1 /**************************************************************************
3 * Copyright 2003 VMware, Inc.
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 VMWARE 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"
41 #include "main/fbobject.h"
43 #include "swrast/swrast.h"
44 #include "swrast_setup/swrast_setup.h"
45 #include "tnl/t_context.h"
46 #include "tnl/t_pipeline.h"
47 #include "tnl/t_vertex.h"
49 #include "intel_screen.h"
50 #include "intel_context.h"
51 #include "intel_tris.h"
52 #include "intel_batchbuffer.h"
53 #include "intel_buffers.h"
54 #include "intel_reg.h"
55 #include "i830_context.h"
57 #include "i915_context.h"
59 static void intelRenderPrimitive(struct gl_context * ctx, GLenum prim);
60 static void intelRasterPrimitive(struct gl_context * ctx, GLenum rprim,
64 intel_flush_inline_primitive(struct intel_context *intel)
66 GLuint used = intel->batch.used - intel->prim.start_ptr;
68 assert(intel->prim.primitive != ~0);
75 intel->batch.map[intel->prim.start_ptr] =
76 _3DPRIMITIVE | intel->prim.primitive | (used - 2);
81 intel->batch.used = intel->prim.start_ptr;
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 = true;
97 /* Emit a slot which will be filled with the inline primitive
102 intel->prim.start_ptr = intel->batch.used;
103 intel->prim.primitive = prim;
104 intel->prim.flush = intel_flush_inline_primitive;
109 intel->no_batch_wrap = false;
113 static void intel_wrap_inline(struct intel_context *intel)
115 GLuint prim = intel->prim.primitive;
117 intel_flush_inline_primitive(intel);
118 intel_batchbuffer_flush(intel);
119 intel_start_inline(intel, prim); /* ??? */
122 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
126 assert(intel->prim.flush == intel_flush_inline_primitive);
128 if (intel_batchbuffer_space(intel) < dwords * sizeof(GLuint))
129 intel_wrap_inline(intel);
133 intel->vtbl.assert_not_dirty(intel);
135 ptr = intel->batch.map + intel->batch.used;
136 intel->batch.used += dwords;
141 /** Sets the primitive type for a primitive sequence, flushing as needed. */
142 void intel_set_prim(struct intel_context *intel, uint32_t prim)
144 /* if we have no VBOs */
146 if (intel->intelScreen->no_vbo) {
147 intel_start_inline(intel, prim);
150 if (prim != intel->prim.primitive) {
151 INTEL_FIREVERTICES(intel);
152 intel->prim.primitive = prim;
156 /** Returns mapped VB space for the given number of vertices */
157 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
161 if (intel->intelScreen->no_vbo) {
162 return intel_extend_inline(intel, count * intel->vertex_size);
165 /* Check for space in the existing VB */
166 if (intel->prim.vb_bo == NULL ||
167 (intel->prim.current_offset +
168 count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
169 (intel->prim.count + count) >= (1 << 16)) {
170 /* Flush existing prim if any */
171 INTEL_FIREVERTICES(intel);
173 intel_finish_vb(intel);
176 if (intel->prim.vb == NULL)
177 intel->prim.vb = malloc(INTEL_VB_SIZE);
178 intel->prim.vb_bo = drm_intel_bo_alloc(intel->bufmgr, "vb",
180 intel->prim.start_offset = 0;
181 intel->prim.current_offset = 0;
184 intel->prim.flush = intel_flush_prim;
186 addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
187 intel->prim.current_offset += intel->vertex_size * 4 * count;
188 intel->prim.count += count;
193 /** Dispatches the accumulated primitive to the batchbuffer. */
194 void intel_flush_prim(struct intel_context *intel)
196 drm_intel_bo *aper_array[2];
198 unsigned int offset, count;
201 /* Must be called after an intel_start_prim. */
202 assert(intel->prim.primitive != ~0);
204 if (intel->prim.count == 0)
207 /* Clear the current prims out of the context state so that a batch flush
208 * flush triggered by emit_state doesn't loop back to flush_prim again.
