// tiny_heap_box.h - TinyHeap front (mimalloc 風) の共通 Box
// 役割:
//   - クラスごとの TinyHeap コンテキスト（current/partial/full + freelist）を TLS で保持。
//   - ホットパス alloc/free は TinyHeap 内で完結させ、ページ枯渇/全 free の境界でのみ
//     Superslab/Tier/Warm などの下層 Box に触れる。
//   - C7HotBox はこの TinyHeapBox 上のクラス7ラッパとして実装する。
#pragma once

#include <stdint.h>
#include <stddef.h>
#include <stdatomic.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>

#include "../hakmem_build_flags.h"
#include "../hakmem_tiny_superslab.h"    // SuperSlab, TinySlabMeta, superslab_refill
#include "../hakmem_super_registry.h"    // hak_super_lookup
#include "../superslab/superslab_inline.h"  // slab_index_for
#include "../tiny_tls.h"                 // TinyTLSSlab
#include "../tiny_box_geometry.h"        // tiny_stride_for_class
#include "../tiny_region_id.h"           // tiny_region_id_write_header
#include "tiny_layout_box.h"             // tiny_user_offset
#include "tiny_next_ptr_box.h"           // tiny_next_read/write

// Forward decls for SuperSlab active counters (definitions in hakmem_tiny_ss_active_box.inc)
void ss_active_add(SuperSlab* ss, uint32_t n);

#ifndef TINY_HEAP_MAX_PAGES_PER_CLASS
#define TINY_HEAP_MAX_PAGES_PER_CLASS 4
#endif

typedef struct tiny_heap_page_t {
    void* free_list;            // ページ内 free list 先頭 (BASE)
    uint16_t used;              // 現在の使用数（active count）
    uint16_t capacity;          // ページ内の総ブロック数
    uint16_t slab_idx;          // 対応する slab index
    uint16_t _pad;
    uint8_t* base;              // slab 先頭（データ領域先頭）
    TinySlabMeta* meta;         // Superslab メタ（owner/Tier 判定用）
    SuperSlab* ss;              // 所有する Superslab
    struct tiny_heap_page_t* next;
    int32_t c7_active_delta;    // C7 meta-light 用: total_active_blocks の差分
    int32_t c7_used_delta;      // C7 meta-light 用: meta->used の差分
} tiny_heap_page_t;

typedef struct tiny_heap_class_t {
    tiny_heap_page_t* current_page;
    tiny_heap_page_t* partial_pages;
    tiny_heap_page_t* full_pages;
    tiny_heap_page_t  nodes[TINY_HEAP_MAX_PAGES_PER_CLASS];
    uint8_t           node_in_use[TINY_HEAP_MAX_PAGES_PER_CLASS];
    uint16_t          stride;   // cached block stride for this class
    uint16_t          _pad;
} tiny_heap_class_t;

typedef struct tiny_heap_ctx_t {
    tiny_heap_class_t cls[TINY_NUM_CLASSES];
    uint8_t initialized;
} tiny_heap_ctx_t;

// TLS state (定義は core/hakmem_tiny.c)
extern __thread tiny_heap_ctx_t g_tiny_heap_ctx;
extern __thread int g_tiny_heap_ctx_init;
extern __thread TinyTLSSlab g_tls_slabs[TINY_NUM_CLASSES];

static inline int tiny_c7_heap_stats_enabled(void) {
    static int g = -1;
    if (__builtin_expect(g == -1, 0)) {
        const char* e = getenv("HAKMEM_TINY_C7_HEAP_STATS");
        g = (e && *e && *e != '0') ? 1 : 0;
    }
    return g;
}

static inline int tiny_c7_meta_light_enabled(void) {
    static int g = -1;
    if (__builtin_expect(g == -1, 0)) {
        const char* e = getenv("HAKMEM_TINY_C7_META_LIGHT");
        g = (e && *e && *e != '0') ? 1 : 0;
    }
    return g;
}

static inline int tiny_c7_delta_debug_enabled(void) {
    static int g = -1;
    if (__builtin_expect(g == -1, 0)) {
        const char* e = getenv("HAKMEM_TINY_C7_DELTA_DEBUG");
        g = (e && *e && *e != '0') ? 1 : 0;
    }
    return g;
}

