clean up

2004-05-04 17:04:31 +00:00
parent 099a207814
commit 8a883aaec9
4 changed files with 102 additions and 148 deletions
--- a/glide3x/h5/glide3/src/xdraw2.asm
+++ b/glide3x/h5/glide3/src/xdraw2.asm
@@ -19,6 +19,9 @@
 ;; $Header$
 ;; $Revision$
 ;; $Log$
+;; Revision 1.1.8.9  2003/11/07 13:38:39  dborca
+;; unite the clans
+;;
 ;; Revision 1.1.8.8  2003/09/12 05:11:45  koolsmoky
 ;; preparing for graphic context checks
 ;; fixed jmp errors
@@ -392,9 +395,15 @@ PROC_TYPE win_nocull_valid

 endp

-%IFDEF       GL_AMD3D
-            ALIGN   32
-proc _trisetup_clip_coor_thunk, 12
+%IFDEF GL_AMD3D GL_SSE
+
+  ALIGN   32
+
+%ifdef GL_AMD3D
+  proc _trisetup_clip_coor_thunk, 12
+%else ; GL_SSE
+  proc _trisetup_SSE_clip_coor_thunk, 12
+%endif

 %define procPtr eax
 %define vPtr    ecx
@@ -436,50 +445,4 @@ __clipSpace:
    ret                         ; pop 3 dwords (vertex addrs) and return
 endp

-%ENDIF ; GL_AMD3D
-
-%IFDEF GL_SSE
-            ALIGN   32
-proc _trisetup_SSE_clip_coor_thunk, 12
-
-%define procPtr eax
-%define vPtr    ecx
-%define gc      edx           ; Current graphics context passed implicitly through edx
-
-%IFDEF GLIDE_ALT_TAB
-    test    gc, gc
-    je      .__contextLost
-;    mov     eax, [gc + windowed]
-;    test    eax, 1
-;    jnz     .pastContextTest
-    mov     eax, DWORD [gc+lostContext]
-    mov     ecx, [eax]
-    test    ecx, 1
-    jnz     .__contextLost
-;.pastContextTest:
-%ENDIF
-    
-    ;; Call through to the gc->curArchProcs.drawTrianglesProc w/o
-    ;; adding extra stuff to the stack. I wish we could actually
-    ;; do a direct return here w/o too much work.
-    lea     vPtr, [esp + _va$ - STKOFF]         ; Get vertex pointer address
-    mov     procPtr, [gc + drawTrianglesProc]   ; Prefetch drawTriangles proc addr
-
-    ;; If debugging make sure that we're in clip coordinates
-%IFDEF GLIDE_DEBUG
-    test    dword [gc + CoordinateSpace], 1
-    jnz     __clipSpace
-    xor     eax, eax
-    mov     [eax], eax
-__clipSpace:    
-%ENDIF ; GLIDE_DEBUG
-
-    invoke  procPtr, 1, 3, vPtr ; (*gc->curArchProcs.drawTrianglesProc)(grDrawVertexArray, 3, vPtr)
-
-%IFDEF GLIDE_ALT_TAB
-.__contextLost:
-%ENDIF
-    ret                         ; pop 3 dwords (vertex addrs) and return
-endp
-
-%ENDIF ; GL_SSE
+%ENDIF ; GL_AMD3D GL_SSE
--- a/glide3x/h5/glide3/src/xdraw2.inc
+++ b/glide3x/h5/glide3/src/xdraw2.inc
@@ -20,6 +20,9 @@
 ;; $Header$
 ;; $Revision$
 ;; $Log$
+;; Revision 1.1.8.8  2003/11/07 13:38:39  dborca
+;; unite the clans
+;;
 ;; Revision 1.1.8.7  2003/09/12 05:11:45  koolsmoky
 ;; preparing for graphic context checks
 ;; fixed jmp errors
@@ -52,19 +55,6 @@
    ;;
    ;;      YOU HAVE BEEN WARNED    

-; Ugly, but seems to workaround the problem with locally defined
-; data segment globals not getting relocated properly when using
-; djgpp.
-
-%define zArea   One+04h
-%define zdxAB   One+08h
-%define zdxBC   One+0ch
-%define zdyAB   One+10h
-%define zdyBC   One+14h
-%define zculltest One+18h
-
-%define gc      esi       ; points to graphics context ;; ZZZ double defined?
-
 %ifdef GL_AMD3D

 ;---------------------------------------------------------------------------
@@ -75,6 +65,7 @@
 %define fifo    ebp           ; points to fifo entries
 %define tempVal esi

+
 %IF GLIDE_CLIP_COORDS
    ;; NB:  Currently, the 3DNow!(tm) clip coordinate stuff
    ;;      thunks through to the grDrawTriangles functions
@@ -197,6 +188,12 @@

    movd      intArea, mm4          ; vectored !

