Files
hakmem/core/smallobject_segment_mid_v3.c

307 lines
10 KiB
C
Raw Normal View History

Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
// smallobject_segment_mid_v3.c
// Phase v11a-2: Multi-class MID v3.5 segment implementation (L2 physical layer)
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>
#include <stdio.h>
#include "box/smallobject_segment_mid_v3_box.h"
#include "box/region_id_v6_box.h"
#include "box/smallobject_mid_v35_geom_box.h" // Phase MID-V35-HOTPATH-OPT-1: geometry SSOT
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
// SmallPageMeta is now defined in smallobject_segment_mid_v3_box.h
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
// ============================================================================
// Geometry Constants
// ============================================================================
#define SMALL_MID_SEGMENT_SIZE (2 * 1024 * 1024) // 2 MiB
#define SMALL_MID_PAGE_SIZE (64 * 1024) // 64 KiB
#define SMALL_MID_PAGES_PER_SEG (SMALL_MID_SEGMENT_SIZE / SMALL_MID_PAGE_SIZE) // 32
// ============================================================================
// Helper: class_idx to slots mapping
// ============================================================================
// Phase MID-V35-HOTPATH-OPT-1: Use geom_box as Single Source of Truth
// See: core/box/smallobject_mid_v35_geom_box.h
// (Removed local class_idx_to_slots() which had wrong C6 value: 102 instead of 128)
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
// ============================================================================
// Segment Lifecycle
// ============================================================================
// Helper: mmap-based internal allocation to avoid recursion
// (malloc() would route back through hakmem → MID v3.5 → infinite loop)
static inline void* internal_mmap_alloc(size_t size) {
void* p = mmap(NULL, size, PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
return (p == MAP_FAILED) ? NULL : p;
}
static inline void internal_mmap_free(void* p, size_t size) {
if (p) munmap(p, size);
}
// Size constants for internal allocations
#define SEG_STRUCT_SIZE (sizeof(SmallSegment_MID_v3))
#define PAGES_ARRAY_SIZE (sizeof(SmallPageMeta*) * SMALL_MID_PAGES_PER_SEG)
#define PAGE_META_SIZE (sizeof(SmallPageMeta))
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
SmallSegment_MID_v3* small_segment_mid_v3_create(void) {
// 1. Allocate SmallSegment_MID_v3 structure via mmap (avoid malloc recursion)
SmallSegment_MID_v3 *seg = internal_mmap_alloc(SEG_STRUCT_SIZE);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (!seg) return NULL;
// 2. mmap 2MiB contiguous memory
seg->start = mmap(NULL, SMALL_MID_SEGMENT_SIZE,
PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
if (seg->start == MAP_FAILED) {
internal_mmap_free(seg, SEG_STRUCT_SIZE);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
return NULL;
}
seg->total_size = SMALL_MID_SEGMENT_SIZE;
seg->page_size = SMALL_MID_PAGE_SIZE;
seg->num_pages = SMALL_MID_PAGES_PER_SEG;
// 3. Initialize per-class free page stacks
for (int i = 0; i < 8; i++) {
seg->free_pages[i] = NULL;
seg->free_count[i] = 0;
seg->current_page[i] = NULL;
seg->page_offset[i] = 0;
}
// 4. Allocate pages array via mmap
seg->pages = internal_mmap_alloc(PAGES_ARRAY_SIZE);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (!seg->pages) {
munmap(seg->start, SMALL_MID_SEGMENT_SIZE);
internal_mmap_free(seg, SEG_STRUCT_SIZE);
return NULL;
}
// 5. Allocate page metadata as single block via mmap
// Layout: 32 SmallPageMeta structs contiguously
SmallPageMeta *meta_block = internal_mmap_alloc(PAGE_META_SIZE * SMALL_MID_PAGES_PER_SEG);
if (!meta_block) {
internal_mmap_free(seg->pages, PAGES_ARRAY_SIZE);
munmap(seg->start, SMALL_MID_SEGMENT_SIZE);
internal_mmap_free(seg, SEG_STRUCT_SIZE);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
return NULL;
}
// Initialize individual page metadata structures
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
for (uint32_t i = 0; i < SMALL_MID_PAGES_PER_SEG; i++) {
SmallPageMeta *meta = &meta_block[i];
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
meta->ptr = (char*)seg->start + (i * SMALL_MID_PAGE_SIZE);
meta->capacity = 0;
meta->class_idx = 0xFF; // Unassigned
meta->alloc_count = 0;