210 vb_bo = intel->prim.vb_bo;
211 drm_intel_bo_reference(vb_bo);
212 count = intel->prim.count;
213 intel->prim.count = 0;
214 offset = intel->prim.start_offset;
215 intel->prim.start_offset = intel->prim.current_offset;
217 intel->prim.current_offset = intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
218 intel->prim.flush = NULL;
220 intel->vtbl.emit_state(intel);
222 aper_array[0] = intel->batch.bo;
223 aper_array[1] = vb_bo;
224 if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
225 intel_batchbuffer_flush(intel);
226 intel->vtbl.emit_state(intel);
229 /* Ensure that we don't start a new batch for the following emit, which
230 * depends on the state just emitted. emit_state should be making sure we
231 * have the space for this.
233 intel->no_batch_wrap = true;
235 if (intel->always_flush_cache) {
236 intel_batchbuffer_emit_mi_flush(intel);
240 printf("emitting %d..%d=%d vertices size %d\n", offset,
241 intel->prim.current_offset, count,
242 intel->vertex_size * 4);
245 if (intel->gen >= 3) {
246 struct i915_context *i915 = i915_context(&intel->ctx);
247 unsigned int cmd = 0, len = 0;
249 if (vb_bo != i915->current_vb_bo) {
254 if (intel->vertex_size != i915->current_vertex_size) {
263 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | cmd | (len - 2));
264 if (vb_bo != i915->current_vb_bo) {
265 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, 0);
266 i915->current_vb_bo = vb_bo;
268 if (intel->vertex_size != i915->current_vertex_size) {
269 OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
270 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
271 i915->current_vertex_size = intel->vertex_size;
273 OUT_BATCH(_3DPRIMITIVE |
275 PRIM_INDIRECT_SEQUENTIAL |
276 intel->prim.primitive |
278 OUT_BATCH(offset / (intel->vertex_size * 4));
281 struct i830_context *i830 = i830_context(&intel->ctx);
284 OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
285 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
287 assert((offset & ~S0_VB_OFFSET_MASK_830) == 0);
288 OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
289 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
292 * This is somewhat unfortunate -- VB width is tied up with
293 * vertex format data that we've already uploaded through
294 * _3DSTATE_VFT[01]_CMD. We may want to replace emits of VFT state with
295 * STATE_IMMEDIATE_1 like this to avoid duplication.
297 OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
298 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
299 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
300 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
302 OUT_BATCH(_3DPRIMITIVE |
304 PRIM_INDIRECT_SEQUENTIAL |
305 intel->prim.primitive |
307 OUT_BATCH(0); /* Beginning vertex index */
311 if (intel->always_flush_cache) {
312 intel_batchbuffer_emit_mi_flush(intel);
315 intel->no_batch_wrap = false;
317 drm_intel_bo_unreference(vb_bo);
321 * Uploads the locally-accumulated VB into the buffer object.
323 * This avoids us thrashing the cachelines in and out as the buffer gets
324 * filled, dispatched, then reused as the hardware completes rendering from it,
325 * and also lets us clflush less if we dispatch with a partially-filled VB.
327 * This is called normally from get_space when we're finishing a BO, but also
328 * at batch flush time so that we don't try accessing the contents of a
329 * just-dispatched buffer.
331 void intel_finish_vb(struct intel_context *intel)
333 if (intel->prim.vb_bo == NULL)
336 drm_intel_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
338 drm_intel_bo_unreference(intel->prim.vb_bo);
339 intel->prim.vb_bo = NULL;
342 /***********************************************************************
343 * Emit primitives as inline vertices *
344 ***********************************************************************/
347 #define COPY_DWORDS( j, vb, vertsize, v ) \
350 __asm__ __volatile__( "rep ; movsl" \
351 : "=%c" (j), "=D" (vb), "=S" (__tmp) \
357 #define COPY_DWORDS( j, vb, vertsize, v ) \
359 for ( j = 0 ; j < vertsize ; j++ ) { \
360 vb[j] = ((GLuint *)v)[j]; \
367 intel_draw_quad(struct intel_context *intel,
369 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
371 GLuint vertsize = intel->vertex_size;
372 GLuint *vb = intel_get_prim_space(intel, 6);
375 COPY_DWORDS(j, vb, vertsize, v0);
376 COPY_DWORDS(j, vb, vertsize, v1);
378 /* If smooth shading, draw like a trifan which gives better
379 * rasterization. Otherwise draw as two triangles with provoking
380 * vertex in third position as required for flat shading.
