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 "intel_chipset.h"
56 #include "i830_context.h"
59 static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
60 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
64 intel_flush_inline_primitive(struct intel_context *intel)
66 GLuint used = intel->batch->ptr - intel->prim.start_ptr;
68 assert(intel->prim.primitive != ~0);
70 /* _mesa_printf("/\n"); */
75 *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
76 intel->prim.primitive | (used / 4 - 2));
81 intel->batch->ptr -= used;
84 intel->prim.primitive = ~0;
85 intel->prim.start_ptr = 0;
86 intel->prim.flush = 0;
89 static void intel_start_inline(struct intel_context *intel, uint32_t prim)
93 intel->vtbl.emit_state(intel);
95 intel->no_batch_wrap = GL_TRUE;
97 /*_mesa_printf("%s *", __progname);*/
99 /* Emit a slot which will be filled with the inline primitive
105 assert((intel->batch->dirty_state & (1<<1)) == 0);
107 intel->prim.start_ptr = intel->batch->ptr;
108 intel->prim.primitive = prim;
109 intel->prim.flush = intel_flush_inline_primitive;
114 intel->no_batch_wrap = GL_FALSE;
115 /* _mesa_printf(">"); */
118 static void intel_wrap_inline(struct intel_context *intel)
120 GLuint prim = intel->prim.primitive;
122 intel_flush_inline_primitive(intel);
123 intel_batchbuffer_flush(intel->batch);
124 intel_start_inline(intel, prim); /* ??? */
127 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
129 GLuint sz = dwords * sizeof(GLuint);
132 assert(intel->prim.flush == intel_flush_inline_primitive);
134 if (intel_batchbuffer_space(intel->batch) < sz)
135 intel_wrap_inline(intel);
137 /* _mesa_printf("."); */
139 intel->vtbl.assert_not_dirty(intel);
141 ptr = (GLuint *) intel->batch->ptr;
142 intel->batch->ptr += sz;
147 /** Sets the primitive type for a primitive sequence, flushing as needed. */
148 void intel_set_prim(struct intel_context *intel, uint32_t prim)
150 /* if we have no VBOs */
152 if (intel->intelScreen->no_vbo) {
153 intel_start_inline(intel, prim);
156 if (prim != intel->prim.primitive) {
157 INTEL_FIREVERTICES(intel);
158 intel->prim.primitive = prim;
162 /** Returns mapped VB space for the given number of vertices */
163 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
167 if (intel->intelScreen->no_vbo) {
168 return intel_extend_inline(intel, count * intel->vertex_size);
171 /* Check for space in the existing VB */
172 if (intel->prim.vb_bo == NULL ||
173 (intel->prim.current_offset +
174 count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
175 (intel->prim.count + count) >= (1 << 16)) {
176 /* Flush existing prim if any */
177 INTEL_FIREVERTICES(intel);
179 intel_finish_vb(intel);
182 if (intel->prim.vb == NULL)
183 intel->prim.vb = malloc(INTEL_VB_SIZE);
184 intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
186 intel->prim.start_offset = 0;
187 intel->prim.current_offset = 0;
190 intel->prim.flush = intel_flush_prim;
192 addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
193 intel->prim.current_offset += intel->vertex_size * 4 * count;
194 intel->prim.count += count;
199 /** Dispatches the accumulated primitive to the batchbuffer. */
200 void intel_flush_prim(struct intel_context *intel)
202 dri_bo *aper_array[2];
204 unsigned int offset, count;
207 /* Must be called after an intel_start_prim. */
208 assert(intel->prim.primitive != ~0);
210 if (intel->prim.count == 0)
213 /* Clear the current prims out of the context state so that a batch flush
214 * flush triggered by emit_state doesn't loop back to flush_prim again.
