1 /* -*- mode: c; c-basic-offset: 3 -*- */
3 * Copyright 2000 Gareth Hughes
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice (including the next
14 * paragraph) shall be included in all copies or substantial portions of the
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * GARETH HUGHES BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
21 * IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
27 * Gareth Hughes <gareth@valinux.com>
28 * Leif Delgass <ldelgass@retinalburn.net>
29 * José Fonseca <j_r_fonseca@yahoo.co.uk>
32 #include "main/glheader.h"
33 #include "main/mtypes.h"
34 #include "main/colormac.h"
35 #include "main/macros.h"
37 #include "swrast/swrast.h"
38 #include "swrast_setup/swrast_setup.h"
40 #include "tnl/t_context.h"
41 #include "tnl/t_pipeline.h"
43 #include "mach64_tris.h"
44 #include "mach64_state.h"
45 #include "mach64_context.h"
46 #include "mach64_vb.h"
47 #include "mach64_ioctl.h"
49 static const GLuint hw_prim[GL_POLYGON+1] = {
52 MACH64_PRIM_LINE_LOOP,
53 MACH64_PRIM_LINE_STRIP,
54 MACH64_PRIM_TRIANGLES,
55 MACH64_PRIM_TRIANGLE_STRIP,
56 MACH64_PRIM_TRIANGLE_FAN,
58 MACH64_PRIM_QUAD_STRIP,
62 static void mach64RasterPrimitive( struct gl_context *ctx, GLuint hwprim );
63 static void mach64RenderPrimitive( struct gl_context *ctx, GLenum prim );
66 /* FIXME: Remove this when native template is finished. */
67 #define MACH64_PRINT_BUFFER 0
69 /***********************************************************************
70 * Emit primitives as inline vertices *
71 ***********************************************************************/
73 #if defined(USE_X86_ASM)
74 #define DO_COPY_VERTEX( vb, vertsize, v, n, m ) \
76 register const CARD32 *__p __asm__( "esi" ) = (CARD32 *)v + 10 - vertsize; \
77 register int __s __asm__( "ecx" ) = vertsize; \
78 if ( vertsize > 7 ) { \
79 *vb++ = (2 << 16) | ADRINDEX( MACH64_VERTEX_##n##_SECONDARY_S ); \
80 __asm__ __volatile__( "movsl ; movsl ; movsl" \
81 : "=D" (vb), "=S" (__p) \
82 : "0" (vb), "1" (__p) ); \
85 *vb++ = ((__s - 1 + m) << 16) | \
86 (ADRINDEX( MACH64_VERTEX_##n##_X_Y ) - (__s - 1) ); \
87 __asm__ __volatile__( "rep ; movsl" \
88 : "=%c" (__s), "=D" (vb), "=S" (__p) \
89 : "0" (__s), "1" (vb), "2" (__p) ); \
92 #define DO_COPY_VERTEX( vb, vertsize, v, n, m ) \
94 CARD32 *__p = (CARD32 *)v + 10 - vertsize; \
96 if ( vertsize > 7 ) { \
97 LE32_OUT( vb++, (2 << 16) | \
98 ADRINDEX( MACH64_VERTEX_##n##_SECONDARY_S ) ); \
104 LE32_OUT( vb++, ((__s - 1 + m) << 16) | \
105 (ADRINDEX( MACH64_VERTEX_##n##_X_Y ) - (__s - 1)) ); \
112 #define COPY_VERTEX( vb, vertsize, v, n ) DO_COPY_VERTEX( vb, vertsize, v, n, 0 )
113 #define COPY_VERTEX_OOA( vb, vertsize, v, n ) DO_COPY_VERTEX( vb, vertsize, v, n, 1 )
116 static INLINE void mach64_draw_quad( mach64ContextPtr mmesa,
122 #if MACH64_NATIVE_VTXFMT
123 struct gl_context *ctx = mmesa->glCtx;
124 const GLuint vertsize = mmesa->vertex_size;
128 const GLuint xyoffset = 9;
129 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
130 unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
133 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
134 fprintf(stderr, "%s:\n", __FUNCTION__);
135 fprintf(stderr,"Vertex 1:\n");
136 mach64_print_vertex( ctx, v0 );
137 fprintf(stderr,"Vertex 2:\n");
138 mach64_print_vertex( ctx, v1 );
139 fprintf(stderr,"Vertex 3:\n");
140 mach64_print_vertex( ctx, v2 );
141 fprintf(stderr,"Vertex 4:\n");
142 mach64_print_vertex( ctx, v3 );
145 xy = LE32_IN( &v0->ui[xyoffset] );
146 xx[0] = (GLshort)( xy >> 16 );
147 yy[0] = (GLshort)( xy & 0xffff );
149 xy = LE32_IN( &v1->ui[xyoffset] );
150 xx[1] = (GLshort)( xy >> 16 );
151 yy[1] = (GLshort)( xy & 0xffff );
153 xy = LE32_IN( &v3->ui[xyoffset] );
154 xx[2] = (GLshort)( xy >> 16 );
155 yy[2] = (GLshort)( xy & 0xffff );
157 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
158 (yy[0] - yy[2]) * (xx[1] - xx[2]);
160 if ( (mmesa->backface_sign &&
161 ((a < 0 && !signbit( mmesa->backface_sign )) ||
162 (a > 0 && signbit( mmesa->backface_sign )))) ) {
164 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
165 fprintf(stderr,"Quad culled\n");
171 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
174 COPY_VERTEX( vb, vertsize, v0, 1 );
175 COPY_VERTEX( vb, vertsize, v1, 2 );
176 COPY_VERTEX_OOA( vb, vertsize, v3, 3 );
177 LE32_OUT( vb++, *(CARD32 *)&ooa );
179 xy = LE32_IN( &v2->ui[xyoffset] );
180 xx[0] = (GLshort)( xy >> 16 );
181 yy[0] = (GLshort)( xy & 0xffff );
183 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
184 (yy[0] - yy[2]) * (xx[1] - xx[2]);
188 COPY_VERTEX_OOA( vb, vertsize, v2, 1 );
189 LE32_OUT( vb++, *(CARD32 *)&ooa );
191 assert( vb == vbchk );
193 #if MACH64_PRINT_BUFFER
196 fprintf(stderr, "quad:\n");
197 for (i = 0; i < vbsiz; i++)
198 fprintf(stderr, " %08lx\n", *(vb - vbsiz + i));
199 fprintf(stderr, "\n");
203 GLuint vertsize = mmesa->vertex_size;
206 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
210 (vertsize > 6 ? 2 : 0) +
211 (vertsize > 4 ? 2 : 0) +
213 (mmesa->multitex ? 4 : 0)
218 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
219 fprintf(stderr, "%s:\n", __FUNCTION__);
220 fprintf(stderr,"Vertex 1: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
221 v0->v.x, v0->v.y, v0->v.z, v0->v.w, v0->v.u0, v0->v.v0, v0->v.u1, v0->v.v1);
222 fprintf(stderr,"Vertex 2: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
223 v1->v.x, v1->v.y, v1->v.z, v1->v.w, v1->v.u0, v1->v.v0, v1->v.u1, v1->v.v1);
224 fprintf(stderr,"Vertex 3: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
225 v2->v.x, v2->v.y, v2->v.z, v2->v.w, v2->v.u0, v2->v.v0, v2->v.u1, v2->v.v1);
226 fprintf(stderr,"Vertex 4: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
227 v3->v.x, v3->v.y, v3->v.z, v3->v.w, v3->v.u0, v3->v.v0, v3->v.u1, v3->v.v1);
230 #if MACH64_CLIENT_STATE_EMITS
231 /* Enable for interleaved client-side state emits */
232 LOCK_HARDWARE( mmesa );
233 if ( mmesa->dirty ) {
234 mach64EmitHwStateLocked( mmesa );
236 if ( mmesa->sarea->dirty ) {
237 mach64UploadHwStateLocked( mmesa );
239 UNLOCK_HARDWARE( mmesa );
242 xx[0] = (GLint)(v0->v.x * 4);
243 yy[0] = (GLint)(v0->v.y * 4);
245 xx[1] = (GLint)(v1->v.x * 4);
246 yy[1] = (GLint)(v1->v.y * 4);
248 xx[2] = (GLint)(v3->v.x * 4);
249 yy[2] = (GLint)(v3->v.y * 4);
251 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
252 (yy[0] - yy[2]) * (xx[1] - xx[2]));
254 if ( ooa * mmesa->backface_sign < 0 ) {
256 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
257 fprintf(stderr,"Quad culled\n");
261 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * 4 );
265 coloridx = (vertsize > 4) ? 4: 3;
267 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
270 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
273 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
275 default: /* vertsize >= 8 */
276 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
280 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_1_S */