typedef struct {
    _Atomic uint64_t alloc_fast_current;
    _Atomic uint64_t alloc_slow_prepare;
    _Atomic uint64_t free_fast_local;
    _Atomic uint64_t free_slow_fallback;
    _Atomic uint64_t alloc_prepare_fail;
    _Atomic uint64_t alloc_fail;
} TinyC7HeapStats;

extern TinyC7HeapStats g_c7_heap_stats;

static inline int tiny_heap_cold_drain_and_free(int class_idx, void* base) {
    (void)class_idx;
    (void)base;
    return 0;
}

static inline tiny_heap_ctx_t* tiny_heap_ctx_for_thread(void) {
    tiny_heap_ctx_t* ctx = &g_tiny_heap_ctx;
    if (!g_tiny_heap_ctx_init) {
        memset(ctx, 0, sizeof(*ctx));
        g_tiny_heap_ctx_init = 1;
        ctx->initialized = 1;
        for (int c = 0; c < TINY_NUM_CLASSES; c++) {
            tiny_heap_class_t* hcls = &ctx->cls[c];
            hcls->stride = (uint16_t)tiny_stride_for_class(c);
        }
    }
    return ctx;
}

static inline tiny_heap_class_t* tiny_heap_class(tiny_heap_ctx_t* ctx, int class_idx) {
    if (!ctx || class_idx < 0 || class_idx >= TINY_NUM_CLASSES) return NULL;
    return &ctx->cls[class_idx];
}

static inline size_t tiny_heap_block_stride(int class_idx) {
    return tiny_stride_for_class(class_idx);
}

static inline void tiny_heap_page_clear(tiny_heap_page_t* page) {
    if (!page) return;
    page->free_list = NULL;
    page->used = 0;
    page->capacity = 0;
    page->slab_idx = 0;
    page->base = NULL;
    page->meta = NULL;
    page->ss = NULL;
    page->next = NULL;
    page->c7_active_delta = 0;
    page->c7_used_delta = 0;
}

static inline void tiny_c7_meta_flush_page(tiny_heap_page_t* page) {
    if (!page || !tiny_c7_meta_light_enabled()) return;
    if (!page->meta || !page->ss) return;

    int32_t active_delta = page->c7_active_delta;
    int32_t used_delta = page->c7_used_delta;
    if (active_delta == 0 && used_delta == 0) {
        return;
    }

    if (used_delta != 0) {
        atomic_fetch_add_explicit(&page->meta->used, used_delta, memory_order_relaxed);
        page->c7_used_delta = 0;
    }
    if (active_delta != 0) {
        if (active_delta > 0) {
            ss_active_add(page->ss, (uint32_t)active_delta);
        } else {
            int32_t n = -active_delta;
            for (int32_t i = 0; i < n; i++) {
                ss_active_dec_one(page->ss);
            }
        }
        page->c7_active_delta = 0;
    }
}

static inline int tiny_c7_delta_should_flush(tiny_heap_page_t* page) {
    if (!page) return 0;
    if (!tiny_c7_meta_light_enabled()) return 0;

    int32_t ud = page->c7_used_delta;
    int32_t ad = page->c7_active_delta;
    int32_t abs_ud = (ud >= 0) ? ud : -ud;
    int32_t abs_ad = (ad >= 0) ? ad : -ad;
    int32_t abs_max = (abs_ud > abs_ad) ? abs_ud : abs_ad;
    if (abs_max == 0) {
        return 0;
    }

    uint16_t cap = page->capacity;
    int32_t base_th = (cap > 0) ? ((int32_t)cap * 16) : 256;
    if (base_th < 256) {
        base_th = 256;
    }
    return abs_max >= base_th;
}

static __attribute__((noinline, unused)) void tiny_c7_heap_debug_dump_deltas(void) {
    if (!tiny_c7_meta_light_enabled() || !tiny_c7_delta_debug_enabled()) {
        return;
    }

    tiny_heap_ctx_t* ctx = tiny_heap_ctx_for_thread();
    int class_idx = 7;
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls) return;

    uint64_t nonzero_pages = 0;
    int64_t used_delta_sum = 0;
    int64_t active_delta_sum = 0;

