1 /****************************************************************************
2 * Copyright (C) 2015 Intel Corporation. All Rights Reserved.
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
22 ***************************************************************************/
24 // llvm redefines DEBUG
25 #pragma push_macro("DEBUG")
27 #include "JitManager.h"
28 #pragma pop_macro("DEBUG")
30 #include "common/os.h"
32 #include "gen_state_llvm.h"
33 #include "core/multisample.h"
34 #include "core/state_funcs.h"
36 #include "gallivm/lp_bld_tgsi.h"
37 #include "util/u_format.h"
39 #include "util/u_memory.h"
40 #include "util/u_inlines.h"
41 #include "util/u_helpers.h"
42 #include "util/u_framebuffer.h"
43 #include "util/u_viewport.h"
44 #include "util/u_prim.h"
46 #include "swr_state.h"
47 #include "swr_context.h"
48 #include "gen_swr_context_llvm.h"
49 #include "swr_screen.h"
50 #include "swr_resource.h"
51 #include "swr_tex_sample.h"
52 #include "swr_scratch.h"
53 #include "swr_shader.h"
54 #include "swr_fence.h"
56 /* These should be pulled out into separate files as necessary
57 * Just initializing everything here to get going. */
60 swr_create_blend_state(struct pipe_context
*pipe
,
61 const struct pipe_blend_state
*blend
)
63 struct swr_blend_state
*state
= CALLOC_STRUCT(swr_blend_state
);
65 memcpy(&state
->pipe
, blend
, sizeof(*blend
));
67 struct pipe_blend_state
*pipe_blend
= &state
->pipe
;
70 target
< std::min(SWR_NUM_RENDERTARGETS
, PIPE_MAX_COLOR_BUFS
);
73 struct pipe_rt_blend_state
*rt_blend
= &pipe_blend
->rt
[target
];
74 SWR_RENDER_TARGET_BLEND_STATE
&blendState
=
75 state
->blendState
.renderTarget
[target
];
76 RENDER_TARGET_BLEND_COMPILE_STATE
&compileState
=
77 state
->compileState
[target
];
79 if (target
!= 0 && !pipe_blend
->independent_blend_enable
) {
81 &state
->compileState
[0],
82 sizeof(RENDER_TARGET_BLEND_COMPILE_STATE
));
86 compileState
.blendEnable
= rt_blend
->blend_enable
;
87 if (compileState
.blendEnable
) {
88 compileState
.sourceAlphaBlendFactor
=
89 swr_convert_blend_factor(rt_blend
->alpha_src_factor
);
90 compileState
.destAlphaBlendFactor
=
91 swr_convert_blend_factor(rt_blend
->alpha_dst_factor
);
92 compileState
.sourceBlendFactor
=
93 swr_convert_blend_factor(rt_blend
->rgb_src_factor
);
94 compileState
.destBlendFactor
=
95 swr_convert_blend_factor(rt_blend
->rgb_dst_factor
);
97 compileState
.colorBlendFunc
=
98 swr_convert_blend_func(rt_blend
->rgb_func
);
99 compileState
.alphaBlendFunc
=
100 swr_convert_blend_func(rt_blend
->alpha_func
);
102 compileState
.logicOpEnable
= state
->pipe
.logicop_enable
;
103 if (compileState
.logicOpEnable
) {
104 compileState
.logicOpFunc
=
105 swr_convert_logic_op(state
->pipe
.logicop_func
);
108 blendState
.writeDisableRed
=
109 (rt_blend
->colormask
& PIPE_MASK_R
) ? 0 : 1;
110 blendState
.writeDisableGreen
=
111 (rt_blend
->colormask
& PIPE_MASK_G
) ? 0 : 1;
112 blendState
.writeDisableBlue
=
113 (rt_blend
->colormask
& PIPE_MASK_B
) ? 0 : 1;
114 blendState
.writeDisableAlpha
=
115 (rt_blend
->colormask
& PIPE_MASK_A
) ? 0 : 1;
117 if (rt_blend
->colormask
== 0)
118 compileState
.blendEnable
= false;
125 swr_bind_blend_state(struct pipe_context
*pipe
, void *blend
)
127 struct swr_context
*ctx
= swr_context(pipe
);
129 if (ctx
->blend
== blend
)
132 ctx
->blend
= (swr_blend_state
*)blend
;
134 ctx
->dirty
|= SWR_NEW_BLEND
;
138 swr_delete_blend_state(struct pipe_context
*pipe
, void *blend
)
144 swr_set_blend_color(struct pipe_context
*pipe
,
145 const struct pipe_blend_color
*color
)
147 struct swr_context
*ctx
= swr_context(pipe
);
149 ctx
->blend_color
= *color
;
151 ctx
->dirty
|= SWR_NEW_BLEND
;
155 swr_set_stencil_ref(struct pipe_context
*pipe
,
156 const struct pipe_stencil_ref
*ref
)
158 struct swr_context
*ctx
= swr_context(pipe
);
160 ctx
->stencil_ref
= *ref
;
162 ctx
->dirty
|= SWR_NEW_DEPTH_STENCIL_ALPHA
;
166 swr_create_depth_stencil_state(
167 struct pipe_context
*pipe
,
168 const struct pipe_depth_stencil_alpha_state
*depth_stencil
)
170 struct pipe_depth_stencil_alpha_state
*state
;
172 state
= (pipe_depth_stencil_alpha_state
*)mem_dup(depth_stencil
,
173 sizeof *depth_stencil
);
179 swr_bind_depth_stencil_state(struct pipe_context
*pipe
, void *depth_stencil
)
181 struct swr_context
*ctx
= swr_context(pipe
);
183 if (ctx
->depth_stencil
== (pipe_depth_stencil_alpha_state
*)depth_stencil
)
186 ctx
->depth_stencil
= (pipe_depth_stencil_alpha_state
*)depth_stencil
;
188 ctx
->dirty
|= SWR_NEW_DEPTH_STENCIL_ALPHA
;
192 swr_delete_depth_stencil_state(struct pipe_context
*pipe
, void *depth
)
199 swr_create_rasterizer_state(struct pipe_context
*pipe
,
200 const struct pipe_rasterizer_state
*rast
)
202 struct pipe_rasterizer_state
*state
;
203 state
= (pipe_rasterizer_state
*)mem_dup(rast
, sizeof *rast
);
209 swr_bind_rasterizer_state(struct pipe_context
*pipe
, void *handle
)
211 struct swr_context
*ctx
= swr_context(pipe
);
212 const struct pipe_rasterizer_state
*rasterizer
=
213 (const struct pipe_rasterizer_state
*)handle
;
215 if (ctx
->rasterizer
== (pipe_rasterizer_state
*)rasterizer
)
218 ctx
->rasterizer
= (pipe_rasterizer_state
*)rasterizer
;
220 ctx
->dirty
|= SWR_NEW_RASTERIZER
;
224 swr_delete_rasterizer_state(struct pipe_context
*pipe
, void *rasterizer
)
231 swr_create_sampler_state(struct pipe_context
*pipe
,
232 const struct pipe_sampler_state
*sampler
)
234 struct pipe_sampler_state
*state
=
235 (pipe_sampler_state
*)mem_dup(sampler
, sizeof *sampler
);
241 swr_bind_sampler_states(struct pipe_context
*pipe
,
242 enum pipe_shader_type shader
,
247 struct swr_context
*ctx
= swr_context(pipe
);
250 assert(shader
< PIPE_SHADER_TYPES
);
251 assert(start
+ num
<= ARRAY_SIZE(ctx
->samplers
[shader
]));
253 /* set the new samplers */
254 ctx
->num_samplers
[shader
] = num
;
255 for (i
= 0; i
< num
; i
++) {
256 ctx
->samplers
[shader
][start
+ i
] = (pipe_sampler_state
*)samplers
[i
];
259 ctx
->dirty
|= SWR_NEW_SAMPLER
;
263 swr_delete_sampler_state(struct pipe_context
*pipe
, void *sampler
)
269 static struct pipe_sampler_view
*
270 swr_create_sampler_view(struct pipe_context
*pipe
,
271 struct pipe_resource
*texture
,
272 const struct pipe_sampler_view
*templ
)
274 struct pipe_sampler_view
*view
= CALLOC_STRUCT(pipe_sampler_view
);
278 view
->reference
.count
= 1;
279 view
->texture
= NULL
;
280 pipe_resource_reference(&view
->texture
, texture
);
281 view
->context
= pipe
;
288 swr_set_sampler_views(struct pipe_context
*pipe
,
289 enum pipe_shader_type shader
,
292 struct pipe_sampler_view
**views
)
294 struct swr_context
*ctx
= swr_context(pipe
);
297 assert(num
<= PIPE_MAX_SHADER_SAMPLER_VIEWS
);
299 assert(shader
< PIPE_SHADER_TYPES
);
300 assert(start
+ num
<= ARRAY_SIZE(ctx
->sampler_views
[shader
]));
302 /* set the new sampler views */
303 ctx
->num_sampler_views
[shader
] = num
;
304 for (i
= 0; i
< num
; i
++) {
305 /* Note: we're using pipe_sampler_view_release() here to work around
306 * a possible crash when the old view belongs to another context that
307 * was already destroyed.
