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 "state_llvm.h"
34 #include "gallivm/lp_bld_tgsi.h"
35 #include "util/u_format.h"
37 #include "util/u_memory.h"
38 #include "util/u_inlines.h"
39 #include "util/u_helpers.h"
40 #include "util/u_framebuffer.h"
41 #include "util/u_viewport.h"
43 #include "swr_state.h"
44 #include "swr_context.h"
45 #include "swr_context_llvm.h"
46 #include "swr_screen.h"
47 #include "swr_resource.h"
48 #include "swr_tex_sample.h"
49 #include "swr_scratch.h"
50 #include "swr_shader.h"
51 #include "swr_fence.h"
53 /* These should be pulled out into separate files as necessary
54 * Just initializing everything here to get going. */
57 swr_create_blend_state(struct pipe_context
*pipe
,
58 const struct pipe_blend_state
*blend
)
60 struct swr_blend_state
*state
= CALLOC_STRUCT(swr_blend_state
);
62 memcpy(&state
->pipe
, blend
, sizeof(*blend
));
64 struct pipe_blend_state
*pipe_blend
= &state
->pipe
;
67 target
< std::min(SWR_NUM_RENDERTARGETS
, PIPE_MAX_COLOR_BUFS
);
70 struct pipe_rt_blend_state
*rt_blend
= &pipe_blend
->rt
[target
];
71 SWR_RENDER_TARGET_BLEND_STATE
&blendState
=
72 state
->blendState
.renderTarget
[target
];
73 RENDER_TARGET_BLEND_COMPILE_STATE
&compileState
=
74 state
->compileState
[target
];
76 if (target
!= 0 && !pipe_blend
->independent_blend_enable
) {
78 &state
->compileState
[0],
79 sizeof(RENDER_TARGET_BLEND_COMPILE_STATE
));
83 compileState
.blendEnable
= rt_blend
->blend_enable
;
84 if (compileState
.blendEnable
) {
85 compileState
.sourceAlphaBlendFactor
=
86 swr_convert_blend_factor(rt_blend
->alpha_src_factor
);
87 compileState
.destAlphaBlendFactor
=
88 swr_convert_blend_factor(rt_blend
->alpha_dst_factor
);
89 compileState
.sourceBlendFactor
=
90 swr_convert_blend_factor(rt_blend
->rgb_src_factor
);
91 compileState
.destBlendFactor
=
92 swr_convert_blend_factor(rt_blend
->rgb_dst_factor
);
94 compileState
.colorBlendFunc
=
95 swr_convert_blend_func(rt_blend
->rgb_func
);
96 compileState
.alphaBlendFunc
=
97 swr_convert_blend_func(rt_blend
->alpha_func
);
99 compileState
.logicOpEnable
= state
->pipe
.logicop_enable
;
100 if (compileState
.logicOpEnable
) {
101 compileState
.logicOpFunc
=
102 swr_convert_logic_op(state
->pipe
.logicop_func
);
105 blendState
.writeDisableRed
=
106 (rt_blend
->colormask
& PIPE_MASK_R
) ? 0 : 1;
107 blendState
.writeDisableGreen
=
108 (rt_blend
->colormask
& PIPE_MASK_G
) ? 0 : 1;
109 blendState
.writeDisableBlue
=
110 (rt_blend
->colormask
& PIPE_MASK_B
) ? 0 : 1;
111 blendState
.writeDisableAlpha
=
112 (rt_blend
->colormask
& PIPE_MASK_A
) ? 0 : 1;
114 if (rt_blend
->colormask
== 0)
115 compileState
.blendEnable
= false;
122 swr_bind_blend_state(struct pipe_context
*pipe
, void *blend
)
124 struct swr_context
*ctx
= swr_context(pipe
);
126 if (ctx
->blend
== blend
)
129 ctx
->blend
= (swr_blend_state
*)blend
;
131 ctx
->dirty
|= SWR_NEW_BLEND
;
135 swr_delete_blend_state(struct pipe_context
*pipe
, void *blend
)
141 swr_set_blend_color(struct pipe_context
*pipe
,
142 const struct pipe_blend_color
*color
)
144 struct swr_context
*ctx
= swr_context(pipe
);
146 ctx
->blend_color
= *color
;
148 ctx
->dirty
|= SWR_NEW_BLEND
;
152 swr_set_stencil_ref(struct pipe_context
*pipe
,
153 const struct pipe_stencil_ref
*ref
)
155 struct swr_context
*ctx
= swr_context(pipe
);
157 ctx
->stencil_ref
= *ref
;
159 ctx
->dirty
|= SWR_NEW_DEPTH_STENCIL_ALPHA
;
163 swr_create_depth_stencil_state(
164 struct pipe_context
*pipe
,
165 const struct pipe_depth_stencil_alpha_state
*depth_stencil
)
167 struct pipe_depth_stencil_alpha_state
*state
;
169 state
= (pipe_depth_stencil_alpha_state
*)mem_dup(depth_stencil
,
170 sizeof *depth_stencil
);
176 swr_bind_depth_stencil_state(struct pipe_context
*pipe
, void *depth_stencil
)
178 struct swr_context
*ctx
= swr_context(pipe
);
180 if (ctx
->depth_stencil
== (pipe_depth_stencil_alpha_state
*)depth_stencil
)
183 ctx
->depth_stencil
= (pipe_depth_stencil_alpha_state
*)depth_stencil
;
185 ctx
->dirty
|= SWR_NEW_DEPTH_STENCIL_ALPHA
;
189 swr_delete_depth_stencil_state(struct pipe_context
*pipe
, void *depth
)
196 swr_create_rasterizer_state(struct pipe_context
*pipe
,
197 const struct pipe_rasterizer_state
*rast
)
199 struct pipe_rasterizer_state
*state
;
200 state
= (pipe_rasterizer_state
*)mem_dup(rast
, sizeof *rast
);
206 swr_bind_rasterizer_state(struct pipe_context
*pipe
, void *handle
)
208 struct swr_context
*ctx
= swr_context(pipe
);
209 const struct pipe_rasterizer_state
*rasterizer
=
210 (const struct pipe_rasterizer_state
*)handle
;
212 if (ctx
->rasterizer
== (pipe_rasterizer_state
*)rasterizer
)
215 ctx
->rasterizer
= (pipe_rasterizer_state
*)rasterizer
;
217 ctx
->dirty
|= SWR_NEW_RASTERIZER
;
221 swr_delete_rasterizer_state(struct pipe_context
*pipe
, void *rasterizer
)
228 swr_create_sampler_state(struct pipe_context
*pipe
,
229 const struct pipe_sampler_state
*sampler
)
231 struct pipe_sampler_state
*state
=
232 (pipe_sampler_state
*)mem_dup(sampler
, sizeof *sampler
);
238 swr_bind_sampler_states(struct pipe_context
*pipe
,
239 enum pipe_shader_type shader
,
244 struct swr_context
*ctx
= swr_context(pipe
);
247 assert(shader
< PIPE_SHADER_TYPES
);
248 assert(start
+ num
<= ARRAY_SIZE(ctx
->samplers
[shader
]));
250 /* set the new samplers */
251 ctx
->num_samplers
[shader
] = num
;
252 for (i
= 0; i
< num
; i
++) {
253 ctx
->samplers
[shader
][start
+ i
] = (pipe_sampler_state
*)samplers
[i
];
256 ctx
->dirty
|= SWR_NEW_SAMPLER
;
260 swr_delete_sampler_state(struct pipe_context
*pipe
, void *sampler
)
266 static struct pipe_sampler_view
*
267 swr_create_sampler_view(struct pipe_context
*pipe
,
268 struct pipe_resource
*texture
,
269 const struct pipe_sampler_view
*templ
)
271 struct pipe_sampler_view
*view
= CALLOC_STRUCT(pipe_sampler_view
);
275 view
->reference
.count
= 1;
276 view
->texture
= NULL
;
277 pipe_resource_reference(&view
->texture
, texture
);
278 view
->context
= pipe
;
285 swr_set_sampler_views(struct pipe_context
*pipe
,
286 enum pipe_shader_type shader
,
289 struct pipe_sampler_view
**views
)
291 struct swr_context
*ctx
= swr_context(pipe
);
294 assert(num
<= PIPE_MAX_SHADER_SAMPLER_VIEWS
);
296 assert(shader
< PIPE_SHADER_TYPES
);
297 assert(start
+ num
<= ARRAY_SIZE(ctx
->sampler_views
[shader
]));
299 /* set the new sampler views */
300 ctx
->num_sampler_views
[shader
] = num
;
301 for (i
= 0; i
< num
; i
++) {
302 /* Note: we're using pipe_sampler_view_release() here to work around
303 * a possible crash when the old view belongs to another context that
304 * was already destroyed.
