Linux GLX, a lot of code shrinkage/cleanup, assembly support.

[xonotic/darkplaces.git] / cpu_x86.nasm

diff --git a/cpu_x86.nasm b/cpu_x86.nasm
new file mode 100644 (file)
index 0000000..aa68a74
--- /dev/null
+++ b/cpu_x86.nasm
@@ -0,0 +1,201 @@
+          BITS 32
+          GLOBAL _Mod_PointInLeaf
+          GLOBAL _SV_HullPointContents
+          SECTION .text
+_Mod_PointInLeaf
+
+;{
+
+        mov     eax, dword [esp+12-4] ; model
+       sub     esp, 4
+       push    ebx
+       push    esi
+
+;       mnode_t         *node;
+;       node = model->nodes;
+
+        mov     esi, dword [eax+200]
+
+;       if (node->contents < 0)
+
+        cmp     dword [esi], 0
+        jge     .firstvalid
+
+;               return (mleaf_t *)node;
+
+       mov     eax, esi
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0
+.firstvalid
+        mov     edx, dword [esp+8+8] ; p
+.loop
+
+;       while (1)
+
+        mov     eax, dword [esi+36]
+        mov     cl, byte [eax+16]
+
+;       {
+;               node = node->children[(node->plane->type < 3 ? p[node->plane->type] : DotProduct (p,node->plane->normal)) < node->plane->dist];
+
+       cmp     cl, 3
+        jb      .axisplane
+        fld     dword [eax+4]
+        fmul    dword [edx+4]
+        fld     dword [eax+8]
+        fmul    dword [edx+8]
+        fld     dword [eax]
+        fmul    dword [edx]
+        faddp   st1, st0
+        faddp   st1, st0
+        fld     dword [eax+12]
+        fcompp
+       xor     ecx, ecx
+       fnstsw  ax
+       test    ah, 65                                  ; 00000041H
+       sete    cl
+        mov     esi, dword [esi+ecx*4+40]
+
+;               if (node->contents < 0)
+
+        cmp     dword [esi], 0
+        jge     .loop
+
+;                       return (mleaf_t *)node;
+
+       mov     eax, esi
+
+;       }
+;       return NULL;    // never reached
+;}
+
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0
+.axisplane:
+       xor     ebx, ebx
+       mov     bl, cl
+        fld     dword [edx+ebx*4]
+        fld     dword [eax+12]
+        fcompp
+       xor     ecx, ecx
+       fnstsw  ax
+       test    ah, 65                                  ; 00000041H
+       sete    cl
+        mov     esi, dword [esi+ecx*4+40]
+
+;               if (node->contents < 0)
+
+        cmp     dword [esi], 0
+        jge     .loop
+
+;                       return (mleaf_t *)node;
+
+       mov     eax, esi
+
+;       }
+;       return NULL;    // never reached
+;}
+
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0
+
+
+_SV_HullPointContents
+
+;{
+        mov     ecx, [esp+12-4] ; num
+       sub     esp, 4
+       test    ecx, ecx
+        nop                     ; padding
+       push    ebx
+       push    esi
+       push    edi
+       push    ebp
+
+;       while (num >= 0)
+
+        jge     .firstvalid
+;       return num;
+       mov     eax, ecx
+       pop     ebp
+;}
+       pop     edi
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0
+.firstvalid
+        mov     eax, [esp+8+16] ; hull
+        mov     edx, [esp+16+16] ; p
+        mov     esi, [eax]
+        mov     edi, [eax+4]
+.loop
+        mov     eax, [esi+ecx*8]
+        lea     ebx, [eax+eax*2]
+        xor     eax, eax
+        mov     al, [edi+ebx*8+16]
+        lea     ebp, [edi+ebx*8]
+
+;               num = hull->clipnodes[num].children[(hull->planes[hull->clipnodes[num].planenum].type < 3 ? p[hull->planes[hull->clipnodes[num].planenum].type] : DotProduct (hull->planes[hull->clipnodes[num].planenum].normal, p)) < hull->planes[hull->clipnodes[num].planenum].dist];
+
+       cmp     al, 3
+        jb      .axisplane
+        fld     dword [edx+8]
+        fmul    dword [ebp+8]
+        fld     dword [edx+4]
+        fmul    dword [ebp+4]
+        fld     dword [edx]
+        fmul    dword [ebp]
+        faddp   st1, st0
+        faddp   st1, st0
+        fstp    dword [esp-4+20]
+
+        fld     dword [ebp+12]
+        fcomp   dword [esp-4+20]
+       xor     ebx, ebx
+       fnstsw  ax
+       test    ah, 65                                  ; 00000041H
+       sete    bl
+        lea     eax, [ebx+ecx*4]
+        movsx   ecx, word [esi+eax*2+4]
+       test    ecx, ecx
+        jge     .loop
+;       return num;
+       mov     eax, ecx
+       pop     ebp
+;}
+       pop     edi
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0
+
+.axisplane
+        mov     eax, [edx+eax*4]
+        mov     [esp-4+20], eax
+
+        fld     dword [ebp+12]
+        fcomp   dword [esp-4+20]
+       xor     ebx, ebx
+       fnstsw  ax
+       test    ah, 65                                  ; 00000041H
+       sete    bl
+        lea     eax, [ebx+ecx*4]
+        movsx   ecx, word [esi+eax*2+4]
+       test    ecx, ecx
+        jge     .loop
+;       return num;
+       mov     eax, ecx
+       pop     ebp
+;}
+       pop     edi
+       pop     esi
+       pop     ebx
+       add     esp, 4
+       ret     0