Phase FREE-LEGACY-OPT-4-2/4-3: C6 ULTRA-free TLS cache + segment learning
Phase 4-2: - Add TinyC6UltraFreeTLS structure with 128-slot TLS freelist - Implement tiny_c6_ultra_free_fast/slow for C6 free hot path - Add c6_ultra_free_fast counter to FreePathStats - ENV gate: HAKMEM_TINY_C6_ULTRA_FREE_ENABLED (default: OFF) Phase 4-3: - Add segment learning on first C6 free via ss_fast_lookup() - Learn seg_base/seg_end from SuperSlab for range check - Increase cache capacity from 32 to 128 blocks Results: - Segment learning works: fast path captures blocks in segment - However, without alloc integration, cache fills up and overflows to legacy - Net effect: +1-3% (within noise range) - Drain strategy also tested: no benefit (equal overhead) Conclusion: - Free-only TLS cache is limited without alloc-side integration - Core v6 already has alloc/free integrated TLS (but -12% slower) - Keep as research box (ENV default OFF) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Moe Charm (CI)
8
Makefile
8
Makefile
@ -218,12 +218,12 @@ LDFLAGS += $(EXTRA_LDFLAGS)
|
||||
|
||||
# Targets
|
||||
TARGET = test_hakmem
|
||||
OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/free_publish_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o
|
||||
OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/free_publish_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/free_front_v3_env_box.o core/box/free_path_stats_box.o core/box/tiny_c6_ultra_free_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o core/smallobject_hotbox_v3.o core/smallobject_hotbox_v4.o core/smallobject_hotbox_v5.o core/smallsegment_v5.o core/smallobject_cold_iface_v5.o core/smallsegment_v6.o core/smallobject_cold_iface_v6.o core/smallobject_core_v6.o
|
||||
OBJS = $(OBJS_BASE)
|
||||
|
||||
# Shared library
|
||||
SHARED_LIB = libhakmem.so
|
||||
SHARED_OBJS = hakmem_shared.o hakmem_config_shared.o hakmem_tiny_config_shared.o hakmem_ucb1_shared.o hakmem_bigcache_shared.o hakmem_pool_shared.o hakmem_l25_pool_shared.o hakmem_site_rules_shared.o hakmem_tiny_shared.o core/box/ss_allocation_box_shared.o superslab_stats_shared.o superslab_cache_shared.o superslab_ace_shared.o superslab_slab_shared.o superslab_backend_shared.o core/superslab_head_stub_shared.o hakmem_smallmid_shared.o core/box/superslab_expansion_box_shared.o core/box/integrity_box_shared.o core/box/mailbox_box_shared.o core/box/front_gate_box_shared.o core/box/front_gate_classifier_shared.o core/box/free_publish_box_shared.o core/box/capacity_box_shared.o core/box/carve_push_box_shared.o core/box/prewarm_box_shared.o core/box/ss_hot_prewarm_box_shared.o core/box/front_metrics_box_shared.o core/box/bench_fast_box_shared.o core/box/ss_addr_map_box_shared.o core/box/slab_recycling_box_shared.o core/box/pagefault_telemetry_box_shared.o core/box/tiny_sizeclass_hist_box_shared.o core/box/tiny_env_box_shared.o core/box/tiny_route_box_shared.o core/box/tiny_page_box_shared.o core/box/tiny_class_policy_box_shared.o core/box/tiny_class_stats_box_shared.o core/box/tiny_policy_learner_box_shared.o core/box/ss_budget_box_shared.o core/box/tiny_mem_stats_box_shared.o core/box/wrapper_env_box_shared.o core/box/madvise_guard_box_shared.o core/box/libm_reloc_guard_box_shared.o core/page_arena_shared.o core/front/tiny_unified_cache_shared.o core/tiny_alloc_fast_push_shared.o core/tiny_c7_ultra_segment_shared.o core/tiny_c7_ultra_shared.o core/link_stubs_shared.o core/tiny_failfast_shared.o tiny_sticky_shared.o tiny_remote_shared.o tiny_publish_shared.o tiny_debug_ring_shared.o hakmem_tiny_magazine_shared.o hakmem_tiny_stats_shared.o hakmem_tiny_sfc_shared.o hakmem_tiny_query_shared.o hakmem_tiny_rss_shared.o hakmem_tiny_registry_shared.o hakmem_tiny_remote_target_shared.o hakmem_tiny_bg_spill_shared.o tiny_adaptive_sizing_shared.o hakmem_super_registry_shared.o