306 pipe_sampler_view_release(pipe
, &ctx
->sampler_views
[shader
][start
+ i
]);
307 pipe_sampler_view_reference(&ctx
->sampler_views
[shader
][start
+ i
],
311 ctx
->dirty
|= SWR_NEW_SAMPLER_VIEW
;
315 swr_sampler_view_destroy(struct pipe_context
*pipe
,
316 struct pipe_sampler_view
*view
)
318 pipe_resource_reference(&view
->texture
, NULL
);
323 swr_create_vs_state(struct pipe_context
*pipe
,
324 const struct pipe_shader_state
*vs
)
326 struct swr_vertex_shader
*swr_vs
= new swr_vertex_shader
;
330 swr_vs
->pipe
.tokens
= tgsi_dup_tokens(vs
->tokens
);
331 swr_vs
->pipe
.stream_output
= vs
->stream_output
;
333 lp_build_tgsi_info(vs
->tokens
, &swr_vs
->info
);
335 swr_vs
->soState
= {0};
337 if (swr_vs
->pipe
.stream_output
.num_outputs
) {
338 pipe_stream_output_info
*stream_output
= &swr_vs
->pipe
.stream_output
;
340 swr_vs
->soState
.soEnable
= true;
341 // soState.rasterizerDisable set on state dirty
342 // soState.streamToRasterizer not used
344 for (uint32_t i
= 0; i
< stream_output
->num_outputs
; i
++) {
345 swr_vs
->soState
.streamMasks
[stream_output
->output
[i
].stream
] |=
346 1 << (stream_output
->output
[i
].register_index
- 1);
348 for (uint32_t i
= 0; i
< MAX_SO_STREAMS
; i
++) {
349 swr_vs
->soState
.streamNumEntries
[i
] =
350 _mm_popcnt_u32(swr_vs
->soState
.streamMasks
[i
]);
358 swr_bind_vs_state(struct pipe_context
*pipe
, void *vs
)
360 struct swr_context
*ctx
= swr_context(pipe
);
365 ctx
->vs
= (swr_vertex_shader
*)vs
;
366 ctx
->dirty
|= SWR_NEW_VS
;
370 swr_delete_vs_state(struct pipe_context
*pipe
, void *vs
)
372 struct swr_vertex_shader
*swr_vs
= (swr_vertex_shader
*)vs
;
373 FREE((void *)swr_vs
->pipe
.tokens
);
374 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
375 if (!swr_is_fence_pending(screen
->flush_fence
))
376 swr_fence_submit(swr_context(pipe
), screen
->flush_fence
);
377 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
382 swr_create_fs_state(struct pipe_context
*pipe
,
383 const struct pipe_shader_state
*fs
)
385 struct swr_fragment_shader
*swr_fs
= new swr_fragment_shader
;
389 swr_fs
->pipe
.tokens
= tgsi_dup_tokens(fs
->tokens
);
391 lp_build_tgsi_info(fs
->tokens
, &swr_fs
->info
);
398 swr_bind_fs_state(struct pipe_context
*pipe
, void *fs
)
400 struct swr_context
*ctx
= swr_context(pipe
);
405 ctx
->fs
= (swr_fragment_shader
*)fs
;
406 ctx
->dirty
|= SWR_NEW_FS
;
410 swr_delete_fs_state(struct pipe_context
*pipe
, void *fs
)
412 struct swr_fragment_shader
*swr_fs
= (swr_fragment_shader
*)fs
;
413 FREE((void *)swr_fs
->pipe
.tokens
);
414 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
415 if (!swr_is_fence_pending(screen
->flush_fence
))
416 swr_fence_submit(swr_context(pipe
), screen
->flush_fence
);
417 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
423 swr_set_constant_buffer(struct pipe_context
*pipe
,
426 const struct pipe_constant_buffer
*cb
)
428 struct swr_context
*ctx
= swr_context(pipe
);
429 struct pipe_resource
*constants
= cb
? cb
->buffer
: NULL
;
431 assert(shader
< PIPE_SHADER_TYPES
);
432 assert(index
< ARRAY_SIZE(ctx
->constants
[shader
]));
434 /* note: reference counting */
435 util_copy_constant_buffer(&ctx
->constants
[shader
][index
], cb
);
437 if (shader
== PIPE_SHADER_VERTEX
|| shader
== PIPE_SHADER_GEOMETRY
) {
438 ctx
->dirty
|= SWR_NEW_VSCONSTANTS
;
439 } else if (shader
== PIPE_SHADER_FRAGMENT
) {
440 ctx
->dirty
|= SWR_NEW_FSCONSTANTS
;
443 if (cb
&& cb
->user_buffer
) {
444 pipe_resource_reference(&constants
, NULL
);
450 swr_create_vertex_elements_state(struct pipe_context
*pipe
,
451 unsigned num_elements
,
452 const struct pipe_vertex_element
*attribs
)
454 struct swr_vertex_element_state
*velems
;
455 assert(num_elements
<= PIPE_MAX_ATTRIBS
);
456 velems
= CALLOC_STRUCT(swr_vertex_element_state
);
458 velems
->fsState
.bVertexIDOffsetEnable
= true;
459 velems
->fsState
.numAttribs
= num_elements
;
460 for (unsigned i
= 0; i
< num_elements
; i
++) {
461 // XXX: we should do this keyed on the VS usage info
463 const struct util_format_description
*desc
=
464 util_format_description(attribs
[i
].src_format
);
466 velems
->fsState
.layout
[i
].AlignedByteOffset
= attribs
[i
].src_offset
;
467 velems
->fsState
.layout
[i
].Format
=
468 mesa_to_swr_format(attribs
[i
].src_format
);
469 velems
->fsState
.layout
[i
].StreamIndex
=
470 attribs
[i
].vertex_buffer_index
;
471 velems
->fsState
.layout
[i
].InstanceEnable
=
472 attribs
[i
].instance_divisor
!= 0;
473 velems
->fsState
.layout
[i
].ComponentControl0
=
474 desc
->channel
[0].type
!= UTIL_FORMAT_TYPE_VOID
475 ? ComponentControl::StoreSrc
476 : ComponentControl::Store0
;
477 velems
->fsState
.layout
[i
].ComponentControl1
=
478 desc
->channel
[1].type
!= UTIL_FORMAT_TYPE_VOID
479 ? ComponentControl::StoreSrc
480 : ComponentControl::Store0
;
481 velems
->fsState
.layout
[i
].ComponentControl2
=
482 desc
->channel
[2].type
!= UTIL_FORMAT_TYPE_VOID
483 ? ComponentControl::StoreSrc
484 : ComponentControl::Store0
;
485 velems
->fsState
.layout
[i
].ComponentControl3
=
486 desc
->channel
[3].type
!= UTIL_FORMAT_TYPE_VOID
487 ? ComponentControl::StoreSrc
488 : ComponentControl::Store1Fp
;
489 velems
->fsState
.layout
[i
].ComponentPacking
= ComponentEnable::XYZW
;
490 velems
->fsState
.layout
[i
].InstanceDataStepRate
=
491 attribs
[i
].instance_divisor
;
493 /* Calculate the pitch of each stream */
494 const SWR_FORMAT_INFO
&swr_desc
= GetFormatInfo(
495 mesa_to_swr_format(attribs
[i
].src_format
));
496 velems
->stream_pitch
[attribs
[i
].vertex_buffer_index
] += swr_desc
.Bpp
;
504 swr_bind_vertex_elements_state(struct pipe_context
*pipe
, void *velems
)
506 struct swr_context
*ctx
= swr_context(pipe
);
507 struct swr_vertex_element_state
*swr_velems
=
508 (struct swr_vertex_element_state
