From feb5575abd810ae0a21de9c3f8dd3033cd4cf573 Mon Sep 17 00:00:00 2001
From: Miguel Freitas <miguelfreitas@users.sourceforge.net>
Date: Sat, 4 Jan 2003 13:26:39 +0000
Subject: really really bad coding. that thing worked by luck, i guess.
 assuming a parameter to be on a specific stack position is absolutely
 non-portable and bad programing pratice. putting emms inside the loop?
 someone must read a mmx manual.

CVS patchset: 3770
CVS date: 2003/01/04 13:26:39
---
 src/post/goom/ifs_display.c     | 15 +-------
 src/post/goom/zoom_filter_mmx.c | 81 ++++++++++++++++++-----------------------
 2 files changed, 37 insertions(+), 59 deletions(-)

(limited to 'src')

diff --git a/src/post/goom/ifs_display.c b/src/post/goom/ifs_display.c
index 657a74063..53530ccf6 100644
--- a/src/post/goom/ifs_display.c
+++ b/src/post/goom/ifs_display.c
@@ -18,25 +18,12 @@ ifs_fun_mmx(guint32 * data, guint32 * back, int width, int height,
 
 		if ((x < width) && (y < height) && (x > 0) && (y > 0)) {
 			int     pos = x + (y * width);
-/*			register int b asm ("eax");
-
-			b = back[pos];
-			__asm__
-				("
-				 movd %%eax, %%mm0
-				 movd %%edx, %%mm1
-				 paddusb %%mm1, %%mm0
-				 movd %%mm0, %%eax
-				 "::
-				 "edx" (couleursl));
-			data[pos] = b;
-*/
 			movd_m2r(back[pos],mm0);
 			paddusb_r2r(mm1,mm0);
 			movd_r2m(mm0,data[pos]);
 		}
 	}
-	__asm__ __volatile__ ("emms");
+	emms();
 }
 #endif
 
diff --git a/src/post/goom/zoom_filter_mmx.c b/src/post/goom/zoom_filter_mmx.c
index ff7a44214..a7a5ffd8f 100644
--- a/src/post/goom/zoom_filter_mmx.c
+++ b/src/post/goom/zoom_filter_mmx.c
@@ -1,4 +1,6 @@
 #include "goom_config.h"
+#include "xineutils.h"
+
 #ifdef MMX
 #define BUFFPOINTNB 16
 #define BUFFPOINTMASK 0xffff
@@ -20,7 +22,7 @@ void zoom_filter_mmx (int prevX, int prevY,
   int bufsize = prevX * prevY;
   int loop;
 
-  __asm__ ("pxor %mm7,%mm7");
+  pxor_r2r(mm7,mm7);
   
   for (loop=0; loop<bufsize; loop++)
 	{
@@ -45,75 +47,64 @@ void zoom_filter_mmx (int prevX, int prevY,
 		/* coef en modulo 15 */
 		coeffs = precalCoef [px & PERTEMASK][py & PERTEMASK];
 	  }
-
-	  __asm__ __volatile__ ("
-               movd %%eax,%%mm6
+  
+		movd_m2r(coeffs, mm6);
 			   ;/* recuperation des deux premiers pixels dans mm0 et mm1 */
-	   	       movq (%%edx,%%ebx,4), %%mm0		/* b1-v1-r1-a1-b2-v2-r2-a2 */
-			   movq %%mm0, %%mm1				/* b1-v1-r1-a1-b2-v2-r2-a2 */
+		movq_m2r(expix1[pos], mm0);		/* b1-v1-r1-a1-b2-v2-r2-a2 */
+		movq_r2r(mm0, mm1);			/* b1-v1-r1-a1-b2-v2-r2-a2 */
 			   
 			   ;/* depackage du premier pixel */
-			   punpcklbw %%mm7, %%mm0	/* 00-b2-00-v2-00-r2-00-a2 */
+		punpcklbw_r2r(mm7, mm0);		/* 00-b2-00-v2-00-r2-00-a2 */
 			   
-			   movq %%mm6, %%mm5			/* ??-??-??-??-c4-c3-c2-c1 */
+		movq_r2r(mm6, mm5);			/* xx-xx-xx-xx-c4-c3-c2-c1 */
 			   ;/* depackage du 2ieme pixel */
-			   punpckhbw %%mm7, %%mm1	/* 00-b1-00-v1-00-r1-00-a1 */
+		punpckhbw_r2r(mm7, mm1);		/* 00-b1-00-v1-00-r1-00-a1 */
 							   