281 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_1_T */
284 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_W */
285 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
287 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
288 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
289 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
291 if (mmesa->multitex) {
292 /* setup for 3 sequential reg writes */
293 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_SECONDARY_S) );
294 LE32_OUT( &vb[vbidx++], v0->ui[8] ); /* MACH64_VERTEX_1_SECONDARY_S */
295 LE32_OUT( &vb[vbidx++], v0->ui[9] ); /* MACH64_VERTEX_1_SECONDARY_T */
296 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_SECONDARY_W */
299 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
302 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_2_W) );
305 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_Z) );
307 default: /* vertsize >= 8 */
308 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_2_S) );
312 LE32_OUT( &vb[vbidx++], v1->ui[6] ); /* MACH64_VERTEX_2_S */
313 LE32_OUT( &vb[vbidx++], v1->ui[7] ); /* MACH64_VERTEX_2_T */
316 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_2_W */
317 LE32_OUT( &vb[vbidx++], v1->ui[5] ); /* MACH64_VERTEX_2_SPEC_ARGB */
319 LE32_OUT( &vb[vbidx++], ((GLint)(v1->v.z) << 15) ); /* MACH64_VERTEX_2_Z */
320 vb[vbidx++] = v1->ui[coloridx]; /* MACH64_VERTEX_2_ARGB */
321 LE32_OUT( &vb[vbidx++], (xx[1] << 16) | (yy[1] & 0xffff) ); /* MACH64_VERTEX_2_X_Y */
323 if (mmesa->multitex) {
324 /* setup for 3 sequential reg writes */
325 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_SECONDARY_S) );
326 LE32_OUT( &vb[vbidx++], v1->ui[8] ); /* MACH64_VERTEX_2_SECONDARY_S */
327 LE32_OUT( &vb[vbidx++], v1->ui[9] ); /* MACH64_VERTEX_2_SECONDARY_T */
328 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_2_SECONDARY_W */
331 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
334 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_3_W) );
337 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_Z) );
339 default: /* vertsize >= 8 */
340 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_3_S) );
344 LE32_OUT( &vb[vbidx++], v3->ui[6] ); /* MACH64_VERTEX_3_S */
345 LE32_OUT( &vb[vbidx++], v3->ui[7] ); /* MACH64_VERTEX_3_T */
348 LE32_OUT( &vb[vbidx++], v3->ui[3] ); /* MACH64_VERTEX_3_W */
349 LE32_OUT( &vb[vbidx++], v3->ui[5] ); /* MACH64_VERTEX_3_SPEC_ARGB */
351 LE32_OUT( &vb[vbidx++], ((GLint)(v3->v.z) << 15) ); /* MACH64_VERTEX_3_Z */
352 vb[vbidx++] = v3->ui[coloridx]; /* MACH64_VERTEX_3_ARGB */
353 LE32_OUT( &vb[vbidx++], (xx[2] << 16) | (yy[2] & 0xffff) ); /* MACH64_VERTEX_3_X_Y */
355 if (mmesa->multitex) {
356 /* setup for 3 sequential reg writes */
357 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_SECONDARY_S) );
358 LE32_OUT( &vb[vbidx++], v3->ui[8] ); /* MACH64_VERTEX_3_SECONDARY_S */
359 LE32_OUT( &vb[vbidx++], v3->ui[9] ); /* MACH64_VERTEX_3_SECONDARY_T */
360 LE32_OUT( &vb[vbidx++], v3->ui[3] ); /* MACH64_VERTEX_3_SECONDARY_W */
363 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
364 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
366 xx[0] = (GLint)(v2->v.x * 4);
367 yy[0] = (GLint)(v2->v.y * 4);
369 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
370 (yy[0] - yy[2]) * (xx[1] - xx[2]));
373 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
376 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
379 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
381 default: /* vertsize >= 8 */
382 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
386 LE32_OUT( &vb[vbidx++], v2->ui[6] ); /* MACH64_VERTEX_1_S */
387 LE32_OUT( &vb[vbidx++], v2->ui[7] ); /* MACH64_VERTEX_1_T */
390 LE32_OUT( &vb[vbidx++], v2->ui[3] ); /* MACH64_VERTEX_1_W */
391 LE32_OUT( &vb[vbidx++], v2->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
393 LE32_OUT( &vb[vbidx++], ((GLint)(v2->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
394 vb[vbidx++] = v2->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
395 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
397 if (mmesa->multitex) {
398 /* setup for 3 sequential reg writes */
399 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_SECONDARY_S) );
400 LE32_OUT( &vb[vbidx++], v2->ui[8] ); /* MACH64_VERTEX_1_SECONDARY_S */
401 LE32_OUT( &vb[vbidx++], v2->ui[9] ); /* MACH64_VERTEX_1_SECONDARY_T */
402 LE32_OUT( &vb[vbidx++], v2->ui[3] ); /* MACH64_VERTEX_1_SECONDARY_W */
405 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
406 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
408 assert(vbsiz == vbidx);
410 #if MACH64_PRINT_BUFFER
413 fprintf(stderr, "quad:\n");
414 for (i = 0; i < vbsiz; i++)
415 fprintf(stderr, " %08lx\n", *(vb + i));
416 fprintf(stderr, "\n");
422 static INLINE void mach64_draw_triangle( mach64ContextPtr mmesa,
427 #if MACH64_NATIVE_VTXFMT
428 struct gl_context *ctx = mmesa->glCtx;
429 GLuint vertsize = mmesa->vertex_size;
433 const GLuint xyoffset = 9;
434 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
435 unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 3 + 1;
438 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
439 fprintf(stderr, "%s:\n", __FUNCTION__);
440 fprintf(stderr,"Vertex 1:\n");
441 mach64_print_vertex( ctx, v0 );
442 fprintf(stderr,"Vertex 2:\n");
443 mach64_print_vertex( ctx, v1 );
444 fprintf(stderr,"Vertex 3:\n");
445 mach64_print_vertex( ctx, v2 );
448 xy = LE32_IN( &v0->ui[xyoffset] );
449 xx[0] = (GLshort)( xy >> 16 );
450 yy[0] = (GLshort)( xy & 0xffff );
452 xy = LE32_IN( &v1->ui[xyoffset] );
453 xx[1] = (GLshort)( xy >> 16 );
454 yy[1] = (GLshort)( xy & 0xffff );
456 xy = LE32_IN( &v2->ui[xyoffset] );
457 xx[2] = (GLshort)( xy >> 16 );
458 yy[2] = (GLshort)( xy & 0xffff );
460 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
461 (yy[0] - yy[2]) * (xx[1] - xx[2]);
463 if ( mmesa->backface_sign &&
464 ((a < 0 && !signbit( mmesa->backface_sign )) ||
465 (a > 0 && signbit( mmesa->backface_sign ))) ) {
467 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
468 fprintf(stderr,"Triangle culled\n");
474 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
477 COPY_VERTEX( vb, vertsize, v0, 1 );
478 COPY_VERTEX( vb, vertsize, v1, 2 );
479 COPY_VERTEX_OOA( vb, vertsize, v2, 3 );
480 LE32_OUT( vb++, *(CARD32 *)&ooa );
482 assert( vb == vbchk );
484 #if MACH64_PRINT_BUFFER
487 fprintf(stderr, "tri:\n");
488 for (i = 0; i < vbsiz; i++)
489 fprintf(stderr, " %08lx\n", *(vb - vbsiz + i));
490 fprintf(stderr, "\n");
494 GLuint vertsize = mmesa->vertex_size;
497 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
501 (vertsize > 6 ? 2 : 0) +
502 (vertsize > 4 ? 2 : 0) +
504 (mmesa->multitex ? 4 : 0)
509 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
510 fprintf(stderr, "%s:\n", __FUNCTION__);
511 fprintf(stderr,"Vertex 1: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
512 v0->v.x, v0->v.y, v0->v.z, v0->v.w, v0->v.u0, v0->v.v0, v0->v.u1, v0->v.v1);
