
||
#ifndef likely
|
||
#define likely(x) __builtin_expect(!!(x), 1)
|
||
#define unlikely(x) __builtin_expect(!!(x), 0)
|
||
#endif
|
||
|
||
// TLS segment slot (metadata stored in TLS, not in mapped memory)
|
||
typedef struct {
|
||
SmallSegmentV6 seg;
|
||
int in_use;
|
||
void* mmap_base; // Actual mmap base (for munmap)
|
||
size_t mmap_size; // Actual mmap size (for munmap)
|
||
} TLSSegmentSlotV6;
|
||
|
||
static __thread TLSSegmentSlotV6 g_tls_segment_v6;
|
||
|
||
/// Acquire 2MiB aligned segment for current thread (called once per thread)
|
||
/// Allocates a 2MiB segment with proper alignment via mmap
|
||
/// @return: Segment pointer on success, NULL on failure
|
||
SmallSegmentV6* small_segment_v6_acquire_for_thread(void) {
|
||
TLSSegmentSlotV6* slot = &g_tls_segment_v6;
|
||
|
||
if (slot->in_use) {
|
||
return &slot->seg; // Already acquired
|
||
}
|
||
|
||
// Allocate 2MiB aligned segment
|
||
// Use mmap with MAP_ANONYMOUS which typically gives aligned addresses for large allocations
|
||
void* mem = mmap(NULL, SMALL_SEGMENT_V6_SIZE,
|
||
PROT_READ | PROT_WRITE,
|
||
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
|
||
|
||
if (mem == MAP_FAILED || mem == NULL) {
|
||
return NULL;
|
||
}
|
||
|
||
uintptr_t addr = (uintptr_t)mem;
|
||
void* mmap_base = mem;
|
||
size_t mmap_size = SMALL_SEGMENT_V6_SIZE;
|
||
|
||
// Check if we got 2MiB alignment
|
||
if ((addr & (SMALL_SEGMENT_V6_SIZE - 1)) != 0) {
|
||
// Not aligned, need to reallocate with overallocation
|
||
munmap(mem, SMALL_SEGMENT_V6_SIZE);
|
||
|
||
// Allocate 4MiB to ensure we can find a 2MiB aligned region
|
||
size_t alloc_size = SMALL_SEGMENT_V6_SIZE * 2;
|
||
mem = mmap(NULL, alloc_size,
|
||
PROT_READ | PROT_WRITE,
|
||
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
|
||
|
||
if (mem == MAP_FAILED || mem == NULL) {
|
||
return NULL;
|
||
}
|
||
|
||
// Find the aligned address within this region
|
||
uintptr_t raw_addr = (uintptr_t)mem;
|
||
addr = (raw_addr + SMALL_SEGMENT_V6_SIZE - 1) & ~(SMALL_SEGMENT_V6_SIZE - 1);
|
||
|
||
// Verify the aligned address is within our mapping
|
||
if (addr < raw_addr || addr + SMALL_SEGMENT_V6_SIZE > raw_addr + alloc_size) {
|
||
munmap(mem, alloc_size);
|
||
return NULL;
|
||
}
|
||
|
||
// Keep track of actual mmap base for munmap
|
||
mmap_base = mem;
|
||
mmap_size = alloc_size;
|
||
}
|
||
|
||
// Initialize segment structure in TLS (not in mapped memory)
|
||
SmallSegmentV6* seg = &slot->seg;
|
||
slot->in_use = 1;
|
||
slot->mmap_base = mmap_base;
|
||
slot->mmap_size = mmap_size;
|
||
|
||
seg->base = addr;
|
||
seg->num_pages = SMALL_PAGES_PER_SEGMENT;
|
||
seg->owner_tid = (uint32_t)getpid(); // Simple owner tracking
|
||
seg->magic = SMALL_SEGMENT_V6_MAGIC;
|
||
|
||
// Initialize all page metadata
|
||
for (uint32_t i = 0; i < seg->num_pages; i++) {
|
||
SmallPageMetaV6* m = &seg->page_meta[i];
|
||
m->free_list = NULL;
|
||
m->used = 0;
|
||
m->capacity = 0;
|
||
m->class_idx = 0;
|
||
m->flags = 0;
|
||
m->page_idx = (uint16_t)i;
|
||
m->segment = seg;
|
||
}
|
||
|
||
return seg;
|
||
}
|
||
|
||
/// Release segment and unmap memory (cleanup)
|
||
/// @param seg: Segment to release
|
||
void small_segment_v6_release(SmallSegmentV6* seg) {
|
||
if (!seg) return;
|
||
if (seg->magic != SMALL_SEGMENT_V6_MAGIC) return;
|
||
|
||
TLSSegmentSlotV6* slot = &g_tls_segment_v6;
|
||
if (seg != &slot->seg) return; // Not our segment
|
||
|
||
seg->magic = 0; // Invalidate
|
||
munmap(slot->mmap_base, slot->mmap_size);
|
||
|
||
slot->in_use = 0;
|
||
slot->mmap_base = NULL;
|
||
slot->mmap_size = 0;
|
||
}
|
||
|
||
/// O(1) ptr -> page_meta lookup using segment metadata in TLS
|
||
/// Maps any pointer within TLS segment to its page metadata
|
||
/// @param ptr: Pointer to lookup (USER or BASE pointer)
|
||
/// @return: Page metadata pointer if found and valid, NULL otherwise
|
||
SmallPageMetaV6* small_page_meta_v6_of(void* ptr) {
|
||
if (unlikely(!ptr)) {
|
||
return NULL;
|
||
}
|
||
|
||
TLSSegmentSlotV6* slot = &g_tls_segment_v6;
|
||
|
||
// Check if segment is initialized
|
||
if (unlikely(!slot->in_use)) {
|
||
return NULL;
|
||
}
|
||
|
||
SmallSegmentV6* seg = &slot->seg;
|
||
|
||
// Check if ptr is within our segment range
|
||
if (unlikely(!small_ptr_in_segment_v6(seg, ptr))) {
|
||
return NULL;
|
||
}
|
||
|
||
// Calculate page index using macro
|
||
uintptr_t addr = (uintptr_t)ptr;
|
||
size_t page_idx = SMALL_V6_PAGE_IDX(seg, addr);
|
||
if (unlikely(page_idx >= seg->num_pages)) {
|
||
return NULL;
|
||
}
|
||
|
||
SmallPageMetaV6* page = &seg->page_meta[page_idx];
|
||
|
||
// Validate that this page is actually in use
|
||
if (unlikely(!small_page_v6_valid(page))) {
|
||
return NULL;
|
||
}
|
||
|
||
return page;
|
||
}
|