+;; E3DN
+;    pswapd    mm2, mm2              ; dxBC | dyBC
+;    pfmul     mm0, mm2              ; dxBC*dyAB | dxAB*dyBC
+;    pfnacc    mm0, mm0              ; dxAB*dyBC - dxBC*dyAB | dxAB*dyBC - dxBC*dyAB
+;    movd      intArea, mm0          ; vectored !
+
    ; Zero Area Triangle Check

    test      intArea, 7fffffffh    ; if ((j & 0x7FFFFFFF) == 0)
@@ -313,9 +310,7 @@
    ALIGN 32                        

 .__triBegin:
-    mov       eax, [gc+triPacketHdr]; Packet 3 header
    lea       dlp,[gc + tsuDataList]; Reset the dataList
-
    mov       fifo, [gc + fifoPtr]  ; Fetch Fifo Ptr
    mov       vertex, [esp + _va$]  ; Current vertex = A

@@ -323,6 +318,7 @@
    test      fifo, 4               ; is fifo pointer qword aligned ?

    jz        .__fifo_aligned       ; yes, it is qword aligned
+    mov       eax, [gc+triPacketHdr]; Packet 3 header
    movq      mm1, [vertex+X]       ; y | x

    GR_FIFO_WRITE fifo, 0, eax      ; write header to fifo; now qword aligned
@@ -640,10 +636,7 @@
 ; end 3DNow! version
 ;---------------------------------------------------------------------------

-%endif ; GL_AMD3D
-
-
-%ifdef GL_SSE
+%elifdef GL_SSE

 ;---------------------------------------------------------------------------
 ; start SSE version - note: CSIM incomplete.
@@ -653,6 +646,7 @@
 %define gc      edi           ; points to graphics context
 %define fifo    ebp           ; points to fifo entries
 %define tempVal esi
+%define zArea   One+04h

 %IF GLIDE_CLIP_COORDS
    ;; NB:  Currently, the 3DNow!(tm) clip coordinate stuff
@@ -746,31 +740,24 @@

    jz        .nocull               ; nope, no culling
    mov       fa, [esp + _va$]      ; get base address of vertex A
-
-    movlps    xmm2,[fc + X]         ; 0 | 0 | yc | xc
    shl       cull, 31              ; culltest << 31

-    movlps    xmm1,[fb + X]         ; 0 | 0 | yb | xb
-    add       tempVal, 4            ; space required in fifo
+    movhps    xmm1,[fc + X]         ; yc | xc | * | *
+    movhps    xmm0,[fb + X]         ; yb | xb | * | *
+    movlps    xmm1,[fb + X]         ; yc | xc | yb | xb
+    movlps    xmm0,[fa + X]         ; yb | xb | ya | xa

-    movlps    xmm0,[fa + X]         ; 0 | 0 | ya | xa
+    add       tempVal, 4            ; space required in fifo
    mov       ebx, [gc + fifoRoom]  ; space available in fifo

    ;; Area_Computation

-    subps     xmm0,xmm1             ; 0 | 0 | dyAB | dxAB
-    subps     xmm1,xmm2             ; 0 | 0 | dyBC | dxBC
-
-    unpcklps  xmm0,xmm3             ;  0  | dyAB | 0  | dxAB
-    movhlps   xmm3,xmm0             ;  0  | 0    | 0  | dyAB
-    unpcklps  xmm1,xmm2             ;  yc | dyBC | xc | dxBC
-    movhlps   xmm2,xmm1             ;  0  | 0    | yc | dyBC
-
-    mulss     xmm0,xmm2             ; 0 | 0 | 0 | dxAB*dyBC
-    mulss     xmm1,xmm3             ; 0 | 0 | 0 | dyAB*dxBC
-
-    subss     xmm0,xmm1             ; 0 | 0 | 0 | dxAB*dyBC - dxBC*dyAB
-
+    subps     xmm0,xmm1             ; dyBC | dxBC | dyAB | dxAB
+    movaps    xmm1,xmm0             ; dyBC | dxBC | dyAB | dxAB
+    shufps    xmm0,xmm0,27          ; dxAB | dyAB | dxBC | dyBC
+    mulps     xmm0,xmm1             ; dxAB*dyBC | dxBC*dyAB | dxBC*dyAB | dxAB*dyBC
+    movhlps   xmm1,xmm0             ; dyBC | dxBC | dxAB*dyBC | dxBC*dyAB
+    subss     xmm0,xmm1             ; dxAB*dyBC | dxBC*dyAB | dxBC*dyAB | dxAB*dyBC - dxBC*dyAB
    movss     [zArea],xmm0          ; dxAB*dyBC - dxBC*dyAB
    mov       intArea,[zArea]       ; vectored !

