// hakmem_tiny_sfc.c - Box 5-NEW: Super Front Cache (SFC) Implementation
// Purpose: Slow path (refill/spill/config/stats), not inline
// Fast path is in tiny_alloc_fast_sfc.inc.h (inline)

#include "tiny_alloc_fast_sfc.inc.h"
#include "hakmem_tiny.h"
#include "hakmem_tiny_config.h"
#include "hakmem_tiny_superslab.h"
#include "hakmem_stats_master.h"  // Phase 4d: Master stats control
#include "tiny_tls.h"
#include "box/tls_sll_box.h"  // static inline tls_sll_pop/push API (Box TLS-SLL)
#include "hakmem_env_cache.h"      // Priority-2: ENV cache (eliminate syscalls)
#include <stdlib.h>
#include <string.h>
#include <stdio.h>

// ============================================================================
// Box 5-NEW: TLS Variables (defined here, extern in header)
// ============================================================================

__thread void* g_sfc_head[TINY_NUM_CLASSES] = {NULL};
__thread uint32_t g_sfc_count[TINY_NUM_CLASSES] = {0};
uint32_t g_sfc_capacity[TINY_NUM_CLASSES] = {0};  // Non-TLS: shared read-only config

// ============================================================================
// Box 5-NEW: Statistics (compile-time gated)
// ============================================================================

#if HAKMEM_DEBUG_COUNTERS
sfc_stats_t g_sfc_stats[TINY_NUM_CLASSES] = {0};
#endif

// ============================================================================
// Box 5-NEW: Global Config (from ENV)
// ============================================================================

int g_sfc_enabled = 1;  // Default: ON (bench-focused; A/B via HAKMEM_SFC_ENABLE)
int g_sfc_debug = 0;    // Set once at init from HAKMEM_SFC_DEBUG

static int g_sfc_default_capacity = SFC_DEFAULT_CAPACITY;
static int g_sfc_default_refill = SFC_DEFAULT_REFILL_COUNT;
static int g_sfc_default_spill_thresh = SFC_DEFAULT_SPILL_THRESH;

// Per-class overrides (0 = use default)
static int g_sfc_capacity_override[TINY_NUM_CLASSES] = {0};
static int g_sfc_refill_override[TINY_NUM_CLASSES] = {0};

// ============================================================================
// Box 5-NEW: Initialization
// ============================================================================

void sfc_init(void) {
    // Priority-2: Use cached ENV (eliminate init syscall overhead)
    g_sfc_debug = HAK_ENV_SFC_DEBUG();
    g_sfc_enabled = HAK_ENV_SFC_ENABLE();

    if (!g_sfc_enabled) {
        // SFC disabled, skip initialization
        return;
    }

    // Priority-2: Use cached ENV (eliminate config syscall overhead)
    int cap = HAK_ENV_SFC_CAPACITY();
    if (cap >= SFC_MIN_CAPACITY && cap <= SFC_MAX_CAPACITY) {
        g_sfc_default_capacity = cap;
    }

    int refill = HAK_ENV_SFC_REFILL_COUNT();
    if (refill >= 8 && refill <= 256) {
        g_sfc_default_refill = refill;
    }

    // Parse ENV: HAKMEM_SFC_CAPACITY_CLASS{0..7} (per-class capacity override)
    for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
        char var[64];
        snprintf(var, sizeof(var), "HAKMEM_SFC_CAPACITY_CLASS%d", cls);
        const char* env_cls_cap = getenv(var);
        if (env_cls_cap && *env_cls_cap) {
            int cap = atoi(env_cls_cap);
            if (cap >= SFC_MIN_CAPACITY && cap <= SFC_MAX_CAPACITY) {
                g_sfc_capacity_override[cls] = cap;
            }
        }
    }

    // Parse ENV: HAKMEM_SFC_REFILL_COUNT_CLASS{0..7} (per-class refill override)
    for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
        char var[64];
        snprintf(var, sizeof(var), "HAKMEM_SFC_REFILL_COUNT_CLASS%d", cls);
        const char* env_cls_refill = getenv(var);
        if (env_cls_refill && *env_cls_refill) {
            int refill = atoi(env_cls_refill);
            if (refill >= 8 && refill <= 256) {
                g_sfc_refill_override[cls] = refill;
            }
        }
    }

    // Initialize per-class capacities (use override or default)
    for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
        if (g_sfc_capacity_override[cls] > 0) {
            g_sfc_capacity[cls] = g_sfc_capacity_override[cls];
        } else {
            g_sfc_capacity[cls] = g_sfc_default_capacity;
        }
    }


    // Register shutdown hook for optional stats dump
    atexit(sfc_shutdown);

