5 #include "goom_config.h"
10 #define BUFFPOINTMASK 0xffff
14 #include "goom_graphic.h"
17 // faire : a % sqrtperte <=> a & pertemask
19 // faire : a / sqrtperte <=> a >> PERTEDEC
25 return (mm_support () & 0x1);
29 zoom_filter_mmx (int prevX, int prevY,
30 Pixel * expix1, Pixel * expix2,
31 int *brutS, int *brutD, int buffratio, int precalCoef[16][16])
33 unsigned int ax = (prevX - 1) << PERTEDEC, ay = (prevY - 1) << PERTEDEC;
35 int bufsize = prevX * prevY;
38 __asm__ __volatile__ ("pxor %mm7,%mm7");
40 for (loop = 0; loop < bufsize; loop++) {
46 int myPos = loop << 1, myPos2 = myPos + 1;
47 int brutSmypos = brutS[myPos];
49 px = brutSmypos + (((brutD[myPos] -
50 brutSmypos) * buffratio) >> BUFFPOINTNB);
51 brutSmypos = brutS[myPos2];
52 py = brutSmypos + (((brutD[myPos2] -
53 brutSmypos) * buffratio) >> BUFFPOINTNB);
55 if ((py >= ay) || (px >= ax)) {
58 pos = ((px >> PERTEDEC) + prevX * (py >> PERTEDEC));
60 coeffs = precalCoef[px & PERTEMASK][py & PERTEMASK];
63 __asm__ __volatile__ ("movd %2, %%mm6 \n\t"
64 /* recuperation des deux premiers pixels dans mm0 et mm1 */
65 "movq (%3,%1,4), %%mm0 \n\t" /* b1-v1-r1-a1-b2-v2-r2-a2 */
66 "movq %%mm0, %%mm1 \n\t" /* b1-v1-r1-a1-b2-v2-r2-a2 */
67 /* depackage du premier pixel */
68 "punpcklbw %%mm7, %%mm0 \n\t" /* 00-b2-00-v2-00-r2-00-a2 */
69 "movq %%mm6, %%mm5 \n\t" /* ??-??-??-??-c4-c3-c2-c1 */
70 /* depackage du 2ieme pixel */
71 "punpckhbw %%mm7, %%mm1 \n\t" /* 00-b1-00-v1-00-r1-00-a1 */
72 /* extraction des coefficients... */
73 "punpcklbw %%mm5, %%mm6 \n\t" /* c4-c4-c3-c3-c2-c2-c1-c1 */
74 "movq %%mm6, %%mm4 \n\t" /* c4-c4-c3-c3-c2-c2-c1-c1 */
75 "movq %%mm6, %%mm5 \n\t" /* c4-c4-c3-c3-c2-c2-c1-c1 */
76 "punpcklbw %%mm5, %%mm6 \n\t" /* c2-c2-c2-c2-c1-c1-c1-c1 */
77 "punpckhbw %%mm5, %%mm4 \n\t" /* c4-c4-c4-c4-c3-c3-c3-c3 */
78 "movq %%mm6, %%mm3 \n\t" /* c2-c2-c2-c2-c1-c1-c1-c1 */
79 "punpcklbw %%mm7, %%mm6 \n\t" /* 00-c1-00-c1-00-c1-00-c1 */
80 "punpckhbw %%mm7, %%mm3 \n\t" /* 00-c2-00-c2-00-c2-00-c2 */
81 /* multiplication des pixels par les coefficients */
82 "pmullw %%mm6, %%mm0 \n\t" /* c1*b2-c1*v2-c1*r2-c1*a2 */
83 "pmullw %%mm3, %%mm1 \n\t" /* c2*b1-c2*v1-c2*r1-c2*a1 */
84 "paddw %%mm1, %%mm0 \n\t"
85 /* ...extraction des 2 derniers coefficients */
86 "movq %%mm4, %%mm5 \n\t" /* c4-c4-c4-c4-c3-c3-c3-c3 */
87 "punpcklbw %%mm7, %%mm4 \n\t" /* 00-c3-00-c3-00-c3-00-c3 */
88 "punpckhbw %%mm7, %%mm5 \n\t" /* 00-c4-00-c4-00-c4-00-c4 */
89 /* ajouter la longueur de ligne a esi */
90 "addl 8(%%ebp),%1 \n\t"
91 /* recuperation des 2 derniers pixels */
92 "movq (%3,%1,4), %%mm1 \n\t" "movq %%mm1, %%mm2 \n\t"
93 /* depackage des pixels */
94 "punpcklbw %%mm7, %%mm1 \n\t" "punpckhbw %%mm7, %%mm2 \n\t"
95 /* multiplication pas les coeffs */
96 "pmullw %%mm4, %%mm1 \n\t" "pmullw %%mm5, %%mm2 \n\t"
97 /* ajout des valeurs obtenues ? la valeur finale */
98 "paddw %%mm1, %%mm0 \n\t" "paddw %%mm2, %%mm0 \n\t"
99 /* division par 256 = 16+16+16+16, puis repackage du pixel final */
100 "psrlw $8, %%mm0 \n\t"
101 "packuswb %%mm7, %%mm0 \n\t" "movd %%mm0,%0 \n\t":"=g" (expix2[loop])
102 :"r" (pos), "r" (coeffs), "r" (expix1)
110 #define DRAWMETHOD_PLUS_MMX(_out,_backbuf,_col) \
112 movd_m2r(_backbuf, mm0); \
113 paddusb_m2r(_col, mm0); \
114 movd_r2m(mm0, _out); \
117 #define DRAWMETHOD DRAWMETHOD_PLUS_MMX(*p,*p,col)
120 draw_line_mmx (Pixel * data, int x1, int y1, int x2, int y2, int col,
121 int screenx, int screeny)
123 int x, y, dx, dy, yy, xx;
126 if ((y1 < 0) || (y2 < 0) || (x1 < 0) || (x2 < 0) || (y1 >= screeny)
127 || (y2 >= screeny) || (x1 >= screenx) || (x2 >= screenx))
148 p = &(data[(screenx * y1) + x1]);
149 for (y = y1; y <= y2; y++) {
154 p = &(data[(screenx * y2) + x1]);
155 for (y = y2; y <= y1; y++) {
162 /* horizontal line */
165 p = &(data[(screenx * y1) + x1]);
166 for (x = x1; x <= x2; x++) {
172 p = &(data[(screenx * y1) + x2]);
173 for (x = x2; x <= x1; x++) {
187 dx = ((dx << 16) / dy);
189 for (y = y1; y <= y2; y++) {
191 p = &(data[(screenx * y) + xx]);
193 if (xx < (screenx - 1)) {
203 dy = ((dy << 16) / dx);
205 for (x = x1; x <= x2; x++) {
207 p = &(data[(screenx * yy) + x]);
209 if (yy < (screeny - 1)) {
224 dx = ((dx << 16) / -dy);
226 for (y = y1; y >= y2; y--) {
228 p = &(data[(screenx * y) + xx]);
230 if (xx < (screenx - 1)) {
240 dy = ((dy << 16) / dx);
242 for (x = x1; x <= x2; x++) {
244 p = &(data[(screenx * yy) + x]);
246 if (yy < (screeny - 1)) {
257 /* __asm__ __volatile__ ("emms"); */
265 #endif /* HAVE_MMX */