meta->free_count = 0;
meta->segment = seg;
meta->next = NULL;
seg->pages[i] = meta;
}
// 6. Register with RegionIdBox
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
seg->region_id = region_id_register_v6(seg->start, seg->total_size,
REGION_KIND_MID_V3, seg);
seg->total_allocations = 0;
seg->total_frees = 0;
seg->last_refill_count = 0;
return seg;
}
void small_segment_mid_v3_destroy(SmallSegment_MID_v3 *seg) {
if (!seg) return;
// Unregister from RegionIdBox
region_id_unregister_v6(seg->region_id);
// Free page metadata (allocated as single block)
if (seg->pages && seg->pages[0]) {
// All metadata is in a single contiguous block starting at pages[0]
internal_mmap_free(seg->pages[0], PAGE_META_SIZE * SMALL_MID_PAGES_PER_SEG);
}
// Free pages array
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (seg->pages) {
internal_mmap_free(seg->pages, PAGES_ARRAY_SIZE);
}
// Free per-class free page stacks (if any were allocated)
for (int i = 0; i < 8; i++) {
if (seg->free_pages[i]) {
internal_mmap_free(seg->free_pages[i], sizeof(void*) * SMALL_MID_PAGES_PER_SEG);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
}
}
// Unmap segment memory
if (seg->start && seg->start != MAP_FAILED) {
munmap(seg->start, seg->total_size);
}
internal_mmap_free(seg, SEG_STRUCT_SIZE);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
}
// ============================================================================
// Page Management
// ============================================================================
// Take a page from the free stack (LIFO)
// If class-specific stack is empty, allocate from unassigned pages
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
void* small_segment_mid_v3_take_page(SmallSegment_MID_v3 *seg, uint32_t class_idx) {
if (!seg || class_idx >= 8) return NULL;
// First: try class-specific free stack
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (seg->free_count[class_idx] > 0) {
void **stack = seg->free_pages[class_idx];
if (stack && seg->free_count[class_idx] > 0) {
void *page = stack[seg->free_count[class_idx] - 1];
seg->free_count[class_idx]--;
return page;
}
}
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
// Second: find an unassigned page from the pages array
for (uint32_t i = 0; i < seg->num_pages; i++) {
SmallPageMeta *meta = seg->pages[i];
if (meta && meta->class_idx == 0xFF) {
// Found unassigned page - assign to this class
meta->class_idx = class_idx;
return meta->ptr;
}
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
}
// No free pages available
return NULL;
}
// Release a page back to the free stack (LIFO)
void small_segment_mid_v3_release_page(SmallSegment_MID_v3 *seg, void *page, uint32_t class_idx) {
if (!seg || !page || class_idx >= 8) return;
// Allocate stack if needed (lazy initialization) - use mmap to avoid malloc recursion
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (!seg->free_pages[class_idx]) {
seg->free_pages[class_idx] = internal_mmap_alloc(sizeof(void*) * SMALL_MID_PAGES_PER_SEG);
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
if (!seg->free_pages[class_idx]) return;
}
// Push to free stack
seg->free_pages[class_idx][seg->free_count[class_idx]] = page;
seg->free_count[class_idx]++;
}
// Get region ID
uint32_t small_segment_mid_v3_region_id(SmallSegment_MID_v3 *seg) {
return seg ? seg->region_id : 0;
}
// Set region ID (called by RegionIdBox)
void small_segment_mid_v3_set_region_id(SmallSegment_MID_v3 *seg, uint32_t region_id) {
if (seg) {
seg->region_id = region_id;
}
}
// Check if page is within segment bounds
bool small_segment_mid_v3_contains_page(SmallSegment_MID_v3 *seg, void *page) {
if (!seg || !page) return false;
uintptr_t page_addr = (uintptr_t)page;
uintptr_t seg_start = (uintptr_t)seg->start;
uintptr_t seg_end = seg_start + seg->total_size;
return (page_addr >= seg_start && page_addr < seg_end);
}
// Get page metadata for pointer
SmallPageMeta_MID_v3* small_segment_mid_v3_get_page_meta(
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
SmallSegment_MID_v3 *seg,
void *page
) {
if (!seg || !page || !seg->pages) return NULL;
// Calculate page index
uintptr_t page_addr = (uintptr_t)page;
uintptr_t seg_start = (uintptr_t)seg->start;
if (page_addr < seg_start) return NULL;
uintptr_t offset = page_addr - seg_start;