    for (int i = 0; i < TINY_HEAP_MAX_PAGES_PER_CLASS; i++) {
        tiny_heap_page_t* p = &hcls->nodes[i];
        if (p->c7_used_delta != 0 || p->c7_active_delta != 0) {
            nonzero_pages++;
            used_delta_sum += (int64_t)p->c7_used_delta;
            active_delta_sum += (int64_t)p->c7_active_delta;
            fprintf(stderr,
                    "[C7_DELTA_PAGE] idx=%d used_delta=%d active_delta=%d used=%u cap=%u ss=%p slab=%u\n",
                    i,
                    p->c7_used_delta,
                    p->c7_active_delta,
                    (unsigned)p->used,
                    (unsigned)p->capacity,
                    (void*)p->ss,
                    (unsigned)p->slab_idx);
        }
    }

    fprintf(stderr,
            "[C7_DELTA_SUMMARY] nonzero_pages=%llu used_delta_sum=%lld active_delta_sum=%lld\n",
            (unsigned long long)nonzero_pages,
            (long long)used_delta_sum,
            (long long)active_delta_sum);
}

static inline tiny_heap_page_t* tiny_heap_class_acquire_node(tiny_heap_class_t* hcls) {
    if (!hcls) return NULL;
    for (int i = 0; i < TINY_HEAP_MAX_PAGES_PER_CLASS; i++) {
        if (hcls->node_in_use[i] == 0) {
            hcls->node_in_use[i] = 1;
            tiny_heap_page_t* p = &hcls->nodes[i];
            tiny_heap_page_clear(p);
            return p;
        }
    }
    return NULL;
}

static inline void tiny_heap_class_release_node(tiny_heap_class_t* hcls, tiny_heap_page_t* page) {
    if (!hcls || !page) return;
    if (tiny_c7_meta_light_enabled()) {
        tiny_c7_meta_flush_page(page);
    }
    intptr_t idx = page - hcls->nodes;
    if (idx >= 0 && idx < TINY_HEAP_MAX_PAGES_PER_CLASS) {
        hcls->node_in_use[idx] = 0;
        tiny_heap_page_clear(page);
    }
}

static inline tiny_heap_page_t* tiny_heap_class_find_page(tiny_heap_class_t* hcls, SuperSlab* ss, int slab_idx) {
    if (!hcls || !ss || slab_idx < 0) return NULL;
    tiny_heap_page_t* lists[] = {hcls->current_page, hcls->partial_pages, hcls->full_pages};
    for (size_t li = 0; li < sizeof(lists) / sizeof(lists[0]); li++) {
        tiny_heap_page_t* p = lists[li];
        while (p) {
            if (p->ss == ss && (int)p->slab_idx == slab_idx) {
                return p;
            }
            p = p->next;
        }
    }
    return NULL;
}

static inline tiny_heap_page_t* tiny_heap_attach_page(tiny_heap_ctx_t* ctx,
                                                      int class_idx,
                                                      SuperSlab* ss,
                                                      int slab_idx) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls || !ss || slab_idx < 0) return NULL;
    TinySlabMeta* meta = &ss->slabs[slab_idx];
    if (meta->class_idx != class_idx) {
        return NULL;
    }

    tiny_heap_page_t* page = tiny_heap_class_find_page(hcls, ss, slab_idx);
    if (!page) {
        page = tiny_heap_class_acquire_node(hcls);
        if (!page) return NULL;
        page->ss = ss;
        page->slab_idx = (uint16_t)slab_idx;
        page->meta = meta;
        page->base = tiny_slab_base_for(ss, slab_idx);
        page->capacity = meta->capacity;
        page->used = atomic_load_explicit(&meta->used, memory_order_relaxed);
        page->free_list = atomic_load_explicit(&meta->freelist, memory_order_acquire);
        page->next = hcls->partial_pages;
        hcls->partial_pages = page;
    } else {
        page->capacity = meta->capacity;
        page->used = atomic_load_explicit(&meta->used, memory_order_relaxed);
        page->free_list = atomic_load_explicit(&meta->freelist, memory_order_acquire);
        if (!page->base) {
            page->base = tiny_slab_base_for(ss, slab_idx);
        }
    }
    if (class_idx == 7 && tiny_c7_meta_light_enabled()) {
        if (page->capacity > 0 && page->used > page->capacity) {
            page->used = page->capacity;
        }
        page->c7_used_delta = 0;
        page->c7_active_delta = 0;
    }
    return page;
}