#if !HAKMEM_BUILD_RELEASE
    // One-shot debug log
    static int debug_printed = 0;
    if (!debug_printed) {
        debug_printed = 1;
        if (g_sfc_debug) {
            fprintf(stderr, "[SFC] Initialized: enabled=%d, default_cap=%d, default_refill=%d\n",
                    g_sfc_enabled, g_sfc_default_capacity, g_sfc_default_refill);
            for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
                if (g_sfc_capacity_override[cls] > 0 || g_sfc_refill_override[cls] > 0) {
                    fprintf(stderr, "[SFC] Class %d: cap=%u, refill_override=%d\n",
                            cls, g_sfc_capacity[cls], g_sfc_refill_override[cls]);
                }
            }
        }
    }
#endif

    // Ensure stats (if requested) are printed at process exit.
    // This is inexpensive and guarded inside sfc_shutdown by HAKMEM_SFC_STATS_DUMP.
    atexit(sfc_shutdown);
}

void sfc_shutdown(void) {
    // Optional: Print stats at exit (full stats when counters enabled)
    // Phase 4d: Now uses hak_stats_check() for unified stats control
    if (hak_stats_check("HAKMEM_SFC_STATS_DUMP", "sfc")) {
    #if HAKMEM_DEBUG_COUNTERS
        sfc_print_stats();
    #else
        // Minimal summary in release builds (no counters): capacity and current counts
        fprintf(stderr, "\n=== SFC Minimal Summary (release) ===\n");
        for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
            if (g_sfc_capacity[cls] == 0) continue;
            fprintf(stderr, "Class %d: cap=%u, count=%u\n",
                    cls, g_sfc_capacity[cls], g_sfc_count[cls]);
        }
        fprintf(stderr, "===========================\n\n");
    #endif
    }

    // No cleanup needed (TLS memory freed by OS)
}

// Cascade a first batch from TLS SLL into SFC after TLS prewarm.
// Hot classes only (0..3 and 5) to focus on 256B/小サイズ。
void sfc_cascade_from_tls_initial(void) {
    if (!g_sfc_enabled) return;
    // TLS SLL extern
    extern __thread TinyTLSSLL g_tls_sll[];
    for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
        if (!(cls <= 3 || cls == 5)) continue;  // focus: 8..64B and 256B
        uint32_t cap = g_sfc_capacity[cls];
        if (cap == 0) continue;
        // target: max half of SFC cap or available SLL count
        uint32_t avail = g_tls_sll[cls].count;
        if (avail == 0) continue;
        // Target: 75% of cap by default, bounded by available
        uint32_t target = (cap * 75u) / 100u;
        if (target == 0) target = (avail < 16 ? avail : 16);
        if (target > avail) target = avail;
        // transfer
        while (target-- > 0 && g_tls_sll[cls].count > 0 && g_sfc_count[cls] < g_sfc_capacity[cls]) {
            void* ptr = NULL;
            // pop one from SLL via Box TLS-SLL API (static inline)
            if (!tls_sll_pop(cls, &ptr)) break;
            // Phase E1-CORRECT: Use Box API for next pointer write
            tiny_next_write(cls, ptr, g_sfc_head[cls]);
            g_sfc_head[cls] = ptr;
            g_sfc_count[cls]++;
        }
    }
}

// ============================================================================
// Box 5-NEW: Refill (Slow Path) - STUB (real logic in hakmem.c)
// ============================================================================

// Stub - real implementation is inline in hakmem.c malloc() to avoid LTO issues
// This is just a placeholder for future modular refactoring
int sfc_refill(int cls, int target_count) {
    if (cls < 0 || cls >= TINY_NUM_CLASSES) return 0;
    if (!g_sfc_enabled) return 0;
    (void)target_count;

#if HAKMEM_DEBUG_COUNTERS
    g_sfc_stats[cls].refill_calls++;
#endif

    return 0;  // Actual refill happens inline in hakmem.c
}

// ============================================================================
// Box 5-NEW: Spill (Slow Path) - STUB (real logic in hakmem.c)
// ============================================================================

// Stub - real implementation is inline in hakmem.c free() to avoid LTO issues
// This is just a placeholder for future modular refactoring
int sfc_spill(int cls, int spill_count) {
    if (cls < 0 || cls >= TINY_NUM_CLASSES) return 0;
    if (!g_sfc_enabled) return 0;
    (void)spill_count;

#if HAKMEM_DEBUG_COUNTERS
    g_sfc_stats[cls].spill_calls++;
#endif

    return 0;  // Actual spill happens inline in hakmem.c
}

// ============================================================================
// Box 5-NEW: Configuration API
// ============================================================================

sfc_config_t sfc_get_config(int cls) {
    sfc_config_t cfg = {0};

    if (cls >= 0 && cls < TINY_NUM_CLASSES) {
        cfg.capacity = g_sfc_capacity[cls];

