feat: major engine overhaul - scene demo with lighting, sphere, ground

- Scene Demo: cube + sphere + ground plane with proper lighting
- Audio: XAudio2 dynamic loading + WAV playback
- Physics: expanded collision system
- Renderer: improved pipeline with materials
- UI: extended widget system
- VisualScript: enhanced serializer + block interactions
- ASM: optimized math/memory/profiling routines
- Window: mouse input + scroll support

Engine/ASM/string_ops.asm

@@ -255,42 +255,21 @@ fast_memset ENDP
 ; uint64_t fast_strlen(const char* str) -> RAX
 ; 使用 SSE4.2 PCMPESTRI 搜索 null terminator
 fast_strlen PROC
-    mov     rax, rcx                        ; rax = str
-    xor     rcx, rcx                        ; rcx = 计数器
-
-    ; 每次检查 16 字节
+    mov     rdx, rcx                        ; rdx = original str
     pxor    xmm0, xmm0                      ; 搜索 0
 
 strlen_sse_loop:
-    ; 检查是否接近字符串末尾 (避免越界读取)
-    ; 简化: 逐 16 字节检查
-    movdqu  xmm1, xmmword ptr [rax + rcx]
-    pcmpestrm xmm0, xmm1, 00001000b        ; 搜索 0, 返回掩码
+    mov     rax, rcx                        ; rax = 当前 16 字节块起始
+    movdqu  xmm1, xmmword ptr [rcx]
+    pcmpestri xmm0, xmm1, 00001000b        ; 搜索 0, CF=1 表示找到, ECX=偏移
     jc      strlen_found
-
     add     rcx, 16
     jmp     strlen_sse_loop
 
 strlen_found:
-    ; 获取第一个匹配位置
-    pcmpestri xmm0, xmm1, 00001000b       ; ECX = 匹配位置
+    ; rax = 块起始, rcx (ecx) = 块内偏移
     add     rax, rcx
-    add     rax, rcx
-    ; 需要计算实际位置
-    ; 简化: 使用标量回退
-    mov     rax, rcx
-    ret
-
-    ; 标量回退
-strlen_scalar:
-    mov     rax, rcx
-strlen_scalar_loop:
-    cmp     byte ptr [rax], 0
-    je      strlen_scalar_done
-    inc     rax
-    jmp     strlen_scalar_loop
-strlen_scalar_done:
-    sub     rax, rcx
+    sub     rax, rdx
     ret
 fast_strlen ENDP