hakmem_shared_pool_shared.o hakmem_shared_pool_acquire_shared.o hakmem_shared_pool_release_shared.o hakmem_elo_shared.o hakmem_batch_shared.o hakmem_p2_shared.o hakmem_sizeclass_dist_shared.o hakmem_evo_shared.o hakmem_debug_shared.o hakmem_sys_shared.o hakmem_whale_shared.o hakmem_policy_shared.o hakmem_ace_shared.o hakmem_ace_stats_shared.o hakmem_ace_controller_shared.o hakmem_ace_metrics_shared.o hakmem_ace_ucb1_shared.o hakmem_prof_shared.o hakmem_learner_shared.o hakmem_size_hist_shared.o hakmem_learn_log_shared.o hakmem_syscall_shared.o tiny_fastcache_shared.o core/box/super_reg_box_shared.o core/box/shared_pool_box_shared.o core/box/remote_side_box_shared.o core/tiny_destructors_shared.o
|
||||
SHARED_OBJS = hakmem_shared.o hakmem_config_shared.o hakmem_tiny_config_shared.o hakmem_ucb1_shared.o hakmem_bigcache_shared.o hakmem_pool_shared.o hakmem_l25_pool_shared.o hakmem_site_rules_shared.o hakmem_tiny_shared.o core/box/ss_allocation_box_shared.o superslab_stats_shared.o superslab_cache_shared.o superslab_ace_shared.o superslab_slab_shared.o superslab_backend_shared.o core/superslab_head_stub_shared.o hakmem_smallmid_shared.o core/box/superslab_expansion_box_shared.o core/box/integrity_box_shared.o core/box/mailbox_box_shared.o core/box/front_gate_box_shared.o core/box/front_gate_classifier_shared.o core/box/free_publish_box_shared.o core/box/capacity_box_shared.o core/box/carve_push_box_shared.o core/box/prewarm_box_shared.o core/box/ss_hot_prewarm_box_shared.o core/box/front_metrics_box_shared.o core/box/bench_fast_box_shared.o core/box/ss_addr_map_box_shared.o core/box/slab_recycling_box_shared.o core/box/pagefault_telemetry_box_shared.o core/box/tiny_sizeclass_hist_box_shared.o core/box/tiny_env_box_shared.o core/box/tiny_route_box_shared.o core/box/free_front_v3_env_box_shared.o core/box/free_path_stats_box_shared.o core/box/tiny_page_box_shared.o core/box/tiny_class_policy_box_shared.o core/box/tiny_class_stats_box_shared.o core/box/tiny_policy_learner_box_shared.o core/box/ss_budget_box_shared.o core/box/tiny_mem_stats_box_shared.o core/box/wrapper_env_box_shared.o core/box/madvise_guard_box_shared.o core/box/libm_reloc_guard_box_shared.o core/page_arena_shared.o core/front/tiny_unified_cache_shared.o core/tiny_alloc_fast_push_shared.o core/tiny_c7_ultra_segment_shared.o core/tiny_c7_ultra_shared.o core/link_stubs_shared.o core/tiny_failfast_shared.o tiny_sticky_shared.o tiny_remote_shared.o tiny_publish_shared.o tiny_debug_ring_shared.o hakmem_tiny_magazine_shared.o hakmem_tiny_stats_shared.o hakmem_tiny_sfc_shared.o hakmem_tiny_query_shared.o hakmem_tiny_rss_shared.o hakmem_tiny_registry_shared.o hakmem_tiny_remote_target_shared.o hakmem_tiny_bg_spill_shared.o tiny_adaptive_sizing_shared.o hakmem_super_registry_shared.o hakmem_shared_pool_shared.o hakmem_shared_pool_acquire_shared.o hakmem_shared_pool_release_shared.o hakmem_elo_shared.o hakmem_batch_shared.o hakmem_p2_shared.o hakmem_sizeclass_dist_shared.o hakmem_evo_shared.o hakmem_debug_shared.o hakmem_sys_shared.o hakmem_whale_shared.o hakmem_policy_shared.o hakmem_ace_shared.o hakmem_ace_stats_shared.o hakmem_ace_controller_shared.o hakmem_ace_metrics_shared.o hakmem_ace_ucb1_shared.o hakmem_prof_shared.o hakmem_learner_shared.o hakmem_size_hist_shared.o hakmem_learn_log_shared.o hakmem_syscall_shared.o tiny_fastcache_shared.o core/box/super_reg_box_shared.o core/box/shared_pool_box_shared.o core/box/remote_side_box_shared.o core/tiny_destructors_shared.o
|
||||
|
||||
# Pool TLS Phase 1 (enable with POOL_TLS_PHASE1=1)
|
||||
ifeq ($(POOL_TLS_PHASE1),1)
|
||||
@ -250,7 +250,7 @@ endif
|
||||
# Benchmark targets
|
||||
BENCH_HAKMEM = bench_allocators_hakmem
|
||||
BENCH_SYSTEM = bench_allocators_system
|
||||