216 vb_bo = intel->prim.vb_bo;
217 dri_bo_reference(vb_bo);
218 count = intel->prim.count;
219 intel->prim.count = 0;
220 offset = intel->prim.start_offset;
221 intel->prim.start_offset = intel->prim.current_offset;
222 if (!intel->gen >= 3)
223 intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
224 intel->prim.flush = NULL;
226 intel->vtbl.emit_state(intel);
228 aper_array[0] = intel->batch->buf;
229 aper_array[1] = vb_bo;
230 if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
231 intel_batchbuffer_flush(intel->batch);
232 intel->vtbl.emit_state(intel);
235 /* Ensure that we don't start a new batch for the following emit, which
236 * depends on the state just emitted. emit_state should be making sure we
237 * have the space for this.
239 intel->no_batch_wrap = GL_TRUE;
241 /* Check that we actually emitted the state into this batch, using the
242 * UPLOAD_CTX bit as the signal.
244 assert((intel->batch->dirty_state & (1<<1)) == 0);
247 printf("emitting %d..%d=%d vertices size %d\n", offset,
248 intel->prim.current_offset, count,
249 intel->vertex_size * 4);
252 if (intel->gen >= 3) {
254 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
255 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
256 assert((offset & !S0_VB_OFFSET_MASK) == 0);
257 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
258 OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
259 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
261 OUT_BATCH(_3DPRIMITIVE |
263 PRIM_INDIRECT_SEQUENTIAL |
264 intel->prim.primitive |
266 OUT_BATCH(0); /* Beginning vertex index */
269 struct i830_context *i830 = i830_context(&intel->ctx);
272 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
273 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
275 assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
276 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
277 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
280 * This is somewhat unfortunate -- VB width is tied up with
281 * vertex format data that we've already uploaded through
282 * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with
283 * STATE_IMMEDIATE_1 like this to avoid duplication.
285 OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
286 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
287 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
288 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
290 OUT_BATCH(_3DPRIMITIVE |
292 PRIM_INDIRECT_SEQUENTIAL |
293 intel->prim.primitive |
295 OUT_BATCH(0); /* Beginning vertex index */
299 intel->no_batch_wrap = GL_FALSE;
301 dri_bo_unreference(vb_bo);
305 * Uploads the locally-accumulated VB into the buffer object.
307 * This avoids us thrashing the cachelines in and out as the buffer gets
308 * filled, dispatched, then reused as the hardware completes rendering from it,
309 * and also lets us clflush less if we dispatch with a partially-filled VB.
311 * This is called normally from get_space when we're finishing a BO, but also
312 * at batch flush time so that we don't try accessing the contents of a
313 * just-dispatched buffer.
315 void intel_finish_vb(struct intel_context *intel)
317 if (intel->prim.vb_bo == NULL)
320 dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
322 dri_bo_unreference(intel->prim.vb_bo);
323 intel->prim.vb_bo = NULL;
326 /***********************************************************************
327 * Emit primitives as inline vertices *
328 ***********************************************************************/
331 #define COPY_DWORDS( j, vb, vertsize, v ) \
334 __asm__ __volatile__( "rep ; movsl" \
335 : "=%c" (j), "=D" (vb), "=S" (__tmp) \
341 #define COPY_DWORDS( j, vb, vertsize, v ) \
343 for ( j = 0 ; j < vertsize ; j++ ) { \
344 vb[j] = ((GLuint *)v)[j]; \
351 intel_draw_quad(struct intel_context *intel,
353 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
355 GLuint vertsize = intel->vertex_size;
356 GLuint *vb = intel_get_prim_space(intel, 6);
359 COPY_DWORDS(j, vb, vertsize, v0);
360 COPY_DWORDS(j, vb, vertsize, v1);
362 /* If smooth shading, draw like a trifan which gives better
363 * rasterization. Otherwise draw as two triangles with provoking
364 * vertex in third position as required for flat shading.