309 pipe_sampler_view_release(pipe
, &ctx
->sampler_views
[shader
][start
+ i
]);
310 pipe_sampler_view_reference(&ctx
->sampler_views
[shader
][start
+ i
],
314 ctx
->dirty
|= SWR_NEW_SAMPLER_VIEW
;
318 swr_sampler_view_destroy(struct pipe_context
*pipe
,
319 struct pipe_sampler_view
*view
)
321 pipe_resource_reference(&view
->texture
, NULL
);
326 swr_create_vs_state(struct pipe_context
*pipe
,
327 const struct pipe_shader_state
*vs
)
329 struct swr_vertex_shader
*swr_vs
= new swr_vertex_shader
;
333 swr_vs
->pipe
.tokens
= tgsi_dup_tokens(vs
->tokens
);
334 swr_vs
->pipe
.stream_output
= vs
->stream_output
;
336 lp_build_tgsi_info(vs
->tokens
, &swr_vs
->info
);
338 swr_vs
->soState
= {0};
340 if (swr_vs
->pipe
.stream_output
.num_outputs
) {
341 pipe_stream_output_info
*stream_output
= &swr_vs
->pipe
.stream_output
;
343 swr_vs
->soState
.soEnable
= true;
344 // soState.rasterizerDisable set on state dirty
345 // soState.streamToRasterizer not used
347 for (uint32_t i
= 0; i
< stream_output
->num_outputs
; i
++) {
348 unsigned attrib_slot
= stream_output
->output
[i
].register_index
;
349 attrib_slot
= swr_so_adjust_attrib(attrib_slot
, swr_vs
);
350 swr_vs
->soState
.streamMasks
[stream_output
->output
[i
].stream
] |=
353 for (uint32_t i
= 0; i
< MAX_SO_STREAMS
; i
++) {
354 swr_vs
->soState
.streamNumEntries
[i
] =
355 _mm_popcnt_u32(swr_vs
->soState
.streamMasks
[i
]);
363 swr_bind_vs_state(struct pipe_context
*pipe
, void *vs
)
365 struct swr_context
*ctx
= swr_context(pipe
);
370 ctx
->vs
= (swr_vertex_shader
*)vs
;
371 ctx
->dirty
|= SWR_NEW_VS
;
375 swr_delete_vs_state(struct pipe_context
*pipe
, void *vs
)
377 struct swr_vertex_shader
*swr_vs
= (swr_vertex_shader
*)vs
;
378 FREE((void *)swr_vs
->pipe
.tokens
);
379 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
381 /* Defer deletion of vs state */
382 swr_fence_work_delete_vs(screen
->flush_fence
, swr_vs
);
386 swr_create_fs_state(struct pipe_context
*pipe
,
387 const struct pipe_shader_state
*fs
)
389 struct swr_fragment_shader
*swr_fs
= new swr_fragment_shader
;
393 swr_fs
->pipe
.tokens
= tgsi_dup_tokens(fs
->tokens
);
395 lp_build_tgsi_info(fs
->tokens
, &swr_fs
->info
);
402 swr_bind_fs_state(struct pipe_context
*pipe
, void *fs
)
404 struct swr_context
*ctx
= swr_context(pipe
);
409 ctx
->fs
= (swr_fragment_shader
*)fs
;
410 ctx
->dirty
|= SWR_NEW_FS
;
414 swr_delete_fs_state(struct pipe_context
*pipe
, void *fs
)
416 struct swr_fragment_shader
*swr_fs
= (swr_fragment_shader
*)fs
;
417 FREE((void *)swr_fs
->pipe
.tokens
);
418 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
420 /* Defer deleton of fs state */
421 swr_fence_work_delete_fs(screen
->flush_fence
, swr_fs
);
425 swr_create_gs_state(struct pipe_context
*pipe
,
426 const struct pipe_shader_state
*gs
)
428 struct swr_geometry_shader
*swr_gs
= new swr_geometry_shader
;
432 swr_gs
->pipe
.tokens
= tgsi_dup_tokens(gs
->tokens
);
434 lp_build_tgsi_info(gs
->tokens
, &swr_gs
->info
);
441 swr_bind_gs_state(struct pipe_context
*pipe
, void *gs
)
443 struct swr_context
*ctx
= swr_context(pipe
);
448 ctx
->gs
= (swr_geometry_shader
*)gs
;
449 ctx
->dirty
|= SWR_NEW_GS
;
453 swr_delete_gs_state(struct pipe_context
*pipe
, void *gs
)
455 struct swr_geometry_shader
*swr_gs
= (swr_geometry_shader
*)gs
;
456 FREE((void *)swr_gs
->pipe
.tokens
);
457 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
459 /* Defer deleton of fs state */
460 swr_fence_work_delete_gs(screen
->flush_fence
, swr_gs
);
464 swr_set_constant_buffer(struct pipe_context
*pipe
,
465 enum pipe_shader_type shader
,
467 const struct pipe_constant_buffer
*cb
)
469 struct swr_context
*ctx
= swr_context(pipe
);
470 struct pipe_resource
*constants
= cb
? cb
->buffer
: NULL
;
472 assert(shader
< PIPE_SHADER_TYPES
);
473 assert(index
< ARRAY_SIZE(ctx
->constants
[shader
]));
475 /* note: reference counting */
476 util_copy_constant_buffer(&ctx
->constants
[shader
][index
], cb
);
478 if (shader
== PIPE_SHADER_VERTEX
) {
479 ctx
->dirty
|= SWR_NEW_VSCONSTANTS
;
480 } else if (shader
== PIPE_SHADER_FRAGMENT
) {
481 ctx
->dirty
|= SWR_NEW_FSCONSTANTS
;
482 } else if (shader
== PIPE_SHADER_GEOMETRY
) {
483 ctx
->dirty
|= SWR_NEW_GSCONSTANTS
;
486 if (cb
&& cb
->user_buffer
) {
487 pipe_resource_reference(&constants
, NULL
);
493 swr_create_vertex_elements_state(struct pipe_context
*pipe
,
494 unsigned num_elements
,
495 const struct pipe_vertex_element
*attribs
)
497 struct swr_vertex_element_state
*velems
;
498 assert(num_elements
<= PIPE_MAX_ATTRIBS
);
499 velems
= new swr_vertex_element_state
;
501 memset(&velems
->fsState
, 0, sizeof(velems
->fsState
));
502 velems
->fsState
.bVertexIDOffsetEnable
= true;
503 velems
->fsState
.numAttribs
= num_elements
;
504 for (unsigned i
= 0; i
< num_elements
; i
++) {
505 // XXX: we should do this keyed on the VS usage info
507 const struct util_format_description
*desc
=
508 util_format_description(attribs
[i
].src_format
);
510 velems
->fsState
.layout
[i
].AlignedByteOffset
= attribs
[i
].src_offset
;
511 velems
->fsState
.layout
[i
].Format
=
512 mesa_to_swr_format(attribs
[i
].src_format
);
513 velems
->fsState
.layout
[i
].StreamIndex
=
514 attribs
[i
].vertex_buffer_index
;
515 velems
->fsState
.layout
[i
].InstanceEnable
=
516 attribs
[i
].instance_divisor
!= 0;
517 velems
->fsState
.layout
[i
].ComponentControl0
=
518 desc
->channel
[0].type
!= UTIL_FORMAT_TYPE_VOID
519 ? ComponentControl::StoreSrc
520 : ComponentControl::Store0
;
521 velems
->fsState
.layout
[i
].ComponentControl1
=
522 desc
->channel
[1].type
!= UTIL_FORMAT_TYPE_VOID
523 ? ComponentControl::StoreSrc
524 : ComponentControl::Store0
;
525 velems
->fsState
.layout
[i
].ComponentControl2
=
526 desc
->channel
[2].type
!= UTIL_FORMAT_TYPE_VOID
527 ? ComponentControl::StoreSrc
528 : ComponentControl::Store0
;
529 velems
->fsState
.layout
[i
].ComponentControl3
=
530 desc
->channel
[3].type
!= UTIL_FORMAT_TYPE_VOID
531 ? ComponentControl::StoreSrc
532 : ComponentControl::Store1Fp
;
533 velems
->fsState
.layout
[i
].ComponentPacking
= ComponentEnable::XYZW
;
534 velems
->fsState
.layout
[i
].InstanceAdvancementState
=
535 attribs
[i
].instance_divisor
;
537 /* Calculate the pitch of each stream */
538 const SWR_FORMAT_INFO
&swr_desc
= GetFormatInfo(
539 mesa_to_swr_format(attribs
[i
].src_format
));
540 velems
->stream_pitch
[attribs
[i
].vertex_buffer_index
] += swr_desc
.Bpp
;
542 if (attribs
[i
].instance_divisor
!= 0) {
543 velems
->instanced_bufs
|= 1U << attribs
[i
].vertex_buffer_index
;
544 uint32_t *min_instance_div
=
545 &velems
->min_instance_div
[attribs
[i
].vertex_buffer_index
];
546 if (!*min_instance_div
||
547 attribs
[i
].instance_divisor
< *min_instance_div
)
548 *min_instance_div
= attribs
[i
].instance_divisor
;
557 swr_bind_vertex_elements_state(struct pipe_context
*pipe
, void *velems
)
559 struct swr_context
*ctx
= swr_context(pipe
);