*)velems
;
510 ctx
->velems
= swr_velems
;
511 ctx
->dirty
|= SWR_NEW_VERTEX
;
515 swr_delete_vertex_elements_state(struct pipe_context
*pipe
, void *velems
)
517 /* XXX Need to destroy fetch shader? */
523 swr_set_vertex_buffers(struct pipe_context
*pipe
,
525 unsigned num_elements
,
526 const struct pipe_vertex_buffer
*buffers
)
528 struct swr_context
*ctx
= swr_context(pipe
);
530 assert(num_elements
<= PIPE_MAX_ATTRIBS
);
532 util_set_vertex_buffers_count(ctx
->vertex_buffer
,
533 &ctx
->num_vertex_buffers
,
538 ctx
->dirty
|= SWR_NEW_VERTEX
;
543 swr_set_index_buffer(struct pipe_context
*pipe
,
544 const struct pipe_index_buffer
*ib
)
546 struct swr_context
*ctx
= swr_context(pipe
);
549 memcpy(&ctx
->index_buffer
, ib
, sizeof(ctx
->index_buffer
));
551 memset(&ctx
->index_buffer
, 0, sizeof(ctx
->index_buffer
));
553 ctx
->dirty
|= SWR_NEW_VERTEX
;
557 swr_set_polygon_stipple(struct pipe_context
*pipe
,
558 const struct pipe_poly_stipple
*stipple
)
560 struct swr_context
*ctx
= swr_context(pipe
);
562 ctx
->poly_stipple
= *stipple
; /* struct copy */
563 ctx
->dirty
|= SWR_NEW_STIPPLE
;
567 swr_set_clip_state(struct pipe_context
*pipe
,
568 const struct pipe_clip_state
*clip
)
570 struct swr_context
*ctx
= swr_context(pipe
);
573 /* XXX Unimplemented, but prevents crash */
575 ctx
->dirty
|= SWR_NEW_CLIP
;
580 swr_set_scissor_states(struct pipe_context
*pipe
,
582 unsigned num_viewports
,
583 const struct pipe_scissor_state
*scissor
)
585 struct swr_context
*ctx
= swr_context(pipe
);
587 ctx
->scissor
= *scissor
;
588 ctx
->swr_scissor
.xmin
= scissor
->minx
;
589 ctx
->swr_scissor
.xmax
= scissor
->maxx
;
590 ctx
->swr_scissor
.ymin
= scissor
->miny
;
591 ctx
->swr_scissor
.ymax
= scissor
->maxy
;
592 ctx
->dirty
|= SWR_NEW_SCISSOR
;
596 swr_set_viewport_states(struct pipe_context
*pipe
,
598 unsigned num_viewports
,
599 const struct pipe_viewport_state
*vpt
)
601 struct swr_context
*ctx
= swr_context(pipe
);
603 ctx
->viewport
= *vpt
;
604 ctx
->dirty
|= SWR_NEW_VIEWPORT
;
609 swr_set_framebuffer_state(struct pipe_context
*pipe
,
610 const struct pipe_framebuffer_state
*fb
)
612 struct swr_context
*ctx
= swr_context(pipe
);
614 boolean changed
= !util_framebuffer_state_equal(&ctx
->framebuffer
, fb
);
616 assert(fb
->width
<= KNOB_GUARDBAND_WIDTH
);
617 assert(fb
->height
<= KNOB_GUARDBAND_HEIGHT
);
620 util_copy_framebuffer_state(&ctx
->framebuffer
, fb
);
622 ctx
->dirty
|= SWR_NEW_FRAMEBUFFER
;
628 swr_set_sample_mask(struct pipe_context
*pipe
, unsigned sample_mask
)
630 struct swr_context
*ctx
= swr_context(pipe
);
632 if (sample_mask
!= ctx
->sample_mask
) {
633 ctx
->sample_mask
= sample_mask
;
634 ctx
->dirty
|= SWR_NEW_RASTERIZER
;
639 * Update resource in-use status
640 * All resources bound to color or depth targets marked as WRITE resources.
641 * VBO Vertex/index buffers and texture views marked as READ resources.
644 swr_update_resource_status(struct pipe_context
*pipe
,
645 const struct pipe_draw_info
*p_draw_info
)
647 struct swr_context
*ctx
= swr_context(pipe
);
648 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
650 /* colorbuffer targets */
652 for (uint32_t i
= 0; i
< fb
->nr_cbufs
; ++i
)
654 swr_resource_write(fb
->cbufs
[i
]->texture
);
656 /* depth/stencil target */
658 swr_resource_write(fb
->zsbuf
->texture
);
660 /* VBO vertex buffers */
661 for (uint32_t i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
662 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
663 if (!vb
->user_buffer
)
664 swr_resource_read(vb
->buffer
);
667 /* VBO index buffer */
668 if (p_draw_info
&& p_draw_info
->indexed
) {
669 struct pipe_index_buffer
*ib
= &ctx
->index_buffer
;
670 if (!ib
->user_buffer
)
671 swr_resource_read(ib
->buffer
);
674 /* transform feedback buffers */
675 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
676 struct pipe_stream_output_target
*target
= ctx
->so_targets
[i
];
677 if (target
&& target
->buffer
)
678 swr_resource_write(target
->buffer
);
681 /* texture sampler views */
682 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
683 for (uint32_t i
= 0; i
< ctx
->num_sampler_views
[j
]; i
++) {
684 struct pipe_sampler_view
*view
= ctx
->sampler_views
[j
][i
];
686 swr_resource_read(view
->texture
);
690 /* constant buffers */
691 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
692 for (uint32_t i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
693 struct pipe_constant_buffer
*cb
= &ctx
->constants
[j
][i
];
695 swr_resource_read(cb
->buffer
);
701 swr_update_texture_state(struct swr_context
*ctx
,
702 unsigned shader_type
,
703 unsigned num_sampler_views
,
704 swr_jit_texture
*textures
)
706 for (unsigned i
= 0; i
< num_sampler_views
; i
++) {
707 struct pipe_sampler_view
*view
=
708 ctx
->sampler_views
[shader_type
][i
];
709 struct swr_jit_texture
*jit_tex
= &textures
[i
];
711 memset(jit_tex
, 0, sizeof(*jit_tex
));
713 struct pipe_resource
*res
= view
->texture
;
714 struct swr_resource
*swr_res
= swr_resource(res
);
715 SWR_SURFACE_STATE
*swr
= &swr_res
->swr
;
716 size_t *mip_offsets
= swr_res
->mip_offsets
;
717 if (swr_res
->has_depth
&& swr_res
->has_stencil
&&
718 !util_format_has_depth(util_format_description(view
->format
))) {
719 swr
= &swr_res
->secondary
;
720 mip_offsets
= swr_res
->secondary_mip_offsets
;
723 jit_tex
->width
= res
->width0
;
724 jit_tex
->height
= res
->height0
;
725 jit_tex
->base_ptr
= swr
->pBaseAddress
;
726 if (view
->target
!= PIPE_BUFFER
) {
727 jit_tex
->first_level
= view
->u
.tex
.first_level
;
728 jit_tex
->last_level
= view
->u
.tex
.last_level
;
729 if (view
->target
== PIPE_TEXTURE_3D
)
730 jit_tex
->depth
= res
->depth0
;
733 view
->u
.tex
.last_layer
- view
->u
.tex
.first_layer
+ 1;
734 jit_tex
->base_ptr
+= view
->u
.tex