366 if (intel->ctx.Light.ShadeModel == GL_FLAT) {
367 COPY_DWORDS(j, vb, vertsize, v3);
368 COPY_DWORDS(j, vb, vertsize, v1);
371 COPY_DWORDS(j, vb, vertsize, v2);
372 COPY_DWORDS(j, vb, vertsize, v0);
375 COPY_DWORDS(j, vb, vertsize, v2);
376 COPY_DWORDS(j, vb, vertsize, v3);
380 intel_draw_triangle(struct intel_context *intel,
381 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
383 GLuint vertsize = intel->vertex_size;
384 GLuint *vb = intel_get_prim_space(intel, 3);
387 COPY_DWORDS(j, vb, vertsize, v0);
388 COPY_DWORDS(j, vb, vertsize, v1);
389 COPY_DWORDS(j, vb, vertsize, v2);
394 intel_draw_line(struct intel_context *intel,
395 intelVertexPtr v0, intelVertexPtr v1)
397 GLuint vertsize = intel->vertex_size;
398 GLuint *vb = intel_get_prim_space(intel, 2);
401 COPY_DWORDS(j, vb, vertsize, v0);
402 COPY_DWORDS(j, vb, vertsize, v1);
407 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
409 GLuint vertsize = intel->vertex_size;
410 GLuint *vb = intel_get_prim_space(intel, 1);
413 /* Adjust for sub pixel position -- still required for conform. */
414 *(float *) &vb[0] = v0->v.x;
415 *(float *) &vb[1] = v0->v.y;
416 for (j = 2; j < vertsize; j++)
422 /***********************************************************************
423 * Fixup for ARB_point_parameters *
424 ***********************************************************************/
426 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
427 * represented in the fragment program InputsRead field.
430 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
432 GLcontext *ctx = &intel->ctx;
433 GLfloat psz[4], col[4], restore_psz, restore_alpha;
435 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
436 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
438 restore_psz = psz[0];
439 restore_alpha = col[3];
441 if (psz[0] >= ctx->Point.Threshold) {
442 psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
445 GLfloat dsize = psz[0] / ctx->Point.Threshold;
446 psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
447 col[3] *= dsize * dsize;
453 if (restore_psz != psz[0] || restore_alpha != col[3]) {
454 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
455 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
457 intel_draw_point(intel, v0);
459 psz[0] = restore_psz;
460 col[3] = restore_alpha;
462 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
463 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
466 intel_draw_point(intel, v0);
473 /***********************************************************************
474 * Fixup for I915 WPOS texture coordinate *
475 ***********************************************************************/
480 intel_wpos_triangle(struct intel_context *intel,
481 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
483 GLuint offset = intel->wpos_offset;
484 GLuint size = intel->wpos_size;
485 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
486 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
487 GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
489 __memcpy(v0_wpos, v0, size);
490 __memcpy(v1_wpos, v1, size);
491 __memcpy(v2_wpos, v2, size);
493 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
494 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
495 v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
498 intel_draw_triangle(intel, v0, v1, v2);
503 intel_wpos_line(struct intel_context *intel,
504 intelVertexPtr v0, intelVertexPtr v1)
506 GLuint offset = intel->wpos_offset;
507 GLuint size = intel->wpos_size;
508 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
509 GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
511 __memcpy(v0_wpos, v0, size);
512 __memcpy(v1_wpos, v1, size);
514 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
515 v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
517 intel_draw_line(intel, v0, v1);
522 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
524 GLuint offset = intel->wpos_offset;
525 GLuint size = intel->wpos_size;
526 GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
528 __memcpy(v0_wpos, v0, size);
529 v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
531 intel_draw_point(intel, v0);
539 /***********************************************************************
540 * Macros for t_dd_tritmp.h to draw basic primitives *
541 ***********************************************************************/
543 #define TRI( a, b, c ) \
546 intel->draw_tri( intel, a, b, c ); \
