1 /**************************************************************************
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
28 /** @file intel_tris.c
30 * This file contains functions for managing the vertex buffer and emitting
34 #include "main/glheader.h"
35 #include "main/context.h"
36 #include "main/macros.h"
37 #include "main/enums.h"
38 #include "main/texobj.h"
39 #include "main/state.h"
42 #include "swrast/swrast.h"
43 #include "swrast_setup/swrast_setup.h"
44 #include "tnl/t_context.h"
45 #include "tnl/t_pipeline.h"
46 #include "tnl/t_vertex.h"
48 #include "intel_screen.h"
49 #include "intel_context.h"
50 #include "intel_tris.h"
51 #include "intel_batchbuffer.h"
52 #include "intel_buffers.h"
53 #include "intel_reg.h"
54 #include "intel_span.h"
55 #include "i830_context.h"
58 static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
59 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
63 intel_flush_inline_primitive(struct intel_context *intel)
65 GLuint used = intel->batch->ptr - intel->prim.start_ptr;
67 assert(intel->prim.primitive != ~0);
69 /* _mesa_printf("/\n"); */
74 *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
75 intel->prim.primitive | (used / 4 - 2));
80 intel->batch->ptr -= used;
83 intel->prim.primitive = ~0;
84 intel->prim.start_ptr = 0;
85 intel->prim.flush = 0;
88 static void intel_start_inline(struct intel_context *intel, uint32_t prim)
92 intel->vtbl.emit_state(intel);
94 intel->no_batch_wrap = GL_TRUE;
96 /*_mesa_printf("%s *", __progname);*/
98 /* Emit a slot which will be filled with the inline primitive
104 assert((intel->batch->dirty_state & (1<<1)) == 0);
106 intel->prim.start_ptr = intel->batch->ptr;
107 intel->prim.primitive = prim;
108 intel->prim.flush = intel_flush_inline_primitive;
113 intel->no_batch_wrap = GL_FALSE;
114 /* _mesa_printf(">"); */
117 static void intel_wrap_inline(struct intel_context *intel)
119 GLuint prim = intel->prim.primitive;
121 intel_flush_inline_primitive(intel);
122 intel_batchbuffer_flush(intel->batch);
123 intel_start_inline(intel, prim); /* ??? */
126 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
128 GLuint sz = dwords * sizeof(GLuint);
131 assert(intel->prim.flush == intel_flush_inline_primitive);
133 if (intel_batchbuffer_space(intel->batch) < sz)
134 intel_wrap_inline(intel);
136 /* _mesa_printf("."); */
138 intel->vtbl.assert_not_dirty(intel);
140 ptr = (GLuint *) intel->batch->ptr;
141 intel->batch->ptr += sz;
146 /** Sets the primitive type for a primitive sequence, flushing as needed. */
147 void intel_set_prim(struct intel_context *intel, uint32_t prim)
149 /* if we have no VBOs */
151 if (intel->intelScreen->no_vbo) {
152 intel_start_inline(intel, prim);
155 if (prim != intel->prim.primitive) {
156 INTEL_FIREVERTICES(intel);
157 intel->prim.primitive = prim;
161 /** Returns mapped VB space for the given number of vertices */
162 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
166 if (intel->intelScreen->no_vbo) {
167 return intel_extend_inline(intel, count * intel->vertex_size);
170 /* Check for space in the existing VB */
171 if (intel->prim.vb_bo == NULL ||
172 (intel->prim.current_offset +
173 count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
174 (intel->prim.count + count) >= (1 << 16)) {
175 /* Flush existing prim if any */
176 INTEL_FIREVERTICES(intel);
178 intel_finish_vb(intel);
181 if (intel->prim.vb == NULL)
182 intel->prim.vb = malloc(INTEL_VB_SIZE);
183 intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
185 intel->prim.start_offset = 0;
186 intel->prim.current_offset = 0;
189 intel->prim.flush = intel_flush_prim;
191 addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
192 intel->prim.current_offset += intel->vertex_size * 4 * count;
193 intel->prim.count += count;
198 /** Dispatches the accumulated primitive to the batchbuffer. */
199 void intel_flush_prim(struct intel_context *intel)
201 dri_bo *aper_array[2];
203 unsigned int offset, count;
206 /* Must be called after an intel_start_prim. */
207 assert(intel->prim.primitive != ~0);
209 if (intel->prim.count == 0)
212 /* Clear the current prims out of the context state so that a batch flush
213 * flush triggered by emit_state doesn't loop back to flush_prim again.