        // Refill count (use override or default)
        cfg.refill_count = (g_sfc_refill_override[cls] > 0)
                           ? g_sfc_refill_override[cls]
                           : g_sfc_default_refill;

        cfg.spill_thresh = g_sfc_default_spill_thresh;
    }

    return cfg;
}

void sfc_set_config(int cls, sfc_config_t cfg) {
    if (cls < 0 || cls >= TINY_NUM_CLASSES) return;

    // Validate capacity
    if (cfg.capacity >= SFC_MIN_CAPACITY && cfg.capacity <= SFC_MAX_CAPACITY) {
        g_sfc_capacity[cls] = cfg.capacity;
    }

    // Validate refill count
    if (cfg.refill_count >= 8 && cfg.refill_count <= 256) {
        g_sfc_refill_override[cls] = cfg.refill_count;
    }

    // Spill threshold (future use)
    if (cfg.spill_thresh > 0 && cfg.spill_thresh <= 100) {
        // Currently unused
    }
}

// ============================================================================
// Box 5-NEW: Statistics API
// ============================================================================

#if HAKMEM_DEBUG_COUNTERS

sfc_stats_t sfc_get_stats(int cls) {
    sfc_stats_t stats = {0};

    if (cls >= 0 && cls < TINY_NUM_CLASSES) {
        stats = g_sfc_stats[cls];
    }

    return stats;
}

void sfc_reset_stats(int cls) {
    if (cls >= 0 && cls < TINY_NUM_CLASSES) {
        memset(&g_sfc_stats[cls], 0, sizeof(sfc_stats_t));
    }
}

void sfc_print_stats(void) {
    fprintf(stderr, "\n=== SFC Statistics (Box 5-NEW) ===\n");

    uint64_t total_alloc_hits = 0;
    uint64_t total_alloc_misses = 0;
    uint64_t total_refill_calls = 0;
    uint64_t total_refill_blocks = 0;

    for (int cls = 0; cls < TINY_NUM_CLASSES; cls++) {
        sfc_stats_t* s = &g_sfc_stats[cls];

        uint64_t total_allocs = s->alloc_hits + s->alloc_misses;
        if (total_allocs == 0) continue;  // Skip unused classes

        total_alloc_hits += s->alloc_hits;
        total_alloc_misses += s->alloc_misses;
        total_refill_calls += s->refill_calls;
        total_refill_blocks += s->refill_blocks;

        double hit_rate = (double)s->alloc_hits / total_allocs * 100.0;
        double refill_freq = (double)s->refill_calls / total_allocs * 100.0;

        fprintf(stderr, "Class %d (%3zu B): allocs=%llu, hit_rate=%.2f%%, "
                        "refills=%llu (%.4f%%), spills=%llu, cap=%u\n",
                cls, g_tiny_class_sizes[cls],
                (unsigned long long)total_allocs, hit_rate,
                (unsigned long long)s->refill_calls, refill_freq,
                (unsigned long long)s->spill_calls,
                g_sfc_capacity[cls]);
    }

    // Summary
    uint64_t grand_total = total_alloc_hits + total_alloc_misses;
    if (grand_total > 0) {
        double overall_hit_rate = (double)total_alloc_hits / grand_total * 100.0;
        double overall_refill_freq = (double)total_refill_calls / grand_total * 100.0;

        fprintf(stderr, "\n=== SFC Summary ===\n");
        fprintf(stderr, "Total allocs:    %llu\n", (unsigned long long)grand_total);
        fprintf(stderr, "Overall hit rate: %.2f%% (target: >95%%)\n", overall_hit_rate);
        fprintf(stderr, "Refill frequency: %.4f%% (target: <0.03%%)\n", overall_refill_freq);
        fprintf(stderr, "Refill calls:    %llu (target: <50K for 4M ops/s workload)\n",
                (unsigned long long)total_refill_calls);
        fprintf(stderr, "Refill blocks:   %llu (avg %.1f blocks/refill)\n",
                (unsigned long long)total_refill_blocks,
                total_refill_calls > 0 ? (double)total_refill_blocks / total_refill_calls : 0.0);

        // Check targets
        if (overall_hit_rate >= 95.0) {
            fprintf(stderr, "✅ Hit rate target achieved!\n");
        } else {
            fprintf(stderr, "⚠️  Hit rate below target (increase capacity?)\n");
        }

        if (overall_refill_freq < 0.03) {
            fprintf(stderr, "✅ Refill frequency target achieved (-98.5%% reduction)!\n");
        } else {
            fprintf(stderr, "⚠️  Refill frequency above target (increase refill_count?)\n");
        }
    }

    fprintf(stderr, "===========================\n\n");
}

#endif // HAKMEM_DEBUG_COUNTERS

// ============================================================================
// End of hakmem_tiny_sfc.c
// ============================================================================