uint32_t page_idx = offset / SMALL_MID_PAGE_SIZE;
if (page_idx >= SMALL_MID_PAGES_PER_SEG) return NULL;
return seg->pages[page_idx];
Phase v11a-2: Core MID v3.5 implementation - segment, cold iface, stats, learner Implement 5-layer infrastructure for multi-class MID v3.5 (C5-C7, 257-1KiB): 1. SegmentBox_mid_v3 (L2 Physical) - core/smallobject_segment_mid_v3.c (9.5 KB) - 2MiB segments, 64KiB pages (32 per segment) - Per-class free page stacks (LIFO) - RegionIdBox registration - Slots: C5→170, C6→102, C7→64 2. ColdIface_mid_v3 (L2→L1) - core/box/smallobject_cold_iface_mid_v3_box.h (NEW) - core/smallobject_cold_iface_mid_v3.c (3.5 KB) - refill: get page from free stack or new segment - retire: calculate free_hit_ratio, publish stats, return to stack - Clean separation: TLS cache for hot path, ColdIface for cold path 3. StatsBox_mid_v3 (L2→L3) - core/smallobject_stats_mid_v3.c (7.2 KB) - Circular buffer history (1000 events) - Per-page metrics: class_idx, allocs, frees, free_hit_ratio_bps - Periodic aggregation (every 100 retires) - Learner notification callback 4. Learner v2 (L3) - core/smallobject_learner_v2.c (11 KB) - Multi-class aggregation: allocs[8], retire_count[8], avg_free_hit_bps[8] - Exponential smoothing (90% history + 10% new) - Per-class efficiency tracking - Stats snapshot API - Route decision disabled for v11a-2 (v11b feature) 5. Build Integration - Modified Makefile: added 4 new .o files (segment, cold_iface, stats, learner) - Updated box header prototypes - Clean compilation, all dependencies resolved Architecture Decision Implementation: - v7 remains frozen (C5/C6 research preset) - MID v3.5 becomes unified 257-1KiB main path - Multi-class isolation: per-class free stacks - Dormant infrastructure: linked but not active (zero overhead) Performance: - Build: clean compilation - Sanity benchmark: 27.3M ops/s (no regression vs v10) - Memory: ~30MB RSS (baseline maintained) Design Compliance: ✅ Layer separation: L2 (segment) → L2 (cold iface) → L3 (stats) → L3 (learner) ✅ Hot path clean: alloc/free never touch stats/learner ✅ Backward compatible: existing MID v3 routes unchanged ✅ Transparent: v11a-2 is dormant (no behavior change) Next Phase (v11a-3): - Activate C5/C6/C7 routing through MID v3.5 - Connect TLS cache to segment refill - Verify performance under load - Then Phase v11a-4: dynamic C5 ratio routing 🤖 Generated with Claude Code Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-12 06:37:06 +09:00
}
// ============================================================================
// Statistics
// ============================================================================
SmallSegmentStatsMID_v3 small_segment_mid_v3_get_stats(SmallSegment_MID_v3 *seg) {
SmallSegmentStatsMID_v3 stats = {0};
if (!seg) return stats;
stats.total_allocations = seg->total_allocations;
stats.total_frees = seg->total_frees;
// Count active pages per class
for (uint32_t i = 0; i < SMALL_MID_PAGES_PER_SEG; i++) {
SmallPageMeta *meta = seg->pages[i];
if (meta && meta->class_idx < 8 && meta->class_idx != 0xFF) {
stats.active_pages[meta->class_idx]++;
}
}
return stats;
}
void small_segment_mid_v3_reset_stats(SmallSegment_MID_v3 *seg) {
if (!seg) return;
seg->total_allocations = 0;
seg->total_frees = 0;
seg->last_refill_count = 0;
}
// ============================================================================
// Validation & Debugging
// ============================================================================
bool small_segment_mid_v3_validate(SmallSegment_MID_v3 *seg) {
if (!seg) return false;
if (!seg->start || seg->start == MAP_FAILED) return false;
if (seg->total_size != SMALL_MID_SEGMENT_SIZE) return false;
if (seg->num_pages != SMALL_MID_PAGES_PER_SEG) return false;
if (!seg->pages) return false;
return true;
}
void small_segment_mid_v3_debug_print(SmallSegment_MID_v3 *seg) {
if (!seg) {
fprintf(stderr, "[MID_v3] Segment: NULL\n");
return;
}
fprintf(stderr, "[MID_v3] Segment %p:\n", (void*)seg);
fprintf(stderr, " start=%p size=%zu pages=%u region_id=%u\n",
seg->start, seg->total_size, seg->num_pages, seg->region_id);
fprintf(stderr, " allocs=%lu frees=%lu\n",
seg->total_allocations, seg->total_frees);
for (int i = 0; i < 8; i++) {
if (seg->free_count[i] > 0) {
fprintf(stderr, " C%d: free_pages=%u\n", i, seg->free_count[i]);
}
}
}