513 fprintf(stderr,"Vertex 2: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
514 v1->v.x, v1->v.y, v1->v.z, v1->v.w, v1->v.u0, v1->v.v0, v1->v.u1, v1->v.v1);
515 fprintf(stderr,"Vertex 3: x: %.2f, y: %.2f, z: %.2f, w: %f\n\ts0: %f, t0: %f\n\ts1: %f, t1: %f\n",
516 v2->v.x, v2->v.y, v2->v.z, v2->v.w, v2->v.u0, v2->v.v0, v2->v.u1, v2->v.v1);
519 #if MACH64_CLIENT_STATE_EMITS
520 /* Enable for interleaved client-side state emits */
521 LOCK_HARDWARE( mmesa );
522 if ( mmesa->dirty ) {
523 mach64EmitHwStateLocked( mmesa );
525 if ( mmesa->sarea->dirty ) {
526 mach64UploadHwStateLocked( mmesa );
528 UNLOCK_HARDWARE( mmesa );
531 xx[0] = (GLint)(v0->v.x * 4);
532 yy[0] = (GLint)(v0->v.y * 4);
534 xx[1] = (GLint)(v1->v.x * 4);
535 yy[1] = (GLint)(v1->v.y * 4);
537 xx[2] = (GLint)(v2->v.x * 4);
538 yy[2] = (GLint)(v2->v.y * 4);
540 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
541 (yy[0] - yy[2]) * (xx[1] - xx[2]));
543 if ( ooa * mmesa->backface_sign < 0 ) {
545 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
546 fprintf(stderr,"Triangle culled\n");
550 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * 4 );
554 coloridx = (vertsize > 4) ? 4: 3;
556 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
559 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
562 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
564 default: /* vertsize >= 8 */
565 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
569 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_1_S */
570 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_1_T */
573 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_W */
574 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
576 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
577 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
578 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
580 if (mmesa->multitex) {
581 /* setup for 3 sequential reg writes */
582 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_SECONDARY_S) );
583 LE32_OUT( &vb[vbidx++], v0->ui[8] ); /* MACH64_VERTEX_1_SECONDARY_S */
584 LE32_OUT( &vb[vbidx++], v0->ui[9] ); /* MACH64_VERTEX_1_SECONDARY_T */
585 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_SECONDARY_W */
588 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
591 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_2_W) );
594 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_Z) );
596 default: /* vertsize >= 8 */
597 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_2_S) );
601 LE32_OUT( &vb[vbidx++], v1->ui[6] ); /* MACH64_VERTEX_2_S */
602 LE32_OUT( &vb[vbidx++], v1->ui[7] ); /* MACH64_VERTEX_2_T */
605 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_2_W */
606 LE32_OUT( &vb[vbidx++], v1->ui[5] ); /* MACH64_VERTEX_2_SPEC_ARGB */
608 LE32_OUT( &vb[vbidx++], ((GLint)(v1->v.z) << 15) ); /* MACH64_VERTEX_2_Z */
609 vb[vbidx++] = v1->ui[coloridx]; /* MACH64_VERTEX_2_ARGB */
610 LE32_OUT( &vb[vbidx++], (xx[1] << 16) | (yy[1] & 0xffff) ); /* MACH64_VERTEX_2_X_Y */
612 if (mmesa->multitex) {
613 /* setup for 3 sequential reg writes */
614 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_SECONDARY_S) );
615 LE32_OUT( &vb[vbidx++], v1->ui[8] ); /* MACH64_VERTEX_2_SECONDARY_S */
616 LE32_OUT( &vb[vbidx++], v1->ui[9] ); /* MACH64_VERTEX_2_SECONDARY_T */
617 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_2_SECONDARY_W */
620 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
623 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_3_W) );
626 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_Z) );
628 default: /* vertsize >= 8 */
629 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_3_S) );
633 LE32_OUT( &vb[vbidx++], v2->ui[6] ); /* MACH64_VERTEX_3_S */
634 LE32_OUT( &vb[vbidx++], v2->ui[7] ); /* MACH64_VERTEX_3_T */
637 LE32_OUT( &vb[vbidx++], v2->ui[3] ); /* MACH64_VERTEX_3_W */
638 LE32_OUT( &vb[vbidx++], v2->ui[5] ); /* MACH64_VERTEX_3_SPEC_ARGB */
640 LE32_OUT( &vb[vbidx++], ((GLint)(v2->v.z) << 15) ); /* MACH64_VERTEX_3_Z */
641 vb[vbidx++] = v2->ui[coloridx]; /* MACH64_VERTEX_3_ARGB */
642 LE32_OUT( &vb[vbidx++], (xx[2] << 16) | (yy[2] & 0xffff) ); /* MACH64_VERTEX_3_X_Y */
644 if (mmesa->multitex) {
645 /* setup for 3 sequential reg writes */
646 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_SECONDARY_S) );
647 LE32_OUT( &vb[vbidx++], v2->ui[8] ); /* MACH64_VERTEX_3_SECONDARY_S */
648 LE32_OUT( &vb[vbidx++], v2->ui[9] ); /* MACH64_VERTEX_3_SECONDARY_T */
649 LE32_OUT( &vb[vbidx++], v2->ui[3] ); /* MACH64_VERTEX_3_SECONDARY_W */
652 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
653 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
655 assert(vbsiz == vbidx);
657 #if MACH64_PRINT_BUFFER
660 fprintf(stderr, "tri:\n");
661 for (i = 0; i < vbsiz; ++i)
662 fprintf(stderr, " %08lx\n", *(vb + i));
663 fprintf(stderr, "\n");
669 static INLINE void mach64_draw_line( mach64ContextPtr mmesa,
673 #if MACH64_NATIVE_VTXFMT
674 struct gl_context *ctx = mmesa->glCtx;
675 const GLuint vertsize = mmesa->vertex_size;
676 /* 2 fractional bits for hardware: */
677 const int width = (int) (2.0 * CLAMP(mmesa->glCtx->Line.Width,
678 mmesa->glCtx->Const.MinLineWidth,
679 mmesa->glCtx->Const.MaxLineWidth));
682 GLuint xy0old, xy0, xy1old, xy1;
683 const GLuint xyoffset = 9;
684 GLint x0, y0, x1, y1;
685 GLint dx, dy, ix, iy;
686 unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
689 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
690 fprintf(stderr, "%s:\n", __FUNCTION__);
691 fprintf(stderr,"Vertex 1:\n");
692 mach64_print_vertex( ctx, v0 );
693 fprintf(stderr,"Vertex 2:\n");
694 mach64_print_vertex( ctx, v1 );
697 pxy0 = &v0->ui[xyoffset];
699 xy0 = LE32_IN( &xy0old );
700 x0 = (GLshort)( xy0 >> 16 );
701 y0 = (GLshort)( xy0 & 0xffff );
703 pxy1 = &v1->ui[xyoffset];
705 xy1 = LE32_IN( &xy1old );
706 x1 = (GLshort)( xy1 >> 16 );
707 y1 = (GLshort)( xy1 & 0xffff );
709 if ( (dx = x1 - x0) < 0 ) {
712 if ( (dy = y1 - y0) < 0 ) {
716 /* adjust vertices depending on line direction */
720 ooa = 8.0 / ((x1 - x0) * width);
724 ooa = 8.0 / ((y0 - y1) * width);
727 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
730 LE32_OUT( pxy0, (( x0 - ix ) << 16) | (( y0 - iy ) & 0xffff) );
731 COPY_VERTEX( vb, vertsize, v0, 1 );
732 LE32_OUT( pxy1, (( x1 - ix ) << 16) | (( y1 - iy ) & 0xffff) );
733 COPY_VERTEX( vb, vertsize, v1, 2 );
734 LE32_OUT( pxy0, (( x0 + ix ) << 16) | (( y0 + iy ) & 0xffff) );
735 COPY_VERTEX_OOA( vb, vertsize, v0, 3 );
736 LE32_OUT( vb++, *(CARD32 *)&ooa );
740 LE32_OUT( pxy1, (( x1 + ix ) << 16) | (( y1 + iy ) & 0xffff) );
741 COPY_VERTEX_OOA( vb, vertsize, v1, 1 );
742 LE32_OUT( vb++, *(CARD32 *)&ooa );
746 #else /* !MACH64_NATIVE_VTXFMT */
747 GLuint vertsize = mmesa->vertex_size;
749 float width = 1.0; /* Only support 1 pix lines now */
751 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
755 (vertsize > 6 ? 2 : 0) +