BENCH_HAKMEM_OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/free_publish_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/c7_meta_used_counter_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o bench_allocators_hakmem.o
|
||||
BENCH_HAKMEM_OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/free_publish_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/free_front_v3_env_box.o core/box/free_path_stats_box.o core/box/tiny_c6_ultra_free_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/c7_meta_used_counter_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o core/smallobject_hotbox_v3.o core/smallobject_hotbox_v4.o core/smallobject_hotbox_v5.o core/smallsegment_v5.o core/smallobject_cold_iface_v5.o core/smallsegment_v6.o core/smallobject_cold_iface_v6.o core/smallobject_core_v6.o bench_allocators_hakmem.o
|
||||
BENCH_HAKMEM_OBJS = $(BENCH_HAKMEM_OBJS_BASE)
|
||||
ifeq ($(POOL_TLS_PHASE1),1)
|
||||
BENCH_HAKMEM_OBJS += pool_tls.o pool_refill.o pool_tls_arena.o pool_tls_registry.o pool_tls_remote.o
|
||||
@ -427,7 +427,7 @@ test-box-refactor: box-refactor
|
||||
./larson_hakmem 10 8 128 1024 1 12345 4
|
||||
|
||||
# Phase 4: Tiny Pool benchmarks (properly linked with hakmem)
|
||||
TINY_BENCH_OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/free_publish_box.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/c7_meta_used_counter_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o core/smallobject_hotbox_v3.o core/smallobject_hotbox_v4.o core/smallobject_hotbox_v5.o core/smallsegment_v5.o core/smallobject_cold_iface_v5.o core/smallsegment_v6.o core/smallobject_cold_iface_v6.o core/smallobject_core_v6.o
|
||||
TINY_BENCH_OBJS_BASE = hakmem.o hakmem_config.o hakmem_tiny_config.o hakmem_ucb1.o hakmem_bigcache.o hakmem_pool.o hakmem_l25_pool.o hakmem_site_rules.o hakmem_tiny.o core/box/ss_allocation_box.o superslab_stats.o superslab_cache.o superslab_ace.o superslab_slab.o superslab_backend.o core/superslab_head_stub.o hakmem_smallmid.o core/box/superslab_expansion_box.o core/box/integrity_box.o core/box/mailbox_box.o core/box/front_gate_box.o core/box/front_gate_classifier.o core/box/free_publish_box.o core/box/capacity_box.o core/box/carve_push_box.o core/box/prewarm_box.o core/box/ss_hot_prewarm_box.o core/box/front_metrics_box.o core/box/bench_fast_box.o core/box/ss_addr_map_box.o core/box/slab_recycling_box.o core/box/pagefault_telemetry_box.o core/box/tiny_sizeclass_hist_box.o core/box/tiny_env_box.o core/box/tiny_route_box.o core/box/free_front_v3_env_box.o core/box/free_path_stats_box.o core/box/tiny_c6_ultra_free_box.o core/box/tiny_page_box.o core/box/tiny_class_policy_box.o core/box/tiny_class_stats_box.o core/box/tiny_policy_learner_box.o core/box/ss_budget_box.o core/box/tiny_mem_stats_box.o core/box/c7_meta_used_counter_box.o core/box/wrapper_env_box.o core/box/madvise_guard_box.o core/box/libm_reloc_guard_box.o core/box/ptr_trace_box.o core/box/link_missing_stubs.o core/box/super_reg_box.o core/box/shared_pool_box.o core/box/remote_side_box.o core/page_arena.o core/front/tiny_unified_cache.o tiny_sticky.o tiny_remote.o tiny_publish.o tiny_debug_ring.o hakmem_tiny_magazine.o hakmem_tiny_stats.o hakmem_tiny_sfc.o hakmem_tiny_query.o hakmem_tiny_rss.o hakmem_tiny_registry.o hakmem_tiny_remote_target.o hakmem_tiny_bg_spill.o tiny_adaptive_sizing.o hakmem_super_registry.o hakmem_shared_pool.o hakmem_shared_pool_acquire.o hakmem_shared_pool_release.o hakmem_elo.o hakmem_batch.o hakmem_p2.o hakmem_sizeclass_dist.o hakmem_evo.o hakmem_debug.o hakmem_sys.o hakmem_whale.o hakmem_policy.o hakmem_ace.o hakmem_ace_stats.o hakmem_prof.o hakmem_learner.o hakmem_size_hist.o hakmem_learn_log.o hakmem_syscall.o hakmem_ace_metrics.o hakmem_ace_ucb1.o hakmem_ace_controller.o tiny_fastcache.o core/tiny_alloc_fast_push.o core/tiny_c7_ultra_segment.o core/tiny_c7_ultra.o core/link_stubs.o core/tiny_failfast.o core/tiny_destructors.o core/smallobject_hotbox_v3.o core/smallobject_hotbox_v4.o core/smallobject_hotbox_v5.o core/smallsegment_v5.o core/smallobject_cold_iface_v5.o core/smallsegment_v6.o core/smallobject_cold_iface_v6.o core/smallobject_core_v6.o