382 if (intel->ctx.Light.ShadeModel == GL_FLAT) {
383 COPY_DWORDS(j, vb, vertsize, v3);
384 COPY_DWORDS(j, vb, vertsize, v1);
387 COPY_DWORDS(j, vb, vertsize, v2);
388 COPY_DWORDS(j, vb, vertsize, v0);
391 COPY_DWORDS(j, vb, vertsize, v2);
392 COPY_DWORDS(j, vb, vertsize, v3);
396 intel_draw_triangle(struct intel_context *intel,
397 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
399 GLuint vertsize = intel->vertex_size;
400 GLuint *vb = intel_get_prim_space(intel, 3);
403 COPY_DWORDS(j, vb, vertsize, v0);
404 COPY_DWORDS(j, vb, vertsize, v1);
405 COPY_DWORDS(j, vb, vertsize, v2);
410 intel_draw_line(struct intel_context *intel,
411 intelVertexPtr v0, intelVertexPtr v1)
413 GLuint vertsize = intel->vertex_size;
414 GLuint *vb = intel_get_prim_space(intel, 2);
417 COPY_DWORDS(j, vb, vertsize, v0);
418 COPY_DWORDS(j, vb, vertsize, v1);
423 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
425 GLuint vertsize = intel->vertex_size;
426 GLuint *vb = intel_get_prim_space(intel, 1);
429 /* Adjust for sub pixel position -- still required for conform. */
430 *(float *) &vb[0] = v0->v.x;
431 *(float *) &vb[1] = v0->v.y;
432 for (j = 2; j < vertsize; j++)
438 /***********************************************************************
439 * Fixup for ARB_point_parameters *
440 ***********************************************************************/
442 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
443 * represented in the fragment program InputsRead field.
446 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
448 struct gl_context *ctx = &intel->ctx;
449 GLfloat psz[4], col[4], restore_psz, restore_alpha;
451 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
452 _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
454 restore_psz = psz[0];
455 restore_alpha = col[3];
457 if (psz[0] >= ctx->Point.Threshold) {
458 psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
461 GLfloat dsize = psz[0] / ctx->Point.Threshold;
462 psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
463 col[3] *= dsize * dsize;
469 if (restore_psz != psz[0] || restore_alpha != col[3]) {
470 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
471 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
473 intel_draw_point(intel, v0);
475 psz[0] = restore_psz;
476 col[3] = restore_alpha;
478 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
479 _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
482 intel_draw_point(intel, v0);
489 /***********************************************************************
490 * Fixup for I915 WPOS texture coordinate *
491 ***********************************************************************/
494 intel_emit_fragcoord(struct intel_context *intel, intelVertexPtr v)
496 struct gl_context *ctx = &intel->ctx;
497 struct gl_framebuffer *fb = ctx->DrawBuffer;
498 GLuint offset = intel->wpos_offset;
499 float *vertex_position = (float *)v;
500 float *fragcoord = (float *)((char *)v + offset);
502 fragcoord[0] = vertex_position[0];
504 if (_mesa_is_user_fbo(fb))
505 fragcoord[1] = vertex_position[1];
507 fragcoord[1] = fb->Height - vertex_position[1];
509 fragcoord[2] = vertex_position[2];
510 fragcoord[3] = vertex_position[3];
514 intel_wpos_triangle(struct intel_context *intel,
515 intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
517 intel_emit_fragcoord(intel, v0);
518 intel_emit_fragcoord(intel, v1);
519 intel_emit_fragcoord(intel, v2);
521 intel_draw_triangle(intel, v0, v1, v2);
526 intel_wpos_line(struct intel_context *intel,
527 intelVertexPtr v0, intelVertexPtr v1)
529 intel_emit_fragcoord(intel, v0);
530 intel_emit_fragcoord(intel, v1);
531 intel_draw_line(intel, v0, v1);
536 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
538 intel_emit_fragcoord(intel, v0);
539 intel_draw_point(intel, v0);