756 (vertsize > 4 ? 2 : 0) +
758 (mmesa->multitex ? 4 : 0)
763 GLfloat hw, dx, dy, ix, iy;
764 GLfloat x0 = v0->v.x;
765 GLfloat y0 = v0->v.y;
766 GLfloat x1 = v1->v.x;
767 GLfloat y1 = v1->v.y;
769 #if MACH64_CLIENT_STATE_EMITS
770 /* Enable for interleaved client-side state emits */
771 LOCK_HARDWARE( mmesa );
772 if ( mmesa->dirty ) {
773 mach64EmitHwStateLocked( mmesa );
775 if ( mmesa->sarea->dirty ) {
776 mach64UploadHwStateLocked( mmesa );
778 UNLOCK_HARDWARE( mmesa );
781 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
782 fprintf(stderr, "%s:\n", __FUNCTION__);
783 fprintf(stderr,"Vertex 1: x: %.2f, y: %.2f, z: %.2f, w: %f\n",
784 v0->v.x, v0->v.y, v0->v.z, v0->v.w);
785 fprintf(stderr,"Vertex 2: x: %.2f, y: %.2f, z: %.2f, w: %f\n",
786 v1->v.x, v1->v.y, v1->v.z, v1->v.w);
790 if (hw > 0.1F && hw < 0.5F) {
794 /* adjust vertices depending on line direction */
795 dx = v0->v.x - v1->v.x;
796 dy = v0->v.y - v1->v.y;
797 if (dx * dx > dy * dy) {
820 xx[0] = (GLint)((x0 - ix) * 4);
821 yy[0] = (GLint)((y0 - iy) * 4);
823 xx[1] = (GLint)((x1 - ix) * 4);
824 yy[1] = (GLint)((y1 - iy) * 4);
826 xx[2] = (GLint)((x0 + ix) * 4);
827 yy[2] = (GLint)((y0 + iy) * 4);
829 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
830 (yy[0] - yy[2]) * (xx[1] - xx[2]));
832 if ( ooa * mmesa->backface_sign < 0 ) {
834 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
835 fprintf(stderr,"Line culled\n");
839 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * 4 );
843 coloridx = (vertsize > 4) ? 4: 3;
845 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
848 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
851 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
853 default: /* vertsize >= 8 */
854 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
858 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_1_S */
859 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_1_T */
862 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_W */
863 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
865 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
866 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
867 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
869 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
872 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_2_W) );
875 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_Z) );
877 default: /* vertsize >= 8 */
878 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_2_S) );
882 LE32_OUT( &vb[vbidx++], v1->ui[6] ); /* MACH64_VERTEX_2_S */
883 LE32_OUT( &vb[vbidx++], v1->ui[7] ); /* MACH64_VERTEX_2_T */
886 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_2_W */
887 LE32_OUT( &vb[vbidx++], v1->ui[5] ); /* MACH64_VERTEX_2_SPEC_ARGB */
889 LE32_OUT( &vb[vbidx++], ((GLint)(v1->v.z) << 15) ); /* MACH64_VERTEX_2_Z */
890 vb[vbidx++] = v1->ui[coloridx]; /* MACH64_VERTEX_2_ARGB */
891 LE32_OUT( &vb[vbidx++], (xx[1] << 16) | (yy[1] & 0xffff) ); /* MACH64_VERTEX_2_X_Y */
893 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
896 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_3_W) );
899 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_Z) );
901 default: /* vertsize >= 8 */
902 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_3_S) );
906 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_3_S */
907 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_3_T */
910 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_3_W */
911 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_3_SPEC_ARGB */
913 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_3_Z */
914 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_3_ARGB */
915 LE32_OUT( &vb[vbidx++], (xx[2] << 16) | (yy[2] & 0xffff) ); /* MACH64_VERTEX_3_X_Y */
917 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
918 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
920 xx[0] = (GLint)((x1 + ix) * 4);
921 yy[0] = (GLint)((y1 + iy) * 4);
923 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
924 (yy[0] - yy[2]) * (xx[1] - xx[2]));
927 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
930 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
933 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
935 default: /* vertsize >= 8 */
936 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
940 LE32_OUT( &vb[vbidx++], v1->ui[6] ); /* MACH64_VERTEX_1_S */
941 LE32_OUT( &vb[vbidx++], v1->ui[7] ); /* MACH64_VERTEX_1_T */
944 LE32_OUT( &vb[vbidx++], v1->ui[3] ); /* MACH64_VERTEX_1_W */
945 LE32_OUT( &vb[vbidx++], v1->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
947 LE32_OUT( &vb[vbidx++], ((GLint)(v1->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
948 vb[vbidx++] = v1->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
949 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
951 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
952 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
954 assert(vbsiz == vbidx);
958 static INLINE void mach64_draw_point( mach64ContextPtr mmesa,
961 #if MACH64_NATIVE_VTXFMT
962 struct gl_context *ctx = mmesa->glCtx;
963 const GLuint vertsize = mmesa->vertex_size;
964 /* 2 fractional bits for hardware: */
965 GLint sz = (GLint) (2.0 * CLAMP(mmesa->glCtx->Point.Size,
966 ctx->Const.MinPointSize,
967 ctx->Const.MaxPointSize));
971 const GLuint xyoffset = 9;
973 unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * 4 + 2;
976 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
977 fprintf(stderr, "%s:\n", __FUNCTION__);
978 fprintf(stderr,"Vertex 1:\n");
979 mach64_print_vertex( ctx, v0 );
983 sz = 1; /* round to the nearest supported size */
985 pxy = &v0->ui[xyoffset];
987 xy = LE32_IN( &xyold );
988 x = (GLshort)( xy >> 16 );
989 y = (GLshort)( xy & 0xffff );
991 ooa = 4.0 / (sz * sz);
993 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
996 LE32_OUT( pxy, (( x - sz ) << 16) | (( y - sz ) & 0xffff) );
997 COPY_VERTEX( vb, vertsize, v0, 1 );
998 LE32_OUT( pxy, (( x + sz ) << 16) | (( y - sz ) & 0xffff) );
999 COPY_VERTEX( vb, vertsize, v0, 2 );
1000 LE32_OUT( pxy, (( x - sz ) << 16) | (( y + sz ) & 0xffff) );
1001 COPY_VERTEX_OOA( vb, vertsize, v0, 3 );
1002 LE32_OUT( vb++, *(CARD32 *)&ooa );
1006 LE32_OUT( pxy, (( x + sz ) << 16) | (( y + sz ) & 0xffff) );
1007 COPY_VERTEX_OOA( vb, vertsize, v0, 1 );
1008 LE32_OUT( vb++, *(CARD32 *)&ooa );
1011 #else /* !MACH64_NATIVE_VTXFMT */
1012 GLuint vertsize = mmesa->vertex_size;
1014 float sz = 1.0; /* Only support 1 pix points now */
1016 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
1020 (vertsize > 6 ? 2 : 0) +
1021 (vertsize > 4 ? 2 : 0) +
1023 (mmesa->multitex ? 4 : 0)
1028 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS ) {
1029 fprintf(stderr, "%s:\n", __FUNCTION__);
1030 fprintf(stderr,"Vertex 1: x: %.2f, y: %.2f, z: %.2f, w: %f\n",
1031 v0->v.x, v0->v.y, v0->v.z, v0->v.w);
1034 #if MACH64_CLIENT_STATE_EMITS
1035 /* Enable for interleaved client-side state emits */