|
||||
TINY_BENCH_OBJS = $(TINY_BENCH_OBJS_BASE)
|
||||
ifeq ($(POOL_TLS_PHASE1),1)
|
||||
TINY_BENCH_OBJS += pool_tls.o pool_refill.o core/pool_tls_arena.o pool_tls_registry.o pool_tls_remote.o
|
||||
|
||||
@ -16,9 +16,10 @@ static void free_path_stats_dump(void) {
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(stderr, "[FREE_PATH_STATS] total=%lu c7_ultra=%lu small_v3=%lu v6=%lu tiny_v1=%lu pool_v1=%lu remote=%lu super_lookup=%lu legacy_fb=%lu\n",
|
||||
fprintf(stderr, "[FREE_PATH_STATS] total=%lu c7_ultra=%lu c6_ultra_free=%lu small_v3=%lu v6=%lu tiny_v1=%lu pool_v1=%lu remote=%lu super_lookup=%lu legacy_fb=%lu\n",
|
||||
g_free_path_stats.total_calls,
|
||||
g_free_path_stats.c7_ultra_fast,
|
||||
g_free_path_stats.c6_ultra_free_fast, // Phase 4-2
|
||||
g_free_path_stats.smallheap_v3_fast,
|
||||
g_free_path_stats.smallheap_v6_fast,
|
||||
g_free_path_stats.tiny_heap_v1_fast,
|
||||
|
||||
@ -9,6 +9,7 @@ typedef struct FreePathStats {
|
||||
uint64_t total_calls;
|
||||
|
||||
uint64_t c7_ultra_fast;
|
||||
uint64_t c6_ultra_free_fast; // Phase 4-2: C6 ULTRA-free
|
||||
uint64_t smallheap_v3_fast;
|
||||
uint64_t smallheap_v6_fast;
|
||||
uint64_t tiny_heap_v1_fast;
|
||||
|
||||
72
core/box/tiny_c6_ultra_free_box.c
Normal file
72
core/box/tiny_c6_ultra_free_box.c
Normal file
@ -0,0 +1,72 @@
|
||||
#include "tiny_c6_ultra_free_box.h"
|
||||
#include "free_path_stats_box.h"
|
||||
#include "tiny_front_v3_env_box.h"
|
||||
#include "../hakmem.h" // For HAK_BASE_FROM_RAW
|
||||
#include "../front/tiny_unified_cache.h"
|
||||
#include "tiny_front_hot_box.h"
|
||||
#include "../superslab/superslab_inline.h" // For ss_fast_lookup
|
||||
#include <string.h>
|
||||
|
||||
// TLS context
|
||||
static __thread TinyC6UltraFreeTLS g_c6_ultra_free_tls = {0};
|
||||
|
||||
TinyC6UltraFreeTLS* tiny_c6_ultra_free_tls(void) {
|
||||
return &g_c6_ultra_free_tls;
|
||||
}
|
||||
|
||||
// Legacy free helper (shared with slow path)
|
||||
static void hak_tiny_free_legacy_impl(void* base, uint32_t class_idx) {
|
||||
const TinyFrontV3Snapshot* front_snap =
|
||||
__builtin_expect(tiny_front_v3_enabled(), 0) ? tiny_front_v3_snapshot_get() : NULL;
|
||||
|
||||
// Legacy fallback - Unified Cache push
|
||||
if (!front_snap || front_snap->unified_cache_on) {
|
||||
if (unified_cache_push(class_idx, HAK_BASE_FROM_RAW(base))) {
|
||||
FREE_PATH_STAT_INC(legacy_fallback);
|
||||
|
||||
// Phase 4-1: Legacy per-class breakdown
|
||||
if (__builtin_expect(free_path_stats_enabled(), 0)) {
|
||||
if (class_idx < 8) {
|
||||
g_free_path_stats.legacy_by_class[class_idx]++;
|
||||
}
|
||||
}
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
// Final fallback
|
||||
tiny_hot_free_fast(class_idx, base);
|
||||
}
|
||||
|
||||
// Fast path: TLS cache push
|
||||
void tiny_c6_ultra_free_fast(void* base, uint32_t class_idx) {
|
||||
TinyC6UltraFreeTLS* ctx = &g_c6_ultra_free_tls;
|
||||
|
||||
// Phase 4-3: Learn segment on first C6 free
|
||||
if (unlikely(ctx->seg_base == 0)) {
|
||||
SuperSlab* ss = ss_fast_lookup(base);
|
||||
if (ss != NULL) {
|
||||
ctx->seg_base = (uintptr_t)ss;
|
||||
ctx->seg_end = ctx->seg_base + (1u << ss->lg_size);
|
||||
}
|
||||
}
|
||||
|
||||
// Check if ptr is in our segment AND cache has room
|
||||
if (likely(ctx->seg_base != 0 &&
|
||||
(uintptr_t)base >= ctx->seg_base &&
|
||||
(uintptr_t)base < ctx->seg_end &&
|
||||
ctx->count < TINY_C6_ULTRA_FREE_CAP)) {
|
||||
// Push to TLS cache
|
||||
ctx->freelist[ctx->count++] = base;
|
||||
FREE_PATH_STAT_INC(c6_ultra_free_fast);
|
||||
return;
|
||||
}
|
||||
|
||||
