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
);
621 for (i
= 0; i
< fb
->nr_cbufs
; ++i
)
622 pipe_surface_reference(&ctx
->framebuffer
.cbufs
[i
], fb
->cbufs
[i
]);
623 for (; i
< ctx
->framebuffer
.nr_cbufs
; ++i
)
624 pipe_surface_reference(&ctx
->framebuffer
.cbufs
[i
], NULL
);
626 ctx
->framebuffer
.nr_cbufs
= fb
->nr_cbufs
;
628 ctx
->framebuffer
.width
= fb
->width
;
629 ctx
->framebuffer
.height
= fb
->height
;
631 pipe_surface_reference(&ctx
->framebuffer
.zsbuf
, fb
->zsbuf
);
633 ctx
->dirty
|= SWR_NEW_FRAMEBUFFER
;
639 swr_set_sample_mask(struct pipe_context
*pipe
, unsigned sample_mask
)
641 struct swr_context
*ctx
= swr_context(pipe
);
643 if (sample_mask
!= ctx
->sample_mask
) {
644 ctx
->sample_mask
= sample_mask
;
645 ctx
->dirty
|= SWR_NEW_RASTERIZER
;
650 * Update resource in-use status
651 * All resources bound to color or depth targets marked as WRITE resources.
652 * VBO Vertex/index buffers and texture views marked as READ resources.
655 swr_update_resource_status(struct pipe_context
*pipe
,
656 const struct pipe_draw_info
*p_draw_info
)
658 struct swr_context
*ctx
= swr_context(pipe
);
659 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
661 /* colorbuffer targets */
663 for (uint32_t i
= 0; i
< fb
->nr_cbufs
; ++i
)
665 swr_resource_write(fb
->cbufs
[i
]->texture
);
667 /* depth/stencil target */
669 swr_resource_write(fb
->zsbuf
->texture
);
671 /* VBO vertex buffers */
672 for (uint32_t i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
673 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
674 if (!vb
->user_buffer
)
675 swr_resource_read(vb
->buffer
);
678 /* VBO index buffer */
679 if (p_draw_info
&& p_draw_info
->indexed
) {
680 struct pipe_index_buffer
*ib
= &ctx
->index_buffer
;
681 if (!ib
->user_buffer
)
682 swr_resource_read(ib
->buffer
);
685 /* transform feedback buffers */
686 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
687 struct pipe_stream_output_target
*target
= ctx
->so_targets
[i
];
688 if (target
&& target
->buffer
)
689 swr_resource_write(target
->buffer
);
692 /* texture sampler views */
693 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
694 for (uint32_t i
= 0; i
< ctx
->num_sampler_views
[j
]; i
++) {
695 struct pipe_sampler_view
*view
= ctx
->sampler_views
[j
][i
];
697 swr_resource_read(view
->texture
);
701 /* constant buffers */
702 for (uint32_t j
: {PIPE_SHADER_VERTEX
, PIPE_SHADER_FRAGMENT
}) {
703 for (uint32_t i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
704 struct pipe_constant_buffer
*cb
= &ctx
->constants
[j
][i
];
706 swr_resource_read(cb
->buffer
);
712 swr_update_texture_state(struct swr_context
*ctx
,
713 unsigned shader_type
,
714 unsigned num_sampler_views
,
715 swr_jit_texture
*textures
)
717 for (unsigned i
= 0; i
< num_sampler_views
; i
++) {
718 struct pipe_sampler_view
*view
=
719 ctx
->sampler_views
[shader_type
][i
];
722 struct pipe_resource
*res
= view
->texture
;
723 struct swr_resource
*swr_res
= swr_resource(res
);
724 struct swr_jit_texture
*jit_tex
= &textures
[i
];
725 memset(jit_tex
, 0, sizeof(*jit_tex
));
726 jit_tex
->width
= res
->width0
;
727 jit_tex
->height
= res
->height0
;
728 jit_tex
->depth
= res
->depth0
;
729 jit_tex
->first_level
= view
->u
.tex
.first_level
;
730 jit_tex
->last_level
= view
->u
.tex
.last_level
;
731 jit_tex
->base_ptr
= swr_res
->swr
.pBaseAddress
;
733 for (unsigned level
= jit_tex
->first_level
;
734 level
<= jit_tex
->last_level
;
736 jit_tex
->row_stride
[level
] = swr_res
->row_stride
[level
];
737 jit_tex
->img_stride
[level
] = swr_res
->img_stride
[level
];
738 jit_tex
->mip_offsets
[level
] = swr_res