560 struct swr_vertex_element_state
*swr_velems
=
561 (struct swr_vertex_element_state
*)velems
;
563 ctx
->velems
= swr_velems
;
564 ctx
->dirty
|= SWR_NEW_VERTEX
;
568 swr_delete_vertex_elements_state(struct pipe_context
*pipe
, void *velems
)
570 struct swr_vertex_element_state
*swr_velems
=
571 (struct swr_vertex_element_state
*) velems
;
572 /* XXX Need to destroy fetch shader? */
578 swr_set_vertex_buffers(struct pipe_context
*pipe
,
580 unsigned num_elements
,
581 const struct pipe_vertex_buffer
*buffers
)
583 struct swr_context
*ctx
= swr_context(pipe
);
585 assert(num_elements
<= PIPE_MAX_ATTRIBS
);
587 util_set_vertex_buffers_count(ctx
->vertex_buffer
,
588 &ctx
->num_vertex_buffers
,
593 ctx
->dirty
|= SWR_NEW_VERTEX
;
598 swr_set_polygon_stipple(struct pipe_context
*pipe
,
599 const struct pipe_poly_stipple
*stipple
)
601 struct swr_context
*ctx
= swr_context(pipe
);
603 ctx
->poly_stipple
.pipe
= *stipple
; /* struct copy */
604 ctx
->dirty
|= SWR_NEW_STIPPLE
;
608 swr_set_clip_state(struct pipe_context
*pipe
,
609 const struct pipe_clip_state
*clip
)
611 struct swr_context
*ctx
= swr_context(pipe
);
614 /* XXX Unimplemented, but prevents crash */
616 ctx
->dirty
|= SWR_NEW_CLIP
;
621 swr_set_scissor_states(struct pipe_context
*pipe
,
623 unsigned num_viewports
,
624 const struct pipe_scissor_state
*scissor
)
626 struct swr_context
*ctx
= swr_context(pipe
);
628 ctx
->scissor
= *scissor
;
629 ctx
->swr_scissor
.xmin
= scissor
->minx
;
630 ctx
->swr_scissor
.xmax
= scissor
->maxx
;
631 ctx
->swr_scissor
.ymin
= scissor
->miny
;
632 ctx
->swr_scissor
.ymax
= scissor
->maxy
;
633 ctx
->dirty
|= SWR_NEW_SCISSOR
;
637 swr_set_viewport_states(struct pipe_context
*pipe
,
639 unsigned num_viewports
,
640 const struct pipe_viewport_state
*vpt
)
642 struct swr_context
*ctx
= swr_context(pipe
);
644 ctx
->viewport
= *vpt
;
645 ctx
->dirty
|= SWR_NEW_VIEWPORT
;
650 swr_set_framebuffer_state(struct pipe_context
*pipe
,
651 const struct pipe_framebuffer_state
*fb
)
653 struct swr_context
*ctx
= swr_context(pipe
);
655 boolean changed
= !util_framebuffer_state_equal(&ctx
->framebuffer
, fb
);
657 assert(fb
->width
<= KNOB_GUARDBAND_WIDTH
);
658 assert(fb
->height
<= KNOB_GUARDBAND_HEIGHT
);
661 util_copy_framebuffer_state(&ctx
->framebuffer
, fb
);
663 /* 0 and 1 both indicate no msaa. Core doesn't understand 0 samples */
664 ctx
->framebuffer
.samples
= std::max((ubyte
)1, ctx
->framebuffer
.samples
);
666 ctx
->dirty
|= SWR_NEW_FRAMEBUFFER
;
672 swr_set_sample_mask(struct pipe_context
*pipe
, unsigned sample_mask
)
674 struct swr_context
*ctx
= swr_context(pipe
);
676 if (sample_mask
!= ctx
->sample_mask
) {
677 ctx
->sample_mask
= sample_mask
;
678 ctx
->dirty
|= SWR_NEW_RASTERIZER
;
683 * MSAA fixed sample position table
684 * used by update_derived and get_sample_position
685 * (integer locations on a 16x16 grid)
687 static const uint8_t swr_sample_positions
[][2] =
689 /* 2x*/ {12,12},{ 4, 4},
690 /* 4x*/ { 6, 2},{14, 6},{ 2,10},{10,14},
691 /* 8x*/ { 9, 5},{ 7,11},{13, 9},{ 5, 3},
692 { 3,13},{ 1, 7},{11,15},{15, 1},
693 /*16x*/ { 9, 9},{ 7, 5},{ 5,10},{12, 7},
694 { 3, 6},{10,13},{13,11},{11, 3},
695 { 6,14},{ 8, 1},{ 4, 2},{ 2,12},
696 { 0, 8},{15, 4},{14,15},{ 1, 0} };
699 swr_get_sample_position(struct pipe_context
*pipe
,
700 unsigned sample_count
, unsigned sample_index
,
703 /* validate sample_count */
704 sample_count
= GetNumSamples(GetSampleCount(sample_count
));
706 const uint8_t *sample
= swr_sample_positions
[sample_count
-1 + sample_index
];
707 out_value
[0] = sample
[0] / 16.0f
;
708 out_value
[1] = sample
[1] / 16.0f
;
713 * Update resource in-use status
714 * All resources bound to color or depth targets marked as WRITE resources.
715 * VBO Vertex/index buffers and texture views marked as READ resources.
718 swr_update_resource_status(struct pipe_context
*pipe
,
719 const struct pipe_draw_info
*p_draw_info
)
721 struct swr_context
*ctx
= swr_context(pipe
);
722 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
724 /* colorbuffer targets */
726 for (uint32_t i
= 0; i
< fb
->nr_cbufs
; ++i
)
728 swr_resource_write(fb
->cbufs
[i
]->texture
);
730 /* depth/stencil target */
732 swr_resource_write(fb
->zsbuf
->texture
);
734 /* VBO vertex buffers */
735 for (uint32_t i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
736 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
737 if (!vb
->is_user_buffer
)
738 swr_resource_read(vb
->buffer
.resource
);
741 /* VBO index buffer */
742 if (p_draw_info
&& p_draw_info
->index_size
) {
743 if (!p_draw_info
->has_user_indices
)
744 swr_resource_read(p_draw_info
->index
.resource
);
747 /* transform feedback buffers */
748 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
749 struct pipe_stream_output_target
*target
= ctx
->so_targets
[i
];
750 if (target
&& target
->buffer
)
751 swr_resource_write(target
->buffer
);
754 /* texture sampler views */
755 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
756 for (uint32_t i
= 0; i
< ctx
->num_sampler_views
[j
]; i
++) {
757 struct pipe_sampler_view
*view
= ctx
->sampler_views
[j
][i
];
759 swr_resource_read(view
->texture
);
763 /* constant buffers */
764 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
765 for (uint32_t i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
766 struct pipe_constant_buffer
*cb
= &ctx
->constants
[j
][i
];
768 swr_resource_read(cb
->buffer
);
774 swr_update_texture_state(struct swr_context
*ctx
,
775 enum pipe_shader_type shader_type
,
776 unsigned num_sampler_views
,
777 swr_jit_texture
*textures
)
779 for (unsigned i
= 0; i
< num_sampler_views
; i
++) {
780 struct pipe_sampler_view
*view
=
781 ctx
->sampler_views
[shader_type
][i
];
782 struct swr_jit_texture
*jit_tex
= &textures
[i
];
784 memset(jit_tex
, 0, sizeof(*jit_tex
));
786 struct pipe_resource
*res
= view
->texture
;
787 struct swr_resource
*swr_res
= swr_resource(res
);
788 SWR_SURFACE_STATE
*swr
= &swr_res
->swr
;
789 size_t *mip_offsets
= swr_res
->mip_offsets
;
790 if (swr_res
->has_depth
&& swr_res
->has_stencil
&&
791 !util_format_has_depth(util_format_description(view
->format
))) {
792 swr
= &swr_res
->secondary
;
793 mip_offsets
= swr_res
->secondary_mip_offsets
;
796 jit_tex
->width
= res
->width0
;
797 jit_tex
->height
= res
->height0
;
798 jit_tex
->base_ptr
= (uint8_t*)swr
->xpBaseAddress
;
799 if (view
->target
!= PIPE_BUFFER
) {
800 jit_tex
->first_level
= view
->u
.tex
.first_level
;
801 jit_tex
->last_level
= view
->u
.tex
.last_level
;
802 if (view
->target
== PIPE_TEXTURE_3D
)
803 jit_tex
->depth
= res
->depth0
;
806 view
->u
.tex
.last_layer
- view
->u
.tex
.first_layer
+ 1;
807 jit_tex
->base_ptr
+= view
->u
.tex
.first_layer
*
808 swr
->qpitch
* swr
->pitch
;
810 unsigned view_blocksize
= util_format_get_blocksize(view
->format
);
811 jit_tex
->base_ptr
+= view
->u
.buf
.offset
;
812 jit_tex
->width
= view
->u
.buf
.size
/ view_blocksize
;