215 vb_bo = intel->prim.vb_bo;
216 dri_bo_reference(vb_bo);
217 count = intel->prim.count;
218 intel->prim.count = 0;
219 offset = intel->prim.start_offset;
220 intel->prim.start_offset = intel->prim.current_offset;
221 if (!intel->gen >= 3)
222 intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
223 intel->prim.flush = NULL;
225 intel->vtbl.emit_state(intel);
227 aper_array[0] = intel->batch->buf;
228 aper_array[1] = vb_bo;
229 if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
230 intel_batchbuffer_flush(intel->batch);
231 intel->vtbl.emit_state(intel);
234 /* Ensure that we don't start a new batch for the following emit, which
235 * depends on the state just emitted. emit_state should be making sure we
236 * have the space for this.
238 intel->no_batch_wrap = GL_TRUE;
240 /* Check that we actually emitted the state into this batch, using the
241 * UPLOAD_CTX bit as the signal.
243 assert((intel->batch->dirty_state & (1<<1)) == 0);
246 printf("emitting %d..%d=%d vertices size %d\n", offset,
247 intel->prim.current_offset, count,
248 intel->vertex_size * 4);
251 if (intel->gen >= 3) {
253 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
254 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
255 assert((offset & !S0_VB_OFFSET_MASK) == 0);
256 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
257 OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
258 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
260 OUT_BATCH(_3DPRIMITIVE |
262 PRIM_INDIRECT_SEQUENTIAL |
263 intel->prim.primitive |
265 OUT_BATCH(0); /* Beginning vertex index */
268 struct i830_context *i830 = i830_context(&intel->ctx);
271 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
272 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
274 assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
275 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
276 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
279 * This is somewhat unfortunate -- VB width is tied up with
280 * vertex format data that we've already uploaded through
281 * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with
282 * STATE_IMMEDIATE_1 like this to avoid duplication.
284 OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
285 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
286 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
287 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
289 OUT_BATCH(_3DPRIMITIVE |
291 PRIM_INDIRECT_SEQUENTIAL |
292 intel->prim.primitive |
294 OUT_BATCH(0); /* Beginning vertex index */
298 intel->no_batch_wrap = GL_FALSE;
300 dri_bo_unreference(vb_bo);
304 * Uploads the locally-accumulated VB into the buffer object.
306 * This avoids us thrashing the cachelines in and out as the buffer gets
307 * filled, dispatched, then reused as the hardware completes rendering from it,
308 * and also lets us clflush less if we dispatch with a partially-filled VB.
310 * This is called normally from get_space when we're finishing a BO, but also
311 * at batch flush time so that we don't try accessing the contents of a
312 * just-dispatched buffer.
314 void intel_finish_vb(struct intel_context *intel)
316 if (intel->prim.vb_bo == NULL)
319 dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
321 dri_bo_unreference(intel->prim.vb_bo);
322 intel->prim.vb_bo = NULL;
325 /***********************************************************************
326 * Emit primitives as inline vertices *
327 ***********************************************************************/
330 #define COPY_DWORDS( j, vb, vertsize, v ) \
333 __asm__ __volatile__( "rep ; movsl" \
334 : "=%c" (j), "=D" (vb), "=S" (__tmp) \
340 #define COPY_DWORDS( j, vb, vertsize, v ) \
342 for ( j = 0 ; j < vertsize ; j++ ) { \
343 vb[j] = ((GLuint *)v)[j]; \
350 intel_draw_quad(struct intel_context *intel,
352 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
354 GLuint vertsize = intel->vertex_size;
355 GLuint *vb = intel_get_prim_space(intel, 6);
358 COPY_DWORDS(j, vb, vertsize, v0);
359 COPY_DWORDS(j, vb, vertsize, v1);
361 /* If smooth shading, draw like a trifan which gives better
362 * rasterization. Otherwise draw as two triangles with provoking
363 * vertex in third position as required for flat shading.