->mip_offsets
[level
];
745 swr_update_sampler_state(struct swr_context
*ctx
,
746 unsigned shader_type
,
747 unsigned num_samplers
,
748 swr_jit_sampler
*samplers
)
750 for (unsigned i
= 0; i
< num_samplers
; i
++) {
751 const struct pipe_sampler_state
*sampler
=
752 ctx
->samplers
[shader_type
][i
];
755 samplers
[i
].min_lod
= sampler
->min_lod
;
756 samplers
[i
].max_lod
= sampler
->max_lod
;
757 samplers
[i
].lod_bias
= sampler
->lod_bias
;
758 COPY_4V(samplers
[i
].border_color
, sampler
->border_color
.f
);
764 swr_update_constants(struct swr_context
*ctx
, enum pipe_shader_type shaderType
)
766 swr_draw_context
*pDC
= &ctx
->swrDC
;
768 const float **constant
;
769 uint32_t *num_constants
;
770 struct swr_scratch_space
*scratch
;
772 switch (shaderType
) {
773 case PIPE_SHADER_VERTEX
:
774 constant
= pDC
->constantVS
;
775 num_constants
= pDC
->num_constantsVS
;
776 scratch
= &ctx
->scratch
->vs_constants
;
778 case PIPE_SHADER_FRAGMENT
:
779 constant
= pDC
->constantFS
;
780 num_constants
= pDC
->num_constantsFS
;
781 scratch
= &ctx
->scratch
->fs_constants
;
784 debug_printf("Unsupported shader type constants\n");
788 for (UINT i
= 0; i
< PIPE_MAX_CONSTANT_BUFFERS
; i
++) {
789 const pipe_constant_buffer
*cb
= &ctx
->constants
[shaderType
][i
];
790 num_constants
[i
] = cb
->buffer_size
;
793 (const float *)(swr_resource_data(cb
->buffer
) +
796 /* Need to copy these constants to scratch space */
797 if (cb
->user_buffer
&& cb
->buffer_size
) {
799 ((const uint8_t *)cb
->user_buffer
+ cb
->buffer_offset
);
800 uint32_t size
= AlignUp(cb
->buffer_size
, 4);
801 ptr
= swr_copy_to_scratch_space(ctx
, scratch
, ptr
, size
);
802 constant
[i
] = (const float *)ptr
;
809 swr_update_derived(struct pipe_context
*pipe
,
810 const struct pipe_draw_info
*p_draw_info
)
812 struct swr_context
*ctx
= swr_context(pipe
);
813 struct swr_screen
*screen
= swr_screen(ctx
->pipe
.screen
);
815 /* Update screen->pipe to current pipe context. */
816 if (screen
->pipe
!= pipe
)
819 /* Any state that requires dirty flags to be re-triggered sets this mask */
820 /* For example, user_buffer vertex and index buffers. */
821 unsigned post_update_dirty_flags
= 0;
824 if (ctx
->dirty
& SWR_NEW_FRAMEBUFFER
) {
825 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
826 SWR_SURFACE_STATE
*new_attachment
[SWR_NUM_ATTACHMENTS
] = {0};
829 /* colorbuffer targets */
831 for (i
= 0; i
< fb
->nr_cbufs
; ++i
)
833 struct swr_resource
*colorBuffer
=
834 swr_resource(fb
->cbufs
[i
]->texture
);
835 new_attachment
[SWR_ATTACHMENT_COLOR0
+ i
] = &colorBuffer
->swr
;
838 /* depth/stencil target */
840 struct swr_resource
*depthStencilBuffer
=
841 swr_resource(fb
->zsbuf
->texture
);
842 if (depthStencilBuffer
->has_depth
) {
843 new_attachment
[SWR_ATTACHMENT_DEPTH
] = &depthStencilBuffer
->swr
;
845 if (depthStencilBuffer
->has_stencil
)
846 new_attachment
[SWR_ATTACHMENT_STENCIL
] =
847 &depthStencilBuffer
->secondary
;
849 } else if (depthStencilBuffer
->has_stencil
)
850 new_attachment
[SWR_ATTACHMENT_STENCIL
] = &depthStencilBuffer
->swr
;
853 /* Make the attachment updates */
854 swr_draw_context
*pDC
= &ctx
->swrDC
;
855 SWR_SURFACE_STATE
*renderTargets
= pDC
->renderTargets
;
856 unsigned need_fence
= FALSE
;
857 for (i
= 0; i
< SWR_NUM_ATTACHMENTS
; i
++) {
858 void *new_base
= nullptr;
859 if (new_attachment
[i
])
860 new_base
= new_attachment
[i
]->pBaseAddress
;
862 /* StoreTile for changed target */
863 if (renderTargets
[i
].pBaseAddress
!= new_base
) {
864 if (renderTargets
[i
].pBaseAddress
) {
865 /* If changing attachment to a new target, mark tiles as
866 * INVALID so they are reloaded from surface.