548 intel_draw_triangle( intel, a, b, c ); \
551 #define QUAD( a, b, c, d ) \
554 intel->draw_tri( intel, a, b, d ); \
555 intel->draw_tri( intel, b, c, d ); \
557 intel_draw_quad( intel, a, b, c, d ); \
560 #define LINE( v0, v1 ) \
563 intel->draw_line( intel, v0, v1 ); \
565 intel_draw_line( intel, v0, v1 ); \
568 #define POINT( v0 ) \
571 intel->draw_point( intel, v0 ); \
573 intel_draw_point( intel, v0 ); \
577 /***********************************************************************
578 * Build render functions from dd templates *
579 ***********************************************************************/
581 #define INTEL_OFFSET_BIT 0x01
582 #define INTEL_TWOSIDE_BIT 0x02
583 #define INTEL_UNFILLED_BIT 0x04
584 #define INTEL_FALLBACK_BIT 0x08
585 #define INTEL_MAX_TRIFUNC 0x10
590 tnl_points_func points;
592 tnl_triangle_func triangle;
594 } rast_tab[INTEL_MAX_TRIFUNC];
597 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
598 #define DO_OFFSET (IND & INTEL_OFFSET_BIT)
599 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
600 #define DO_TWOSIDE (IND & INTEL_TWOSIDE_BIT)
606 #define DO_FULL_QUAD 1
610 #define HAVE_BACK_COLORS 0
611 #define HAVE_HW_FLATSHADE 1
612 #define VERTEX intelVertex
615 /* Only used to pull back colors into vertices (ie, we know color is
618 #define INTEL_COLOR( dst, src ) \
620 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
621 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
622 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
623 UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \
626 #define INTEL_SPEC( dst, src ) \
628 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
629 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
630 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
634 #define DEPTH_SCALE intel->polygon_offset_scale
635 #define UNFILLED_TRI unfilled_tri
636 #define UNFILLED_QUAD unfilled_quad
637 #define VERT_X(_v) _v->v.x
638 #define VERT_Y(_v) _v->v.y
639 #define VERT_Z(_v) _v->v.z
640 #define AREA_IS_CCW( a ) (a > 0)
641 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
643 #define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
644 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
645 #define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
646 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
648 #define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
649 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
650 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
651 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
653 #define LOCAL_VARS(n) \
654 struct intel_context *intel = intel_context(ctx); \
655 GLuint color[n] = { 0, }, spec[n] = { 0, }; \
656 GLuint coloroffset = intel->coloroffset; \
657 GLboolean specoffset = intel->specoffset; \
658 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
661 /***********************************************************************
662 * Helpers for rendering unfilled primitives *
663 ***********************************************************************/
665 static const GLuint hw_prim[GL_POLYGON + 1] = {
678 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
679 #define RENDER_PRIMITIVE intel->render_primitive
681 #define IND INTEL_FALLBACK_BIT
682 #include "tnl_dd/t_dd_unfilled.h"
685 /***********************************************************************
686 * Generate GL render functions *
687 ***********************************************************************/
691 #include "tnl_dd/t_dd_tritmp.h"
693 #define IND (INTEL_OFFSET_BIT)
694 #define TAG(x) x##_offset
695 #include "tnl_dd/t_dd_tritmp.h"
697 #define IND (INTEL_TWOSIDE_BIT)
698 #define TAG(x) x##_twoside
699 #include "tnl_dd/t_dd_tritmp.h"
701 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
702 #define TAG(x) x##_twoside_offset
703 #include "tnl_dd/t_dd_tritmp.h"
705 #define IND (INTEL_UNFILLED_BIT)
706 #define TAG(x) x##_unfilled
707 #include "tnl_dd/t_dd_tritmp.h"
709 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
710 #define TAG(x) x##_offset_unfilled
711 #include "tnl_dd/t_dd_tritmp.h"
713 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
714 #define TAG(x) x##_twoside_unfilled
715 #include "tnl_dd/t_dd_tritmp.h"
717 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
718 #define TAG(x) x##_twoside_offset_unfilled
719 #include "tnl_dd/t_dd_tritmp.h"
721 #define IND (INTEL_FALLBACK_BIT)
722 #define TAG(x) x##_fallback