365 if (intel->ctx.Light.ShadeModel == GL_FLAT) {
366 COPY_DWORDS(j, vb, vertsize, v3);
367 COPY_DWORDS(j, vb, vertsize, v1);
370 COPY_DWORDS(j, vb, vertsize, v2);
371 COPY_DWORDS(j, vb, vertsize, v0);
374 COPY_DWORDS(j, vb, vertsize, v2);
375 COPY_DWORDS(j, vb, vertsize, v3);
379 intel_draw_triangle(struct intel_context *intel,
380 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
382 GLuint vertsize = intel->vertex_size;
383 GLuint *vb = intel_get_prim_space(intel, 3);
386 COPY_DWORDS(j, vb, vertsize, v0);
387 COPY_DWORDS(j, vb, vertsize, v1);
388 COPY_DWORDS(j, vb, vertsize, v2);
393 intel_draw_line(struct intel_context *intel,
394 intelVertexPtr v0, intelVertexPtr v1)
396 GLuint vertsize = intel->vertex_size;
397 GLuint *vb = intel_get_prim_space(intel, 2);
400 COPY_DWORDS(j, vb, vertsize, v0);
401 COPY_DWORDS(j, vb, vertsize, v1);
406 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
408 GLuint vertsize = intel->vertex_size;
409 GLuint *vb = intel_get_prim_space(intel, 1);
412 /* Adjust for sub pixel position -- still required for conform. */
413 *(float *) &vb[0] = v0->v.x;
414 *(float *) &vb[1] = v0->v.y;
415 for (j = 2; j < vertsize; j++)
421 /***********************************************************************
422 * Fixup for ARB_point_parameters *
423 ***********************************************************************/
425 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
426 * represented in the fragment program InputsRead field.
429 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
431 GLcontext *ctx = &intel->ctx;
432 GLfloat psz[4], col[4], restore_psz, restore_alpha;
434 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
435 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
437 restore_psz = psz[0];
438 restore_alpha = col[3];
440 if (psz[0] >= ctx->Point.Threshold) {
441 psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
444 GLfloat dsize = psz[0] / ctx->Point.Threshold;
445 psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
446 col[3] *= dsize * dsize;
452 if (restore_psz != psz[0] || restore_alpha != col[3]) {
453 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
454 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
456 intel_draw_point(intel, v0);
458 psz[0] = restore_psz;
459 col[3] = restore_alpha;
461 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
462 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
465 intel_draw_point(intel, v0);
472 /***********************************************************************
473 * Fixup for I915 WPOS texture coordinate *
474 ***********************************************************************/
479 intel_wpos_triangle(struct intel_context *intel,
480 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
482 GLuint offset = intel->wpos_offset;
483 GLuint size = intel->wpos_size;
484 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
485 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
486 GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
488 __memcpy(v0_wpos, v0, size);
489 __memcpy(v1_wpos, v1, size);
490 __memcpy(v2_wpos, v2, size);
492 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
493 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
494 v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
497 intel_draw_triangle(intel, v0, v1, v2);
502 intel_wpos_line(struct intel_context *intel,
503 intelVertexPtr v0, intelVertexPtr v1)
505 GLuint offset = intel->wpos_offset;
506 GLuint size = intel->wpos_size;
507 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
508 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
510 __memcpy(v0_wpos, v0, size);
511 __memcpy(v1_wpos, v1, size);
513 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
514 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
516 intel_draw_line(intel, v0, v1);
521 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
523 GLuint offset = intel->wpos_offset;
524 GLuint size = intel->wpos_size;
525 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
527 __memcpy(v0_wpos, v0, size);
528 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
530 intel_draw_point(intel, v0);
538 /***********************************************************************
539 * Macros for t_dd_tritmp.h to draw basic primitives *
540 ***********************************************************************/
542 #define TRI( a, b, c ) \
545 intel->draw_tri( intel, a, b, c ); \