867 * If detaching attachment, mark tiles as RESOLVED so core
868 * won't try to load from non-existent target. */
869 enum SWR_TILE_STATE post_state
= (new_attachment
[i
]
870 ? SWR_TILE_INVALID
: SWR_TILE_RESOLVED
);
871 swr_store_render_target(pipe
, i
, post_state
);
876 /* Make new attachment */
877 if (new_attachment
[i
])
878 renderTargets
[i
] = *new_attachment
[i
];
880 if (renderTargets
[i
].pBaseAddress
)
881 renderTargets
[i
] = {0};
885 /* This fence ensures any attachment changes are resolved before the
888 swr_fence_submit(ctx
, screen
->flush_fence
);
892 if (ctx
->dirty
& (SWR_NEW_RASTERIZER
|
893 SWR_NEW_VS
| // clipping
894 SWR_NEW_FRAMEBUFFER
)) {
895 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
896 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
898 SWR_RASTSTATE
*rastState
= &ctx
->derived
.rastState
;
899 rastState
->cullMode
= swr_convert_cull_mode(rasterizer
->cull_face
);
900 rastState
->frontWinding
= rasterizer
->front_ccw
901 ? SWR_FRONTWINDING_CCW
902 : SWR_FRONTWINDING_CW
;
903 rastState
->scissorEnable
= rasterizer
->scissor
;
904 rastState
->pointSize
= rasterizer
->point_size
> 0.0f
905 ? rasterizer
->point_size
907 rastState
->lineWidth
= rasterizer
->line_width
> 0.0f
908 ? rasterizer
->line_width
911 rastState
->pointParam
= rasterizer
->point_size_per_vertex
;
913 rastState
->pointSpriteEnable
= rasterizer
->sprite_coord_enable
;
914 rastState
->pointSpriteTopOrigin
=
915 rasterizer
->sprite_coord_mode
== PIPE_SPRITE_COORD_UPPER_LEFT
;
917 /* XXX TODO: Add multisample */
918 rastState
->msaaRastEnable
= false;
919 rastState
->rastMode
= SWR_MSAA_RASTMODE_OFF_PIXEL
;
920 rastState
->sampleCount
= SWR_MULTISAMPLE_1X
;
921 rastState
->forcedSampleCount
= false;
923 bool do_offset
= false;
924 switch (rasterizer
->fill_front
) {
925 case PIPE_POLYGON_MODE_FILL
:
926 do_offset
= rasterizer
->offset_tri
;
928 case PIPE_POLYGON_MODE_LINE
:
929 do_offset
= rasterizer
->offset_line
;
931 case PIPE_POLYGON_MODE_POINT
:
932 do_offset
= rasterizer
->offset_point
;
937 rastState
->depthBias
= rasterizer
->offset_units
;
938 rastState
->slopeScaledDepthBias
= rasterizer
->offset_scale
;
939 rastState
->depthBiasClamp
= rasterizer
->offset_clamp
;
941 rastState
->depthBias
= 0;
942 rastState
->slopeScaledDepthBias
= 0;
943 rastState
->depthBiasClamp
= 0;
945 struct pipe_surface
*zb
= fb
->zsbuf
;
946 if (zb
&& swr_resource(zb
->texture
)->has_depth
)
947 rastState
->depthFormat
= swr_resource(zb
->texture
)->swr
.format
;
949 rastState
->depthClipEnable
= rasterizer
->depth_clip
;
950 rastState
->clipHalfZ
= rasterizer
->clip_halfz
;
952 rastState
->clipDistanceMask
=
953 ctx
->vs
->info
.base
.num_written_clipdistance
?