723 #include "tnl_dd/t_dd_tritmp.h"
725 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
726 #define TAG(x) x##_offset_fallback
727 #include "tnl_dd/t_dd_tritmp.h"
729 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
730 #define TAG(x) x##_twoside_fallback
731 #include "tnl_dd/t_dd_tritmp.h"
733 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
734 #define TAG(x) x##_twoside_offset_fallback
735 #include "tnl_dd/t_dd_tritmp.h"
737 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
738 #define TAG(x) x##_unfilled_fallback
739 #include "tnl_dd/t_dd_tritmp.h"
741 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
742 #define TAG(x) x##_offset_unfilled_fallback
743 #include "tnl_dd/t_dd_tritmp.h"
745 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
746 #define TAG(x) x##_twoside_unfilled_fallback
747 #include "tnl_dd/t_dd_tritmp.h"
749 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
751 #define TAG(x) x##_twoside_offset_unfilled_fallback
752 #include "tnl_dd/t_dd_tritmp.h"
761 init_twoside_offset();
763 init_offset_unfilled();
764 init_twoside_unfilled();
765 init_twoside_offset_unfilled();
767 init_offset_fallback();
768 init_twoside_fallback();
769 init_twoside_offset_fallback();
770 init_unfilled_fallback();
771 init_offset_unfilled_fallback();
772 init_twoside_unfilled_fallback();
773 init_twoside_offset_unfilled_fallback();
777 /***********************************************************************
778 * Rasterization fallback helpers *
779 ***********************************************************************/
782 /* This code is hit only when a mix of accelerated and unaccelerated
783 * primitives are being drawn, and only for the unaccelerated
787 intel_fallback_tri(struct intel_context *intel,
788 intelVertex * v0, intelVertex * v1, intelVertex * v2)
790 GLcontext *ctx = &intel->ctx;
794 fprintf(stderr, "\n%s\n", __FUNCTION__);
796 INTEL_FIREVERTICES(intel);
798 _swsetup_Translate(ctx, v0, &v[0]);
799 _swsetup_Translate(ctx, v1, &v[1]);
800 _swsetup_Translate(ctx, v2, &v[2]);
801 intelSpanRenderStart(ctx);
802 _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
803 intelSpanRenderFinish(ctx);
808 intel_fallback_line(struct intel_context *intel,
809 intelVertex * v0, intelVertex * v1)
811 GLcontext *ctx = &intel->ctx;
815 fprintf(stderr, "\n%s\n", __FUNCTION__);
817 INTEL_FIREVERTICES(intel);
819 _swsetup_Translate(ctx, v0, &v[0]);
820 _swsetup_Translate(ctx, v1, &v[1]);
821 intelSpanRenderStart(ctx);
822 _swrast_Line(ctx, &v[0], &v[1]);
823 intelSpanRenderFinish(ctx);
827 intel_fallback_point(struct intel_context *intel,
830 GLcontext *ctx = &intel->ctx;
834 fprintf(stderr, "\n%s\n", __FUNCTION__);
836 INTEL_FIREVERTICES(intel);
838 _swsetup_Translate(ctx, v0, &v[0]);
839 intelSpanRenderStart(ctx);
840 _swrast_Point(ctx, &v[0]);
841 intelSpanRenderFinish(ctx);
845 /**********************************************************************/
846 /* Render unclipped begin/end objects */
847 /**********************************************************************/
850 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
851 #define RENDER_POINTS( start, count ) \
852 for ( ; start < count ; start++) POINT( V(ELT(start)) );
853 #define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) )
854 #define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) )
855 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
856 #define INIT(x) intelRenderPrimitive( ctx, x )
859 struct intel_context *intel = intel_context(ctx); \
860 GLubyte *vertptr = (GLubyte *)intel->verts; \
861 const GLuint vertsize = intel->vertex_size; \
862 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
864 #define RESET_STIPPLE
865 #define RESET_OCCLUSION
866 #define PRESERVE_VB_DEFS
868 #define TAG(x) intel_##x##_verts
869 #include "tnl/t_vb_rendertmp.h"
872 #define TAG(x) intel_##x##_elts
873 #define ELT(x) elt[x]
874 #include "tnl/t_vb_rendertmp.h"
876 /**********************************************************************/
877 /* Render clipped primitives */
878 /**********************************************************************/
883 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
885 struct intel_context *intel = intel_context(ctx);
886 TNLcontext *tnl = TNL_CONTEXT(ctx);
887 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
888 GLuint prim = intel->render_primitive;
890 /* Render the new vertices as an unclipped polygon.