547 intel_draw_triangle( intel, a, b, c ); \
550 #define QUAD( a, b, c, d ) \
553 intel->draw_tri( intel, a, b, d ); \
554 intel->draw_tri( intel, b, c, d ); \
556 intel_draw_quad( intel, a, b, c, d ); \
559 #define LINE( v0, v1 ) \
562 intel->draw_line( intel, v0, v1 ); \
564 intel_draw_line( intel, v0, v1 ); \
567 #define POINT( v0 ) \
570 intel->draw_point( intel, v0 ); \
572 intel_draw_point( intel, v0 ); \
576 /***********************************************************************
577 * Build render functions from dd templates *
578 ***********************************************************************/
580 #define INTEL_OFFSET_BIT 0x01
581 #define INTEL_TWOSIDE_BIT 0x02
582 #define INTEL_UNFILLED_BIT 0x04
583 #define INTEL_FALLBACK_BIT 0x08
584 #define INTEL_MAX_TRIFUNC 0x10
589 tnl_points_func points;
591 tnl_triangle_func triangle;
593 } rast_tab[INTEL_MAX_TRIFUNC];
596 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
597 #define DO_OFFSET (IND & INTEL_OFFSET_BIT)
598 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
599 #define DO_TWOSIDE (IND & INTEL_TWOSIDE_BIT)
605 #define DO_FULL_QUAD 1
609 #define HAVE_BACK_COLORS 0
610 #define HAVE_HW_FLATSHADE 1
611 #define VERTEX intelVertex
614 /* Only used to pull back colors into vertices (ie, we know color is
617 #define INTEL_COLOR( dst, src ) \
619 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
620 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
621 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
622 UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \
625 #define INTEL_SPEC( dst, src ) \
627 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
628 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
629 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
633 #define DEPTH_SCALE intel->polygon_offset_scale
634 #define UNFILLED_TRI unfilled_tri
635 #define UNFILLED_QUAD unfilled_quad
636 #define VERT_X(_v) _v->v.x
637 #define VERT_Y(_v) _v->v.y
638 #define VERT_Z(_v) _v->v.z
639 #define AREA_IS_CCW( a ) (a > 0)
640 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
642 #define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
643 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
644 #define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
645 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
647 #define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
648 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
649 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
650 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
652 #define LOCAL_VARS(n) \
653 struct intel_context *intel = intel_context(ctx); \
654 GLuint color[n] = { 0, }, spec[n] = { 0, }; \
655 GLuint coloroffset = intel->coloroffset; \
656 GLboolean specoffset = intel->specoffset; \
657 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
660 /***********************************************************************
661 * Helpers for rendering unfilled primitives *
662 ***********************************************************************/
664 static const GLuint hw_prim[GL_POLYGON + 1] = {
677 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
678 #define RENDER_PRIMITIVE intel->render_primitive
680 #define IND INTEL_FALLBACK_BIT
681 #include "tnl_dd/t_dd_unfilled.h"
684 /***********************************************************************
685 * Generate GL render functions *
686 ***********************************************************************/
690 #include "tnl_dd/t_dd_tritmp.h"
692 #define IND (INTEL_OFFSET_BIT)
693 #define TAG(x) x##_offset
694 #include "tnl_dd/t_dd_tritmp.h"
696 #define IND (INTEL_TWOSIDE_BIT)
697 #define TAG(x) x##_twoside
698 #include "tnl_dd/t_dd_tritmp.h"
700 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
701 #define TAG(x) x##_twoside_offset
702 #include "tnl_dd/t_dd_tritmp.h"
704 #define IND (INTEL_UNFILLED_BIT)
705 #define TAG(x) x##_unfilled
706 #include "tnl_dd/t_dd_tritmp.h"
708 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
709 #define TAG(x) x##_offset_unfilled
710 #include "tnl_dd/t_dd_tritmp.h"
712 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
713 #define TAG(x) x##_twoside_unfilled
714 #include "tnl_dd/t_dd_tritmp.h"
716 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
717 #define TAG(x) x##_twoside_offset_unfilled
718 #include "tnl_dd/t_dd_tritmp.h"
720 #define IND (INTEL_FALLBACK_BIT)
721 #define TAG(x) x##_fallback