816 for (unsigned level
= jit_tex
->first_level
;
817 level
<= jit_tex
->last_level
;
819 jit_tex
->row_stride
[level
] = swr
->pitch
;
820 jit_tex
->img_stride
[level
] = swr
->qpitch
* swr
->pitch
;
821 jit_tex
->mip_offsets
[level
] = mip_offsets
[level
];
828 swr_update_sampler_state(struct swr_context
*ctx
,
829 enum pipe_shader_type shader_type
,
830 unsigned num_samplers
,
831 swr_jit_sampler
*samplers
)
833 for (unsigned i
= 0; i
< num_samplers
; i
++) {
834 const struct pipe_sampler_state
*sampler
=
835 ctx
->samplers
[shader_type
][i
];
838 samplers
[i
].min_lod
= sampler
->min_lod
;
839 samplers
[i
].max_lod
= sampler
->max_lod
;
840 samplers
[i
].lod_bias
= sampler
->lod_bias
;
841 COPY_4V(samplers
[i
].border_color
, sampler
->border_color
.f
);
847 swr_update_constants(struct swr_context
*ctx
, enum pipe_shader_type shaderType
)
849 swr_draw_context
*pDC
= &ctx
->swrDC
;
851 const float **constant
;
852 uint32_t *num_constants
;
853 struct swr_scratch_space
*scratch
;
855 switch (shaderType
) {
856 case PIPE_SHADER_VERTEX
:
857 constant
= pDC
->constantVS
;
858 num_constants
= pDC
->num_constantsVS
;
859 scratch
= &ctx
->scratch
->vs_constants
;
861 case PIPE_SHADER_FRAGMENT
:
862 constant
= pDC
->constantFS
;
863 num_constants
= pDC
->num_constantsFS
;
864 scratch
= &ctx
->scratch
->fs_constants
;
866 case PIPE_SHADER_GEOMETRY
:
867 constant
= pDC
->constantGS
;
868 num_constants
= pDC
->num_constantsGS
;
869 scratch
= &ctx
->scratch
->gs_constants
;
872 debug_printf("Unsupported shader type constants\n");
876 for (UINT i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
877 const pipe_constant_buffer
*cb
= &ctx
->constants
[shaderType
][i
];
878 num_constants
[i
] = cb
->buffer_size
;
881 (const float *)(swr_resource_data(cb
->buffer
) +
884 /* Need to copy these constants to scratch space */
885 if (cb
->user_buffer
&& cb
->buffer_size
) {
887 ((const uint8_t *)cb
->user_buffer
+ cb
->buffer_offset
);
888 uint32_t size
= AlignUp(cb
->buffer_size
, 4);
889 ptr
= swr_copy_to_scratch_space(ctx
, scratch
, ptr
, size
);
890 constant
[i
] = (const float *)ptr
;
897 swr_change_rt(struct swr_context
*ctx
,
899 const struct pipe_surface
*sf
)
901 swr_draw_context
*pDC
= &ctx
->swrDC
;
902 struct SWR_SURFACE_STATE
*rt
= &pDC
->renderTargets
[attachment
];
904 /* Do nothing if the render target hasn't changed */
905 if ((!sf
|| !sf
->texture
) && (void*)(rt
->xpBaseAddress
) == nullptr)
908 /* Deal with disabling RT up front */
909 if (!sf
|| !sf
->texture
) {
910 /* If detaching attachment, mark tiles as RESOLVED so core
911 * won't try to load from non-existent target. */
912 swr_store_render_target(&ctx
->pipe
, attachment
, SWR_TILE_RESOLVED
);
917 const struct swr_resource
*swr
= swr_resource(sf
->texture
);
918 const SWR_SURFACE_STATE
*swr_surface
= &swr
->swr
;
919 SWR_FORMAT fmt
= mesa_to_swr_format(sf
->format
);
921 if (attachment
== SWR_ATTACHMENT_STENCIL
&& swr
->secondary
.xpBaseAddress
) {
922 swr_surface
= &swr
->secondary
;
923 fmt
= swr_surface
->format
;
926 if (rt
->xpBaseAddress
== swr_surface
->xpBaseAddress
&&
928 rt
->lod
== sf
->u
.tex
.level
&&
929 rt
->arrayIndex
== sf
->u
.tex
.first_layer
)
932 bool need_fence
= false;
934 /* StoreTile for changed target */
935 if (rt
->xpBaseAddress
) {
936 /* If changing attachment to a new target, mark tiles as
937 * INVALID so they are reloaded from surface. */
938 swr_store_render_target(&ctx
->pipe
, attachment
, SWR_TILE_INVALID
);
941 /* if no previous attachment, invalidate tiles that may be marked
942 * RESOLVED because of an old attachment */
943 swr_invalidate_render_target(&ctx
->pipe
, attachment
, sf
->width
, sf
->height
);
944 /* no need to set fence here */
947 /* Make new attachment */
950 rt
->lod
= sf
->u
.tex
.level
;
951 rt
->arrayIndex
= sf
->u
.tex
.first_layer
;
957 * for cases where resources are shared between contexts, invalidate
958 * this ctx's resource. so it can be fetched fresh. Old ctx's resource
959 * is already stored during a flush
962 swr_invalidate_buffers_after_ctx_change(struct pipe_context
*pipe
)
964 struct swr_context
*ctx
= swr_context(pipe
);
966 for (uint32_t i
= 0; i
< ctx
->framebuffer
.nr_cbufs
; i
++) {
967 struct pipe_surface
*cb
= ctx
->framebuffer
.cbufs
[i
];
969 struct swr_resource
*res
= swr_resource(cb
->texture
);
970 if (res
->curr_pipe
!= pipe
) {
971 /* if curr_pipe is NULL (first use), status should not be WRITE */
972 assert(res
->curr_pipe
|| !(res
->status
& SWR_RESOURCE_WRITE
));
973 if (res
->status
& SWR_RESOURCE_WRITE
) {
974 swr_invalidate_render_target(pipe
, i
, cb
->width
, cb
->height
);
977 res
->curr_pipe
= pipe
;
980 if (ctx
->framebuffer
.zsbuf
) {
981 struct pipe_surface
*zb
= ctx
->framebuffer
.zsbuf
;
983 struct swr_resource
*res
= swr_resource(zb
->texture
);
984 if (res
->curr_pipe
!= pipe
) {
985 /* if curr_pipe is NULL (first use), status should not be WRITE */
986 assert(res
->curr_pipe
|| !(res
->status
& SWR_RESOURCE_WRITE
));
987 if (res
->status
& SWR_RESOURCE_WRITE
) {
988 swr_invalidate_render_target(pipe
, SWR_ATTACHMENT_DEPTH
, zb
->width
, zb
->height
);
989 swr_invalidate_render_target(pipe
, SWR_ATTACHMENT_STENCIL
, zb
->width
, zb
->height
);
992 res
->curr_pipe
= pipe
;
998 swr_user_vbuf_range(const struct pipe_draw_info
*info
,
999 const struct swr_vertex_element_state
*velems
,
1000 const struct pipe_vertex_buffer
*vb
,
1006 /* FIXME: The size is too large - we don't access the full extra stride. */
1008 if (velems
->instanced_bufs
& (1U << i
)) {
1009 elems
= info
->instance_count
/ velems
->min_instance_div
[i
] + 1;
1010 *totelems
= info
->start_instance
+ elems
;
1011 *base
= info
->start_instance
* vb
->stride
;
1012 *size
= elems
* vb
->stride
;
1013 } else if (vb
->stride
) {
1014 elems
= info
->max_index
- info
->min_index
+ 1;
1015 *totelems
= (info
->max_index
+ info
->index_bias
) + 1;
1016 *base
= (info
->min_index
+ info
->index_bias
) * vb
->stride
;
1017 *size
= elems
* vb
->stride
;
1021 *size
= velems
->stream_pitch
[i
];
1026 swr_update_poly_stipple(struct swr_context
*ctx
)
1028 struct swr_draw_context
*pDC
= &ctx
->swrDC
;
1030 assert(sizeof(ctx
->poly_stipple
.pipe
.stipple
) == sizeof(pDC
->polyStipple
));
1031 memcpy(pDC
->polyStipple
,
1032 ctx
->poly_stipple
.pipe
.stipple
,
1033 sizeof(ctx
->poly_stipple
.pipe
.stipple
));
1037 swr_update_derived(struct pipe_context
*pipe
,
1038 const struct pipe_draw_info
*p_draw_info
)
1040 struct swr_context
*ctx
= swr_context(pipe
);
1041 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
1043 /* When called from swr_clear (p_draw_info = null), set any null
1044 * state-objects to the dummy state objects to prevent nullptr dereference
1045 * in validation below.
1047 * Important that this remains static for zero initialization. These
1048 * aren't meant to be proper state objects, just empty structs. They will
1049 * not be written to.