.first_layer
*
735 swr
->qpitch
* swr
->pitch
;
737 unsigned view_blocksize
= util_format_get_blocksize(view
->format
);
738 jit_tex
->base_ptr
+= view
->u
.buf
.offset
;
739 jit_tex
->width
= view
->u
.buf
.size
/ view_blocksize
;
743 for (unsigned level
= jit_tex
->first_level
;
744 level
<= jit_tex
->last_level
;
746 jit_tex
->row_stride
[level
] = swr
->pitch
;
747 jit_tex
->img_stride
[level
] = swr
->qpitch
* swr
->pitch
;
748 jit_tex
->mip_offsets
[level
] = mip_offsets
[level
];
755 swr_update_sampler_state(struct swr_context
*ctx
,
756 unsigned shader_type
,
757 unsigned num_samplers
,
758 swr_jit_sampler
*samplers
)
760 for (unsigned i
= 0; i
< num_samplers
; i
++) {
761 const struct pipe_sampler_state
*sampler
=
762 ctx
->samplers
[shader_type
][i
];
765 samplers
[i
].min_lod
= sampler
->min_lod
;
766 samplers
[i
].max_lod
= sampler
->max_lod
;
767 samplers
[i
].lod_bias
= sampler
->lod_bias
;
768 COPY_4V(samplers
[i
].border_color
, sampler
->border_color
.f
);
774 swr_update_constants(struct swr_context
*ctx
, enum pipe_shader_type shaderType
)
776 swr_draw_context
*pDC
= &ctx
->swrDC
;
778 const float **constant
;
779 uint32_t *num_constants
;
780 struct swr_scratch_space
*scratch
;
782 switch (shaderType
) {
783 case PIPE_SHADER_VERTEX
:
784 constant
= pDC
->constantVS
;
785 num_constants
= pDC
->num_constantsVS
;
786 scratch
= &ctx
->scratch
->vs_constants
;
788 case PIPE_SHADER_FRAGMENT
:
789 constant
= pDC
->constantFS
;
790 num_constants
= pDC
->num_constantsFS
;
791 scratch
= &ctx
->scratch
->fs_constants
;
794 debug_printf("Unsupported shader type constants\n");
798 for (UINT i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
799 const pipe_constant_buffer
*cb
= &ctx
->constants
[shaderType
][i
];
800 num_constants
[i
] = cb
->buffer_size
;
803 (const float *)(swr_resource_data(cb
->buffer
) +
806 /* Need to copy these constants to scratch space */
807 if (cb
->user_buffer
&& cb
->buffer_size
) {
809 ((const uint8_t *)cb
->user_buffer
+ cb
->buffer_offset
);
810 uint32_t size
= AlignUp(cb
->buffer_size
, 4);
811 ptr
= swr_copy_to_scratch_space(ctx
, scratch
, ptr
, size
);
812 constant
[i
] = (const float *)ptr
;
819 swr_change_rt(struct swr_context
*ctx
,
821 const struct pipe_surface
*sf
)
823 swr_draw_context
*pDC
= &ctx
->swrDC
;
824 struct SWR_SURFACE_STATE
*rt
= &pDC
->renderTargets
[attachment
];
826 /* Do nothing if the render target hasn't changed */
827 if ((!sf
|| !sf
->texture
) && rt
->pBaseAddress
== nullptr)
830 /* Deal with disabling RT up front */
831 if (!sf
|| !sf
->texture
) {
832 /* If detaching attachment, mark tiles as RESOLVED so core
833 * won't try to load from non-existent target. */
834 swr_store_render_target(&ctx
->pipe
, attachment
, SWR_TILE_RESOLVED
);
839 const struct swr_resource
*swr
= swr_resource(sf
->texture
);
840 const SWR_SURFACE_STATE
*swr_surface
= &swr
->swr
;
841 SWR_FORMAT fmt
= mesa_to_swr_format(sf
->format
);
843 if (attachment
== SWR_ATTACHMENT_STENCIL
&& swr
->secondary
.pBaseAddress
) {
844 swr_surface
= &swr
->secondary
;
845 fmt
= swr_surface
->format
;
848 if (rt
->pBaseAddress
== swr_surface
->pBaseAddress
&&
850 rt
->lod
== sf
->u
.tex
.level
&&
851 rt
->arrayIndex
== sf
->u
.tex
.first_layer
)
854 bool need_fence
= false;
856 /* StoreTile for changed target */
857 if (rt
->pBaseAddress
) {
858 /* If changing attachment to a new target, mark tiles as
859 * INVALID so they are reloaded from surface. */
860 swr_store_render_target(&ctx
->pipe
, attachment
, SWR_TILE_INVALID
);
864 /* Make new attachment */
867 rt
->lod
= sf
->u
.tex
.level
;
868 rt
->arrayIndex
= sf
->u
.tex
.first_layer
;
874 swr_update_derived(struct pipe_context
*pipe
,
875 const struct pipe_draw_info
*p_draw_info
)
877 struct swr_context
*ctx
= swr_context(pipe
);
878 struct swr_screen
*screen
= swr_screen(ctx
->pipe
.screen
);
880 /* Update screen->pipe to current pipe context. */
881 if (screen
->pipe
!= pipe
)
884 /* Any state that requires dirty flags to be re-triggered sets this mask */
885 /* For example, user_buffer vertex and index buffers. */
886 unsigned post_update_dirty_flags
= 0;
889 if (ctx
->dirty
& SWR_NEW_FRAMEBUFFER
) {
890 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
891 const struct util_format_description
*desc
= NULL
;
892 bool need_fence
= false;
894 /* colorbuffer targets */
896 for (unsigned i
= 0; i
< fb
->nr_cbufs
; ++i
)
897 need_fence
|= swr_change_rt(
898 ctx
, SWR_ATTACHMENT_COLOR0
+ i
, fb
->cbufs
[i
]);
900 for (unsigned i
= fb
->nr_cbufs
; i
< SWR_NUM_RENDERTARGETS
; ++i
)
901 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_COLOR0
+ i
, NULL
);
903 /* depth/stencil target */
905 desc
= util_format_description(fb
->zsbuf
->format
);
906 if (fb
->zsbuf
&& util_format_has_depth(desc
))
907 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_DEPTH
, fb
->zsbuf
);
909 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_DEPTH
, NULL
);
911 if (fb
->zsbuf
&& util_format_has_stencil(desc
))
912 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_STENCIL
, fb
->zsbuf
);
914 need_fence
|= swr_change_rt(ctx
, SWR_ATTACHMENT_STENCIL
, NULL
);
916 /* This fence ensures any attachment changes are resolved before the
919 swr_fence_submit(ctx
, screen
->flush_fence
);
923 if (ctx
->dirty
& (SWR_NEW_RASTERIZER
|
924 SWR_NEW_VS
| // clipping
925 SWR_NEW_FRAMEBUFFER
)) {
926 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
927 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
929 SWR_RASTSTATE
*rastState
= &ctx
->derived
.rastState
;
930 rastState
->cullMode
= swr_convert_cull_mode(rasterizer
->cull_face
);
931 rastState
->frontWinding
= rasterizer
->front_ccw
932 ? SWR_FRONTWINDING_CCW
933 : SWR_FRONTWINDING_CW
;
934 rastState
->scissorEnable
= rasterizer