722 #include "tnl_dd/t_dd_tritmp.h"
724 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
725 #define TAG(x) x##_offset_fallback
726 #include "tnl_dd/t_dd_tritmp.h"
728 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
729 #define TAG(x) x##_twoside_fallback
730 #include "tnl_dd/t_dd_tritmp.h"
732 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
733 #define TAG(x) x##_twoside_offset_fallback
734 #include "tnl_dd/t_dd_tritmp.h"
736 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
737 #define TAG(x) x##_unfilled_fallback
738 #include "tnl_dd/t_dd_tritmp.h"
740 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
741 #define TAG(x) x##_offset_unfilled_fallback
742 #include "tnl_dd/t_dd_tritmp.h"
744 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
745 #define TAG(x) x##_twoside_unfilled_fallback
746 #include "tnl_dd/t_dd_tritmp.h"
748 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
750 #define TAG(x) x##_twoside_offset_unfilled_fallback
751 #include "tnl_dd/t_dd_tritmp.h"
760 init_twoside_offset();
762 init_offset_unfilled();
763 init_twoside_unfilled();
764 init_twoside_offset_unfilled();
766 init_offset_fallback();
767 init_twoside_fallback();
768 init_twoside_offset_fallback();
769 init_unfilled_fallback();
770 init_offset_unfilled_fallback();
771 init_twoside_unfilled_fallback();
772 init_twoside_offset_unfilled_fallback();
776 /***********************************************************************
777 * Rasterization fallback helpers *
778 ***********************************************************************/
781 /* This code is hit only when a mix of accelerated and unaccelerated
782 * primitives are being drawn, and only for the unaccelerated
786 intel_fallback_tri(struct intel_context *intel,
787 intelVertex * v0, intelVertex * v1, intelVertex * v2)
789 GLcontext *ctx = &intel->ctx;
793 fprintf(stderr, "\n%s\n", __FUNCTION__);
795 INTEL_FIREVERTICES(intel);
797 _swsetup_Translate(ctx, v0, &v[0]);
798 _swsetup_Translate(ctx, v1, &v[1]);
799 _swsetup_Translate(ctx, v2, &v[2]);
800 intelSpanRenderStart(ctx);
801 _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
802 intelSpanRenderFinish(ctx);
807 intel_fallback_line(struct intel_context *intel,
808 intelVertex * v0, intelVertex * v1)
810 GLcontext *ctx = &intel->ctx;
814 fprintf(stderr, "\n%s\n", __FUNCTION__);
816 INTEL_FIREVERTICES(intel);
818 _swsetup_Translate(ctx, v0, &v[0]);
819 _swsetup_Translate(ctx, v1, &v[1]);
820 intelSpanRenderStart(ctx);
821 _swrast_Line(ctx, &v[0], &v[1]);
822 intelSpanRenderFinish(ctx);
826 intel_fallback_point(struct intel_context *intel,
829 GLcontext *ctx = &intel->ctx;
833 fprintf(stderr, "\n%s\n", __FUNCTION__);
835 INTEL_FIREVERTICES(intel);
837 _swsetup_Translate(ctx, v0, &v[0]);
838 intelSpanRenderStart(ctx);
839 _swrast_Point(ctx, &v[0]);
840 intelSpanRenderFinish(ctx);
844 /**********************************************************************/
845 /* Render unclipped begin/end objects */
846 /**********************************************************************/
849 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
850 #define RENDER_POINTS( start, count ) \
851 for ( ; start < count ; start++) POINT( V(ELT(start)) );
852 #define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) )
853 #define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) )
854 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
855 #define INIT(x) intelRenderPrimitive( ctx, x )
858 struct intel_context *intel = intel_context(ctx); \
859 GLubyte *vertptr = (GLubyte *)intel->verts; \
860 const GLuint vertsize = intel->vertex_size; \
861 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
863 #define RESET_STIPPLE
864 #define RESET_OCCLUSION
865 #define PRESERVE_VB_DEFS
867 #define TAG(x) intel_##x##_verts
868 #include "tnl/t_vb_rendertmp.h"
871 #define TAG(x) intel_##x##_elts
872 #define ELT(x) elt[x]
873 #include "tnl/t_vb_rendertmp.h"
875 /**********************************************************************/
876 /* Render clipped primitives */
877 /**********************************************************************/
882 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
884 struct intel_context *intel = intel_context(ctx);
885 TNLcontext *tnl = TNL_CONTEXT(ctx);
886 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
887 GLuint prim = intel->render_primitive;
889 /* Render the new vertices as an unclipped polygon.