static inline tiny_heap_page_t* tiny_heap_take_current(tiny_heap_class_t* hcls) {
    if (!hcls) return NULL;
    if (hcls->current_page) {
        return hcls->current_page;
    }
    if (hcls->partial_pages) {
        tiny_heap_page_t* page = hcls->partial_pages;
        hcls->partial_pages = page->next;
        page->next = NULL;
        hcls->current_page = page;
        return page;
    }
    return NULL;
}

static inline void tiny_heap_class_unlink(tiny_heap_class_t* hcls, tiny_heap_page_t* page) {
    if (!hcls || !page) return;
    if (hcls->current_page == page) {
        hcls->current_page = NULL;
    }
    tiny_heap_page_t** lists[] = {&hcls->partial_pages, &hcls->full_pages};
    for (size_t i = 0; i < sizeof(lists) / sizeof(lists[0]); i++) {
        tiny_heap_page_t** head = lists[i];
        tiny_heap_page_t* prev = NULL;
        tiny_heap_page_t* cur = head ? *head : NULL;
        while (cur) {
            if (cur == page) {
                if (prev) {
                    prev->next = cur->next;
                } else {
                    *head = cur->next;
                }
                break;
            }
            prev = cur;
            cur = cur->next;
        }
    }
}

static inline tiny_heap_page_t* tiny_heap_page_of(tiny_heap_ctx_t* ctx, int class_idx, void* base_ptr) {
    if (!ctx || !base_ptr || class_idx < 0 || class_idx >= TINY_NUM_CLASSES) return NULL;
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls) return NULL;

    // TLS fast path: TLS slab の範囲内なら lookup を避ける
    TinyTLSSlab* tls = &g_tls_slabs[class_idx];
    if (tls->ss && tls->meta && tls->slab_base) {
        size_t stride = hcls->stride;
        if (stride == 0) {
            stride = tiny_heap_block_stride(class_idx);
            hcls->stride = (uint16_t)stride;
        }
        uint16_t cap = tls->meta->capacity;
        if (cap > 0) {
            uint8_t* slab_base = tls->slab_base;
            size_t slab_size = stride * (size_t)cap;
            uint8_t* base_u8 = (uint8_t*)base_ptr;
            if (base_u8 >= slab_base && base_u8 < (slab_base + slab_size)) {
                return tiny_heap_attach_page(ctx, class_idx, tls->ss, tls->slab_idx);
            }
        }
    }

    // fallback: Superslab lookup
    SuperSlab* ss = hak_super_lookup(base_ptr);
    if (!ss || ss->magic != SUPERSLAB_MAGIC) {
        return NULL;
    }
    int slab_idx = slab_index_for(ss, base_ptr);
    if (slab_idx < 0 || slab_idx >= ss_slabs_capacity(ss)) {
        return NULL;
    }
    return tiny_heap_attach_page(ctx, class_idx, ss, slab_idx);
}

static inline void tiny_heap_page_push_to_full(tiny_heap_class_t* hcls, tiny_heap_page_t* page) {
    if (!hcls || !page) return;
    page->next = hcls->full_pages;
    hcls->full_pages = page;
}

static inline void tiny_heap_page_push_to_partial(tiny_heap_class_t* hcls, tiny_heap_page_t* page) {
    if (!hcls || !page) return;
    page->next = hcls->partial_pages;
    hcls->partial_pages = page;
}

static inline void tiny_heap_page_becomes_empty(tiny_heap_ctx_t* ctx, int class_idx, tiny_heap_page_t* page) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls || !page) return;

    if (class_idx == 7 && tiny_c7_meta_light_enabled()) {
        tiny_c7_meta_flush_page(page);
    }
    if (page->meta && page->ss) {
        ss_partial_publish(class_idx, page->ss);
    }
    tiny_heap_class_unlink(hcls, page);
    tiny_heap_class_release_node(hcls, page);
}

static inline void tiny_heap_page_mark_full(tiny_heap_class_t* hcls, tiny_heap_page_t* page) {
    if (!hcls || !page) return;
    tiny_heap_class_unlink(hcls, page);
    tiny_heap_page_push_to_full(hcls, page);
}