// Slow path: fallback to legacy (cache full or ptr not in segment)
|
||||
tiny_c6_ultra_free_slow(base, class_idx);
|
||||
}
|
||||
|
||||
// Slow path: fallback to legacy free
|
||||
void tiny_c6_ultra_free_slow(void* base, uint32_t class_idx) {
|
||||
hak_tiny_free_legacy_impl(base, class_idx);
|
||||
}
|
||||
30
core/box/tiny_c6_ultra_free_box.h
Normal file
30
core/box/tiny_c6_ultra_free_box.h
Normal file
@ -0,0 +1,30 @@
|
||||
#ifndef HAKMEM_TINY_C6_ULTRA_FREE_BOX_H
|
||||
#define HAKMEM_TINY_C6_ULTRA_FREE_BOX_H
|
||||
|
||||
#include <stdint.h>
|
||||
#include <stdbool.h>
|
||||
#include "tiny_c6_ultra_free_env_box.h"
|
||||
|
||||
#ifndef likely
|
||||
#define likely(x) __builtin_expect(!!(x), 1)
|
||||
#define unlikely(x) __builtin_expect(!!(x), 0)
|
||||
#endif
|
||||
|
||||
// TLS cache capacity (Phase 4-3: increased for better coverage)
|
||||
#define TINY_C6_ULTRA_FREE_CAP 128
|
||||
|
||||
// TLS context for C6 ULTRA-free
|
||||
typedef struct TinyC6UltraFreeTLS {
|
||||
void* freelist[TINY_C6_ULTRA_FREE_CAP]; // BASE pointers
|
||||
uint8_t count;
|
||||
uint8_t _pad[7];
|
||||
uintptr_t seg_base; // C6 segment range (0 = not initialized)
|
||||
uintptr_t seg_end;
|
||||
} TinyC6UltraFreeTLS;
|
||||
|
||||
// API
|
||||
TinyC6UltraFreeTLS* tiny_c6_ultra_free_tls(void);
|
||||
void tiny_c6_ultra_free_fast(void* base, uint32_t class_idx);
|
||||
void tiny_c6_ultra_free_slow(void* base, uint32_t class_idx);
|
||||
|
||||
#endif // HAKMEM_TINY_C6_ULTRA_FREE_BOX_H
|
||||
17
core/box/tiny_c6_ultra_free_env_box.h
Normal file
17
core/box/tiny_c6_ultra_free_env_box.h
Normal file
@ -0,0 +1,17 @@
|
||||
#ifndef HAKMEM_TINY_C6_ULTRA_FREE_ENV_BOX_H
|
||||
#define HAKMEM_TINY_C6_ULTRA_FREE_ENV_BOX_H
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdbool.h>
|
||||
|
||||
// ENV: HAKMEM_TINY_C6_ULTRA_FREE_ENABLED (default 0)
|
||||
static inline bool tiny_c6_ultra_free_enabled(void) {
|
||||
static int g_enabled = -1;
|
||||
if (__builtin_expect(g_enabled == -1, 0)) {
|
||||
const char* e = getenv("HAKMEM_TINY_C6_ULTRA_FREE_ENABLED");
|
||||
g_enabled = (e && *e && *e != '0') ? 1 : 0;
|
||||
}
|
||||
return g_enabled;
|
||||
}
|
||||
|
||||
#endif // HAKMEM_TINY_C6_ULTRA_FREE_ENV_BOX_H
|
||||
@ -46,6 +46,7 @@
|
||||
// Phase FREE-LEGACY-BREAKDOWN-1: v6 は型エラーがあるため一時的にコメントアウト（デフォルト OFF なので影響なし）
|
||||
// #include "../box/smallobject_core_v6_box.h" // SmallObject Core v6 (C6-only route stub, Phase v6-1)
|
||||
#include "../box/tiny_c7_ultra_box.h" // C7 ULTRA stub (UF-1, delegates to v3)
|
||||
#include "../box/tiny_c6_ultra_free_box.h" // Phase 4-2: C6 ULTRA-free (free-only, C6-only)
|
||||
#include "../box/tiny_front_v3_env_box.h" // Tiny front v3 snapshot gate
|
||||
#include "../box/tiny_heap_env_box.h" // ENV gate for TinyHeap front (A/B)
|
||||
#include "../box/tiny_route_env_box.h" // Route snapshot (Heap vs Legacy)
|
||||
@ -80,6 +81,35 @@ static inline int front_gate_unified_enabled(void) {
|
||||
return g_enable;
|
||||
}
|
||||
|
||||
// ============================================================================
|
||||
// Phase 4-2: Legacy free helper (shared implementation in tiny_c6_ultra_free_box.c)
|
||||
// ============================================================================
|
||||
|
||||
// Helper function to perform legacy free (used in both regular path and C6 ULTRA fallback)
|
||||
__attribute__((always_inline))
|
||||
static inline void hak_tiny_free_legacy_inline(void* base, uint32_t class_idx) {
|
||||
const TinyFrontV3Snapshot* front_snap =
|
||||
__builtin_expect(tiny_front_v3_enabled(), 0) ? tiny_front_v3_snapshot_get() : NULL;
|
||||
|
||||
// Legacy fallback - Unified Cache push
|
||||
if (!front_snap || front_snap->unified_cache_on) {
|
||||
if (unified_cache_push(class_idx, HAK_BASE_FROM_RAW(base))) {
|
||||