->scissor
;
935 rastState
->pointSize
= rasterizer
->point_size
> 0.0f
936 ? rasterizer
->point_size
938 rastState
->lineWidth
= rasterizer
->line_width
> 0.0f
939 ? rasterizer
->line_width
942 rastState
->pointParam
= rasterizer
->point_size_per_vertex
;
944 rastState
->pointSpriteEnable
= rasterizer
->sprite_coord_enable
;
945 rastState
->pointSpriteTopOrigin
=
946 rasterizer
->sprite_coord_mode
== PIPE_SPRITE_COORD_UPPER_LEFT
;
948 /* XXX TODO: Add multisample */
949 rastState
->msaaRastEnable
= false;
950 rastState
->rastMode
= SWR_MSAA_RASTMODE_OFF_PIXEL
;
951 rastState
->sampleCount
= SWR_MULTISAMPLE_1X
;
952 rastState
->forcedSampleCount
= false;
954 bool do_offset
= false;
955 switch (rasterizer
->fill_front
) {
956 case PIPE_POLYGON_MODE_FILL
:
957 do_offset
= rasterizer
->offset_tri
;
959 case PIPE_POLYGON_MODE_LINE
:
960 do_offset
= rasterizer
->offset_line
;
962 case PIPE_POLYGON_MODE_POINT
:
963 do_offset
= rasterizer
->offset_point
;
968 rastState
->depthBias
= rasterizer
->offset_units
;
969 rastState
->slopeScaledDepthBias
= rasterizer
->offset_scale
;
970 rastState
->depthBiasClamp
= rasterizer
->offset_clamp
;
972 rastState
->depthBias
= 0;
973 rastState
->slopeScaledDepthBias
= 0;
974 rastState
->depthBiasClamp
= 0;
976 struct pipe_surface
*zb
= fb
->zsbuf
;
977 if (zb
&& swr_resource(zb
->texture
)->has_depth
)
978 rastState
->depthFormat
= swr_resource(zb
->texture
)->swr
.format
;
980 rastState
->depthClipEnable
= rasterizer
->depth_clip
;
981 rastState
->clipHalfZ
= rasterizer
->clip_halfz
;
983 rastState
->clipDistanceMask
=
984 ctx
->vs
->info
.base
.num_written_clipdistance
?
985 ctx
->vs
->info
.base
.clipdist_writemask
& rasterizer
->clip_plane_enable
:
986 rasterizer
->clip_plane_enable
;
988 rastState
->cullDistanceMask
=
989 ctx
->vs
->info
.base
.culldist_writemask
<< ctx
->vs
->info
.base
.num_written_clipdistance
;
991 SwrSetRastState(ctx
->swrContext
, rastState
);
995 if (ctx
->dirty
& SWR_NEW_SCISSOR
) {
996 SwrSetScissorRects(ctx
->swrContext
, 1, &ctx
->swr_scissor
);
1000 if (ctx
->dirty
& (SWR_NEW_VIEWPORT
| SWR_NEW_FRAMEBUFFER
1001 | SWR_NEW_RASTERIZER
)) {
1002 pipe_viewport_state
*state
= &ctx
->viewport
;
1003 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1004 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
1006 SWR_VIEWPORT
*vp
= &ctx
->derived
.vp
;
1007 SWR_VIEWPORT_MATRICES
*vpm
= &ctx
->derived
.vpm
;
1009 vp
->x
= state
->translate
[0] - state
->scale
[0];
1010 vp
->width
= 2 * state
->scale
[0];
1011 vp
->y
= state
->translate
[1] - fabs(state
->scale
[1]);
1012 vp
->height
= 2 * fabs(state
->scale
[1]);
1013 util_viewport_zmin_zmax(state
, rasterizer
->clip_halfz
,
1014 &vp
->minZ
, &vp
->maxZ
);
1016 vpm
->m00
[0] = state
->scale
[0];
1017 vpm
->m11
[0] = state
->scale
[1];
1018 vpm
->m22
[0] = state
->scale
[2];
1019 vpm
->m30
[0] = state
->translate
[0];
1020 vpm
->m31
[0] = state
->translate
[1];
1021 vpm
->m32
[0] = state
->translate
[2];
1023 /* Now that the matrix is calculated, clip the view coords to screen
1024 * size. OpenGL allows for -ve x,y in the viewport. */
1030 vp
->height
+= vp
->y
;
1033 vp
->width
= std::min(vp
->width
, (float)fb
->width
- vp
->x
);
1034 vp
->height
= std::min(vp
->height
, (float)fb
->height
- vp
->y
);
1036 SwrSetViewports(ctx
->swrContext
, 1, vp
, vpm
);
1039 /* Set vertex & index buffers */
1040 /* (using draw info if called by swr_draw_vbo) */
1041 if (ctx
->dirty
& SWR_NEW_VERTEX
) {
1042 uint32_t size
, pitch
, max_vertex
, partial_inbounds
, scratch_total
;
1043 const uint8_t *p_data
;
1044 uint8_t *scratch
= NULL
;
1046 /* If being called by swr_draw_vbo, copy draw details */
1047 struct pipe_draw_info info
= {0};
1049 info
= *p_draw_info
;
1051 /* We must get all the scratch space in one go */
1053 for (UINT i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
1054 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
1056 if (!vb
->user_buffer
)
1060 size
= (info
.max_index
- info
.min_index
+ 1) * vb
->stride
;
1062 /* pitch = 0, means constant value
1063 * set size to 1 vertex */
1064 size
= ctx
->velems
->stream_pitch
[i
];
1067 scratch_total
+= AlignUp(size
, 4);
1070 if (scratch_total
) {
1071 scratch
= (uint8_t *)swr_copy_to_scratch_space(
1072 ctx
, &ctx
->scratch
->vertex_buffer
, NULL
, scratch_total
);
1075 /* vertex buffers */
1076 SWR_VERTEX_BUFFER_STATE swrVertexBuffers
[PIPE_MAX_ATTRIBS
];
1077 for (UINT i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
1078 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
1081 if (!vb
->user_buffer
) {
1083 * size is based on buffer->width0 rather than info.max_index
1084 * to prevent having to validate VBO on each draw */
1085 size
= vb
->buffer
->width0
;
1086 max_vertex
= size
/ pitch
;
1087 partial_inbounds
= size
% pitch
;
1089 p_data
= swr_resource_data(vb
->buffer
) + vb
->buffer_offset
;
1092 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1093 * revalidate on each draw */
1094 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1097 size
= (info
.max_index
- info
.min_index
+ 1) * pitch
;
1099 /* pitch = 0, means constant value
1100 * set size to 1 vertex */
1101 size
= ctx
->velems
->stream_pitch
[i
];
1104 max_vertex
= info
.max_index
+ 1;
1105 partial_inbounds
= 0;
1107 /* Copy only needed vertices to scratch space */
1108 size
= AlignUp(size
, 4);
1109 const void *ptr
= (const uint8_t *) vb
->user_buffer
1110 + info
.min_index
* pitch
;
1111 memcpy(scratch
, ptr
, size
);
1114 p_data
= (const uint8_t *)ptr
- info
.min_index
* pitch
;
1117 swrVertexBuffers
[i
] = {0};
1118 swrVertexBuffers
[i
].index
= i
;
1119 swrVertexBuffers
[i
].pitch
= pitch
;
1120 swrVertexBuffers
[i
].pData
= p_data
;
1121 swrVertexBuffers
[i
].size
= size
;
1122 swrVertexBuffers
[i