547 /***********************************************************************
548 * Macros for t_dd_tritmp.h to draw basic primitives *
549 ***********************************************************************/
551 #define TRI( a, b, c ) \
554 intel->draw_tri( intel, a, b, c ); \
556 intel_draw_triangle( intel, a, b, c ); \
559 #define QUAD( a, b, c, d ) \
562 intel->draw_tri( intel, a, b, d ); \
563 intel->draw_tri( intel, b, c, d ); \
565 intel_draw_quad( intel, a, b, c, d ); \
568 #define LINE( v0, v1 ) \
571 intel->draw_line( intel, v0, v1 ); \
573 intel_draw_line( intel, v0, v1 ); \
576 #define POINT( v0 ) \
579 intel->draw_point( intel, v0 ); \
581 intel_draw_point( intel, v0 ); \
585 /***********************************************************************
586 * Build render functions from dd templates *
587 ***********************************************************************/
589 #define INTEL_OFFSET_BIT 0x01
590 #define INTEL_TWOSIDE_BIT 0x02
591 #define INTEL_UNFILLED_BIT 0x04
592 #define INTEL_FALLBACK_BIT 0x08
593 #define INTEL_MAX_TRIFUNC 0x10
598 tnl_points_func points;
600 tnl_triangle_func triangle;
602 } rast_tab[INTEL_MAX_TRIFUNC];
605 #define DO_FALLBACK ((IND & INTEL_FALLBACK_BIT) != 0)
606 #define DO_OFFSET ((IND & INTEL_OFFSET_BIT) != 0)
607 #define DO_UNFILLED ((IND & INTEL_UNFILLED_BIT) != 0)
608 #define DO_TWOSIDE ((IND & INTEL_TWOSIDE_BIT) != 0)
614 #define DO_FULL_QUAD 1
617 #define HAVE_BACK_COLORS 0
618 #define HAVE_HW_FLATSHADE 1
619 #define VERTEX intelVertex
622 /* Only used to pull back colors into vertices (ie, we know color is
625 #define INTEL_COLOR( 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]); \
630 UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \
633 #define INTEL_SPEC( dst, src ) \
635 UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \
636 UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \
637 UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \
641 #define DEPTH_SCALE (ctx->DrawBuffer->Visual.depthBits == 16 ? 1.0 : 2.0)
642 #define UNFILLED_TRI unfilled_tri
643 #define UNFILLED_QUAD unfilled_quad
644 #define VERT_X(_v) _v->v.x
645 #define VERT_Y(_v) _v->v.y
646 #define VERT_Z(_v) _v->v.z
647 #define AREA_IS_CCW( a ) (a > 0)
648 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
650 #define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
651 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
652 #define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
653 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
655 #define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
656 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
657 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
658 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
660 #define LOCAL_VARS(n) \
661 struct intel_context *intel = intel_context(ctx); \
662 GLuint color[n] = { 0, }, spec[n] = { 0, }; \
663 GLuint coloroffset = intel->coloroffset; \
664 GLuint specoffset = intel->specoffset; \
665 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
668 /***********************************************************************
669 * Helpers for rendering unfilled primitives *
670 ***********************************************************************/
672 static const GLuint hw_prim[GL_POLYGON + 1] = {
685 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
686 #define RENDER_PRIMITIVE intel->render_primitive
688 #define IND INTEL_FALLBACK_BIT
689 #include "tnl_dd/t_dd_unfilled.h"
692 /***********************************************************************
693 * Generate GL render functions *
694 ***********************************************************************/
698 #include "tnl_dd/t_dd_tritmp.h"
700 #define IND (INTEL_OFFSET_BIT)
701 #define TAG(x) x##_offset
702 #include "tnl_dd/t_dd_tritmp.h"