FREE_PATH_STAT_INC(legacy_fallback);
|
||||
|
||||
// Phase 4-1: Legacy per-class breakdown
|
||||
if (__builtin_expect(free_path_stats_enabled(), 0)) {
|
||||
if (class_idx < 8) {
|
||||
g_free_path_stats.legacy_by_class[class_idx]++;
|
||||
}
|
||||
}
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
// Final fallback
|
||||
tiny_hot_free_fast(class_idx, base);
|
||||
}
|
||||
|
||||
// ============================================================================
|
||||
// Phase 4-Step2: malloc_tiny_fast() - Hot/Cold Path Box (ACTIVE)
|
||||
// ============================================================================
|
||||
@ -289,6 +319,12 @@ static inline int free_tiny_fast(void* ptr) {
|
||||
return 1;
|
||||
}
|
||||
|
||||
// Phase 4-2: C6 ULTRA-free (C6-only, free-only, ENV gated)
|
||||
if (class_idx == 6 && tiny_c6_ultra_free_enabled()) {
|
||||
tiny_c6_ultra_free_fast(base, class_idx);
|
||||
return 1;
|
||||
}
|
||||
|
||||
// C7 v3 fast classify: bypass classify_ptr/ss_map_lookup for clear hits
|
||||
if (class_idx == 7 &&
|
||||
tiny_front_v3_enabled() &&
|
||||
@ -457,26 +493,9 @@ static inline int free_tiny_fast(void* ptr) {
|
||||
}
|
||||
#endif
|
||||
|
||||
int pushed = 0;
|
||||
if (!front_snap || front_snap->unified_cache_on) {
|
||||
pushed = unified_cache_push(class_idx, HAK_BASE_FROM_RAW(base));
|
||||
}
|
||||
if (__builtin_expect(pushed, 1)) {
|
||||
// Phase FREE-LEGACY-BREAKDOWN-1: カウンタ散布 (10. legacy fallback)
|
||||
FREE_PATH_STAT_INC(legacy_fallback);
|
||||
|
||||
// Phase 4-1: Legacy per-class breakdown
|
||||
if (__builtin_expect(free_path_stats_enabled(), 0)) {
|
||||
if (class_idx >= 0 && class_idx < 8) {
|
||||
g_free_path_stats.legacy_by_class[class_idx]++;
|
||||
}
|
||||
}
|
||||
|
||||
return 1; // Success
|
||||
}
|
||||
|
||||
// Unified Cache full → 通常 free 経路へ
|
||||
return 0;
|
||||
// Phase 4-2: Legacy fallback (use inline helper)
|
||||
hak_tiny_free_legacy_inline(base, class_idx);
|
||||
return 1;
|
||||
#else
|
||||
// No header mode - fall back to normal free
|
||||
return 0;
|
||||
|
||||
@ -304,3 +304,220 @@ Phase FREE-LEGACY-OPT-4-2 で **C6 クラス** に ULTRA-Free lane を実装す
|
||||
|
||||
- **C5 (257-512B)**: Legacy の 25.8%（全体の 12.7%）
|
||||
- C5 も ULTRA-Free の候補だが、C6 の効果を確認してから判断すべき
|
||||
|
||||
### 次のアクション（Phase 4-2）
|
||||
|
||||
**最大ターゲット**: C6 Legacy free（Legacy の 51.4%、全体の 25.3%）
|
||||
|
||||
**実装方針**:
|
||||
1. C6_ULTRA_FREE_BOX（free-only, C6-only）を追加
|
||||
2. C7 ULTRA 風の TLS キャッシュ（`freelist[32]` + `count` + segment range check）
|
||||
3. ENV で opt-in（`HAKMEM_TINY_C6_ULTRA_FREE_ENABLED=0`）
|
||||
4. alloc 側は既存ルートのまま（後続フェーズで検討）
|
||||
|
||||
**期待効果**:
|
||||
- C6 Legacy を TLS キャッシュに逃がす → Legacy 半減
|
||||
- Legacy: 49.2% → 24-27%
|
||||
- Mixed: +5-8% 改善
|
||||
|
||||
## Phase FREE-LEGACY-OPT-4-2: C6_ULTRA_FREE_BOX 実装結果
|
||||
|
||||
### 実装内容
|
||||
|
||||
**コンポーネント**:
|
||||
- `core/box/tiny_c6_ultra_free_env_box.h`: ENV ゲート（HAKMEM_TINY_C6_ULTRA_FREE_ENABLED=0）
|
||||
- `core/box/tiny_c6_ultra_free_box.h/c`: C6 ULTRA-free TLS キャッシュ
|
||||
- `core/front/malloc_tiny_fast.h`: free フロントへの接続（C7 ULTRA 直後に分岐追加）
|
||||
- `core/box/free_path_stats_box.h/c`: c6_ultra_free_fast カウンタ追加
|
||||
|
||||
**設計**:
|
||||
- TLS に `freelist[32]` + `count` + `seg_base/seg_end` (segment range check)
|
||||
- Fast path: seg_base が初期化済み && ptr in range && count < 32 → TLS push
|
||||
- Slow path: Legacy fallback
|
||||
|
||||
### A/B テスト結果（Mixed 16-1024B）
|
||||
|
||||
**ベースライン（C6 ULTRA OFF）**:
|
||||
```
|
||||
[FREE_PATH_STATS] total=542031 c7_ultra=275089 c6_ultra_free=0 small_v3=0 v6=0 tiny_v1=0 pool_v1=8081 remote=0 super_lookup=0 legacy_fb=266942
|
||||
[FREE_PATH_STATS_LEGACY_BY_CLASS] c0=0 c1=0 c2=8746 c3=17279 c4=34727 c5=68871 c6=137319 c7=0
|
||||
Throughput = 42.2M ops/s
|
||||
```
|
||||
|
||||
**C6 ULTRA ON（実測）**:
|
||||
```
|