1051 * Shaders can't be part of the union since they contain std::unordered_map
1055 struct pipe_rasterizer_state rasterizer
;
1056 struct pipe_depth_stencil_alpha_state depth_stencil
;
1057 struct swr_blend_state blend
;
1059 struct swr_vertex_shader vs
;
1060 struct swr_fragment_shader fs
;
1064 if (!ctx
->rasterizer
)
1065 ctx
->rasterizer
= &swr_dummy
.state
.rasterizer
;
1066 if (!ctx
->depth_stencil
)
1067 ctx
->depth_stencil
= &swr_dummy
.state
.depth_stencil
;
1069 ctx
->blend
= &swr_dummy
.state
.blend
;
1071 ctx
->vs
= &swr_dummy
.vs
;
1073 ctx
->fs
= &swr_dummy
.fs
;
1076 /* Update screen->pipe to current pipe context. */
1077 screen
->pipe
= pipe
;
1079 /* Any state that requires dirty flags to be re-triggered sets this mask */
1080 /* For example, user_buffer vertex and index buffers. */
1081 unsigned post_update_dirty_flags
= 0;
1083 /* bring resources that changed context up-to-date */
1084 swr_invalidate_buffers_after_ctx_change(pipe
);
1086 /* Render Targets */
1087 if (ctx
->dirty
& SWR_NEW_FRAMEBUFFER
) {
1088 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1089 const struct util_format_description
*desc
= NULL
;
1090 bool need_fence
= false;
1092 /* colorbuffer targets */
1094 for (unsigned i
= 0; i
< fb
->nr_cbufs
; ++i
)
1095 need_fence
|= swr_change_rt(
1096 ctx
, SWR_ATTACHMENT_COLOR0
+ i
, fb
->cbufs
[i
]);
1098 for (unsigned i
= fb
->nr_cbufs
; i
< SWR_NUM_RENDERTARGETS
; ++i
)
1099 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_COLOR0
+ i
, NULL
);
1101 /* depth/stencil target */
1103 desc
= util_format_description(fb
->zsbuf
->format
);
1104 if (fb
->zsbuf
&& util_format_has_depth(desc
))
1105 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_DEPTH
, fb
->zsbuf
);
1107 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_DEPTH
, NULL
);
1109 if (fb
->zsbuf
&& util_format_has_stencil(desc
))
1110 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_STENCIL
, fb
->zsbuf
);
1112 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_STENCIL
, NULL
);
1114 /* This fence ensures any attachment changes are resolved before the
1117 swr_fence_submit(ctx
, screen
->flush_fence
);
1121 if (ctx
->dirty
& (SWR_NEW_RASTERIZER
|
1122 SWR_NEW_VS
| // clipping
1123 SWR_NEW_FRAMEBUFFER
)) {
1124 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
1125 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1127 SWR_RASTSTATE
*rastState
= &ctx
->derived
.rastState
;
1128 rastState
->cullMode
= swr_convert_cull_mode(rasterizer
->cull_face
);
1129 rastState
->frontWinding
= rasterizer
->front_ccw
1130 ? SWR_FRONTWINDING_CCW
1131 : SWR_FRONTWINDING_CW
;
1132 rastState
->scissorEnable
= rasterizer
->scissor
;
1133 rastState
->pointSize
= rasterizer
->point_size
> 0.0f
1134 ? rasterizer
->point_size
1136 rastState
->lineWidth
= rasterizer
->line_width
> 0.0f
1137 ? rasterizer
->line_width
1140 rastState
->pointParam
= rasterizer
->point_size_per_vertex
;
1142 rastState
->pointSpriteEnable
= rasterizer
->sprite_coord_enable
;
1143 rastState
->pointSpriteTopOrigin
=
1144 rasterizer
->sprite_coord_mode
== PIPE_SPRITE_COORD_UPPER_LEFT
;
1146 /* If SWR_MSAA_FORCE_ENABLE is set, turn msaa on */
1147 if (screen
->msaa_force_enable
&& !rasterizer
->multisample
) {
1148 /* Force enable and use the value the surface was created with */
1149 rasterizer
->multisample
= true;
1150 fb
->samples
= swr_resource(fb
->cbufs
[0]->texture
)->swr
.numSamples
;
1151 fprintf(stderr
,"msaa force enable: %d samples\n", fb
->samples
);
1154 rastState
->sampleCount
= GetSampleCount(fb
->samples
);
1155 rastState
->forcedSampleCount
= false;
1156 rastState
->bIsCenterPattern
= !rasterizer
->multisample
;
1157 rastState
->pixelLocation
= SWR_PIXEL_LOCATION_CENTER
;
1159 /* Only initialize sample positions if msaa is enabled */
1160 if (rasterizer
->multisample
) {
1161 for (uint32_t i
= 0; i
< fb
->samples
; i
++) {
1162 const uint8_t *sample
= swr_sample_positions
[fb
->samples
-1 + i
];
1163 rastState
->samplePositions
.SetXi(i
, sample
[0] << 4);
1164 rastState
->samplePositions
.SetYi(i
, sample
[1] << 4);
1165 rastState
->samplePositions
.SetX (i
, sample
[0] / 16.0f
);
1166 rastState
->samplePositions
.SetY (i
, sample
[1] / 16.0f
);
1168 rastState
->samplePositions
.PrecalcSampleData(fb
->samples
);
1171 bool do_offset
= false;
1172 switch (rasterizer
->fill_front
) {
1173 case PIPE_POLYGON_MODE_FILL
:
1174 do_offset
= rasterizer
->offset_tri
;
1176 case PIPE_POLYGON_MODE_LINE
:
1177 do_offset
= rasterizer
->offset_line
;
1179 case PIPE_POLYGON_MODE_POINT
:
1180 do_offset
= rasterizer
->offset_point
;
1185 rastState
->depthBias
= rasterizer
->offset_units
;
1186 rastState
->slopeScaledDepthBias
= rasterizer
->offset_scale
;
1187 rastState
->depthBiasClamp
= rasterizer
->offset_clamp
;
1189 rastState
->depthBias
= 0;
1190 rastState
->slopeScaledDepthBias
= 0;
1191 rastState
->depthBiasClamp
= 0;
1194 /* translate polygon mode, at least for the front==back case */
1195 rastState
->fillMode
= swr_convert_fill_mode(rasterizer
->fill_front
);
1197 struct pipe_surface
*zb
= fb
->zsbuf
;
1198 if (zb
&& swr_resource(zb
->texture
)->has_depth
)
1199 rastState
->depthFormat
= swr_resource(zb
->texture
)->swr
.format
;
1201 rastState
->depthClipEnable
= rasterizer
->depth_clip
;
1202 rastState
->clipHalfZ
= rasterizer
->clip_halfz
;
1204 ctx
->api
.pfnSwrSetRastState(ctx
->swrContext
, rastState
);
1208 if (ctx
->dirty
& (SWR_NEW_VIEWPORT
| SWR_NEW_FRAMEBUFFER
1209 | SWR_NEW_RASTERIZER
)) {
1210 pipe_viewport_state
*state
= &ctx
->viewport
;
1211 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1212 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
1214 SWR_VIEWPORT
*vp
= &ctx
->derived
.vp
;
1215 SWR_VIEWPORT_MATRICES
*vpm
= &ctx
->derived
.vpm
;
1217 vp
->x
= state
->translate
[0] - state
->scale
[0];
1218 vp
->width
= 2 * state
->scale
[0];
1219 vp
->y
= state
->translate
[1] - fabs(state
->scale
[1]);
1220 vp
->height
= 2 * fabs(state
->scale
[1]);
1221 util_viewport_zmin_zmax(state
, rasterizer
->clip_halfz
,
1222 &vp
->minZ
, &vp
->maxZ
);
1224 vpm
->m00
[0] = state
->scale
[0];
1225 vpm
->m11
[0] = state
->scale
[1];
1226 vpm
->m22
[0] = state
->scale
[2];
1227 vpm
->m30
[0] = state
->translate
[0];
1228 vpm
->m31
[0] = state
->translate
[1];
1229 vpm
->m32
[0] = state
->translate
[2];
1231 /* Now that the matrix is calculated, clip the view coords to screen
1232 * size. OpenGL allows for -ve x,y in the viewport. */
1238 vp
->height
+= vp
->y
;
1241 vp
->width
= std::min(vp
->width
, (float)fb
->width
- vp
->x
);
1242 vp
->height
= std::min(vp
->height
, (float)fb
->height
- vp
->y
);
1244 ctx
->api
.pfnSwrSetViewports(ctx
->swrContext
, 1, vp
, vpm
);
1247 /* When called from swr_clear (p_draw_info = null), render targets,
1248 * rasterState and viewports (dependent on render targets) are the only
1249 * necessary validation. Defer remaining validation by setting
1250 * post_update_dirty_flags and clear all dirty flags. BackendState is
1251 * still unconditionally validated below */
1253 post_update_dirty_flags
= ctx
->dirty
& ~(SWR_NEW_FRAMEBUFFER
|
1254 SWR_NEW_RASTERIZER
|
1260 if (ctx
->dirty
& SWR_NEW_SCISSOR
) {
1261 ctx
->api
.pfnSwrSetScissorRects(ctx
->swrContext
, 1, &ctx
->swr_scissor
);
1264 /* Set vertex & index buffers */
1265 if (ctx
->dirty
& SWR_NEW_VERTEX
) {
1266 const struct pipe_draw_info
&info
= *p_draw_info
;
1268 /* vertex buffers */
1269 SWR_VERTEX_BUFFER_STATE swrVertexBuffers
[PIPE_MAX_ATTRIBS
];
1270 for (UINT i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
1271 uint32_t size
, pitch
, elems
, partial_inbounds
;
1272 uint32_t min_vertex_index
;
1273 const uint8_t *p_data
;
1274 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
1277 if (!vb
->is_user_buffer
) {