@@ -895,9 +882,10 @@
    mov       vertex, [esp + _va$]  ; Current vertex = A

    mov       dlpstrt, dlp          ; save pointer to start of dataList
-    test      fifo, 4               ; is fifo pointer qword aligned ?

-    jz        .__fifo_aligned       ; yes, it is qword aligned
+;    test      fifo, 4               ; is fifo pointer qword aligned ?
+;    jz        .__fifo_aligned       ; yes, it is qword aligned
+
    movlps    xmm1,[vertex+X]       ; 0 | 0 | y | x

    GR_FIFO_WRITE fifo, 0, eax      ; write header to fifo; now qword aligned
@@ -934,40 +922,40 @@

    jmp       .__paramLoopDoneWBzero1; write buffer empty

-.__fifo_aligned:
-    movss     xmm2,[vertex+X]       ; 0 | 0 | 0 | x of vertex A
-
-    GR_FIFO_WRITE fifo, 0, eax      ; write header to fifo
-
-    add       fifo, 8               ; fifoPtr += 2*sizeof(FxU32)
-
-    movss     [fifo-4],xmm2         ; PCI write x - not CSIM compatible!!!!
-    movss     xmm1,[vertex+Y]       ; 0 | 0 | 0 | y of vertex A
-
-    mov       eax, [dlp]            ; get first offset from the data list
-    add       dlp, 4                ; dlp++
-
-    test      eax, eax              ; end of list ?
-    jz        .__paramLoopDoneWBone1; yes, "write buffer" has y data
-
-.__paramLoop1b:
-    movss     xmm2,[eax+vertex]     ; 0 | 0 | 0 | get next parameter
-    add       fifo, 8               ; fifoPtr += 2*sizeof(FxU32)
-
-    mov       eax, [dlp]            ; offset = *(dlp + 1)
-    add       dlp, 8                ; dlp += 2
-
-    unpcklps  xmm1,xmm2             ; 0 | 0 | current param | previous param
-    test      eax, eax              ; at end of offset list (offset == 0) ?
-  
-    WRITE_MM1_FIFO_ALIGNED -8       ; PCI write current param | previous param
-    jz        .__paramLoopDoneWBzero1; exit, "write buffer" empty
-
-    movss     xmm1,[eax+vertex]     ; 0 | 0 | 0 | get next parameter
-    mov       eax, [dlp-4]          ; offset = *(dlp + 1)
-
-    test      eax, eax              ; at end of offset list (offset == 0) ?
-    jnz       .__paramLoop1b        ; nope, copy next parameter
+;.__fifo_aligned:
+;    movss     xmm2,[vertex+X]       ; 0 | 0 | 0 | x of vertex A
+;
+;    GR_FIFO_WRITE fifo, 0, eax      ; write header to fifo
+;
+;    add       fifo, 8               ; fifoPtr += 2*sizeof(FxU32)
+;
+;    movss     [fifo-4],xmm2         ; PCI write x - not CSIM compatible!!!!
+;    movss     xmm1,[vertex+Y]       ; 0 | 0 | 0 | y of vertex A
+;
+;    mov       eax, [dlp]            ; get first offset from the data list
+;    add       dlp, 4                ; dlp++
+;
+;    test      eax, eax              ; end of list ?
+;    jz        .__paramLoopDoneWBone1; yes, "write buffer" has y data
+;
+;.__paramLoop1b:
+;    movss     xmm2,[eax+vertex]     ; 0 | 0 | 0 | get next parameter
+;    add       fifo, 8               ; fifoPtr += 2*sizeof(FxU32)
+;
+;    mov       eax, [dlp]            ; offset = *(dlp + 1)
+;    add       dlp, 8                ; dlp += 2
+;
+;    unpcklps  xmm1,xmm2             ; 0 | 0 | current param | previous param
+;    test      eax, eax              ; at end of offset list (offset == 0) ?
+;  
+;    WRITE_MM1_FIFO_ALIGNED -8       ; PCI write current param | previous param
+;    jz        .__paramLoopDoneWBzero1; exit, "write buffer" empty
+;
+;    movss     xmm1,[eax+vertex]     ; 0 | 0 | 0 | get next parameter
+;    mov       eax, [dlp-4]          ; offset = *(dlp + 1)
+;
+;    test      eax, eax              ; at end of offset list (offset == 0) ?
+;    jnz       .__paramLoop1b        ; nope, copy next parameter