892 GLuint *tmp = VB->Elts;
893 VB->Elts = (GLuint *) elts;
894 tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
895 PRIM_BEGIN | PRIM_END);
899 /* Restore the render primitive
901 if (prim != GL_POLYGON)
902 tnl->Driver.Render.PrimitiveNotify(ctx, prim);
906 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
908 TNLcontext *tnl = TNL_CONTEXT(ctx);
910 tnl->Driver.Render.Line(ctx, ii, jj);
914 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
916 struct intel_context *intel = intel_context(ctx);
917 const GLuint vertsize = intel->vertex_size;
918 GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
919 GLubyte *vertptr = (GLubyte *) intel->verts;
920 const GLuint *start = (const GLuint *) V(elts[0]);
923 for (i = 2; i < n; i++) {
924 COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
925 COPY_DWORDS(j, vb, vertsize, V(elts[i]));
926 COPY_DWORDS(j, vb, vertsize, start);
930 /**********************************************************************/
931 /* Choose render functions */
932 /**********************************************************************/
937 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
938 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
941 intelChooseRenderState(GLcontext * ctx)
943 TNLcontext *tnl = TNL_CONTEXT(ctx);
944 struct intel_context *intel = intel_context(ctx);
945 GLuint flags = ctx->_TriangleCaps;
946 const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
947 GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
950 if (INTEL_DEBUG & DEBUG_STATE)
951 fprintf(stderr, "\n%s\n", __FUNCTION__);
953 if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
955 if (flags & ANY_RASTER_FLAGS) {
956 if (flags & DD_TRI_LIGHT_TWOSIDE)
957 index |= INTEL_TWOSIDE_BIT;
958 if (flags & DD_TRI_OFFSET)
959 index |= INTEL_OFFSET_BIT;
960 if (flags & DD_TRI_UNFILLED)
961 index |= INTEL_UNFILLED_BIT;
965 intel->draw_point = intel_wpos_point;
966 intel->draw_line = intel_wpos_line;
967 intel->draw_tri = intel_wpos_triangle;
969 /* Make sure these get called:
971 index |= INTEL_FALLBACK_BIT;
974 intel->draw_point = intel_draw_point;
975 intel->draw_line = intel_draw_line;
976 intel->draw_tri = intel_draw_triangle;
979 /* Hook in fallbacks for specific primitives.
981 if (flags & ANY_FALLBACK_FLAGS) {
982 if (flags & DD_LINE_STIPPLE)
983 intel->draw_line = intel_fallback_line;
985 if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
986 intel->draw_tri = intel_fallback_tri;
988 if (flags & DD_TRI_SMOOTH) {
989 if (intel->conformance_mode > 0)
990 intel->draw_tri = intel_fallback_tri;
993 if (flags & DD_POINT_ATTEN) {
995 intel->draw_point = intel_atten_point;
997 intel->draw_point = intel_fallback_point;
1000 if (flags & DD_POINT_SMOOTH) {
1001 if (intel->conformance_mode > 0)
1002 intel->draw_point = intel_fallback_point;
1005 index |= INTEL_FALLBACK_BIT;
1009 if (intel->RenderIndex != index) {
1010 intel->RenderIndex = index;
1012 tnl->Driver.Render.Points = rast_tab[index].points;
1013 tnl->Driver.Render.Line = rast_tab[index].line;
1014 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1015 tnl->Driver.Render.Quad = rast_tab[index].quad;
1018 tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1019 tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1020 tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1021 tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1024 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1025 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1026 tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1027 tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1032 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1046 /**********************************************************************/
1047 /* High level hooks for t_vb_render.c */
1048 /**********************************************************************/
1054 intelRunPipeline(GLcontext * ctx)
1056 struct intel_context *intel = intel_context(ctx);
1058 _mesa_lock_context_textures(ctx);
1061 _mesa_update_state_locked(ctx);
1063 if (intel->NewGLState) {
1064 if (intel->NewGLState & _NEW_TEXTURE) {
1065 intel->vtbl.update_texture_state(intel);
1068 if (!intel->Fallback) {
1069 if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1070 intelChooseRenderState(ctx);
1073 intel->NewGLState = 0;
1076 intel_map_vertex_shader_textures(ctx);
1077 _tnl_run_pipeline(ctx);
1078 intel_unmap_vertex_shader_textures(ctx);
1080 _mesa_unlock_context_textures(ctx);
1084 intelRenderStart(GLcontext * ctx)
1086 struct intel_context *intel = intel_context(ctx);
1088 intel_check_front_buffer_rendering(intel);
1089 intel->vtbl.render_start(intel_context(ctx));
1090 intel->vtbl.emit_state(intel);
1094 intelRenderFinish(GLcontext * ctx)
1096 struct intel_context *intel = intel_context(ctx);
1098 if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1101 INTEL_FIREVERTICES(intel);
1107 /* System to flush dma and emit state changes based on the rasterized
1111 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1113 struct intel_context *intel = intel_context(ctx);
1116 fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1117 _mesa_lookup_enum_by_nr(rprim), hwprim);
1119 intel->vtbl.reduced_primitive_state(intel, rprim);
1121 /* Start a new primitive. Arrange to have it flushed later on.