1280 /* If pitch=0 (ie vb->stride), buffer contains a single
1281 * constant attribute. Use the stream_pitch which was
1282 * calculated during creation of vertex_elements_state for the
1283 * size of the attribute. */
1284 size
= ctx
->velems
->stream_pitch
[i
];
1286 partial_inbounds
= 0;
1287 min_vertex_index
= 0;
1289 /* size is based on buffer->width0 rather than info.max_index
1290 * to prevent having to validate VBO on each draw. */
1291 size
= vb
->buffer
.resource
->width0
;
1292 elems
= size
/ pitch
;
1293 partial_inbounds
= size
% pitch
;
1294 min_vertex_index
= 0;
1297 p_data
= swr_resource_data(vb
->buffer
.resource
) + vb
->buffer_offset
;
1300 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1301 * revalidate on each draw */
1302 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1305 swr_user_vbuf_range(&info
, ctx
->velems
, vb
, i
, &elems
, &base
, &size
);
1306 partial_inbounds
= 0;
1307 min_vertex_index
= info
.min_index
+ info
.index_bias
;
1309 size
= AlignUp(size
, 4);
1310 /* If size of client memory copy is too large, don't copy. The
1311 * draw will access user-buffer directly and then block. This is
1312 * faster than queuing many large client draws. */
1313 if (size
>= screen
->client_copy_limit
) {
1314 post_update_dirty_flags
|= SWR_LARGE_CLIENT_DRAW
;
1315 p_data
= (const uint8_t *) vb
->buffer
.user
;
1317 /* Copy only needed vertices to scratch space */
1318 const void *ptr
= (const uint8_t *) vb
->buffer
.user
+ base
;
1319 ptr
= (uint8_t *)swr_copy_to_scratch_space(
1320 ctx
, &ctx
->scratch
->vertex_buffer
, ptr
, size
);
1321 p_data
= (const uint8_t *)ptr
- base
;
1325 swrVertexBuffers
[i
] = {0};
1326 swrVertexBuffers
[i
].index
= i
;
1327 swrVertexBuffers
[i
].pitch
= pitch
;
1328 swrVertexBuffers
[i
].xpData
= (gfxptr_t
) p_data
;
1329 swrVertexBuffers
[i
].size
= size
;
1330 swrVertexBuffers
[i
].minVertex
= min_vertex_index
;
1331 swrVertexBuffers
[i
].maxVertex
= elems
;
1332 swrVertexBuffers
[i
].partialInboundsSize
= partial_inbounds
;
1335 ctx
->api
.pfnSwrSetVertexBuffers(
1336 ctx
->swrContext
, ctx
->num_vertex_buffers
, swrVertexBuffers
);
1338 /* index buffer, if required (info passed in by swr_draw_vbo) */
1339 SWR_FORMAT index_type
= R32_UINT
; /* Default for non-indexed draws */
1340 if (info
.index_size
) {
1341 const uint8_t *p_data
;
1342 uint32_t size
, pitch
;
1344 pitch
= info
.index_size
? info
.index_size
: sizeof(uint32_t);
1345 index_type
= swr_convert_index_type(pitch
);
1347 if (!info
.has_user_indices
) {
1349 * size is based on buffer->width0 rather than info.count
1350 * to prevent having to validate VBO on each draw */
1351 size
= info
.index
.resource
->width0
;
1352 p_data
= swr_resource_data(info
.index
.resource
);
1355 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1356 * revalidate on each draw */
1357 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1359 size
= info
.count
* pitch
;
1360 size
= AlignUp(size
, 4);
1361 /* If size of client memory copy is too large, don't copy. The
1362 * draw will access user-buffer directly and then block. This is
1363 * faster than queuing many large client draws. */
1364 if (size
>= screen
->client_copy_limit
) {
1365 post_update_dirty_flags
|= SWR_LARGE_CLIENT_DRAW
;
1366 p_data
= (const uint8_t *) info
.index
.user
;
1368 /* Copy indices to scratch space */
1369 const void *ptr
= info
.index
.user
;
1370 ptr
= swr_copy_to_scratch_space(
1371 ctx
, &ctx
->scratch
->index_buffer
, ptr
, size
);
1372 p_data
= (const uint8_t *)ptr
;
1376 SWR_INDEX_BUFFER_STATE swrIndexBuffer
;
1377 swrIndexBuffer
.format
= swr_convert_index_type(info
.index_size
);
1378 swrIndexBuffer
.xpIndices
= (gfxptr_t
) p_data
;
1379 swrIndexBuffer
.size
= size
;
1381 ctx
->api
.pfnSwrSetIndexBuffer(ctx
->swrContext
, &swrIndexBuffer
);
1384 struct swr_vertex_element_state
*velems
= ctx
->velems
;
1385 if (velems
&& velems
->fsState
.indexType
!= index_type
) {
1386 velems
->fsFunc
= NULL
;
1387 velems
->fsState
.indexType
= index_type
;
1391 /* GeometryShader */
1392 if (ctx
->dirty
& (SWR_NEW_GS
|
1395 SWR_NEW_SAMPLER_VIEW
)) {
1398 swr_generate_gs_key(key
, ctx
, ctx
->gs
);
1399 auto search
= ctx
->gs
->map
.find(key
);
1401 if (search
!= ctx
->gs
->map
.end()) {
1402 func
= search
->second
->shader
;
1404 func
= swr_compile_gs(ctx
, key
);
1406 ctx
->api
.pfnSwrSetGsFunc(ctx
->swrContext
, func
);
1408 /* JIT sampler state */
1409 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1410 swr_update_sampler_state(ctx
,
1411 PIPE_SHADER_GEOMETRY
,
1413 ctx
->swrDC
.samplersGS
);
1416 /* JIT sampler view state */
1417 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1418 swr_update_texture_state(ctx
,
1419 PIPE_SHADER_GEOMETRY
,
1420 key
.nr_sampler_views
,
1421 ctx
->swrDC
.texturesGS
);
1424 ctx
->api
.pfnSwrSetGsState(ctx
->swrContext
, &ctx
->gs
->gsState
);
1426 SWR_GS_STATE state
= { 0 };
1427 ctx
->api
.pfnSwrSetGsState(ctx
->swrContext
, &state
);
1428 ctx
->api
.pfnSwrSetGsFunc(ctx
->swrContext
, NULL
);
1433 if (ctx
->dirty
& (SWR_NEW_VS
|
1434 SWR_NEW_RASTERIZER
| // for clip planes
1436 SWR_NEW_SAMPLER_VIEW
|
1437 SWR_NEW_FRAMEBUFFER
)) {
1439 swr_generate_vs_key(key
, ctx
, ctx
->vs
);
1440 auto search
= ctx
->vs
->map
.find(key
);
1441 PFN_VERTEX_FUNC func
;
1442 if (search
!= ctx
->vs
->map
.end()) {
1443 func
= search
->second
->shader
;
1445 func
= swr_compile_vs(ctx
, key
);
1447 ctx
->api
.pfnSwrSetVertexFunc(ctx
->swrContext
, func
);
1449 /* JIT sampler state */
1450 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1451 swr_update_sampler_state(ctx
,
1454 ctx
->swrDC
.samplersVS
);
1457 /* JIT sampler view state */
1458 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1459 swr_update_texture_state(ctx
,
1461 key
.nr_sampler_views
,
1462 ctx
->swrDC
.texturesVS
);
1466 /* work around the fact that poly stipple also affects lines */
1467 /* and points, since we rasterize them as triangles, too */
1468 /* Has to be before fragment shader, since it sets SWR_NEW_FS */
1470 bool new_prim_is_poly
=
1471 (u_reduced_prim(p_draw_info
->mode
) == PIPE_PRIM_TRIANGLES
) &&
1472 (ctx
->derived
.rastState
.fillMode
== SWR_FILLMODE_SOLID
);
1473 if (new_prim_is_poly
!= ctx
->poly_stipple
.prim_is_poly
) {
1474 ctx
->dirty
|= SWR_NEW_FS
;
1475 ctx
->poly_stipple
.prim_is_poly
= new_prim_is_poly
;
1479 /* FragmentShader */
1480 if (ctx
->dirty
& (SWR_NEW_FS
|
1483 SWR_NEW_RASTERIZER
|
1485 SWR_NEW_SAMPLER_VIEW
|
1486 SWR_NEW_FRAMEBUFFER
)) {
1488 swr_generate_fs_key(key
, ctx
, ctx
->fs
);
1489 auto search
= ctx
->fs
->map
.find(key
);
1490 PFN_PIXEL_KERNEL func
;
1491 if (search
!= ctx
->fs
->map
.end()) {
1492 func
= search
->second
->shader
;
1494 func
= swr_compile_fs(ctx
, key
);
1496 SWR_PS_STATE psState
= {0};
1497 psState
.pfnPixelShader
= func
;
1498 psState
.killsPixel
= ctx
->fs
->info
.base
.uses_kill
;
1499 psState
.inputCoverage
= SWR_INPUT_COVERAGE_NORMAL
;
1500 psState
.writesODepth
= ctx
->fs
->info
.base
.writes_z
;
1501 psState
.usesSourceDepth
= ctx
->fs
->info
.base
.reads_z
;
1502 psState
.shadingRate
= SWR_SHADING_RATE_PIXEL
;
1503 psState
.renderTargetMask
= (1 << ctx
->framebuffer
.nr_cbufs
) - 1;
1504 psState
.posOffset
= SWR_PS_POSITION_SAMPLE_NONE
;
1505 uint32_t barycentricsMask
= 0;
1507 // when we switch to mesa-master
1508 if (ctx
->fs
->info
.base
.uses_persp_center
||
1509 ctx
->fs
->info
.base
.uses_linear_center
)
1510 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1511 if (ctx
->fs
->info
.base
.uses_persp_centroid
||
1512 ctx
->fs
->info
.base
.uses_linear_centroid
)
1513 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1514 if (ctx
->fs
->info
.base
.uses_persp_sample
||
1515 ctx
->fs
->info
.base
.uses_linear_sample
)
