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/profiling.asm

@@ -28,14 +28,12 @@ prof_rdtsc PROC
 prof_rdtsc ENDP
 
 ; uint64_t prof_get_freq(void) -> RAX
-; 获取 CPU 频率 (通过 OS 或推断)
+; 获取 CPU 频率 (Hz), 使用 QPC 校准 TSC
 prof_get_freq PROC
     ; 检查是否已初始化
     cmp     byte ptr [cpu_freq_init], 0
     jnz     prof_freq_return
 
-    ; 初始化 CPU 频率
-    ; 方法: 使用 QueryPerformanceCounter 测量 RDTSC
     push    rbx
     push    rdi
     push    rsi
@@ -46,7 +44,7 @@ prof_get_freq PROC
     call    QueryPerformanceFrequency
 
     ; 获取起始 QPC 和 TSC
-    lea     rcx, [rsp + 16]
+    lea     rcx, [rsp + 24]
     call    QueryPerformanceCounter
     rdtscp
     shl     rdx, 32
@@ -54,37 +52,31 @@ prof_get_freq PROC
     mov     rbx, rax                        ; rbx = 起始 TSC
 
     ; 等待约 10ms
-    mov     rcx, 10000                      ; 10,000 微秒
+    mov     rcx, 10                         ; 10 毫秒
     call    Sleep
 
     ; 获取结束 QPC 和 TSC
-    lea     rcx, [rsp]
+    lea     rcx, [rsp + 16]
     call    QueryPerformanceCounter
     rdtscp
     shl     rdx, 32
     or      rax, rdx
     mov     rdi, rax                        ; rdi = 结束 TSC
 
-    ; 计算经过的 QPC ticks
-    mov     rax, [rsp]
-    sub     rax, [rsp + 16]                 ; qpc_diff
+    ; qpc_diff = end_qpc - start_qpc
+    mov     rax, [rsp + 16]
+    sub     rax, [rsp + 24]
+    mov     rsi, rax                        ; rsi = qpc_diff
 
-    ; 计算 QPC 频率 (ticks/sec)
-    mov     rsi, [rsp + 32]
+    ; tsc_diff = end_tsc - start_tsc
+    mov     rcx, rdi
+    sub     rcx, rbx                        ; rcx = tsc_diff
 
-    ; TSC diff
-    mov     rax, rdi
-    sub     rax, rbx                        ; tsc_diff
-
-    ; CPU freq = tsc_diff * qpc_freq / qpc_diff
-    ; 但 qpc_diff 对应 10ms, 所以要 * 100
-    mov     rdx, rsi
-    mul     rdx                              ; tsc_diff * qpc_freq
-    ; rdx:rax = result
-    div     qword ptr [rsp + 16]            ; / qpc_diff
-    ; rax = freq * 100 (因为等待了 10ms)
-    mov     rcx, 100
-    mul     rcx
+    ; cpu_freq = tsc_diff * qpc_freq / qpc_diff
+    mov     rax, rcx                        ; rax = tsc_diff
+    mov     rdx, [rsp + 32]                 ; rdx = qpc_freq
+    mul     rdx                             ; rdx:rax = tsc_diff * qpc_freq
+    div     rsi                             ; rax = freq (Hz)
 
     mov     [cpu_freq], rax
     mov     byte ptr [cpu_freq_init], 1
@@ -100,33 +92,17 @@ prof_freq_return:
 prof_get_freq ENDP
 
 ; double prof_cycles_to_ms(uint64_t cycles) -> XMM0
-; 将 TSC cycles 转换为毫秒
+; 将 TSC cycles 转换为毫秒: ms = cycles * 1000 / freq
 prof_cycles_to_ms PROC
-    ; 先获取频率
-    call    prof_get_freq
-    ; rax = CPU freq (Hz)
+    mov     rdx, rcx                        ; 保存 cycles
+    call    prof_get_freq                   ; rax = CPU freq (Hz)
 
-    ; ms = cycles / (freq / 1000)
-    ; 或: ms = cycles * 1000 / freq
-    mov     rdx, 1000
-    mul     rdx                              ; freq * 1000
-    ; rdx:rax = freq * 1000
-
-    ; 转换为 double
-    cvtsi2sd xmm0, qword ptr [rsp - 8]     ; 需要栈存储
-    ; 简化: 使用整数除法再转 double
-    mov     rcx, rax
-    mov     rax, [cpu_freq]
-    ; 重新计算
-    mov     rax, rcx
-    ; cycles * 1000 / freq
-    ; 已经: rax = freq * 1000, 需要 cycles * 1000 / freq
-    ; 调整: cycles * 1000 / freq = cycles / (freq/1000)
-
-    ; 简化实现
-    cvtsi2sd xmm0, rcx                      ; cycles * 1000 (转 double)
-    cvtsi2sd xmm1, rax                      ; freq (转 double)
-    divsd   xmm0, xmm1                      ; (cycles*1000) / freq = ms
+    cvtsi2sd xmm0, rdx                      ; xmm0 = cycles (double)
+    mov     rcx, 1000
+    cvtsi2sd xmm1, rcx                      ; xmm1 = 1000.0
+    mulsd   xmm0, xmm1                      ; cycles * 1000
+    cvtsi2sd xmm1, rax                      ; xmm1 = freq
+    divsd   xmm0, xmm1                      ; ms
     ret
 prof_cycles_to_ms ENDP