893 GLuint *tmp = VB->Elts;
894 VB->Elts = (GLuint *) elts;
895 tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
896 PRIM_BEGIN | PRIM_END);
900 /* Restore the render primitive
902 if (prim != GL_POLYGON)
903 tnl->Driver.Render.PrimitiveNotify(ctx, prim);
907 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
909 TNLcontext *tnl = TNL_CONTEXT(ctx);
911 tnl->Driver.Render.Line(ctx, ii, jj);
915 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
917 struct intel_context *intel = intel_context(ctx);
918 const GLuint vertsize = intel->vertex_size;
919 GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
920 GLubyte *vertptr = (GLubyte *) intel->verts;
921 const GLuint *start = (const GLuint *) V(elts[0]);
924 for (i = 2; i < n; i++) {
925 COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
926 COPY_DWORDS(j, vb, vertsize, V(elts[i]));
927 COPY_DWORDS(j, vb, vertsize, start);
931 /**********************************************************************/
932 /* Choose render functions */
933 /**********************************************************************/
938 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
939 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
942 intelChooseRenderState(GLcontext * ctx)
944 TNLcontext *tnl = TNL_CONTEXT(ctx);
945 struct intel_context *intel = intel_context(ctx);
946 GLuint flags = ctx->_TriangleCaps;
947 const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
948 GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
951 if (INTEL_DEBUG & DEBUG_STATE)
952 fprintf(stderr, "\n%s\n", __FUNCTION__);
954 if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
956 if (flags & ANY_RASTER_FLAGS) {
957 if (flags & DD_TRI_LIGHT_TWOSIDE)
958 index |= INTEL_TWOSIDE_BIT;
959 if (flags & DD_TRI_OFFSET)
960 index |= INTEL_OFFSET_BIT;
961 if (flags & DD_TRI_UNFILLED)
962 index |= INTEL_UNFILLED_BIT;
966 intel->draw_point = intel_wpos_point;
967 intel->draw_line = intel_wpos_line;
968 intel->draw_tri = intel_wpos_triangle;
970 /* Make sure these get called:
972 index |= INTEL_FALLBACK_BIT;
975 intel->draw_point = intel_draw_point;
976 intel->draw_line = intel_draw_line;
977 intel->draw_tri = intel_draw_triangle;
980 /* Hook in fallbacks for specific primitives.
982 if (flags & ANY_FALLBACK_FLAGS) {
983 if (flags & DD_LINE_STIPPLE)
984 intel->draw_line = intel_fallback_line;
986 if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
987 intel->draw_tri = intel_fallback_tri;
989 if (flags & DD_TRI_SMOOTH) {
990 if (intel->conformance_mode > 0)
991 intel->draw_tri = intel_fallback_tri;
994 if (flags & DD_POINT_ATTEN) {
996 intel->draw_point = intel_atten_point;
998 intel->draw_point = intel_fallback_point;
1001 if (flags & DD_POINT_SMOOTH) {
1002 if (intel->conformance_mode > 0)
1003 intel->draw_point = intel_fallback_point;
1006 index |= INTEL_FALLBACK_BIT;
1010 if (intel->RenderIndex != index) {
1011 intel->RenderIndex = index;
1013 tnl->Driver.Render.Points = rast_tab[index].points;
1014 tnl->Driver.Render.Line = rast_tab[index].line;
1015 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1016 tnl->Driver.Render.Quad = rast_tab[index].quad;
1019 tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1020 tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1021 tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1022 tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1025 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1026 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1027 tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1028 tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1033 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1047 /**********************************************************************/
1048 /* High level hooks for t_vb_render.c */
1049 /**********************************************************************/
1055 intelRunPipeline(GLcontext * ctx)
1057 struct intel_context *intel = intel_context(ctx);
1059 _mesa_lock_context_textures(ctx);
1062 _mesa_update_state_locked(ctx);
1064 if (intel->NewGLState) {
1065 if (intel->NewGLState & _NEW_TEXTURE) {
1066 intel->vtbl.update_texture_state(intel);
1069 if (!intel->Fallback) {
1070 if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1071 intelChooseRenderState(ctx);
1074 intel->NewGLState = 0;
1077 intel_map_vertex_shader_textures(ctx);
1078 _tnl_run_pipeline(ctx);
1079 intel_unmap_vertex_shader_textures(ctx);
1081 _mesa_unlock_context_textures(ctx);
1085 intelRenderStart(GLcontext * ctx)
1087 struct intel_context *intel = intel_context(ctx);
1089 intel_check_front_buffer_rendering(intel);
1090 intel->vtbl.render_start(intel_context(ctx));
1091 intel->vtbl.emit_state(intel);
1095 intelRenderFinish(GLcontext * ctx)
1097 struct intel_context *intel = intel_context(ctx);
1099 if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1102 INTEL_FIREVERTICES(intel);
1108 /* System to flush dma and emit state changes based on the rasterized
1112 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1114 struct intel_context *intel = intel_context(ctx);
1117 fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1118 _mesa_lookup_enum_by_nr(rprim), hwprim);
1120 intel->vtbl.reduced_primitive_state(intel, rprim);
1122 /* Start a new primitive. Arrange to have it flushed later on.