 			   ;/* extraction des coefficients... */
-			   punpcklbw %%mm5, %%mm6	/* c4-c4-c3-c3-c2-c2-c1-c1 */
-			   movq %%mm6, %%mm4			/* c4-c4-c3-c3-c2-c2-c1-c1 */
-			   movq %%mm6, %%mm5			/* c4-c4-c3-c3-c2-c2-c1-c1 */
+		punpcklbw_r2r(mm5, mm6);		/* c4-c4-c3-c3-c2-c2-c1-c1 */
+		movq_r2r(mm6, mm4);			/* c4-c4-c3-c3-c2-c2-c1-c1 */
+		movq_r2r(mm6, mm5);			/* c4-c4-c3-c3-c2-c2-c1-c1 */
 											   
-			   punpcklbw %%mm5, %%mm6	/* c2-c2-c2-c2-c1-c1-c1-c1 */
-			   punpckhbw %%mm5, %%mm4	/* c4-c4-c4-c4-c3-c3-c3-c3 */
+		punpcklbw_r2r(mm5, mm6);	/* c2-c2-c2-c2-c1-c1-c1-c1 */
+		punpckhbw_r2r(mm5, mm4);	/* c4-c4-c4-c4-c3-c3-c3-c3 */
 
-			   movq %%mm6, %%mm3			/* c2-c2-c2-c2-c1-c1-c1-c1 */
-			   punpcklbw %%mm7, %%mm6	/* 00-c1-00-c1-00-c1-00-c1 */
-			   punpckhbw %%mm7, %%mm3	/* 00-c2-00-c2-00-c2-00-c2 */
+		movq_r2r(mm6, mm3);			/* c2-c2-c2-c2-c1-c1-c1-c1 */
+		punpcklbw_r2r(mm7, mm6);	/* 00-c1-00-c1-00-c1-00-c1 */
+		punpckhbw_r2r(mm7, mm3);	/* 00-c2-00-c2-00-c2-00-c2 */
 	
 			   ;/* multiplication des pixels par les coefficients */
-			   pmullw %%mm6, %%mm0		/* c1*b2-c1*v2-c1*r2-c1*a2 */
-			   pmullw %%mm3, %%mm1		/* c2*b1-c2*v1-c2*r1-c2*a1 */
-			   paddw %%mm1, %%mm0
+		pmullw_r2r(mm6, mm0);		/* c1*b2-c1*v2-c1*r2-c1*a2 */
+		pmullw_r2r(mm3, mm1);		/* c2*b1-c2*v1-c2*r1-c2*a1 */
+		paddw_r2r(mm1, mm0);
 			   
 			   ;/* ...extraction des 2 derniers coefficients */
-			   movq %%mm4, %%mm5			/* c4-c4-c4-c4-c3-c3-c3-c3 */
-			   punpcklbw %%mm7, %%mm4	/* 00-c3-00-c3-00-c3-00-c3 */
-			   punpckhbw %%mm7, %%mm5	/* 00-c4-00-c4-00-c4-00-c4 */
+		movq_r2r(mm4, mm5);			/* c4-c4-c4-c4-c3-c3-c3-c3 */
+		punpcklbw_r2r(mm7, mm4);	/* 00-c3-00-c3-00-c3-00-c3 */
+		punpckhbw_r2r(mm7, mm5);	/* 00-c4-00-c4-00-c4-00-c4 */
 
 			   /* ajouter la longueur de ligne a esi */
-			   addl 8(%%ebp),%%ebx
-	   
 			   ;/* recuperation des 2 derniers pixels */
-			   movq (%%edx,%%ebx,4), %%mm1
-			   movq %%mm1, %%mm2
+		movq_m2r(expix1[pos+prevX], mm1);
+		movq_r2r(mm1, mm2);
 			
 			   ;/* depackage des pixels */
-			   punpcklbw %%mm7, %%mm1
-			   punpckhbw %%mm7, %%mm2
+		punpcklbw_r2r(mm7, mm1);
+		punpckhbw_r2r(mm7, mm2);
 			
 			   ;/* multiplication pas les coeffs */
-			   pmullw %%mm4, %%mm1
-			   pmullw %%mm5, %%mm2
+		pmullw_r2r(mm4, mm1);
+		pmullw_r2r(mm5, mm2);
 			   
 			   ;/* ajout des valeurs obtenues à la valeur finale */
-			   paddw %%mm1, %%mm0
-			   paddw %%mm2, %%mm0
+		paddw_r2r(mm1, mm0);
+		paddw_r2r(mm2, mm0);
 			   
 			   ;/* division par 256 = 16+16+16+16, puis repackage du pixel final */
-			   psrlw $8, %%mm0
-			   packuswb %%mm7, %%mm0
-
-               movd %%mm0,%%eax
-			   "
-							:"=eax"(expix2[loop])
-							:"ebx"(pos),"eax"(coeffs),"edx"(expix1)
-							
-				);
-	  
-/*	  expix2[loop] = couleur; */
+		psrlw_i2r(8, mm0);
+		packuswb_r2r(mm7, mm0);
+		movd_r2m(mm0,expix2[loop]);
 	  
-	  __asm__ __volatile__ ("emms");
 	}
+	emms();
 }
 #endif
-- 
cgit v1.2.3