 .__paramLoopDoneWBone1:

@@ -1211,15 +1199,24 @@
 ; end SSE version
 ;---------------------------------------------------------------------------

-%endif ; GL_SSE
-
-
-%ifndef GL_AMD3D
-%ifndef GL_SSE
+%else

 ;---------------------------------------------------------------------------
 ; original code
 ;---------------------------------------------------------------------------
+
+; Ugly, but seems to workaround the problem with locally defined
+; data segment globals not getting relocated properly when using
+; djgpp.
+
+%define zArea   One+04h
+%define zdxAB   One+08h
+%define zdxBC   One+0ch
+%define zdyAB   One+10h
+%define zdyBC   One+14h
+%define zculltest One+18h
+
+%define gc      esi       ; points to graphics context
        
 %IF GLIDE_CLIP_COORDS
    ;; NB:  We should never hit this because the proc is
@@ -1461,5 +1458,5 @@
    pop     esi
    ret
 %ENDIF ; !GLIDE_CLIP_COORD
-%ENDIF ; !GL_SSE
-%ENDIF ; !GL_AMD3D
+
+%endif
--- a/glide3x/h5/glide3/src/xdraw3.asm
+++ b/glide3x/h5/glide3/src/xdraw3.asm
@@ -2183,9 +2183,8 @@ endp
 ;;--------------------------------------------------------------------------
 ;; end AMD3D version
 ;;--------------------------------------------------------------------------
-%endif ; GL_AMD3D

-%ifdef GL_SSE
+%elifdef GL_SSE

 ;;--------------------------------------------------------------------------
 ;; start SSE version
@@ -4384,16 +4383,13 @@ endp
 ;;--------------------------------------------------------------------------
 ;; end SSE version
 ;;--------------------------------------------------------------------------
-%endif ; GL_SSE
+
+%else

 ;;--------------------------------------------------------------------------
 ;; start original code
 ;;--------------------------------------------------------------------------

-%ifndef GL_AMD3D
-%ifndef GL_MMX
-%ifndef GL_SSE
-
 ;;; include listing.inc
 %INCLUDE "fxgasm.h"

@@ -5525,6 +5521,4 @@ _vc equ 28
 %ENDIF
 endp

-%endif  ; !GL_SSE
-%endif  ; !GL_MMX
-%endif  ; !GL_AMD3D
+%endif  ; !GL_SSE !GL_MMX !GL_AMD3D
--- a/glide3x/h5/glide3/src/xtexdl.asm
+++ b/glide3x/h5/glide3/src/xtexdl.asm
@@ -19,6 +19,9 @@
 ;; $Header$
 ;; $Revision$
 ;; $Log$
+;; Revision 1.1.8.7  2003/09/12 05:08:35  koolsmoky
+;; preparing for graphic context checks
+;;
 ;; Revision 1.1.8.6  2003/07/07 23:29:06  koolsmoky
 ;; cleaned logs
 ;;
@@ -97,7 +100,7 @@ _texData$   equ 24 + STACKOFFSET

 ;--------------------------------------------------------------------------

-%IFNDEF GL_SSE2
+%IFDEF GL_AMD3D GL_MMX

 ;--------------------------------------------------------------------------
 ;
@@ -111,8 +114,7 @@ segment		TEXT

 %IFDEF GL_AMD3D
 proc _grTexDownload_3DNow_MMX, 24
-%ENDIF
-%IFDEF GL_MMX
+%ELSE ;GL_MMX
 proc _grTexDownload_MMX, 24
 %ENDIF

@@ -149,8 +151,7 @@ proc _grTexDownload_MMX, 24

 %IFDEF GL_AMD3D
    femms                               ; we'll use MMX/3DNow!, make sure FPU register cleared
-%ENDIF
-%IFDEF GL_MMX
+%ELSE ;GL_MMX
    emms                                ; we'll use MMX
 %ENDIF

@@ -299,8 +300,7 @@ proc _grTexDownload_MMX, 24
 .dlDone:           
 %IFDEF GL_AMD3D
    femms                               ; exit 3DNow!(tm) state
-%ENDIF
-%IFDEF GL_MMX
+%ELSE ;GL_MMX
    emms                                ; exit MMX state
 %ENDIF

@@ -313,7 +313,7 @@ proc _grTexDownload_MMX, 24
    ret                                 ; pop 6 DWORD parameters and return
 endp

-%ELSE ; !GL_SSE2
+%ELSE ;GL_AMD3D GL_MMX

 ;--------------------------------------------------------------------------
 ;