1123 if (hwprim != intel->prim.primitive) {
1124 INTEL_FIREVERTICES(intel);
1126 intel_set_prim(intel, hwprim);
1134 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1136 struct intel_context *intel = intel_context(ctx);
1139 fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1141 /* Let some clipping routines know which primitive they're dealing
1144 intel->render_primitive = prim;
1146 /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1147 * triangles. The rasterized primitive will always be reset by
1148 * lower level functions in that case, potentially pingponging the
1151 if (reduced_prim[prim] == GL_TRIANGLES &&
1152 (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1155 /* Set some primitive-dependent state and Start? a new primitive.
1157 intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1161 /**********************************************************************/
1162 /* Transition to/from hardware rasterization. */
1163 /**********************************************************************/
1165 static char *fallbackStrings[] = {
1166 [0] = "Draw buffer",
1167 [1] = "Read buffer",
1168 [2] = "Depth buffer",
1169 [3] = "Stencil buffer",
1170 [4] = "User disable",
1171 [5] = "Render mode",
1174 [13] = "Color mask",
1179 [18] = "Smooth polygon",
1180 [19] = "Smooth point",
1185 getFallbackString(GLuint bit)
1192 return fallbackStrings[i];
1198 * Enable/disable a fallback flag.
1199 * \param bit one of INTEL_FALLBACK_x flags.
1202 intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode)
1204 GLcontext *ctx = &intel->ctx;
1205 TNLcontext *tnl = TNL_CONTEXT(ctx);
1206 const GLbitfield oldfallback = intel->Fallback;
1209 intel->Fallback |= bit;
1210 if (oldfallback == 0) {
1212 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1213 fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1214 bit, getFallbackString(bit));
1215 _swsetup_Wakeup(ctx);
1216 intel->RenderIndex = ~0;
1220 intel->Fallback &= ~bit;
1221 if (oldfallback == bit) {
1223 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1224 fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1225 tnl->Driver.Render.Start = intelRenderStart;
1226 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1227 tnl->Driver.Render.Finish = intelRenderFinish;
1228 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1229 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1230 tnl->Driver.Render.Interp = _tnl_interp;
1232 _tnl_invalidate_vertex_state(ctx, ~0);
1233 _tnl_invalidate_vertices(ctx, ~0);
1234 _tnl_install_attrs(ctx,
1235 intel->vertex_attrs,
1236 intel->vertex_attr_count,
1237 intel->ViewportMatrix.m, 0);
1239 intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1250 /**********************************************************************/
1251 /* Initialization. */
1252 /**********************************************************************/
1256 intelInitTriFuncs(GLcontext * ctx)
1258 TNLcontext *tnl = TNL_CONTEXT(ctx);
1259 static int firsttime = 1;
1266 tnl->Driver.RunPipeline = intelRunPipeline;
1267 tnl->Driver.Render.Start = intelRenderStart;
1268 tnl->Driver.Render.Finish = intelRenderFinish;
1269 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1270 tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1271 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1272 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1273 tnl->Driver.Render.Interp = _tnl_interp;