].maxVertex
= max_vertex
;
1123 swrVertexBuffers
[i
].partialInboundsSize
= partial_inbounds
;
1126 SwrSetVertexBuffers(
1127 ctx
->swrContext
, ctx
->num_vertex_buffers
, swrVertexBuffers
);
1129 /* index buffer, if required (info passed in by swr_draw_vbo) */
1130 SWR_FORMAT index_type
= R32_UINT
; /* Default for non-indexed draws */
1132 struct pipe_index_buffer
*ib
= &ctx
->index_buffer
;
1134 pitch
= ib
->index_size
? ib
->index_size
: sizeof(uint32_t);
1135 index_type
= swr_convert_index_type(pitch
);
1137 if (!ib
->user_buffer
) {
1139 * size is based on buffer->width0 rather than info.count
1140 * to prevent having to validate VBO on each draw */
1141 size
= ib
->buffer
->width0
;
1142 p_data
= swr_resource_data(ib
->buffer
) + ib
->offset
;
1145 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1146 * revalidate on each draw */
1147 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1149 size
= info
.count
* pitch
;
1150 size
= AlignUp(size
, 4);
1152 /* Copy indices to scratch space */
1153 const void *ptr
= ib
->user_buffer
;
1154 ptr
= swr_copy_to_scratch_space(
1155 ctx
, &ctx
->scratch
->index_buffer
, ptr
, size
);
1156 p_data
= (const uint8_t *)ptr
;
1159 SWR_INDEX_BUFFER_STATE swrIndexBuffer
;
1160 swrIndexBuffer
.format
= swr_convert_index_type(ib
->index_size
);
1161 swrIndexBuffer
.pIndices
= p_data
;
1162 swrIndexBuffer
.size
= size
;
1164 SwrSetIndexBuffer(ctx
->swrContext
, &swrIndexBuffer
);
1167 struct swr_vertex_element_state
*velems
= ctx
->velems
;
1168 if (velems
&& velems
->fsState
.indexType
!= index_type
) {
1169 velems
->fsFunc
= NULL
;
1170 velems
->fsState
.indexType
= index_type
;
1175 if (ctx
->dirty
& (SWR_NEW_VS
|
1176 SWR_NEW_RASTERIZER
| // for clip planes
1178 SWR_NEW_SAMPLER_VIEW
|
1179 SWR_NEW_FRAMEBUFFER
)) {
1181 swr_generate_vs_key(key
, ctx
, ctx
->vs
);
1182 auto search
= ctx
->vs
->map
.find(key
);
1183 PFN_VERTEX_FUNC func
;
1184 if (search
!= ctx
->vs
->map
.end()) {
1185 func
= search
->second
->shader
;
1187 func
= swr_compile_vs(ctx
, key
);
1189 SwrSetVertexFunc(ctx
->swrContext
, func
);
1191 /* JIT sampler state */
1192 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1193 swr_update_sampler_state(ctx
,
1196 ctx
->swrDC
.samplersVS
);
1199 /* JIT sampler view state */
1200 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1201 swr_update_texture_state(ctx
,
1203 key
.nr_sampler_views
,
1204 ctx
->swrDC
.texturesVS
);
1208 /* FragmentShader */
1209 if (ctx
->dirty
& (SWR_NEW_FS
| SWR_NEW_SAMPLER
| SWR_NEW_SAMPLER_VIEW
1210 | SWR_NEW_RASTERIZER
| SWR_NEW_FRAMEBUFFER
)) {
1212 swr_generate_fs_key(key
, ctx
, ctx
->fs
);
1213 auto search
= ctx
->fs
->map
.find(key
);
1214 PFN_PIXEL_KERNEL func
;
1215 if (search
!= ctx
->fs
->map
.end()) {
1216 func
= search
->second
->shader
;
1218 func
= swr_compile_fs(ctx
, key
);
1220 SWR_PS_STATE psState
= {0};
1221 psState
.pfnPixelShader
= func
;
1222 psState
.killsPixel
= ctx
->fs
->info
.base
.uses_kill
;
1223 psState
.inputCoverage
= SWR_INPUT_COVERAGE_NORMAL
;
1224 psState
.writesODepth
= ctx
->fs
->info
.base
.writes_z
;
1225 psState
.usesSourceDepth
= ctx
->fs
->info
.base
.reads_z
;
1226 psState
.shadingRate
= SWR_SHADING_RATE_PIXEL
; // XXX
1227 psState
.numRenderTargets
= ctx
->framebuffer
.nr_cbufs
;
1228 psState
.posOffset
= SWR_PS_POSITION_SAMPLE_NONE
; // XXX msaa
1229 uint32_t barycentricsMask
= 0;
1231 // when we switch to mesa-master
1232 if (ctx
->fs
->info
.base
.uses_persp_center
||
1233 ctx
->fs
->info
.base
.uses_linear_center
)
1234 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1235 if (ctx
->fs
->info
.base
.uses_persp_centroid
||
1236 ctx
->fs
->info
.base
.uses_linear_centroid
)
1237 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1238 if (ctx
->fs
->info
.base
.uses_persp_sample
||
1239 ctx
->fs
->info
.base
.uses_linear_sample
)
1240 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1242 for (unsigned i
= 0; i
< ctx
->fs
->info
.base
.num_inputs
; i
++) {
1243 switch (ctx
->fs
->info
.base
.input_interpolate_loc
[i
]) {
1244 case TGSI_INTERPOLATE_LOC_CENTER
:
1245 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1247 case TGSI_INTERPOLATE_LOC_CENTROID
:
1248 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1250 case TGSI_INTERPOLATE_LOC_SAMPLE
:
1251 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1256 psState
.barycentricsMask
= barycentricsMask
;
1257 psState
.usesUAV
= false; // XXX
1258 psState
.forceEarlyZ
= false;
1259 SwrSetPixelShaderState(ctx
->swrContext
, &psState
);
1261 /* JIT sampler state */
1262 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1263 swr_update_sampler_state(ctx
,
1264 PIPE_SHADER_FRAGMENT
,
1266 ctx
->swrDC
.samplersFS
);
1269 /* JIT sampler view state */
1270 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1271 swr_update_texture_state(ctx
,
1272 PIPE_SHADER_FRAGMENT
,
1273 key
.nr_sampler_views
,
1274 ctx
->swrDC
.texturesFS
);
1279 /* VertexShader Constants */
1280 if (ctx
->dirty
& SWR_NEW_VSCONSTANTS
) {
1281 swr_update_constants(ctx
, PIPE_SHADER_VERTEX
);
1284 /* FragmentShader Constants */
1285 if (ctx
->dirty
& SWR_NEW_FSCONSTANTS
) {
1286 swr_update_constants(ctx
, PIPE_SHADER_FRAGMENT
);
1289 /* Depth/stencil state */
1290 if (ctx
->dirty
& (SWR_NEW_DEPTH_STENCIL_ALPHA
| SWR_NEW_FRAMEBUFFER
)) {
1291 struct pipe_depth_state
*depth
= &(ctx
->depth_stencil
->depth
);
1292 struct pipe_stencil_state
*stencil
= ctx
->depth_stencil
->stencil
;
1293 SWR_DEPTH_STENCIL_STATE depthStencilState
= {{0}};
1294 SWR_DEPTH_BOUNDS_STATE depthBoundsState
= {0};
1296 /* XXX, incomplete. Need to flesh out stencil & alpha test state
1297 struct pipe_stencil_state *front_stencil =
1298 ctx->depth_stencil.stencil[0];
1299 struct pipe_stencil_state *back_stencil = ctx->depth_stencil.stencil[1];