1124 if (hwprim != intel->prim.primitive) {
1125 INTEL_FIREVERTICES(intel);
1127 intel_set_prim(intel, hwprim);
1135 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1137 struct intel_context *intel = intel_context(ctx);
1140 fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1142 /* Let some clipping routines know which primitive they're dealing
1145 intel->render_primitive = prim;
1147 /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1148 * triangles. The rasterized primitive will always be reset by
1149 * lower level functions in that case, potentially pingponging the
1152 if (reduced_prim[prim] == GL_TRIANGLES &&
1153 (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1156 /* Set some primitive-dependent state and Start? a new primitive.
1158 intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1162 /**********************************************************************/
1163 /* Transition to/from hardware rasterization. */
1164 /**********************************************************************/
1166 static char *fallbackStrings[] = {
1167 [0] = "Draw buffer",
1168 [1] = "Read buffer",
1169 [2] = "Depth buffer",
1170 [3] = "Stencil buffer",
1171 [4] = "User disable",
1172 [5] = "Render mode",
1175 [13] = "Color mask",
1180 [18] = "Smooth polygon",
1181 [19] = "Smooth point",
1186 getFallbackString(GLuint bit)
1193 return fallbackStrings[i];
1199 * Enable/disable a fallback flag.
1200 * \param bit one of INTEL_FALLBACK_x flags.
1203 intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode)
1205 GLcontext *ctx = &intel->ctx;
1206 TNLcontext *tnl = TNL_CONTEXT(ctx);
1207 const GLbitfield oldfallback = intel->Fallback;
1210 intel->Fallback |= bit;
1211 if (oldfallback == 0) {
1213 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1214 fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1215 bit, getFallbackString(bit));
1216 _swsetup_Wakeup(ctx);
1217 intel->RenderIndex = ~0;
1221 intel->Fallback &= ~bit;
1222 if (oldfallback == bit) {
1224 if (INTEL_DEBUG & DEBUG_FALLBACKS)
1225 fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1226 tnl->Driver.Render.Start = intelRenderStart;
1227 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1228 tnl->Driver.Render.Finish = intelRenderFinish;
1229 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1230 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1231 tnl->Driver.Render.Interp = _tnl_interp;
1233 _tnl_invalidate_vertex_state(ctx, ~0);
1234 _tnl_invalidate_vertices(ctx, ~0);
1235 _tnl_install_attrs(ctx,
1236 intel->vertex_attrs,
1237 intel->vertex_attr_count,
1238 intel->ViewportMatrix.m, 0);
1240 intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1251 /**********************************************************************/
1252 /* Initialization. */
1253 /**********************************************************************/
1257 intelInitTriFuncs(GLcontext * ctx)
1259 TNLcontext *tnl = TNL_CONTEXT(ctx);
1260 static int firsttime = 1;
1267 tnl->Driver.RunPipeline = intelRunPipeline;
1268 tnl->Driver.Render.Start = intelRenderStart;
1269 tnl->Driver.Render.Finish = intelRenderFinish;
1270 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1271 tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1272 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1273 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1274 tnl->Driver.Render.Interp = _tnl_interp;