954 ctx
->vs
->info
.base
.clipdist_writemask
& rasterizer
->clip_plane_enable
:
955 rasterizer
->clip_plane_enable
;
957 rastState
->cullDistanceMask
=
958 ctx
->vs
->info
.base
.culldist_writemask
<< ctx
->vs
->info
.base
.num_written_clipdistance
;
960 SwrSetRastState(ctx
->swrContext
, rastState
);
964 if (ctx
->dirty
& SWR_NEW_SCISSOR
) {
965 SwrSetScissorRects(ctx
->swrContext
, 1, &ctx
->swr_scissor
);
969 if (ctx
->dirty
& (SWR_NEW_VIEWPORT
| SWR_NEW_FRAMEBUFFER
970 | SWR_NEW_RASTERIZER
)) {
971 pipe_viewport_state
*state
= &ctx
->viewport
;
972 pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
973 pipe_rasterizer_state
*rasterizer
= ctx
->rasterizer
;
975 SWR_VIEWPORT
*vp
= &ctx
->derived
.vp
;
976 SWR_VIEWPORT_MATRICES
*vpm
= &ctx
->derived
.vpm
;
978 vp
->x
= state
->translate
[0] - state
->scale
[0];
979 vp
->width
= 2 * state
->scale
[0];
980 vp
->y
= state
->translate
[1] - fabs(state
->scale
[1]);
981 vp
->height
= 2 * fabs(state
->scale
[1]);
982 util_viewport_zmin_zmax(state
, rasterizer
->clip_halfz
,
983 &vp
->minZ
, &vp
->maxZ
);
985 vpm
->m00
[0] = state
->scale
[0];
986 vpm
->m11
[0] = state
->scale
[1];
987 vpm
->m22
[0] = state
->scale
[2];
988 vpm
->m30
[0] = state
->translate
[0];
989 vpm
->m31
[0] = state
->translate
[1];
990 vpm
->m32
[0] = state
->translate
[2];
992 /* Now that the matrix is calculated, clip the view coords to screen
993 * size. OpenGL allows for -ve x,y in the viewport. */
1002 vp
->width
= std::min(vp
->width
, (float)fb
->width
- vp
->x
);
1003 vp
->height
= std::min(vp
->height
, (float)fb
->height
- vp
->y
);
1005 SwrSetViewports(ctx
->swrContext
, 1, vp
, vpm
);
1008 /* Set vertex & index buffers */
1009 /* (using draw info if called by swr_draw_vbo) */
1010 if (ctx
->dirty
& SWR_NEW_VERTEX
) {
1011 uint32_t size
, pitch
, max_vertex
, partial_inbounds
, scratch_total
;
1012 const uint8_t *p_data
;
1013 uint8_t *scratch
= NULL
;
1015 /* If being called by swr_draw_vbo, copy draw details */
1016 struct pipe_draw_info info
= {0};
1018 info
= *p_draw_info
;
1020 /* We must get all the scratch space in one go */
1022 for (UINT i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
1023 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
1025 if (!vb
->user_buffer
)
1029 size
= (info
.max_index
- info
.min_index
+ 1) * vb
->stride
;
1031 /* pitch = 0, means constant value
1032 * set size to 1 vertex */
1033 size
= ctx
->velems
->stream_pitch
[i
];
1036 scratch_total
+= AlignUp(size
, 4);
1039 if (scratch_total
) {
1040 scratch
= (uint8_t *)swr_copy_to_scratch_space(
1041 ctx
, &ctx
->scratch
->vertex_buffer
, NULL
, scratch_total
);
1044 /* vertex buffers */
1045 SWR_VERTEX_BUFFER_STATE swrVertexBuffers
[PIPE_MAX_ATTRIBS
];
1046 for (UINT i
= 0; i
< ctx
->num_vertex_buffers
; i
++) {
1047 struct pipe_vertex_buffer
*vb
= &ctx
->vertex_buffer
[i
];
1050 if (!vb
->user_buffer
) {
1052 * size is based on buffer->width0 rather than info.max_index
1053 * to prevent having to validate VBO on each draw */
1054 size
= vb
->buffer
->width0
;
1055 max_vertex
= size
/ pitch
;
1056 partial_inbounds
= size
% pitch
;
1058 p_data
= swr_resource_data(vb
->buffer
) + vb
->buffer_offset
;
1061 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1062 * revalidate on each draw */