1300 struct pipe_alpha_state alpha;
1302 if (stencil
[0].enabled
) {
1303 depthStencilState
.stencilWriteEnable
= 1;
1304 depthStencilState
.stencilTestEnable
= 1;
1305 depthStencilState
.stencilTestFunc
=
1306 swr_convert_depth_func(stencil
[0].func
);
1308 depthStencilState
.stencilPassDepthPassOp
=
1309 swr_convert_stencil_op(stencil
[0].zpass_op
);
1310 depthStencilState
.stencilPassDepthFailOp
=
1311 swr_convert_stencil_op(stencil
[0].zfail_op
);
1312 depthStencilState
.stencilFailOp
=
1313 swr_convert_stencil_op(stencil
[0].fail_op
);
1314 depthStencilState
.stencilWriteMask
= stencil
[0].writemask
;
1315 depthStencilState
.stencilTestMask
= stencil
[0].valuemask
;
1316 depthStencilState
.stencilRefValue
= ctx
->stencil_ref
.ref_value
[0];
1318 if (stencil
[1].enabled
) {
1319 depthStencilState
.doubleSidedStencilTestEnable
= 1;
1321 depthStencilState
.backfaceStencilTestFunc
=
1322 swr_convert_depth_func(stencil
[1].func
);
1324 depthStencilState
.backfaceStencilPassDepthPassOp
=
1325 swr_convert_stencil_op(stencil
[1].zpass_op
);
1326 depthStencilState
.backfaceStencilPassDepthFailOp
=
1327 swr_convert_stencil_op(stencil
[1].zfail_op
);
1328 depthStencilState
.backfaceStencilFailOp
=
1329 swr_convert_stencil_op(stencil
[1].fail_op
);
1330 depthStencilState
.backfaceStencilWriteMask
= stencil
[1].writemask
;
1331 depthStencilState
.backfaceStencilTestMask
= stencil
[1].valuemask
;
1333 depthStencilState
.backfaceStencilRefValue
=
1334 ctx
->stencil_ref
.ref_value
[1];
1337 depthStencilState
.depthTestEnable
= depth
->enabled
;
1338 depthStencilState
.depthTestFunc
= swr_convert_depth_func(depth
->func
);
1339 depthStencilState
.depthWriteEnable
= depth
->writemask
;
1340 SwrSetDepthStencilState(ctx
->swrContext
, &depthStencilState
);
1342 depthBoundsState
.depthBoundsTestEnable
= depth
->bounds_test
;
1343 depthBoundsState
.depthBoundsTestMinValue
= depth
->bounds_min
;
1344 depthBoundsState
.depthBoundsTestMaxValue
= depth
->bounds_max
;
1345 SwrSetDepthBoundsState(ctx
->swrContext
, &depthBoundsState
);
1349 if (ctx
->dirty
& (SWR_NEW_BLEND
|
1350 SWR_NEW_FRAMEBUFFER
|
1351 SWR_NEW_DEPTH_STENCIL_ALPHA
)) {
1352 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1354 SWR_BLEND_STATE blendState
;
1355 memcpy(&blendState
, &ctx
->blend
->blendState
, sizeof(blendState
));
1356 blendState
.constantColor
[0] = ctx
->blend_color
.color
[0];
1357 blendState
.constantColor
[1] = ctx
->blend_color
.color
[1];
1358 blendState
.constantColor
[2] = ctx
->blend_color
.color
[2];
1359 blendState
.constantColor
[3] = ctx
->blend_color
.color
[3];
1360 blendState
.alphaTestReference
=
1361 *((uint32_t*)&ctx
->depth_stencil
->alpha
.ref_value
);
1364 blendState
.sampleMask
= 0;
1365 blendState
.sampleCount
= SWR_MULTISAMPLE_1X
;
1367 /* If there are no color buffers bound, disable writes on RT0
1369 if (fb
->nr_cbufs
== 0) {
1370 blendState
.renderTarget
[0].writeDisableRed
= 1;
1371 blendState
.renderTarget
[0].writeDisableGreen
= 1;
1372 blendState
.renderTarget
[0].writeDisableBlue
= 1;
1373 blendState
.renderTarget
[0].writeDisableAlpha
= 1;
1374 SwrSetBlendFunc(ctx
->swrContext
, 0, NULL
);
1377 for (int target
= 0;
1378 target
< std::min(SWR_NUM_RENDERTARGETS
,
1379 PIPE_MAX_COLOR_BUFS
);
1381 if (!fb
->cbufs
[target
])
1384 struct swr_resource
*colorBuffer
=
1385 swr_resource(fb
->cbufs
[target
]->texture
);
1387 BLEND_COMPILE_STATE compileState
;
1388 memset(&compileState
, 0, sizeof(compileState
));
1389 compileState
.format
= colorBuffer
->swr
.format
;
1390 memcpy(&compileState
.blendState
,
1391 &ctx
->blend
->compileState
[target
],
1392 sizeof(compileState
.blendState
));
1394 const SWR_FORMAT_INFO
& info
= GetFormatInfo(compileState
.format
);
1395 if (compileState
.blendState
.logicOpEnable
&&
1396 ((info
.type
[0] == SWR_TYPE_FLOAT
) || info
.isSRGB
)) {
1397 compileState
.blendState
.logicOpEnable
= false;
1400 if (info
.type
[0] == SWR_TYPE_SINT
|| info
.type
[0] == SWR_TYPE_UINT
)
1401 compileState
.blendState
.blendEnable
= false;
1403 if (compileState
.blendState
.blendEnable
== false &&
1404 compileState
.blendState
.logicOpEnable
== false &&
1405 ctx
->depth_stencil
->alpha
.enabled
== 0) {
1406 SwrSetBlendFunc(ctx
->swrContext
, target
, NULL
);
1410 compileState
.desc
.alphaTestEnable
=
1411 ctx
->depth_stencil
->alpha
.enabled
;
1412 compileState
.desc
.independentAlphaBlendEnable
=
1413 (compileState
.blendState
.sourceBlendFactor
!=
1414 compileState
.blendState
.sourceAlphaBlendFactor
) ||
1415 (compileState
.blendState
.destBlendFactor
!=
1416 compileState
.blendState
.destAlphaBlendFactor
) ||
1417 (compileState
.blendState
.colorBlendFunc
!=
1418 compileState
.blendState
.alphaBlendFunc
);
1419 compileState
.desc
.alphaToCoverageEnable
=
1420 ctx
->blend
->pipe
.alpha_to_coverage
;
1421 compileState
.desc
.sampleMaskEnable
= 0; // XXX
1422 compileState
.desc
.numSamples
= 1; // XXX
1424 compileState
.alphaTestFunction
=
1425 swr_convert_depth_func(ctx
->depth_stencil
->alpha
.func
);
1426 compileState
.alphaTestFormat
= ALPHA_TEST_FLOAT32
; // xxx
1428 compileState
.Canonicalize();
1430 PFN_BLEND_JIT_FUNC func
= NULL
;
1431 auto search
= ctx
->blendJIT
->find(compileState
);
1432 if (search
!= ctx
->blendJIT
->end()) {
1433 func
= search
->second
;
1435 HANDLE hJitMgr
= screen
->hJitMgr
;
1436 func
= JitCompileBlend(hJitMgr
, compileState
);
1437 debug_printf("BLEND shader %p\n", func
);
1438 assert(func
&& "Error: BlendShader = NULL");
1440 ctx
->blendJIT
->insert(std::make_pair(compileState
, func
));
1442 SwrSetBlendFunc(ctx
->swrContext
, target
, func
);
1445 SwrSetBlendState(ctx
->swrContext
, &blendState
);
1448 if (ctx
->dirty
& SWR_NEW_STIPPLE
) {
1449 /* XXX What to do with this one??? SWR doesn't stipple */