1036 LOCK_HARDWARE( mmesa );
1037 if ( mmesa->dirty ) {
1038 mach64EmitHwStateLocked( mmesa );
1040 if ( mmesa->sarea->dirty ) {
1041 mach64UploadHwStateLocked( mmesa );
1043 UNLOCK_HARDWARE( mmesa );
1046 xx[0] = (GLint)((v0->v.x - sz) * 4);
1047 yy[0] = (GLint)((v0->v.y - sz) * 4);
1049 xx[1] = (GLint)((v0->v.x + sz) * 4);
1050 yy[1] = (GLint)((v0->v.y - sz) * 4);
1052 xx[2] = (GLint)((v0->v.x - sz) * 4);
1053 yy[2] = (GLint)((v0->v.y + sz) * 4);
1055 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
1056 (yy[0] - yy[2]) * (xx[1] - xx[2]));
1058 if ( ooa * mmesa->backface_sign < 0 ) {
1060 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
1061 fprintf(stderr,"Point culled\n");
1065 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * 4 );
1069 coloridx = (vertsize > 4) ? 4: 3;
1071 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
1074 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
1077 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
1079 default: /* vertsize >= 8 */
1080 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
1084 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_1_S */
1085 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_1_T */
1088 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_W */
1089 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
1091 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
1092 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
1093 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
1095 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
1098 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_2_W) );
1101 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_2_Z) );
1103 default: /* vertsize >= 8 */
1104 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_2_S) );
1108 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_2_S */
1109 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_2_T */
1112 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_2_W */
1113 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_2_SPEC_ARGB */
1115 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_2_Z */
1116 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_2_ARGB */
1117 LE32_OUT( &vb[vbidx++], (xx[1] << 16) | (yy[1] & 0xffff) ); /* MACH64_VERTEX_2_X_Y */
1119 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
1122 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_3_W) );
1125 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_3_Z) );
1127 default: /* vertsize >= 8 */
1128 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_3_S) );
1132 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_3_S */
1133 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_3_T */
1136 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_3_W */
1137 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_3_SPEC_ARGB */
1139 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_3_Z */
1140 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_3_ARGB */
1141 LE32_OUT( &vb[vbidx++], (xx[2] << 16) | (yy[2] & 0xffff) ); /* MACH64_VERTEX_3_X_Y */
1143 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
1144 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
1146 xx[0] = (GLint)((v0->v.x + sz) * 4);
1147 yy[0] = (GLint)((v0->v.y + sz) * 4);
1149 ooa = 0.25 * 0.25 * ((xx[0] - xx[2]) * (yy[1] - yy[2]) -
1150 (yy[0] - yy[2]) * (xx[1] - xx[2]));
1153 /* setup for 3,5, or 7 sequential reg writes based on vertex format */
1156 LE32_OUT( &vb[vbidx++], (4 << 16) | ADRINDEX(MACH64_VERTEX_1_W) );
1159 LE32_OUT( &vb[vbidx++], (2 << 16) | ADRINDEX(MACH64_VERTEX_1_Z) );
1161 default: /* vertsize >= 8 */
1162 LE32_OUT( &vb[vbidx++], (6 << 16) | ADRINDEX(MACH64_VERTEX_1_S) );
1166 LE32_OUT( &vb[vbidx++], v0->ui[6] ); /* MACH64_VERTEX_1_S */
1167 LE32_OUT( &vb[vbidx++], v0->ui[7] ); /* MACH64_VERTEX_1_T */
1170 LE32_OUT( &vb[vbidx++], v0->ui[3] ); /* MACH64_VERTEX_1_W */
1171 LE32_OUT( &vb[vbidx++], v0->ui[5] ); /* MACH64_VERTEX_1_SPEC_ARGB */
1173 LE32_OUT( &vb[vbidx++], ((GLint)(v0->v.z) << 15) ); /* MACH64_VERTEX_1_Z */
1174 vb[vbidx++] = v0->ui[coloridx]; /* MACH64_VERTEX_1_ARGB */
1175 LE32_OUT( &vb[vbidx++], (xx[0] << 16) | (yy[0] & 0xffff) ); /* MACH64_VERTEX_1_X_Y */
1177 LE32_OUT( &vb[vbidx++], ADRINDEX(MACH64_ONE_OVER_AREA_UC) );
1178 LE32_OUT( &vb[vbidx++], *(GLuint *)&ooa );
1180 assert(vbsiz == vbidx);
1184 /***********************************************************************
1185 * Macros for t_dd_tritmp.h to draw basic primitives *
1186 ***********************************************************************/
1188 #define TRI( a, b, c ) \
1191 mmesa->draw_tri( mmesa, a, b, c ); \
1193 mach64_draw_triangle( mmesa, a, b, c ); \
1196 #define QUAD( a, b, c, d ) \
1198 if (DO_FALLBACK) { \
1199 mmesa->draw_tri( mmesa, a, b, d ); \
1200 mmesa->draw_tri( mmesa, b, c, d ); \
1202 mach64_draw_quad( mmesa, a, b, c, d ); \
1205 #define LINE( v0, v1 ) \
1208 mmesa->draw_line( mmesa, v0, v1 ); \
1210 mach64_draw_line( mmesa, v0, v1 ); \
1213 #define POINT( v0 ) \
1216 mmesa->draw_point( mmesa, v0 ); \
1218 mach64_draw_point( mmesa, v0 ); \
1222 /***********************************************************************
1223 * Build render functions from dd templates *
1224 ***********************************************************************/
1226 #define MACH64_OFFSET_BIT 0x01
1227 #define MACH64_TWOSIDE_BIT 0x02
1228 #define MACH64_UNFILLED_BIT 0x04
1229 #define MACH64_FALLBACK_BIT 0x08
1230 #define MACH64_MAX_TRIFUNC 0x10
1233 tnl_points_func points;
1235 tnl_triangle_func triangle;
1237 } rast_tab[MACH64_MAX_TRIFUNC];
1240 #define DO_FALLBACK (IND & MACH64_FALLBACK_BIT)
1241 #define DO_OFFSET (IND & MACH64_OFFSET_BIT)
1242 #define DO_UNFILLED (IND & MACH64_UNFILLED_BIT)
1243 #define DO_TWOSIDE (IND & MACH64_TWOSIDE_BIT)
1249 #define DO_FULL_QUAD 1
1252 #define HAVE_BACK_COLORS 0
1253 #define HAVE_HW_FLATSHADE 1
1254 #define VERTEX mach64Vertex
1255 #define TAB rast_tab
1257 #if MACH64_NATIVE_VTXFMT
1259 /* #define DEPTH_SCALE 65536.0 */
1260 #define DEPTH_SCALE 1
1261 #define UNFILLED_TRI unfilled_tri
1262 #define UNFILLED_QUAD unfilled_quad
1263 #define VERT_X(_v) ((GLfloat)(GLshort)(LE32_IN( &(_v)->ui[xyoffset] ) & 0xffff) / 4.0)
1264 #define VERT_Y(_v) ((GLfloat)(GLshort)(LE32_IN( &(_v)->ui[xyoffset] ) >> 16) / 4.0)
1265 #define VERT_Z(_v) ((GLfloat) LE32_IN( &(_v)->ui[zoffset] ))
1266 #define INSANE_VERTICES
1267 #define VERT_SET_Z(_v,val) LE32_OUT( &(_v)->ui[zoffset], (GLuint)(val) )
1268 #define VERT_Z_ADD(_v,val) LE32_OUT( &(_v)->ui[zoffset], LE32_IN( &(_v)->ui[zoffset] ) + (GLuint)(val) )
1269 #define AREA_IS_CCW( a ) ((a) < 0)
1270 #define GET_VERTEX(e) (mmesa->verts + ((e) * mmesa->vertex_size * sizeof(int)))
1272 #define MACH64_COLOR( dst, src ) \
1274 UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]); \
1275 UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]); \