1516 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1518 for (unsigned i
= 0; i
< ctx
->fs
->info
.base
.num_inputs
; i
++) {
1519 switch (ctx
->fs
->info
.base
.input_interpolate_loc
[i
]) {
1520 case TGSI_INTERPOLATE_LOC_CENTER
:
1521 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1523 case TGSI_INTERPOLATE_LOC_CENTROID
:
1524 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1526 case TGSI_INTERPOLATE_LOC_SAMPLE
:
1527 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1532 psState
.barycentricsMask
= barycentricsMask
;
1533 psState
.usesUAV
= false; // XXX
1534 psState
.forceEarlyZ
= false;
1535 ctx
->api
.pfnSwrSetPixelShaderState(ctx
->swrContext
, &psState
);
1537 /* JIT sampler state */
1538 if (ctx
->dirty
& (SWR_NEW_SAMPLER
|
1540 swr_update_sampler_state(ctx
,
1541 PIPE_SHADER_FRAGMENT
,
1543 ctx
->swrDC
.samplersFS
);
1546 /* JIT sampler view state */
1547 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
|
1548 SWR_NEW_FRAMEBUFFER
|
1550 swr_update_texture_state(ctx
,
1551 PIPE_SHADER_FRAGMENT
,
1552 key
.nr_sampler_views
,
1553 ctx
->swrDC
.texturesFS
);
1558 /* VertexShader Constants */
1559 if (ctx
->dirty
& SWR_NEW_VSCONSTANTS
) {
1560 swr_update_constants(ctx
, PIPE_SHADER_VERTEX
);
1563 /* FragmentShader Constants */
1564 if (ctx
->dirty
& SWR_NEW_FSCONSTANTS
) {
1565 swr_update_constants(ctx
, PIPE_SHADER_FRAGMENT
);
1568 /* GeometryShader Constants */
1569 if (ctx
->dirty
& SWR_NEW_GSCONSTANTS
) {
1570 swr_update_constants(ctx
, PIPE_SHADER_GEOMETRY
);
1573 /* Depth/stencil state */
1574 if (ctx
->dirty
& (SWR_NEW_DEPTH_STENCIL_ALPHA
| SWR_NEW_FRAMEBUFFER
)) {
1575 struct pipe_depth_state
*depth
= &(ctx
->depth_stencil
->depth
);
1576 struct pipe_stencil_state
*stencil
= ctx
->depth_stencil
->stencil
;
1577 SWR_DEPTH_STENCIL_STATE depthStencilState
= {{0}};
1578 SWR_DEPTH_BOUNDS_STATE depthBoundsState
= {0};
1580 /* XXX, incomplete. Need to flesh out stencil & alpha test state
1581 struct pipe_stencil_state *front_stencil =
1582 ctx->depth_stencil.stencil[0];
1583 struct pipe_stencil_state *back_stencil = ctx->depth_stencil.stencil[1];
1584 struct pipe_alpha_state alpha;
1586 if (stencil
[0].enabled
) {
1587 depthStencilState
.stencilWriteEnable
= 1;
1588 depthStencilState
.stencilTestEnable
= 1;
1589 depthStencilState
.stencilTestFunc
=
1590 swr_convert_depth_func(stencil
[0].func
);
1592 depthStencilState
.stencilPassDepthPassOp
=
1593 swr_convert_stencil_op(stencil
[0].zpass_op
);
1594 depthStencilState
.stencilPassDepthFailOp
=
1595 swr_convert_stencil_op(stencil
[0].zfail_op
);
1596 depthStencilState
.stencilFailOp
=
1597 swr_convert_stencil_op(stencil
[0].fail_op
);
1598 depthStencilState
.stencilWriteMask
= stencil
[0].writemask
;
1599 depthStencilState
.stencilTestMask
= stencil
[0].valuemask
;
1600 depthStencilState
.stencilRefValue
= ctx
->stencil_ref
.ref_value
[0];
1602 if (stencil
[1].enabled
) {
1603 depthStencilState
.doubleSidedStencilTestEnable
= 1;
1605 depthStencilState
.backfaceStencilTestFunc
=
1606 swr_convert_depth_func(stencil
[1].func
);
1608 depthStencilState
.backfaceStencilPassDepthPassOp
=
1609 swr_convert_stencil_op(stencil
[1].zpass_op
);
1610 depthStencilState
.backfaceStencilPassDepthFailOp
=
1611 swr_convert_stencil_op(stencil
[1].zfail_op
);
1612 depthStencilState
.backfaceStencilFailOp
=
1613 swr_convert_stencil_op(stencil
[1].fail_op
);
1614 depthStencilState
.backfaceStencilWriteMask
= stencil
[1].writemask
;
1615 depthStencilState
.backfaceStencilTestMask
= stencil
[1].valuemask
;
1617 depthStencilState
.backfaceStencilRefValue
=
1618 ctx
->stencil_ref
.ref_value
[1];
1621 depthStencilState
.depthTestEnable
= depth
->enabled
;
1622 depthStencilState
.depthTestFunc
= swr_convert_depth_func(depth
->func
);
1623 depthStencilState
.depthWriteEnable
= depth
->writemask
;
1624 ctx
->api
.pfnSwrSetDepthStencilState(ctx
->swrContext
, &depthStencilState
);
1626 depthBoundsState
.depthBoundsTestEnable
= depth
->bounds_test
;
1627 depthBoundsState
.depthBoundsTestMinValue
= depth
->bounds_min
;
1628 depthBoundsState
.depthBoundsTestMaxValue
= depth
->bounds_max
;
1629 ctx
->api
.pfnSwrSetDepthBoundsState(ctx
->swrContext
, &depthBoundsState
);
1633 if (ctx
->dirty
& (SWR_NEW_BLEND
|
1634 SWR_NEW_RASTERIZER
|
1635 SWR_NEW_FRAMEBUFFER
|
1636 SWR_NEW_DEPTH_STENCIL_ALPHA
)) {
1637 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1639 SWR_BLEND_STATE blendState
;
1640 memcpy(&blendState
, &ctx
->blend
->blendState
, sizeof(blendState
));
1641 blendState
.constantColor
[0] = ctx
->blend_color
.color
[0];
1642 blendState
.constantColor
[1] = ctx
->blend_color
.color
[1];
1643 blendState
.constantColor
[2] = ctx
->blend_color
.color
[2];
1644 blendState
.constantColor
[3] = ctx
->blend_color
.color
[3];
1645 blendState
.alphaTestReference
=
1646 *((uint32_t*)&ctx
->depth_stencil
->alpha
.ref_value
);
1648 blendState
.sampleMask
= ctx
->sample_mask
;
1649 blendState
.sampleCount
= GetSampleCount(fb
->samples
);
1651 /* If there are no color buffers bound, disable writes on RT0
1653 if (fb
->nr_cbufs
== 0) {
1654 blendState
.renderTarget
[0].writeDisableRed
= 1;
1655 blendState
.renderTarget
[0].writeDisableGreen
= 1;
1656 blendState
.renderTarget
[0].writeDisableBlue
= 1;
1657 blendState
.renderTarget
[0].writeDisableAlpha
= 1;
1658 ctx
->api
.pfnSwrSetBlendFunc(ctx
->swrContext
, 0, NULL
);
1661 for (int target
= 0;
1662 target
< std::min(SWR_NUM_RENDERTARGETS
,
1663 PIPE_MAX_COLOR_BUFS
);
1665 if (!fb
->cbufs
[target
])
1668 struct swr_resource
*colorBuffer
=
1669 swr_resource(fb
->cbufs
[target
]->texture
);
1671 BLEND_COMPILE_STATE compileState
;
1672 memset(&compileState
, 0, sizeof(compileState
));
1673 compileState
.format
= colorBuffer
->swr
.format
;
1674 memcpy(&compileState
.blendState
,
1675 &ctx
->blend
->compileState
[target
],
1676 sizeof(compileState
.blendState
));
1678 const SWR_FORMAT_INFO
& info
= GetFormatInfo(compileState
.format
);
1679 if (compileState
.blendState
.logicOpEnable
&&
1680 ((info
.type
[0] == SWR_TYPE_FLOAT
) || info
.isSRGB
)) {
1681 compileState
.blendState
.logicOpEnable
= false;
1684 if (info
.type
[0] == SWR_TYPE_SINT
|| info
.type
[0] == SWR_TYPE_UINT
)
1685 compileState
.blendState
.blendEnable
= false;
1687 if (compileState
.blendState
.blendEnable
== false &&
1688 compileState
.blendState
.logicOpEnable
== false &&
1689 ctx
->depth_stencil
->alpha
.enabled
== 0) {
1690 ctx
->api
.pfnSwrSetBlendFunc(ctx
->swrContext
, target
, NULL
);
1694 compileState
.desc
.alphaTestEnable
=
1695 ctx
->depth_stencil
->alpha
.enabled
;
1696 compileState
.desc
.independentAlphaBlendEnable
=
1697 (compileState
.blendState
.sourceBlendFactor
!=
1698 compileState
.blendState
.sourceAlphaBlendFactor
) ||
1699 (compileState
.blendState
.destBlendFactor
!=
1700 compileState
.blendState
.destAlphaBlendFactor
) ||
1701 (compileState
.blendState
.colorBlendFunc
!=
1702 compileState
.blendState
.alphaBlendFunc
);
1703 compileState
.desc
.alphaToCoverageEnable
=
1704 ctx
->blend
->pipe
.alpha_to_coverage
;
1705 compileState
.desc
.sampleMaskEnable
= (blendState
.sampleMask
!= 0);
1706 compileState
.desc
.numSamples
= fb
->samples
;
1708 compileState
.alphaTestFunction
=
1709 swr_convert_depth_func(ctx
->depth_stencil
->alpha
.func
);
1710 compileState
.alphaTestFormat
= ALPHA_TEST_FLOAT32
; // xxx
1712 compileState
.Canonicalize();
1714 PFN_BLEND_JIT_FUNC func
= NULL
;
1715 auto search
= ctx
->blendJIT
->find(compileState
);
1716 if (search
!= ctx
->blendJIT
->end()) {
1717 func
= search
->second
;
1719 HANDLE hJitMgr
= screen
->hJitMgr
;
1720 func
= JitCompileBlend(hJitMgr
, compileState
);
1721 debug_printf("BLEND shader %p\n", func
);
1722 assert(func
&& "Error: BlendShader = NULL");
1724 ctx
->blendJIT
->insert(std::make_pair(compileState