1063 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1066 size
= (info
.max_index
- info
.min_index
+ 1) * pitch
;
1068 /* pitch = 0, means constant value
1069 * set size to 1 vertex */
1070 size
= ctx
->velems
->stream_pitch
[i
];
1073 max_vertex
= info
.max_index
+ 1;
1074 partial_inbounds
= 0;
1076 /* Copy only needed vertices to scratch space */
1077 size
= AlignUp(size
, 4);
1078 const void *ptr
= (const uint8_t *) vb
->user_buffer
1079 + info
.min_index
* pitch
;
1080 memcpy(scratch
, ptr
, size
);
1083 p_data
= (const uint8_t *)ptr
- info
.min_index
* pitch
;
1086 swrVertexBuffers
[i
] = {0};
1087 swrVertexBuffers
[i
].index
= i
;
1088 swrVertexBuffers
[i
].pitch
= pitch
;
1089 swrVertexBuffers
[i
].pData
= p_data
;
1090 swrVertexBuffers
[i
].size
= size
;
1091 swrVertexBuffers
[i
].maxVertex
= max_vertex
;
1092 swrVertexBuffers
[i
].partialInboundsSize
= partial_inbounds
;
1095 SwrSetVertexBuffers(
1096 ctx
->swrContext
, ctx
->num_vertex_buffers
, swrVertexBuffers
);
1098 /* index buffer, if required (info passed in by swr_draw_vbo) */
1099 SWR_FORMAT index_type
= R32_UINT
; /* Default for non-indexed draws */
1101 struct pipe_index_buffer
*ib
= &ctx
->index_buffer
;
1103 pitch
= ib
->index_size
? ib
->index_size
: sizeof(uint32_t);
1104 index_type
= swr_convert_index_type(pitch
);
1106 if (!ib
->user_buffer
) {
1108 * size is based on buffer->width0 rather than info.count
1109 * to prevent having to validate VBO on each draw */
1110 size
= ib
->buffer
->width0
;
1111 p_data
= swr_resource_data(ib
->buffer
) + ib
->offset
;
1114 * client memory is one-time use, re-trigger SWR_NEW_VERTEX to
1115 * revalidate on each draw */
1116 post_update_dirty_flags
|= SWR_NEW_VERTEX
;
1118 size
= info
.count
* pitch
;
1119 size
= AlignUp(size
, 4);
1121 /* Copy indices to scratch space */
1122 const void *ptr
= ib
->user_buffer
;
1123 ptr
= swr_copy_to_scratch_space(
1124 ctx
, &ctx
->scratch
->index_buffer
, ptr
, size
);
1125 p_data
= (const uint8_t *)ptr
;
1128 SWR_INDEX_BUFFER_STATE swrIndexBuffer
;
1129 swrIndexBuffer
.format
= swr_convert_index_type(ib
->index_size
);
1130 swrIndexBuffer
.pIndices
= p_data
;
1131 swrIndexBuffer
.size
= size
;
1133 SwrSetIndexBuffer(ctx
->swrContext
, &swrIndexBuffer
);
1136 struct swr_vertex_element_state
*velems
= ctx
->velems
;
1137 if (velems
&& velems
->fsState
.indexType
!= index_type
) {
1138 velems
->fsFunc
= NULL
;
1139 velems
->fsState
.indexType
= index_type
;
1144 if (ctx
->dirty
& (SWR_NEW_VS
|
1145 SWR_NEW_RASTERIZER
| // for clip planes
1147 SWR_NEW_SAMPLER_VIEW
|
1148 SWR_NEW_FRAMEBUFFER
)) {
1150 swr_generate_vs_key(key
, ctx
, ctx
->vs
);
1151 auto search
= ctx
->vs
->map
.find(key
);
1152 PFN_VERTEX_FUNC func
;
1153 if (search
!= ctx
->vs
->map
.end()) {
1154 func
= search
->second
->shader
;
1156 func
= swr_compile_vs(ctx
, key
);
1158 SwrSetVertexFunc(ctx
->swrContext
, func
);
1160 /* JIT sampler state */
1161 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1162 swr_update_sampler_state(ctx
,
1165 ctx
->swrDC
.samplersVS
);
1168 /* JIT sampler view state */
1169 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1170 swr_update_texture_state(ctx
,
1172 key
.nr_sampler_views
,
1173 ctx
->swrDC
.texturesVS
);
1177 /* FragmentShader */
1178 if (ctx
->dirty
& (SWR_NEW_FS
| SWR_NEW_SAMPLER
| SWR_NEW_SAMPLER_VIEW