704 #define IND (INTEL_TWOSIDE_BIT)
705 #define TAG(x) x##_twoside
706 #include "tnl_dd/t_dd_tritmp.h"
708 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
709 #define TAG(x) x##_twoside_offset
710 #include "tnl_dd/t_dd_tritmp.h"
712 #define IND (INTEL_UNFILLED_BIT)
713 #define TAG(x) x##_unfilled
714 #include "tnl_dd/t_dd_tritmp.h"
716 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
717 #define TAG(x) x##_offset_unfilled
718 #include "tnl_dd/t_dd_tritmp.h"
720 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
721 #define TAG(x) x##_twoside_unfilled
722 #include "tnl_dd/t_dd_tritmp.h"
724 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
725 #define TAG(x) x##_twoside_offset_unfilled
726 #include "tnl_dd/t_dd_tritmp.h"
728 #define IND (INTEL_FALLBACK_BIT)
729 #define TAG(x) x##_fallback
730 #include "tnl_dd/t_dd_tritmp.h"
732 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
733 #define TAG(x) x##_offset_fallback
734 #include "tnl_dd/t_dd_tritmp.h"
736 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
737 #define TAG(x) x##_twoside_fallback
738 #include "tnl_dd/t_dd_tritmp.h"
740 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
741 #define TAG(x) x##_twoside_offset_fallback
742 #include "tnl_dd/t_dd_tritmp.h"
744 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
745 #define TAG(x) x##_unfilled_fallback
746 #include "tnl_dd/t_dd_tritmp.h"
748 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
749 #define TAG(x) x##_offset_unfilled_fallback
750 #include "tnl_dd/t_dd_tritmp.h"
752 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
753 #define TAG(x) x##_twoside_unfilled_fallback
754 #include "tnl_dd/t_dd_tritmp.h"
756 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
758 #define TAG(x) x##_twoside_offset_unfilled_fallback
759 #include "tnl_dd/t_dd_tritmp.h"
768 init_twoside_offset();
770 init_offset_unfilled();
771 init_twoside_unfilled();
772 init_twoside_offset_unfilled();
774 init_offset_fallback();
775 init_twoside_fallback();
776 init_twoside_offset_fallback();
777 init_unfilled_fallback();
778 init_offset_unfilled_fallback();
779 init_twoside_unfilled_fallback();
780 init_twoside_offset_unfilled_fallback();
784 /***********************************************************************
785 * Rasterization fallback helpers *
786 ***********************************************************************/
789 /* This code is hit only when a mix of accelerated and unaccelerated
790 * primitives are being drawn, and only for the unaccelerated
794 intel_fallback_tri(struct intel_context *intel,
795 intelVertex * v0, intelVertex * v1, intelVertex * v2)
797 struct gl_context *ctx = &intel->ctx;
801 fprintf(stderr, "\n%s\n", __FUNCTION__);
803 INTEL_FIREVERTICES(intel);
805 _swsetup_Translate(ctx, v0, &v[0]);
806 _swsetup_Translate(ctx, v1, &v[1]);
807 _swsetup_Translate(ctx, v2, &v[2]);
808 _swrast_render_start(ctx);
809 _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
810 _swrast_render_finish(ctx);
815 intel_fallback_line(struct intel_context *intel,
816 intelVertex * v0, intelVertex * v1)
818 struct gl_context *ctx = &intel->ctx;
822 fprintf(stderr, "\n%s\n", __FUNCTION__);
824 INTEL_FIREVERTICES(intel);
826 _swsetup_Translate(ctx, v0, &v[0]);
827 _swsetup_Translate(ctx, v1, &v[1]);
828 _swrast_render_start(ctx);
829 _swrast_Line(ctx, &v[0], &v[1]);
830 _swrast_render_finish(ctx);
834 intel_fallback_point(struct intel_context *intel,
837 struct gl_context *ctx = &intel->ctx;
841 fprintf(stderr, "\n%s\n", __FUNCTION__);
843 INTEL_FIREVERTICES(intel);
845 _swsetup_Translate(ctx, v0, &v[0]);
846 _swrast_render_start(ctx);
847 _swrast_Point(ctx, &v[0]);
848 _swrast_render_finish(ctx);
852 /**********************************************************************/
853 /* Render unclipped begin/end objects */