1276 UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]); \
1277 UNCLAMPED_FLOAT_TO_UBYTE(dst[3], src[3]); \
1280 #define MACH64_SPEC( dst, src ) \
1282 UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]); \
1283 UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]); \
1284 UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]); \
1287 #define VERT_SET_RGBA( v, c ) MACH64_COLOR( v->ub4[coloroffset], c )
1288 #define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
1289 #define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset]
1290 #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
1292 #define VERT_SET_SPEC( v, c ) if (havespec) MACH64_SPEC( v->ub4[specoffset], c )
1293 #define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V( v0->ub4[specoffset], v1->ub4[specoffset] )
1294 #define VERT_SAVE_SPEC( idx ) if (havespec) spec[idx] = v[idx]->ui[specoffset]
1295 #define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[specoffset] = spec[idx]
1297 #define LOCAL_VARS(n) \
1298 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx); \
1299 GLuint color[n] = { 0 }; \
1300 GLuint spec[n] = { 0 }; \
1301 GLuint vertex_size = mmesa->vertex_size; \
1302 const GLuint xyoffset = 9; \
1303 const GLuint coloroffset = 8; \
1304 const GLuint zoffset = 7; \
1305 const GLuint specoffset = 6; \
1306 GLboolean havespec = vertex_size >= 4 ? 1 : 0; \
1307 (void) color; (void) spec; (void) vertex_size; \
1308 (void) xyoffset; (void) coloroffset; (void) zoffset; \
1309 (void) specoffset; (void) havespec;
1313 #define DEPTH_SCALE 1.0
1314 #define UNFILLED_TRI unfilled_tri
1315 #define UNFILLED_QUAD unfilled_quad
1316 #define VERT_X(_v) _v->v.x
1317 #define VERT_Y(_v) _v->v.y
1318 #define VERT_Z(_v) _v->v.z
1319 #define AREA_IS_CCW( a ) (a > 0)
1320 #define GET_VERTEX(e) (mmesa->verts + ((e) * mmesa->vertex_size * sizeof(int)))
1322 #define MACH64_COLOR( dst, src ) \
1324 UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]); \
1325 UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]); \
1326 UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]); \
1327 UNCLAMPED_FLOAT_TO_UBYTE(dst[3], src[3]); \
1330 #define MACH64_SPEC( dst, src ) \
1332 UNCLAMPED_FLOAT_TO_UBYTE(dst[0], src[2]); \
1333 UNCLAMPED_FLOAT_TO_UBYTE(dst[1], src[1]); \
1334 UNCLAMPED_FLOAT_TO_UBYTE(dst[2], src[0]); \
1337 #define VERT_SET_RGBA( v, c ) MACH64_COLOR( v->ub4[coloroffset], c )
1338 #define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
1339 #define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset]
1340 #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
1342 #define VERT_SET_SPEC( v, c ) if (havespec) MACH64_SPEC( v->ub4[5], c )
1343 #define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V(v0->ub4[5], v1->ub4[5])
1344 #define VERT_SAVE_SPEC( idx ) if (havespec) spec[idx] = v[idx]->ui[5]
1345 #define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[5] = spec[idx]
1347 #define LOCAL_VARS(n) \
1348 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx); \
1349 GLuint color[n], spec[n]; \
1350 GLuint coloroffset = (mmesa->vertex_size == 4 ? 3 : 4); \
1351 GLboolean havespec = (mmesa->vertex_size == 4 ? 0 : 1); \
1352 (void) color; (void) spec; (void) coloroffset; (void) havespec;
1356 /***********************************************************************
1357 * Helpers for rendering unfilled primitives *
1358 ***********************************************************************/
1360 #define RASTERIZE(x) if (mmesa->hw_primitive != hw_prim[x]) \
1361 mach64RasterPrimitive( ctx, hw_prim[x] )
1362 #define RENDER_PRIMITIVE mmesa->render_primitive
1363 #define IND MACH64_FALLBACK_BIT
1365 #include "tnl_dd/t_dd_unfilled.h"
1369 /***********************************************************************
1370 * Generate GL render functions *
1371 ***********************************************************************/
1376 #include "tnl_dd/t_dd_tritmp.h"
1378 #define IND (MACH64_OFFSET_BIT)
1379 #define TAG(x) x##_offset
1380 #include "tnl_dd/t_dd_tritmp.h"
1382 #define IND (MACH64_TWOSIDE_BIT)
1383 #define TAG(x) x##_twoside
1384 #include "tnl_dd/t_dd_tritmp.h"
1386 #define IND (MACH64_TWOSIDE_BIT|MACH64_OFFSET_BIT)
1387 #define TAG(x) x##_twoside_offset
1388 #include "tnl_dd/t_dd_tritmp.h"
1390 #define IND (MACH64_UNFILLED_BIT)
1391 #define TAG(x) x##_unfilled
1392 #include "tnl_dd/t_dd_tritmp.h"
1394 #define IND (MACH64_OFFSET_BIT|MACH64_UNFILLED_BIT)
1395 #define TAG(x) x##_offset_unfilled
1396 #include "tnl_dd/t_dd_tritmp.h"
1398 #define IND (MACH64_TWOSIDE_BIT|MACH64_UNFILLED_BIT)
1399 #define TAG(x) x##_twoside_unfilled
1400 #include "tnl_dd/t_dd_tritmp.h"
1402 #define IND (MACH64_TWOSIDE_BIT|MACH64_OFFSET_BIT|MACH64_UNFILLED_BIT)
1403 #define TAG(x) x##_twoside_offset_unfilled
1404 #include "tnl_dd/t_dd_tritmp.h"
1406 #define IND (MACH64_FALLBACK_BIT)
1407 #define TAG(x) x##_fallback
1408 #include "tnl_dd/t_dd_tritmp.h"
1410 #define IND (MACH64_OFFSET_BIT|MACH64_FALLBACK_BIT)
1411 #define TAG(x) x##_offset_fallback
1412 #include "tnl_dd/t_dd_tritmp.h"
1414 #define IND (MACH64_TWOSIDE_BIT|MACH64_FALLBACK_BIT)
1415 #define TAG(x) x##_twoside_fallback
1416 #include "tnl_dd/t_dd_tritmp.h"
1418 #define IND (MACH64_TWOSIDE_BIT|MACH64_OFFSET_BIT|MACH64_FALLBACK_BIT)
1419 #define TAG(x) x##_twoside_offset_fallback
1420 #include "tnl_dd/t_dd_tritmp.h"
1422 #define IND (MACH64_UNFILLED_BIT|MACH64_FALLBACK_BIT)
1423 #define TAG(x) x##_unfilled_fallback
1424 #include "tnl_dd/t_dd_tritmp.h"
1426 #define IND (MACH64_OFFSET_BIT|MACH64_UNFILLED_BIT|MACH64_FALLBACK_BIT)
1427 #define TAG(x) x##_offset_unfilled_fallback
1428 #include "tnl_dd/t_dd_tritmp.h"
1430 #define IND (MACH64_TWOSIDE_BIT|MACH64_UNFILLED_BIT|MACH64_FALLBACK_BIT)
1431 #define TAG(x) x##_twoside_unfilled_fallback
1432 #include "tnl_dd/t_dd_tritmp.h"
1434 #define IND (MACH64_TWOSIDE_BIT|MACH64_OFFSET_BIT|MACH64_UNFILLED_BIT| \
1435 MACH64_FALLBACK_BIT)
1436 #define TAG(x) x##_twoside_offset_unfilled_fallback
1437 #include "tnl_dd/t_dd_tritmp.h"
1440 static void init_rast_tab( void )
1445 init_twoside_offset();
1447 init_offset_unfilled();
1448 init_twoside_unfilled();
1449 init_twoside_offset_unfilled();
1451 init_offset_fallback();
1452 init_twoside_fallback();
1453 init_twoside_offset_fallback();
1454 init_unfilled_fallback();
1455 init_offset_unfilled_fallback();
1456 init_twoside_unfilled_fallback();
1457 init_twoside_offset_unfilled_fallback();
1461 /***********************************************************************
1462 * Rasterization fallback helpers *
1463 ***********************************************************************/
1466 /* This code is hit only when a mix of accelerated and unaccelerated
1467 * primitives are being drawn, and only for the unaccelerated
1471 mach64_fallback_tri( mach64ContextPtr mmesa,
1476 struct gl_context *ctx = mmesa->glCtx;
1478 mach64_translate_vertex( ctx, v0, &v[0] );
1479 mach64_translate_vertex( ctx, v1, &v[1] );
