From 3e4e90eadba2f39a4c81351ccadd085e74d4d7e8 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 5 Nov 2025 05:10:02 +0000
Subject: [PATCH] Phase 6-5: Entry Point Optimization (Phase 1) - Unexpected
 results
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Implementation: Move HAKMEM_TINY_FAST_PATH check BEFORE all guard checks
in malloc(), inspired by mimalloc/tcache entry point design.

Strategy:
- tcache has 0 branches before fast path
- mimalloc has 1-2 branches before fast path
- Old HAKMEM had 8+ branches before fast path
- Phase 1: Move fast path to line 1, add branch prediction hints

Changes in core/hakmem.c:
1. Fast Path First: Size check → Init check → Cache hit (3 branches)
2. Slow Path: All guards moved after fast path (rare cases)
3. Branch hints: __builtin_expect() for hot paths

Expected results (from research):
- ST: 0.46M → 1.4-2.3M ops/s (+204-400%)
- MT: 1.86M → 3.7-5.6M ops/s (+99-201%)

Actual results (Larson 2s 8-128B 1024):
- ST: 0.377M → 0.424M ops/s (+12% only)
- MT: 1.856M → 1.453M ops/s (-22% regression!)

Analysis:
- Similar pattern to previous Option A test (+42% ST, -20% MT)
- Entry point reordering alone is insufficient
- True bottleneck may be:
  1. tiny_fast_alloc() internals (size-to-class, cache access)
  2. Refill cost (1,600 cycles for 16 individual calls)
  3. Need Batch Refill optimization (Phase 3) as priority

Next steps:
- Investigate refill bottleneck with perf profiling
- Consider implementing Phase 3 (Batch Refill) before Phase 2
- May need combination of multiple optimizations for breakthrough

Related: LARSON_PERFORMANCE_ANALYSIS_2025_11_05.md
---
 core/hakmem.c | 53 +++++++++++++++++++++++++++++++--------------------
 1 file changed, 32 insertions(+), 21 deletions(-)

diff --git a/core/hakmem.c b/core/hakmem.c
index 7d654c78..71ce5a9b 100644
--- a/core/hakmem.c
+++ b/core/hakmem.c
@@ -1248,6 +1248,38 @@ void* realloc(void* ptr, size_t size) {
 
 // malloc wrapper - intercepts system malloc() calls
 void* malloc(size_t size) {
+    // ========================================================================
+    // Phase 6-5: ULTRA-FAST PATH FIRST (mimalloc/tcache style)
+    // Inspired by research: tcache has 0 branches, mimalloc has 1-2 branches
+    // Key insight: Move fast path BEFORE all guards (common case optimization)
+    // ========================================================================
+#ifdef HAKMEM_TINY_FAST_PATH
+    // Branch 1: Size check (predicted taken for tiny allocations)
+    if (__builtin_expect(size <= TINY_FAST_THRESHOLD, 1)) {
+        extern void* tiny_fast_alloc(size_t);
+        extern void tiny_fast_init(void);
+        extern __thread int g_tiny_fast_initialized;
+
+        // Branch 2: Initialization check (predicted taken after first call)
+        if (__builtin_expect(g_tiny_fast_initialized, 1)) {
+            // Branch 3: Cache hit check (predicted taken ~90% of time)
+            void* ptr = tiny_fast_alloc(size);
+            if (__builtin_expect(ptr != NULL, 1)) {
+                return ptr;  // ✅ FAST PATH: 3 branches total (vs tcache's 0, mimalloc's 1-2)
+            }
+            // Cache miss: fall through to slow path refill
+        } else {
+            // Cold path: initialize once per thread (rare)
+            tiny_fast_init();
+            void* ptr = tiny_fast_alloc(size);
+            if (ptr) return ptr;
+        }
+    }
+#endif
+    // ========================================================================
+    // SLOW PATH: All guards moved here (only executed on fast path miss)
+    // ========================================================================
+
     // Recursion guard: if we're inside the allocator already, fall back to libc
     if (g_hakmem_lock_depth > 0) {
         // Nested call detected - fallback to system malloc
@@ -1288,27 +1320,6 @@ void* malloc(size_t size) {
         }
     }
 
-    // ========================================================================
-    // Phase 6-3: Tiny Fast Path (System tcache style, 3-4 instruction fast path)
-    // ========================================================================
-#ifdef HAKMEM_TINY_FAST_PATH
-    if (size <= TINY_FAST_THRESHOLD) {
-        // Ultra-simple TLS cache pop (bypasses Magazine/SuperSlab)
-        extern void* tiny_fast_alloc(size_t);
-        extern void tiny_fast_init(void);
-        extern __thread int g_tiny_fast_initialized;
-
-        if (__builtin_expect(!g_tiny_fast_initialized, 0)) {
-            tiny_fast_init();
-        }
-
-        void* ptr = tiny_fast_alloc(size);
-        if (ptr) return ptr;
-        // Fall through to slow path on failure
-    }
-#endif
-    // ========================================================================
-
     // First-level call: enter allocator (no global lock)
     g_hakmem_lock_depth++;
     void* ptr = hak_alloc_at(size, HAK_CALLSITE());