854 /**********************************************************************/
857 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
858 #define RENDER_POINTS( start, count ) \
859 for ( ; start < count ; start++) POINT( V(ELT(start)) );
860 #define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) )
861 #define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) )
862 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
863 #define INIT(x) intelRenderPrimitive( ctx, x )
866 struct intel_context *intel = intel_context(ctx); \
867 GLubyte *vertptr = (GLubyte *)intel->verts; \
868 const GLuint vertsize = intel->vertex_size; \
869 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
871 #define RESET_STIPPLE
872 #define RESET_OCCLUSION
873 #define PRESERVE_VB_DEFS
875 #define TAG(x) intel_##x##_verts
876 #include "tnl/t_vb_rendertmp.h"
879 #define TAG(x) intel_##x##_elts
880 #define ELT(x) elt[x]
881 #include "tnl/t_vb_rendertmp.h"
883 /**********************************************************************/
884 /* Render clipped primitives */
885 /**********************************************************************/
890 intelRenderClippedPoly(struct gl_context * ctx, const GLuint * elts, GLuint n)
892 struct intel_context *intel = intel_context(ctx);
893 TNLcontext *tnl = TNL_CONTEXT(ctx);
894 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
895 GLuint prim = intel->render_primitive;
897 /* Render the new vertices as an unclipped polygon.
900 GLuint *tmp = VB->Elts;
901 VB->Elts = (GLuint *) elts;
902 tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
903 PRIM_BEGIN | PRIM_END);
907 /* Restore the render primitive
909 if (prim != GL_POLYGON)
910 tnl->Driver.Render.PrimitiveNotify(ctx, prim);
914 intelRenderClippedLine(struct gl_context * ctx, GLuint ii, GLuint jj)
916 TNLcontext *tnl = TNL_CONTEXT(ctx);
918 tnl->Driver.Render.Line(ctx, ii, jj);
922 intelFastRenderClippedPoly(struct gl_context * ctx, const GLuint * elts, GLuint n)
924 struct intel_context *intel = intel_context(ctx);
925 const GLuint vertsize = intel->vertex_size;
926 GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
927 GLubyte *vertptr = (GLubyte *) intel->verts;
928 const GLuint *start = (const GLuint *) V(elts[0]);
931 for (i = 2; i < n; i++) {
932 COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
933 COPY_DWORDS(j, vb, vertsize, V(elts[i]));
934 COPY_DWORDS(j, vb, vertsize, start);
938 /**********************************************************************/
939 /* Choose render functions */
940 /**********************************************************************/
943 #define DD_TRI_LIGHT_TWOSIDE (1 << 1)
944 #define DD_TRI_UNFILLED (1 << 2)
945 #define DD_TRI_STIPPLE (1 << 4)
946 #define DD_TRI_OFFSET (1 << 5)
947 #define DD_LINE_STIPPLE (1 << 7)
948 #define DD_POINT_ATTEN (1 << 9)
950 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN)
951 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
954 intelChooseRenderState(struct gl_context * ctx)
956 TNLcontext *tnl = TNL_CONTEXT(ctx);
957 struct intel_context *intel = intel_context(ctx);
959 ((ctx->Light.Enabled &&
960 ctx->Light.Model.TwoSide) ? DD_TRI_LIGHT_TWOSIDE : 0) |
961 ((ctx->Polygon.FrontMode != GL_FILL ||
962 ctx->Polygon.BackMode != GL_FILL) ? DD_TRI_UNFILLED : 0) |
963 (ctx->Polygon.StippleFlag ? DD_TRI_STIPPLE : 0) |
964 ((ctx->Polygon.OffsetPoint ||
965 ctx->Polygon.OffsetLine ||
966 ctx->Polygon.OffsetFill) ? DD_TRI_OFFSET : 0) |
967 (ctx->Line.StippleFlag ? DD_LINE_STIPPLE : 0) |
968 (ctx->Point._Attenuated ? DD_POINT_ATTEN : 0);
969 const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
970 bool have_wpos = (fprog && (fprog->Base.InputsRead & VARYING_BIT_POS));
973 if (INTEL_DEBUG & DEBUG_STATE)
974 fprintf(stderr, "\n%s\n", __FUNCTION__);