static inline void* tiny_heap_page_pop(tiny_heap_class_t* hcls, int class_idx, tiny_heap_page_t* page) {
    if (!hcls || !page || !page->meta || !page->ss || !page->base) return NULL;
    const int meta_light = (class_idx == 7 && tiny_c7_meta_light_enabled());
    void* block = NULL;
    if (page->free_list) {
        block = page->free_list;
        void* next = tiny_next_read(class_idx, block);
        page->free_list = next;
        atomic_store_explicit(&page->meta->freelist, next, memory_order_release);
    } else if (page->used < page->capacity) {
        size_t stride = hcls->stride;
        if (stride == 0) {
            stride = tiny_heap_block_stride(class_idx);
            hcls->stride = (uint16_t)stride;
        }
        block = (void*)(page->base + ((size_t)page->used * stride));
        if (page->meta->carved < page->capacity) {
            page->meta->carved++;
        }
    } else {
        return NULL;
    }

    page->used++;
    if (__builtin_expect(meta_light, 0)) {
        page->c7_used_delta++;
        page->c7_active_delta++;
        if (tiny_c7_delta_should_flush(page)) {
            tiny_c7_meta_flush_page(page);
        }
        return tiny_region_id_write_header(block, class_idx);
    }
    atomic_fetch_add_explicit(&page->meta->used, 1, memory_order_relaxed);
    ss_active_add(page->ss, 1u);

    return tiny_region_id_write_header(block, class_idx);
}

static inline void tiny_heap_page_push_free(int class_idx, tiny_heap_page_t* page, void* base_ptr) {
    if (!page || !base_ptr || !page->meta) return;
    tiny_next_write(class_idx, base_ptr, page->free_list);
    page->free_list = base_ptr;
    atomic_store_explicit(&page->meta->freelist, base_ptr, memory_order_release);
}

static inline void tiny_heap_page_free_local(tiny_heap_ctx_t* ctx,
                                             int class_idx,
                                             tiny_heap_page_t* page,
                                             void* base_ptr) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls || !page || !base_ptr) return;
    const int stats_on = (class_idx == 7 && tiny_c7_heap_stats_enabled());
    if (__builtin_expect(stats_on, 0)) {
        atomic_fetch_add_explicit(&g_c7_heap_stats.free_fast_local, 1, memory_order_relaxed);
    }
    const int meta_light = (class_idx == 7 && tiny_c7_meta_light_enabled());
    tiny_heap_page_push_free(class_idx, page, base_ptr);
    if (page->used > 0) {
        page->used--;
        if (!__builtin_expect(meta_light, 0)) {
            atomic_fetch_sub_explicit(&page->meta->used, 1, memory_order_relaxed);
            ss_active_dec_one(page->ss);
        } else {
            page->c7_used_delta--;
            page->c7_active_delta--;
            if (tiny_c7_delta_should_flush(page)) {
                tiny_c7_meta_flush_page(page);
            }
        }
    }

    if (page->used == 0) {
        tiny_heap_page_becomes_empty(ctx, class_idx, page);
    } else if (page->used < page->capacity && page->free_list) {
        tiny_heap_page_t* old_cur = hcls->current_page;
        tiny_heap_class_unlink(hcls, page);
        page->next = NULL;
        if (class_idx == 7) {
            hcls->current_page = page;
            if (old_cur && old_cur != page) {
                tiny_heap_page_push_to_partial(hcls, old_cur);
            }
        } else {
            if (!hcls->current_page) {
                hcls->current_page = page;
            } else {
                tiny_heap_page_push_to_partial(hcls, page);
            }
        }
    }
}

static inline tiny_heap_page_t* tiny_heap_prepare_page(tiny_heap_ctx_t* ctx, int class_idx) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    const int stats_on = (class_idx == 7 && tiny_c7_heap_stats_enabled());
    if (!hcls) {
        if (__builtin_expect(stats_on, 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_prepare_fail, 1, memory_order_relaxed);
        }
        return NULL;
    }

    tiny_heap_page_t* page = tiny_heap_take_current(hcls);
    if (page) return page;

    TinyTLSSlab* tls = &g_tls_slabs[class_idx];
    if (!tls->ss) {
        if (superslab_refill(class_idx) == NULL) {
            if (__builtin_expect(stats_on, 0)) {
                atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_prepare_fail, 1, memory_order_relaxed);
            }
            return NULL;
        }
    }
    tls = &g_tls_slabs[class_idx];  // superslab_refill で更新されるため再取得
    if (!tls->ss || !tls->meta) {
        if (__builtin_expect(stats_on, 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_prepare_fail, 1, memory_order_relaxed);
        }
        return NULL;
    }
    page = tiny_heap_attach_page(ctx, class_idx, tls->ss, tls->slab_idx);
    if (page) {
        tiny_heap_class_unlink(hcls, page);
        page->next = NULL;
        hcls->current_page = page;
    }
    return page;
}