1452 if (ctx
->dirty
& (SWR_NEW_VS
| SWR_NEW_SO
| SWR_NEW_RASTERIZER
)) {
1453 ctx
->vs
->soState
.rasterizerDisable
=
1454 ctx
->rasterizer
->rasterizer_discard
;
1455 SwrSetSoState(ctx
->swrContext
, &ctx
->vs
->soState
);
1457 pipe_stream_output_info
*stream_output
= &ctx
->vs
->pipe
.stream_output
;
1459 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
1460 SWR_STREAMOUT_BUFFER buffer
= {0};
1461 if (!ctx
->so_targets
[i
])
1463 buffer
.enable
= true;
1465 (uint32_t *)swr_resource_data(ctx
->so_targets
[i
]->buffer
);
1466 buffer
.bufferSize
= ctx
->so_targets
[i
]->buffer_size
>> 2;
1467 buffer
.pitch
= stream_output
->stride
[i
];
1468 buffer
.streamOffset
= ctx
->so_targets
[i
]->buffer_offset
>> 2;
1470 SwrSetSoBuffers(ctx
->swrContext
, &buffer
, i
);
1474 if (ctx
->dirty
& SWR_NEW_CLIP
) {
1475 // shader exporting clip distances overrides all user clip planes
1476 if (ctx
->rasterizer
->clip_plane_enable
&&
1477 !ctx
->vs
->info
.base
.num_written_clipdistance
)
1479 swr_draw_context
*pDC
= &ctx
->swrDC
;
1480 memcpy(pDC
->userClipPlanes
,
1482 sizeof(pDC
->userClipPlanes
));
1486 // set up backend state
1487 SWR_BACKEND_STATE backendState
= {0};
1488 backendState
.numAttributes
=
1489 ctx
->vs
->info
.base
.num_outputs
- 1 +
1490 (ctx
->rasterizer
->sprite_coord_enable
? 1 : 0);
1491 for (unsigned i
= 0; i
< backendState
.numAttributes
; i
++)
1492 backendState
.numComponents
[i
] = 4;
1493 backendState
.constantInterpolationMask
= ctx
->fs
->constantMask
|
1494 (ctx
->rasterizer
->flatshade
? ctx
->fs
->flatConstantMask
: 0);
1495 backendState
.pointSpriteTexCoordMask
= ctx
->fs
->pointSpriteMask
;
1497 SwrSetBackendState(ctx
->swrContext
, &backendState
);
1499 /* Ensure that any in-progress attachment change StoreTiles finish */
1500 if (swr_is_fence_pending(screen
->flush_fence
))
1501 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
1503 /* Finally, update the in-use status of all resources involved in draw */
1504 swr_update_resource_status(pipe
, p_draw_info
);
1506 ctx
->dirty
= post_update_dirty_flags
;
1510 static struct pipe_stream_output_target
*
1511 swr_create_so_target(struct pipe_context
*pipe
,
1512 struct pipe_resource
*buffer
,
1513 unsigned buffer_offset
,
1514 unsigned buffer_size
)
1516 struct pipe_stream_output_target
*target
;
1518 target
= CALLOC_STRUCT(pipe_stream_output_target
);
1522 target
->context
= pipe
;
1523 target
->reference
.count
= 1;
1524 pipe_resource_reference(&target
->buffer
, buffer
);
1525 target
->buffer_offset
= buffer_offset
;
1526 target
->buffer_size
= buffer_size
;
1531 swr_destroy_so_target(struct pipe_context
*pipe
,
1532 struct pipe_stream_output_target
*target
)
1534 pipe_resource_reference(&target
->buffer
, NULL
);
1539 swr_set_so_targets(struct pipe_context
*pipe
,
1540 unsigned num_targets
,
1541 struct pipe_stream_output_target
**targets
,
1542 const unsigned *offsets
)
1544 struct swr_context
*swr
= swr_context(pipe
);
1547 assert(num_targets
< MAX_SO_STREAMS
);
1549 for (i
= 0; i
< num_targets
; i
++) {
1550 pipe_so_target_reference(
1551 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
],
1555 for (/* fall-through */; i
< swr
->num_so_targets
; i
++) {
1556 pipe_so_target_reference(
1557 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
], NULL
);
1560 swr
->num_so_targets
= num_targets
;
1562 swr
->dirty
|= SWR_NEW_SO
;
1567 swr_state_init(struct pipe_context
*pipe
)
1569 pipe
->create_blend_state
= swr_create_blend_state
;
1570 pipe
->bind_blend_state
= swr_bind_blend_state
;
1571 pipe
->delete_blend_state
= swr_delete_blend_state
;
1573 pipe
->create_depth_stencil_alpha_state
= swr_create_depth_stencil_state
;
1574 pipe
->bind_depth_stencil_alpha_state
= swr_bind_depth_stencil_state
;
1575 pipe
->delete_depth_stencil_alpha_state
= swr_delete_depth_stencil_state
;
1577 pipe
->create_rasterizer_state
= swr_create_rasterizer_state
;
1578 pipe
->bind_rasterizer_state
= swr_bind_rasterizer_state
;
1579 pipe
->delete_rasterizer_state
= swr_delete_rasterizer_state
;
1581 pipe
->create_sampler_state
= swr_create_sampler_state
;
1582 pipe
->bind_sampler_states
= swr_bind_sampler_states
;
1583 pipe
->delete_sampler_state
= swr_delete_sampler_state
;
1585 pipe
->create_sampler_view
= swr_create_sampler_view
;
1586 pipe
->set_sampler_views
= swr_set_sampler_views
;
1587 pipe
->sampler_view_destroy
= swr_sampler_view_destroy
;
1589 pipe
->create_vs_state
= swr_create_vs_state
;
1590 pipe
->bind_vs_state
= swr_bind_vs_state
;
1591 pipe
->delete_vs_state
= swr_delete_vs_state
;
1593 pipe
->create_fs_state
= swr_create_fs_state
;
1594 pipe
->bind_fs_state
= swr_bind_fs_state
;
1595 pipe
->delete_fs_state
= swr_delete_fs_state
;
1597 pipe
->set_constant_buffer
= swr_set_constant_buffer
;
1599 pipe
->create_vertex_elements_state
= swr_create_vertex_elements_state
;
1600 pipe
->bind_vertex_elements_state
= swr_bind_vertex_elements_state
;
1601 pipe
->delete_vertex_elements_state
= swr_delete_vertex_elements_state
;
1603 pipe
->set_vertex_buffers
= swr_set_vertex_buffers
;
1604 pipe
->set_index_buffer
= swr_set_index_buffer
;
1606 pipe
->set_polygon_stipple
= swr_set_polygon_stipple
;
1607 pipe
->set_clip_state
= swr_set_clip_state
;
1608 pipe
->set_scissor_states
= swr_set_scissor_states
;
1609 pipe
->set_viewport_states
= swr_set_viewport_states
;
1611 pipe
->set_framebuffer_state
= swr_set_framebuffer_state
;
1613 pipe
->set_blend_color
= swr_set_blend_color
;
1614 pipe
->set_stencil_ref
= swr_set_stencil_ref
;
1616 pipe
->set_sample_mask
= swr_set_sample_mask
;
1618 pipe
->create_stream_output_target
= swr_create_so_target
;
1619 pipe
->stream_output_target_destroy
= swr_destroy_so_target
;
1620 pipe
->set_stream_output_targets
= swr_set_so_targets
;