976 if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
978 if (flags & ANY_RASTER_FLAGS) {
979 if (flags & DD_TRI_LIGHT_TWOSIDE)
980 index |= INTEL_TWOSIDE_BIT;
981 if (flags & DD_TRI_OFFSET)
982 index |= INTEL_OFFSET_BIT;
983 if (flags & DD_TRI_UNFILLED)
984 index |= INTEL_UNFILLED_BIT;
988 intel->draw_point = intel_wpos_point;
989 intel->draw_line = intel_wpos_line;
990 intel->draw_tri = intel_wpos_triangle;
992 /* Make sure these get called:
994 index |= INTEL_FALLBACK_BIT;
997 intel->draw_point = intel_draw_point;
998 intel->draw_line = intel_draw_line;
999 intel->draw_tri = intel_draw_triangle;
1002 /* Hook in fallbacks for specific primitives.
1004 if (flags & ANY_FALLBACK_FLAGS) {
1005 if (flags & DD_LINE_STIPPLE)
1006 intel->draw_line = intel_fallback_line;
1008 if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
1009 intel->draw_tri = intel_fallback_tri;
1011 if (flags & DD_POINT_ATTEN) {
1013 intel->draw_point = intel_atten_point;
1015 intel->draw_point = intel_fallback_point;
1018 index |= INTEL_FALLBACK_BIT;
1022 if (intel->RenderIndex != index) {
1023 intel->RenderIndex = index;
1025 tnl->Driver.Render.Points = rast_tab[index].points;
1026 tnl->Driver.Render.Line = rast_tab[index].line;
1027 tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1028 tnl->Driver.Render.Quad = rast_tab[index].quad;
1031 tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1032 tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1033 tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1034 tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1037 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1038 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1039 tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1040 tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1045 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1059 /**********************************************************************/
1060 /* High level hooks for t_vb_render.c */
1061 /**********************************************************************/
1067 intelRunPipeline(struct gl_context * ctx)
1069 struct intel_context *intel = intel_context(ctx);
1071 _mesa_lock_context_textures(ctx);
1074 _mesa_update_state_locked(ctx);
1076 /* We need to get this done before we start the pipeline, or a
1077 * change in the INTEL_FALLBACK() of its intel_draw_buffers() call
1078 * while the pipeline is running will result in mismatched swrast
1079 * map/unmaps, and later assertion failures.
1081 intel_prepare_render(intel);
1083 if (intel->NewGLState) {
1084 if (intel->NewGLState & _NEW_TEXTURE) {
1085 intel->vtbl.update_texture_state(intel);
1088 if (!intel->Fallback) {
1089 if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1090 intelChooseRenderState(ctx);
1093 intel->NewGLState = 0;
1096 intel->tnl_pipeline_running = true;
1097 _tnl_run_pipeline(ctx);
1098 intel->tnl_pipeline_running = false;
1100 _mesa_unlock_context_textures(ctx);
1104 intelRenderStart(struct gl_context * ctx)
1106 struct intel_context *intel = intel_context(ctx);
1108 intel_check_front_buffer_rendering(intel);
1109 intel->vtbl.render_start(intel_context(ctx));
1110 intel->vtbl.emit_state(intel);
1114 intelRenderFinish(struct gl_context * ctx)
1116 struct intel_context *intel = intel_context(ctx);
1118 if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1121 INTEL_FIREVERTICES(intel);
1127 /* System to flush dma and emit state changes based on the rasterized
1131 intelRasterPrimitive(struct gl_context * ctx, GLenum rprim, GLuint hwprim)
1133 struct intel_context *intel = intel_context(ctx);
1136 fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1137 _mesa_lookup_enum_by_nr(rprim), hwprim);
1139 intel->vtbl.reduced_primitive_state(intel, rprim);
1141 /* Start a new primitive. Arrange to have it flushed later on.