// class_idx 固定での alloc ホットパス
static inline void* tiny_heap_alloc_slow_from_class(tiny_heap_ctx_t* ctx, int class_idx) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!hcls) return NULL;

    const int stats_on = (class_idx == 7 && tiny_c7_heap_stats_enabled());
    tiny_heap_page_t* page = tiny_heap_prepare_page(ctx, class_idx);
    if (!page) {
        if (__builtin_expect(stats_on, 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_fail, 1, memory_order_relaxed);
        }
        return NULL;
    }

    void* user = tiny_heap_page_pop(hcls, class_idx, page);
    if (user && page->used >= page->capacity && page->free_list == NULL) {
        tiny_heap_page_mark_full(hcls, page);
    }
    return user;
}

// class_idx 固定での alloc ホットパス
__attribute__((always_inline)) static inline void* tiny_heap_alloc_class_fast(tiny_heap_ctx_t* ctx,
                                                                               int class_idx,
                                                                               size_t size) {
    (void)size;
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    tiny_heap_page_t* page = hcls ? hcls->current_page : NULL;
    const int stats_on = (class_idx == 7 && tiny_c7_heap_stats_enabled());

    if (page) {
        if (page->free_list || page->used < page->capacity) {
            void* user = tiny_heap_page_pop(hcls, class_idx, page);
            if (user) {
                if (page->used >= page->capacity && page->free_list == NULL) {
                    tiny_heap_page_mark_full(hcls, page);
                }
                if (__builtin_expect(stats_on, 0)) {
                    atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_fast_current, 1, memory_order_relaxed);
                }
                return user;
            }
        }
    }

    if (__builtin_expect(stats_on, 0)) {
        atomic_fetch_add_explicit(&g_c7_heap_stats.alloc_slow_prepare, 1, memory_order_relaxed);
    }
    return tiny_heap_alloc_slow_from_class(ctx, class_idx);
}

// Superslab/Slab メタが既に分かっている場合の free（Gate から渡されるホットパス用）
static inline void tiny_heap_free_class_fast_with_meta(tiny_heap_ctx_t* ctx,
                                                       int class_idx,
                                                       SuperSlab* ss,
                                                       int slab_idx,
                                                       void* base) {
    tiny_heap_class_t* hcls = tiny_heap_class(ctx, class_idx);
    if (!base || !ss || slab_idx < 0 || !hcls) return;
    const int stats_on = (class_idx == 7 && tiny_c7_heap_stats_enabled());

    tiny_heap_page_t* page = tiny_heap_attach_page(ctx, class_idx, ss, slab_idx);
    if (!page) {
        if (__builtin_expect(stats_on, 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.free_slow_fallback, 1, memory_order_relaxed);
        }
        tiny_heap_cold_drain_and_free(class_idx, base);
        return;
    }

    tiny_heap_page_free_local(ctx, class_idx, page, base);
}

// class_idx 固定での free ホットパス（ptr は USER ポインタ）
static inline void tiny_heap_free_class_fast(tiny_heap_ctx_t* ctx, int class_idx, void* ptr) {
    if (!ptr) return;

#if HAKMEM_TINY_HEADERLESS
    void* base = ptr;
#else
    void* base = (void*)((uint8_t*)ptr - 1);
#endif
    SuperSlab* ss = hak_super_lookup(base);
    if (!ss || ss->magic != SUPERSLAB_MAGIC) {
        if (__builtin_expect((class_idx == 7 && tiny_c7_heap_stats_enabled()), 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.free_slow_fallback, 1, memory_order_relaxed);
        }
        tiny_heap_cold_drain_and_free(class_idx, base);
        return;
    }
    int slab_idx = slab_index_for(ss, base);
    if (slab_idx < 0 || slab_idx >= ss_slabs_capacity(ss)) {
        if (__builtin_expect((class_idx == 7 && tiny_c7_heap_stats_enabled()), 0)) {
            atomic_fetch_add_explicit(&g_c7_heap_stats.free_slow_fallback, 1, memory_order_relaxed);
        }
        tiny_heap_cold_drain_and_free(class_idx, base);
        return;
    }

    tiny_heap_free_class_fast_with_meta(ctx, class_idx, ss, slab_idx, base);
}