, func
));
1726 ctx
->api
.pfnSwrSetBlendFunc(ctx
->swrContext
, target
, func
);
1729 ctx
->api
.pfnSwrSetBlendState(ctx
->swrContext
, &blendState
);
1732 if (ctx
->dirty
& SWR_NEW_STIPPLE
) {
1733 swr_update_poly_stipple(ctx
);
1736 if (ctx
->dirty
& (SWR_NEW_VS
| SWR_NEW_SO
| SWR_NEW_RASTERIZER
)) {
1737 ctx
->vs
->soState
.rasterizerDisable
=
1738 ctx
->rasterizer
->rasterizer_discard
;
1739 ctx
->api
.pfnSwrSetSoState(ctx
->swrContext
, &ctx
->vs
->soState
);
1741 pipe_stream_output_info
*stream_output
= &ctx
->vs
->pipe
.stream_output
;
1743 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
1744 SWR_STREAMOUT_BUFFER buffer
= {0};
1745 if (!ctx
->so_targets
[i
])
1747 buffer
.enable
= true;
1749 (uint32_t *)(swr_resource_data(ctx
->so_targets
[i
]->buffer
) +
1750 ctx
->so_targets
[i
]->buffer_offset
);
1751 buffer
.bufferSize
= ctx
->so_targets
[i
]->buffer_size
>> 2;
1752 buffer
.pitch
= stream_output
->stride
[i
];
1753 buffer
.streamOffset
= 0;
1755 ctx
->api
.pfnSwrSetSoBuffers(ctx
->swrContext
, &buffer
, i
);
1759 if (ctx
->dirty
& (SWR_NEW_CLIP
| SWR_NEW_RASTERIZER
| SWR_NEW_VS
)) {
1760 // shader exporting clip distances overrides all user clip planes
1761 if (ctx
->rasterizer
->clip_plane_enable
&&
1762 !ctx
->vs
->info
.base
.num_written_clipdistance
)
1764 swr_draw_context
*pDC
= &ctx
->swrDC
;
1765 memcpy(pDC
->userClipPlanes
,
1767 sizeof(pDC
->userClipPlanes
));
1771 // set up backend state
1772 SWR_BACKEND_STATE backendState
= {0};
1774 backendState
.numAttributes
= ctx
->gs
->info
.base
.num_outputs
- 1;
1776 backendState
.numAttributes
= ctx
->vs
->info
.base
.num_outputs
- 1;
1777 if (ctx
->fs
->info
.base
.uses_primid
) {
1778 backendState
.numAttributes
++;
1779 backendState
.swizzleEnable
= true;
1780 for (unsigned i
= 0; i
< sizeof(backendState
.numComponents
); i
++) {
1781 backendState
.swizzleMap
[i
].sourceAttrib
= i
;
1783 backendState
.swizzleMap
[ctx
->vs
->info
.base
.num_outputs
- 1].constantSource
=
1784 SWR_CONSTANT_SOURCE_PRIM_ID
;
1785 backendState
.swizzleMap
[ctx
->vs
->info
.base
.num_outputs
- 1].componentOverrideMask
= 1;
1788 if (ctx
->rasterizer
->sprite_coord_enable
)
1789 backendState
.numAttributes
++;
1791 backendState
.numAttributes
= std::min((size_t)backendState
.numAttributes
,
1792 sizeof(backendState
.numComponents
));
1793 for (unsigned i
= 0; i
< backendState
.numAttributes
; i
++)
1794 backendState
.numComponents
[i
] = 4;
1795 backendState
.constantInterpolationMask
= ctx
->fs
->constantMask
|
1796 (ctx
->rasterizer
->flatshade
? ctx
->fs
->flatConstantMask
: 0);
1797 backendState
.pointSpriteTexCoordMask
= ctx
->fs
->pointSpriteMask
;
1799 struct tgsi_shader_info
*pLastFE
=
1801 &ctx
->gs
->info
.base
:
1802 &ctx
->vs
->info
.base
;
1803 backendState
.readRenderTargetArrayIndex
= pLastFE
->writes_layer
;
1804 backendState
.readViewportArrayIndex
= pLastFE
->writes_viewport_index
;
1805 backendState
.vertexAttribOffset
= VERTEX_ATTRIB_START_SLOT
; // TODO: optimize
1807 backendState
.clipDistanceMask
=
1808 ctx
->vs
->info
.base
.num_written_clipdistance
?
1809 ctx
->vs
->info
.base
.clipdist_writemask
& ctx
->rasterizer
->clip_plane_enable
:
1810 ctx
->rasterizer
->clip_plane_enable
;
1812 backendState
.cullDistanceMask
=
1813 ctx
->vs
->info
.base
.culldist_writemask
<< ctx
->vs
->info
.base
.num_written_clipdistance
;
1815 // Assume old layout of SGV, POSITION, CLIPCULL, ATTRIB
1816 backendState
.vertexClipCullOffset
= backendState
.vertexAttribOffset
- 2;
1818 ctx
->api
.pfnSwrSetBackendState(ctx
->swrContext
, &backendState
);
1820 /* Ensure that any in-progress attachment change StoreTiles finish */
1821 if (swr_is_fence_pending(screen
->flush_fence
))
1822 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
1824 /* Finally, update the in-use status of all resources involved in draw */
1825 swr_update_resource_status(pipe
, p_draw_info
);
1827 ctx
->dirty
= post_update_dirty_flags
;
1831 static struct pipe_stream_output_target
*
1832 swr_create_so_target(struct pipe_context
*pipe
,
1833 struct pipe_resource
*buffer
,
1834 unsigned buffer_offset
,
1835 unsigned buffer_size
)
1837 struct pipe_stream_output_target
*target
;
1839 target
= CALLOC_STRUCT(pipe_stream_output_target
);
1843 target
->context
= pipe
;
1844 target
->reference
.count
= 1;
1845 pipe_resource_reference(&target
->buffer
, buffer
);
1846 target
->buffer_offset
= buffer_offset
;
1847 target
->buffer_size
= buffer_size
;
1852 swr_destroy_so_target(struct pipe_context
*pipe
,
1853 struct pipe_stream_output_target
*target
)
1855 pipe_resource_reference(&target
->buffer
, NULL
);
1860 swr_set_so_targets(struct pipe_context
*pipe
,
1861 unsigned num_targets
,
1862 struct pipe_stream_output_target
**targets
,
1863 const unsigned *offsets
)
1865 struct swr_context
*swr
= swr_context(pipe
);
1868 assert(num_targets
<= MAX_SO_STREAMS
);
1870 for (i
= 0; i
< num_targets
; i
++) {
1871 pipe_so_target_reference(
1872 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
],
1876 for (/* fall-through */; i
< swr
->num_so_targets
; i
++) {
1877 pipe_so_target_reference(
1878 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
], NULL
);
1881 swr
->num_so_targets
= num_targets
;
1883 swr
->dirty
|= SWR_NEW_SO
;
1888 swr_state_init(struct pipe_context
*pipe
)
1890 pipe
->create_blend_state
= swr_create_blend_state
;
1891 pipe
->bind_blend_state
= swr_bind_blend_state
;
1892 pipe
->delete_blend_state
= swr_delete_blend_state
;
1894 pipe
->create_depth_stencil_alpha_state
= swr_create_depth_stencil_state
;
1895 pipe
->bind_depth_stencil_alpha_state
= swr_bind_depth_stencil_state
;
1896 pipe
->delete_depth_stencil_alpha_state
= swr_delete_depth_stencil_state
;
1898 pipe
->create_rasterizer_state
= swr_create_rasterizer_state
;
1899 pipe
->bind_rasterizer_state
= swr_bind_rasterizer_state
;
1900 pipe
->delete_rasterizer_state
= swr_delete_rasterizer_state
;
1902 pipe
->create_sampler_state
= swr_create_sampler_state
;
1903 pipe
->bind_sampler_states
= swr_bind_sampler_states
;
1904 pipe
->delete_sampler_state
= swr_delete_sampler_state
;
1906 pipe
->create_sampler_view
= swr_create_sampler_view
;
1907 pipe
->set_sampler_views
= swr_set_sampler_views
;
1908 pipe
->sampler_view_destroy
= swr_sampler_view_destroy
;
1910 pipe
->create_vs_state
= swr_create_vs_state
;
1911 pipe
->bind_vs_state
= swr_bind_vs_state
;
1912 pipe
->delete_vs_state
= swr_delete_vs_state
;
1914 pipe
->create_fs_state
= swr_create_fs_state
;
1915 pipe
->bind_fs_state
= swr_bind_fs_state
;
1916 pipe
->delete_fs_state
= swr_delete_fs_state
;
1918 pipe
->create_gs_state
= swr_create_gs_state
;
1919 pipe
->bind_gs_state
= swr_bind_gs_state
;
1920 pipe
->delete_gs_state
= swr_delete_gs_state
;
1922 pipe
->set_constant_buffer
= swr_set_constant_buffer
;
1924 pipe
->create_vertex_elements_state
= swr_create_vertex_elements_state
;
1925 pipe
->bind_vertex_elements_state
= swr_bind_vertex_elements_state
;
1926 pipe
->delete_vertex_elements_state
= swr_delete_vertex_elements_state
;
1928 pipe
->set_vertex_buffers
= swr_set_vertex_buffers
;
1930 pipe
->set_polygon_stipple
= swr_set_polygon_stipple
;
1931 pipe
->set_clip_state
= swr_set_clip_state
;
1932 pipe
->set_scissor_states
= swr_set_scissor_states
;
1933 pipe
->set_viewport_states
= swr_set_viewport_states
;
1935 pipe
->set_framebuffer_state
= swr_set_framebuffer_state
;
1937 pipe
->set_blend_color
= swr_set_blend_color
;
1938 pipe
->set_stencil_ref
= swr_set_stencil_ref
;
1940 pipe
->set_sample_mask
= swr_set_sample_mask
;
1941 pipe
->get_sample_position
= swr_get_sample_position
;
1943 pipe
->create_stream_output_target
= swr_create_so_target
;
1944 pipe
->stream_output_target_destroy
= swr_destroy_so_target
;
1945 pipe
->set_stream_output_targets
= swr_set_so_targets
;