1143 if (hwprim != intel->prim.primitive) {
1144 INTEL_FIREVERTICES(intel);
1146 intel_set_prim(intel, hwprim);
1154 intelRenderPrimitive(struct gl_context * ctx, GLenum prim)
1156 struct intel_context *intel = intel_context(ctx);
1157 GLboolean unfilled = (ctx->Polygon.FrontMode != GL_FILL ||
1158 ctx->Polygon.BackMode != GL_FILL);
1161 fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1163 /* Let some clipping routines know which primitive they're dealing
1166 intel->render_primitive = prim;
1168 /* Shortcircuit this when called for unfilled triangles. The rasterized
1169 * primitive will always be reset by lower level functions in that case,
1170 * potentially pingponging the state:
1172 if (reduced_prim[prim] == GL_TRIANGLES && unfilled)
1175 /* Set some primitive-dependent state and Start? a new primitive.
1177 intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1181 /**********************************************************************/
1182 /* Transition to/from hardware rasterization. */
1183 /**********************************************************************/
1185 static char *fallbackStrings[] = {
1186 [0] = "Draw buffer",
1187 [1] = "Read buffer",
1188 [2] = "Depth buffer",
1189 [3] = "Stencil buffer",
1190 [4] = "User disable",
1191 [5] = "Render mode",
1194 [13] = "Color mask",
1199 [18] = "Smooth polygon",
1200 [19] = "Smooth point",
1201 [20] = "point sprite coord origin",
1202 [21] = "depth/color drawing offset",
1203 [22] = "coord replace(SPRITE POINT ENABLE)",
1208 getFallbackString(GLuint bit)
1215 return fallbackStrings[i];
1221 * Enable/disable a fallback flag.
1222 * \param bit one of INTEL_FALLBACK_x flags.
1225 intelFallback(struct intel_context *intel, GLbitfield bit, bool mode)
1227 struct gl_context *ctx = &intel->ctx;
1228 TNLcontext *tnl = TNL_CONTEXT(ctx);
1229 const GLbitfield oldfallback = intel->Fallback;
1232 intel->Fallback |= bit;
1233 if (oldfallback == 0) {
1234 assert(!intel->tnl_pipeline_running);
1237 if (INTEL_DEBUG & DEBUG_PERF)
1238 fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1239 bit, getFallbackString(bit));
1240 _swsetup_Wakeup(ctx);
1241 intel->RenderIndex = ~0;
1245 intel->Fallback &= ~bit;
1246 if (oldfallback == bit) {
1247 assert(!intel->tnl_pipeline_running);
1250 if (INTEL_DEBUG & DEBUG_PERF)
1251 fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1252 tnl->Driver.Render.Start = intelRenderStart;
1253 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1254 tnl->Driver.Render.Finish = intelRenderFinish;
1255 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1256 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1257 tnl->Driver.Render.Interp = _tnl_interp;
1259 _tnl_invalidate_vertex_state(ctx, ~0);
1260 _tnl_invalidate_vertices(ctx, ~0);
1261 _tnl_install_attrs(ctx,
1262 intel->vertex_attrs,
1263 intel->vertex_attr_count,
1264 intel->ViewportMatrix.m, 0);
1266 intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1277 /**********************************************************************/
1278 /* Initialization. */
1279 /**********************************************************************/
1283 intelInitTriFuncs(struct gl_context * ctx)
1285 TNLcontext *tnl = TNL_CONTEXT(ctx);
1286 static int firsttime = 1;
1293 tnl->Driver.RunPipeline = intelRunPipeline;
1294 tnl->Driver.Render.Start = intelRenderStart;
1295 tnl->Driver.Render.Finish = intelRenderFinish;
1296 tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1297 tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1298 tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1299 tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1300 tnl->Driver.Render.Interp = _tnl_interp;