1480 mach64_translate_vertex( ctx, v2, &v[2] );
1481 _swrast_Triangle( ctx, &v[0], &v[1], &v[2] );
1486 mach64_fallback_line( mach64ContextPtr mmesa,
1490 struct gl_context *ctx = mmesa->glCtx;
1492 mach64_translate_vertex( ctx, v0, &v[0] );
1493 mach64_translate_vertex( ctx, v1, &v[1] );
1494 _swrast_Line( ctx, &v[0], &v[1] );
1499 mach64_fallback_point( mach64ContextPtr mmesa,
1502 struct gl_context *ctx = mmesa->glCtx;
1504 mach64_translate_vertex( ctx, v0, &v[0] );
1505 _swrast_Point( ctx, &v[0] );
1510 /**********************************************************************/
1511 /* Render unclipped begin/end objects */
1512 /**********************************************************************/
1514 #define VERT(x) (mach64Vertex *)(mach64verts + ((x) * vertsize * sizeof(int)))
1515 #define RENDER_POINTS( start, count ) \
1516 for ( ; start < count ; start++) \
1517 mach64_draw_point( mmesa, VERT(start) )
1518 #define RENDER_LINE( v0, v1 ) \
1519 mach64_draw_line( mmesa, VERT(v0), VERT(v1) )
1520 #define RENDER_TRI( v0, v1, v2 ) \
1521 mach64_draw_triangle( mmesa, VERT(v0), VERT(v1), VERT(v2) )
1522 #define RENDER_QUAD( v0, v1, v2, v3 ) \
1523 mach64_draw_quad( mmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) )
1524 #define INIT(x) do { \
1525 if (0) fprintf(stderr, "%s\n", __FUNCTION__); \
1526 mach64RenderPrimitive( ctx, x ); \
1529 #define LOCAL_VARS \
1530 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx); \
1531 const GLuint vertsize = mmesa->vertex_size; \
1532 const char *mach64verts = (char *)mmesa->verts; \
1533 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
1535 #define RESET_STIPPLE
1536 #define RESET_OCCLUSION
1537 #define PRESERVE_VB_DEFS
1539 #define TAG(x) mach64_##x##_verts
1540 #include "tnl/t_vb_rendertmp.h"
1543 #define TAG(x) mach64_##x##_elts
1544 #define ELT(x) elt[x]
1545 #include "tnl/t_vb_rendertmp.h"
1548 /**********************************************************************/
1549 /* Render clipped primitives */
1550 /**********************************************************************/
1552 static void mach64RenderClippedPoly( struct gl_context *ctx, const GLuint *elts,
1555 mach64ContextPtr mmesa = MACH64_CONTEXT( ctx );
1556 TNLcontext *tnl = TNL_CONTEXT(ctx);
1557 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
1558 GLuint prim = mmesa->render_primitive;
1560 /* Render the new vertices as an unclipped polygon.
1563 GLuint *tmp = VB->Elts;
1564 VB->Elts = (GLuint *)elts;
1565 tnl->Driver.Render.PrimTabElts[GL_POLYGON]( ctx, 0, n, PRIM_BEGIN|PRIM_END );
1569 /* Restore the render primitive
1571 if (prim != GL_POLYGON)
1572 tnl->Driver.Render.PrimitiveNotify( ctx, prim );
1576 static void mach64RenderClippedLine( struct gl_context *ctx, GLuint ii, GLuint jj )
1578 TNLcontext *tnl = TNL_CONTEXT(ctx);
1579 tnl->Driver.Render.Line( ctx, ii, jj );
1582 #if MACH64_NATIVE_VTXFMT
1583 static void mach64FastRenderClippedPoly( struct gl_context *ctx, const GLuint *elts,
1586 mach64ContextPtr mmesa = MACH64_CONTEXT( ctx );
1587 const GLuint vertsize = mmesa->vertex_size;
1594 const GLuint xyoffset = 9;
1595 GLint xx[3], yy[3]; /* 2 fractional bits for hardware */
1596 unsigned vbsiz = (vertsize + (vertsize > 7 ? 2 : 1)) * n + (n-2);
1598 GLubyte *mach64verts = (GLubyte *)mmesa->verts;
1599 mach64VertexPtr v0, v1, v2;
1602 v0 = (mach64VertexPtr)VERT(elts[1]);
1603 v1 = (mach64VertexPtr)VERT(elts[2]);
1604 v2 = (mach64VertexPtr)VERT(elts[0]);
1606 xy = LE32_IN( &v0->ui[xyoffset] );
1607 xx[0] = (GLshort)( xy >> 16 );
1608 yy[0] = (GLshort)( xy & 0xffff );
1610 xy = LE32_IN( &v1->ui[xyoffset] );
1611 xx[1] = (GLshort)( xy >> 16 );
1612 yy[1] = (GLshort)( xy & 0xffff );
1614 xy = LE32_IN( &v2->ui[xyoffset] );
1615 xx[2] = (GLshort)( xy >> 16 );
1616 yy[2] = (GLshort)( xy & 0xffff );
1618 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
1619 (yy[0] - yy[2]) * (xx[1] - xx[2]);
1621 if ( (mmesa->backface_sign &&
1622 ((a < 0 && !signbit( mmesa->backface_sign )) ||
1623 (a > 0 && signbit( mmesa->backface_sign )))) ) {
1625 if ( MACH64_DEBUG & DEBUG_VERBOSE_PRIMS )
1626 fprintf(stderr,"Polygon culled\n");
1632 vb = (CARD32 *)mach64AllocDmaLow( mmesa, vbsiz * sizeof(CARD32) );
1635 COPY_VERTEX( vb, vertsize, v0, 1 );
1636 COPY_VERTEX( vb, vertsize, v1, 2 );
1637 COPY_VERTEX_OOA( vb, vertsize, v2, 3 );
1638 LE32_OUT( vb++, ooa.u );
1644 v0 = (mach64VertexPtr)VERT(elts[i]);
1647 xy = LE32_IN( &v0->ui[xyoffset] );
1648 xx[0] = (GLshort)( xy >> 16 );
1649 yy[0] = (GLshort)( xy & 0xffff );
1651 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
1652 (yy[0] - yy[2]) * (xx[1] - xx[2]);
1655 COPY_VERTEX_OOA( vb, vertsize, v0, 1 );
1656 LE32_OUT( vb++, ooa.u );
1660 v1 = (mach64VertexPtr)VERT(elts[i]);
1663 xy = LE32_IN( &v1->ui[xyoffset] );
1664 xx[1] = (GLshort)( xy >> 16 );
1665 yy[1] = (GLshort)( xy & 0xffff );
1667 a = (xx[0] - xx[2]) * (yy[1] - yy[2]) -
1668 (yy[0] - yy[2]) * (xx[1] - xx[2]);
1671 COPY_VERTEX_OOA( vb, vertsize, v1, 2 );
1672 LE32_OUT( vb++, ooa.u );
1675 assert( vb == vbchk );
1678 static void mach64FastRenderClippedPoly( struct gl_context *ctx, const GLuint *elts,
1681 mach64ContextPtr mmesa = MACH64_CONTEXT( ctx );
1682 const GLuint vertsize = mmesa->vertex_size;
1683 GLubyte *mach64verts = (GLubyte *)mmesa->verts;
1684 const GLuint *start = (const GLuint *)VERT(elts[0]);
1687 for (i = 2 ; i < n ; i++) {
1688 mach64_draw_triangle( mmesa,
1691 (mach64VertexPtr) start
1695 #endif /* MACH64_NATIVE_VTXFMT */
1697 /**********************************************************************/
1698 /* Choose render functions */
1699 /**********************************************************************/
1701 #define _MACH64_NEW_RENDER_STATE (_DD_NEW_POINT_SMOOTH | \
1702 _DD_NEW_LINE_SMOOTH | \
1703 _DD_NEW_LINE_STIPPLE | \
1704 _DD_NEW_TRI_SMOOTH | \
1705 _DD_NEW_TRI_STIPPLE | \
1706 _NEW_POLYGONSTIPPLE | \
1707 _DD_NEW_TRI_UNFILLED | \
1708 _DD_NEW_TRI_LIGHT_TWOSIDE | \
1709 _DD_NEW_TRI_OFFSET) \
1711 #define POINT_FALLBACK (DD_POINT_SMOOTH)
1712 #define LINE_FALLBACK (DD_LINE_SMOOTH|DD_LINE_STIPPLE)
1713 #define TRI_FALLBACK (DD_TRI_SMOOTH|DD_TRI_STIPPLE)
1714 #define ANY_FALLBACK_FLAGS (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK)
1715 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE|DD_TRI_OFFSET|DD_TRI_UNFILLED)
1718 static void mach64ChooseRenderState(struct gl_context *ctx)
1720 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
1721 GLuint flags = ctx->_TriangleCaps;
1724 if (flags & (ANY_RASTER_FLAGS|ANY_FALLBACK_FLAGS)) {
1725 mmesa->draw_point = mach64_draw_point;
1726 mmesa->draw_line = mach64_draw_line;
1727 mmesa->draw_tri = mach64_draw_triangle;
1729 if (flags & ANY_RASTER_FLAGS) {
1730 if (flags & DD_TRI_LIGHT_TWOSIDE) index |= MACH64_TWOSIDE_BIT;
1731 if (flags & DD_TRI_OFFSET) index |= MACH64_OFFSET_BIT;
1732 if (flags & DD_TRI_UNFILLED) index |= MACH64_UNFILLED_BIT;
1735 /* Hook in fallbacks for specific primitives.