||||
[FREE_PATH_STATS] total=542031 c7_ultra=275089 c6_ultra_free=0 small_v3=0 v6=0 tiny_v1=0 pool_v1=8081 remote=0 super_lookup=0 legacy_fb=266942
|
||||
[FREE_PATH_STATS_LEGACY_BY_CLASS] c0=0 c1=0 c2=8746 c3=17279 c4=34727 c5=68871 c6=137319 c7=0
|
||||
Throughput = 39.9M ops/s (-5.5%)
|
||||
```
|
||||
|
||||
### 問題分析
|
||||
|
||||
**観測された動作**:
|
||||
- `c6_ultra_free=0` → Fast path に全く入っていない
|
||||
- `legacy_by_class[6]=137,319` → C6 はすべて Legacy fallback のまま
|
||||
- Throughput 低下 (42.2M → 39.9M, -5.5%) → ENV チェックのオーバーヘッドのみ
|
||||
|
||||
**根本原因**: **Segment 未初期化**
|
||||
|
||||
C6 ULTRA-free は「free-only」設計のため、TLS の `seg_base/seg_end` を初期化する手段がない:
|
||||
- Fast path の条件: `seg_base != 0 && ptr in range`
|
||||
- しかし、seg_base を設定する alloc 側の実装がない
|
||||
- 結果: 常に seg_base == 0 で slow path（Legacy fallback）に落ちる
|
||||
|
||||
**設計上の矛盾**:
|
||||
1. **Free-only の限界**: TLS キャッシュを使うには、そのポインタが「自分が管理する segment」内かを確認する必要がある
|
||||
2. **Segment 所有権**: C7 ULTRA は alloc 時に segment を割り当て、free 時にその segment 内かを確認する
|
||||
3. **C6 の現状**: alloc は既存ルート（pool_v1）を使い、C6 ULTRA は segment を持たない
|
||||
4. **結論**: Segment range check は不可能 → Fast path は機能しない
|
||||
|
||||
### 技術的洞察
|
||||
|
||||
**C7 ULTRA との違い**:
|
||||
- C7 ULTRA は alloc/free 両方を制御し、専用 segment を持つ
|
||||
- C6 ULTRA-free は free のみを受けるため、segment 所有権がない
|
||||
- 結果: Segment-based ownership check は不可能
|
||||
|
||||
**代替案の検討**:
|
||||
|
||||
1. **Option A: Segment 初期化を追加**
|
||||
- C6 alloc 時に専用 segment を割り当て
|
||||
- しかし、これは「alloc は既存ルートのまま」という要件に反する
|
||||
|
||||
2. **Option B: Segment check を削除**
|
||||
- すべての C6 ポインタを TLS に push（ownership check なし）
|
||||
- リスク: 他スレッドのポインタを TLS に入れる可能性（cross-thread free 問題）
|
||||
|
||||
3. **Option C: Header-based check**
|
||||
- ポインタの header を読み、C6 かを確認（すでに class_idx == 6 で確認済み）
|
||||
- TLS ownership は諦め、class ベースの simple cache とする
|
||||
|
||||
4. **Option D: Phase 中止**
|
||||
- Free-only での TLS キャッシュは設計上困難
|
||||
- C6 の最適化は alloc/free 両方を含む v4/v5 に委ねる
|
||||
|
||||
### 推奨アクション
|
||||
|
||||
**短期**: Phase 4-2 を「研究箱（実装のみ、無効）」として保持
|
||||
- ENV デフォルト OFF のまま
|
||||
- ビルドエラーがないことを確認済み
|
||||
- 実際の効果はなし（seg_base == 0 のため）
|
||||
|
||||
**中期**: C6 最適化は別アプローチで
|
||||
- SmallHeap v4/v5/v6 など、alloc/free 両方を制御する経路で C6 を最適化
|
||||
- または、C6 専用の alloc/free 統合パスを新規実装
|
||||
|
||||
**長期**: Free-only 最適化の限界を認識
|
||||
- TLS キャッシュは alloc/free 統合が前提
|
||||
- Segment ownership なしでは安全な TLS push は困難
|
||||
|
||||
## Phase FREE-LEGACY-OPT-4-3: Segment 初回学習実装
|
||||
|
||||
### 目的
|
||||
|
||||
Phase 4-2 の「seg_base == 0 で fast path に入らない」問題を解決するため、**初回 C6 free で `ss_fast_lookup` を呼び出し、segment 情報を学習**する。
|
||||
|
||||
### 実装内容
|
||||
|
||||
**`core/box/tiny_c6_ultra_free_box.c` の修正**:
|
||||
|
||||
```c
|
||||
void tiny_c6_ultra_free_fast(void* base, uint32_t class_idx) {
|
||||
TinyC6UltraFreeTLS* ctx = &g_c6_ultra_free_tls;
|
||||
|
||||
// Phase 4-3: Learn segment on first C6 free
|
||||
if (unlikely(ctx->seg_base == 0)) {
|
||||
SuperSlab* ss = ss_fast_lookup(base);
|
||||
if (ss != NULL) {
|
||||
ctx->seg_base = (uintptr_t)ss;
|
||||
ctx->seg_end = ctx->seg_base + (1u << ss->lg_size);
|
||||
}
|
||||
}
|
||||
|
||||
// Fast path: segment ownership check + TLS push
|
||||
if (likely(ctx->seg_base != 0 &&
|
||||
(uintptr_t)base >= ctx->seg_base &&
|
||||
(uintptr_t)base < ctx->seg_end &&
|
||||
ctx->count < TINY_C6_ULTRA_FREE_CAP)) {
|
||||
ctx->freelist[ctx->count++] = base;
|
||||
FREE_PATH_STAT_INC(c6_ultra_free_fast);
|
||||
return;
|
||||
}
|
||||
|
||||
// Slow path: fallback to legacy
|
||||
tiny_c6_ultra_free_slow(base, class_idx);
|
||||
}
|
||||
```
|
||||
|
||||
**キャッシュサイズ拡大**:
|
||||
- `TINY_C6_ULTRA_FREE_CAP = 128`（32 → 128 に増加）
|
||||
|
||||
### テスト結果
|
||||
|
||||
**C6 ULTRA ON + Segment 学習（Mixed 16-1024B）**:
|
||||
```
|
||||
[FREE_PATH_STATS] total=542031 c7_ultra=275089 c6_ultra_free=128 ...