1737 if (flags & (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK)) {
1738 if (flags & POINT_FALLBACK) mmesa->draw_point = mach64_fallback_point;
1739 if (flags & LINE_FALLBACK) mmesa->draw_line = mach64_fallback_line;
1740 if (flags & TRI_FALLBACK) mmesa->draw_tri = mach64_fallback_tri;
1741 index |= MACH64_FALLBACK_BIT;
1745 if (index != mmesa->RenderIndex) {
1746 TNLcontext *tnl = TNL_CONTEXT(ctx);
1747 tnl->Driver.Render.Points = rast_tab[index].points;
1748 tnl->Driver.Render.Line = rast_tab[index].line;
1749 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1750 tnl->Driver.Render.Quad = rast_tab[index].quad;
1753 tnl->Driver.Render.PrimTabVerts = mach64_render_tab_verts;
1754 tnl->Driver.Render.PrimTabElts = mach64_render_tab_elts;
1755 tnl->Driver.Render.ClippedLine = rast_tab[index].line;
1756 tnl->Driver.Render.ClippedPolygon = mach64FastRenderClippedPoly;
1758 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1759 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1760 tnl->Driver.Render.ClippedLine = mach64RenderClippedLine;
1761 tnl->Driver.Render.ClippedPolygon = mach64RenderClippedPoly;
1764 mmesa->RenderIndex = index;
1768 /**********************************************************************/
1769 /* Validate state at pipeline start */
1770 /**********************************************************************/
1772 static void mach64RunPipeline( struct gl_context *ctx )
1774 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
1776 if (mmesa->new_state)
1777 mach64DDUpdateHWState( ctx );
1779 if (!mmesa->Fallback && mmesa->NewGLState) {
1780 if (mmesa->NewGLState & _MACH64_NEW_VERTEX_STATE)
1781 mach64ChooseVertexState( ctx );
1783 if (mmesa->NewGLState & _MACH64_NEW_RENDER_STATE)
1784 mach64ChooseRenderState( ctx );
1786 mmesa->NewGLState = 0;
1789 _tnl_run_pipeline( ctx );
1792 /**********************************************************************/
1793 /* High level hooks for t_vb_render.c */
1794 /**********************************************************************/
1796 /* This is called when Mesa switches between rendering triangle
1797 * primitives (such as GL_POLYGON, GL_QUADS, GL_TRIANGLE_STRIP, etc),
1798 * and lines, points and bitmaps.
1801 static void mach64RasterPrimitive( struct gl_context *ctx, GLuint hwprim )
1803 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
1805 mmesa->new_state |= MACH64_NEW_CONTEXT;
1806 mmesa->dirty |= MACH64_UPLOAD_CONTEXT;
1808 if (mmesa->hw_primitive != hwprim) {
1809 FLUSH_BATCH( mmesa );
1810 mmesa->hw_primitive = hwprim;
1814 static void mach64RenderPrimitive( struct gl_context *ctx, GLenum prim )
1816 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
1817 GLuint hw = hw_prim[prim];
1819 mmesa->render_primitive = prim;
1821 if (prim >= GL_TRIANGLES && (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1824 mach64RasterPrimitive( ctx, hw );
1828 static void mach64RenderStart( struct gl_context *ctx )
1830 /* Check for projective texturing. Make sure all texcoord
1831 * pointers point to something. (fix in mesa?)
1833 mach64CheckTexSizes( ctx );
1836 static void mach64RenderFinish( struct gl_context *ctx )
1838 if (MACH64_CONTEXT(ctx)->RenderIndex & MACH64_FALLBACK_BIT)
1839 _swrast_flush( ctx );
1843 /**********************************************************************/
1844 /* Transition to/from hardware rasterization. */
1845 /**********************************************************************/
1847 static const char * const fallbackStrings[] = {
1849 "glDrawBuffer(GL_FRONT_AND_BACK)",
1851 "glEnable(GL_STENCIL) without hw stencil buffer",
1852 "glRenderMode(selection or feedback)",
1853 "glLogicOp (mode != GL_COPY)",
1854 "GL_SEPARATE_SPECULAR_COLOR",
1855 "glBlendEquation (mode != ADD)",
1857 "Rasterization disable",
1861 static const char *getFallbackString(GLuint bit)
1868 return fallbackStrings[i];
1871 void mach64Fallback( struct gl_context *ctx, GLuint bit, GLboolean mode )
1873 TNLcontext *tnl = TNL_CONTEXT(ctx);
1874 mach64ContextPtr mmesa = MACH64_CONTEXT(ctx);
1875 GLuint oldfallback = mmesa->Fallback;
1878 mmesa->Fallback |= bit;
1879 if (oldfallback == 0) {
1880 FLUSH_BATCH( mmesa );
1881 _swsetup_Wakeup( ctx );
1882 mmesa->RenderIndex = ~0;
1883 if (MACH64_DEBUG & DEBUG_VERBOSE_FALLBACK) {
1884 fprintf(stderr, "Mach64 begin rasterization fallback: 0x%x %s\n",
1885 bit, getFallbackString(bit));
1890 mmesa->Fallback &= ~bit;
1891 if (oldfallback == bit) {
1892 _swrast_flush( ctx );
1893 tnl->Driver.Render.Start = mach64RenderStart;
1894 tnl->Driver.Render.PrimitiveNotify = mach64RenderPrimitive;
1895 tnl->Driver.Render.Finish = mach64RenderFinish;
1896 tnl->Driver.Render.BuildVertices = mach64BuildVertices;
1897 mmesa->NewGLState |= (_MACH64_NEW_RENDER_STATE|
1898 _MACH64_NEW_VERTEX_STATE);
1899 if (MACH64_DEBUG & DEBUG_VERBOSE_FALLBACK) {
1900 fprintf(stderr, "Mach64 end rasterization fallback: 0x%x %s\n",
1901 bit, getFallbackString(bit));
1907 /**********************************************************************/
1908 /* Initialization. */
1909 /**********************************************************************/
1911 void mach64InitTriFuncs( struct gl_context *ctx )
1913 TNLcontext *tnl = TNL_CONTEXT(ctx);
1914 static int firsttime = 1;
1921 tnl->Driver.RunPipeline = mach64RunPipeline;
1922 tnl->Driver.Render.Start = mach64RenderStart;
1923 tnl->Driver.Render.Finish = mach64RenderFinish;
1924 tnl->Driver.Render.PrimitiveNotify = mach64RenderPrimitive;
1925 tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1926 tnl->Driver.Render.BuildVertices = mach64BuildVertices;