|
||||
[FREE_PATH_STATS_LEGACY_BY_CLASS] ... c6=137191 ...
|
||||
```
|
||||
|
||||
**観測**:
|
||||
- `c6_ultra_free=128` → Fast path に 128 ブロックがヒット（seg_base 学習成功）
|
||||
- `legacy_by_class[6]=137,191` → 残りの C6 (137K - 128 = ~137K) は Legacy
|
||||
|
||||
### A/B 比較
|
||||
|
||||
| 測定 | OFF (M ops/s) | ON (M ops/s) | 差 |
|
||||
|------|---------------|--------------|-----|
|
||||
| Run 1 | 40.2 | 41.5 | +3.2% |
|
||||
| Run 2 | 42.5 | 42.9 | +1.1% |
|
||||
| 平均 | 41.4 | 42.2 | +1-3% (誤差範囲) |
|
||||
|
||||
### 問題分析
|
||||
|
||||
**根本的限界**:
|
||||
1. **キャッシュは満杯になる**: 128 ブロックがキャッシュされると、以降の C6 free はすべて Legacy へオーバーフロー
|
||||
2. **alloc との連携なし**: キャッシュされたブロックは alloc で再利用されない
|
||||
3. **ドレイン戦略は効果なし**: ドレイン時に Legacy を呼ぶため、結局同じオーバーヘッド
|
||||
|
||||
**数値で見る限界**:
|
||||
- C6 total frees: 137,319
|
||||
- Fast path hits: 128 (0.09%)
|
||||
- Legacy fallback: 137,191 (99.91%)
|
||||
|
||||
### ドレイン戦略の試行（失敗）
|
||||
|
||||
満杯時に半分をドレインする戦略も試行:
|
||||
|
||||
```c
|
||||
static void tiny_c6_ultra_free_drain_half(TinyC6UltraFreeTLS* ctx, uint32_t class_idx) {
|
||||
uint8_t drain_count = ctx->count / 2;
|
||||
for (uint8_t i = 0; i < drain_count; i++) {
|
||||
hak_tiny_free_legacy_impl(ctx->freelist[i], class_idx);
|
||||
}
|
||||
// compact remaining...
|
||||
}
|
||||
```
|
||||
|
||||
**結果**:
|
||||
- 全 137K C6 frees が fast path を通過 (`c6_ultra_free=137319`)
|
||||
- しかし、137K blocks がドレインで Legacy に送られる
|
||||
- **Throughput は baseline と同等**（+0%, 相殺）
|
||||
|
||||
### 結論
|
||||
|
||||
**Free-only TLS キャッシュは alloc 連携なしでは効果限定的**
|
||||
|
||||
**既存の解決策**:
|
||||
- Core v6 (`SmallHeapCtxV6`) に既に alloc/free 統合 TLS freelist が存在
|
||||
- ただし Core v6 は現在 -12% で baseline より遅い（別途最適化が必要）
|
||||
|
||||
**Phase 4-3 の成果**:
|
||||
1. ✅ Segment 初回学習の実装（`ss_fast_lookup` 活用）
|
||||
2. ✅ Fast path が機能することを確認
|
||||
3. ❌ alloc 連携なしでは効果なし（0.09% のブロックのみキャッシュ）
|
||||
4. ❌ ドレイン戦略も相殺で効果なし
|
||||
|
||||
**推奨**:
|
||||
- C6 ULTRA Free は **研究箱として維持**（ENV デフォルト OFF）
|
||||
- 真の最適化には **alloc 側との TLS 連携**が必須
|
||||
- Core v6 の改良、または新規 alloc/free 統合パスの検討が必要
|
||||
|
||||
Reference in New Issue
Block a user