1179 | SWR_NEW_RASTERIZER
| SWR_NEW_FRAMEBUFFER
)) {
1181 swr_generate_fs_key(key
, ctx
, ctx
->fs
);
1182 auto search
= ctx
->fs
->map
.find(key
);
1183 PFN_PIXEL_KERNEL func
;
1184 if (search
!= ctx
->fs
->map
.end()) {
1185 func
= search
->second
->shader
;
1187 func
= swr_compile_fs(ctx
, key
);
1189 SWR_PS_STATE psState
= {0};
1190 psState
.pfnPixelShader
= func
;
1191 psState
.killsPixel
= ctx
->fs
->info
.base
.uses_kill
;
1192 psState
.inputCoverage
= SWR_INPUT_COVERAGE_NORMAL
;
1193 psState
.writesODepth
= ctx
->fs
->info
.base
.writes_z
;
1194 psState
.usesSourceDepth
= ctx
->fs
->info
.base
.reads_z
;
1195 psState
.shadingRate
= SWR_SHADING_RATE_PIXEL
; // XXX
1196 psState
.numRenderTargets
= ctx
->framebuffer
.nr_cbufs
;
1197 psState
.posOffset
= SWR_PS_POSITION_SAMPLE_NONE
; // XXX msaa
1198 uint32_t barycentricsMask
= 0;
1200 // when we switch to mesa-master
1201 if (ctx
->fs
->info
.base
.uses_persp_center
||
1202 ctx
->fs
->info
.base
.uses_linear_center
)
1203 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1204 if (ctx
->fs
->info
.base
.uses_persp_centroid
||
1205 ctx
->fs
->info
.base
.uses_linear_centroid
)
1206 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1207 if (ctx
->fs
->info
.base
.uses_persp_sample
||
1208 ctx
->fs
->info
.base
.uses_linear_sample
)
1209 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1211 for (unsigned i
= 0; i
< ctx
->fs
->info
.base
.num_inputs
; i
++) {
1212 switch (ctx
->fs
->info
.base
.input_interpolate_loc
[i
]) {
1213 case TGSI_INTERPOLATE_LOC_CENTER
:
1214 barycentricsMask
|= SWR_BARYCENTRIC_PER_PIXEL_MASK
;
1216 case TGSI_INTERPOLATE_LOC_CENTROID
:
1217 barycentricsMask
|= SWR_BARYCENTRIC_CENTROID_MASK
;
1219 case TGSI_INTERPOLATE_LOC_SAMPLE
:
1220 barycentricsMask
|= SWR_BARYCENTRIC_PER_SAMPLE_MASK
;
1225 psState
.barycentricsMask
= barycentricsMask
;
1226 psState
.usesUAV
= false; // XXX
1227 psState
.forceEarlyZ
= false;
1228 SwrSetPixelShaderState(ctx
->swrContext
, &psState
);
1230 /* JIT sampler state */
1231 if (ctx
->dirty
& SWR_NEW_SAMPLER
) {
1232 swr_update_sampler_state(ctx
,
1233 PIPE_SHADER_FRAGMENT
,
1235 ctx
->swrDC
.samplersFS
);
1238 /* JIT sampler view state */
1239 if (ctx
->dirty
& (SWR_NEW_SAMPLER_VIEW
| SWR_NEW_FRAMEBUFFER
)) {
1240 swr_update_texture_state(ctx
,
1241 PIPE_SHADER_FRAGMENT
,
1242 key
.nr_sampler_views
,
1243 ctx
->swrDC
.texturesFS
);
1248 /* VertexShader Constants */
1249 if (ctx
->dirty
& SWR_NEW_VSCONSTANTS
) {
1250 swr_update_constants(ctx
, PIPE_SHADER_VERTEX
);
1253 /* FragmentShader Constants */
1254 if (ctx
->dirty
& SWR_NEW_FSCONSTANTS
) {
1255 swr_update_constants(ctx
, PIPE_SHADER_FRAGMENT
);
1258 /* Depth/stencil state */
1259 if (ctx
->dirty
& (SWR_NEW_DEPTH_STENCIL_ALPHA
| SWR_NEW_FRAMEBUFFER
)) {
1260 struct pipe_depth_state
*depth
= &(ctx
->depth_stencil
->depth
);
1261 struct pipe_stencil_state
*stencil
= ctx
->depth_stencil
->stencil
;
1262 SWR_DEPTH_STENCIL_STATE depthStencilState
= {{0}};
1263 SWR_DEPTH_BOUNDS_STATE depthBoundsState
= {0};
1265 /* XXX, incomplete. Need to flesh out stencil & alpha test state
1266 struct pipe_stencil_state *front_stencil =
1267 ctx->depth_stencil.stencil[0];
1268 struct pipe_stencil_state *back_stencil = ctx->depth_stencil.stencil[1];
1269 struct pipe_alpha_state alpha;
1271 if (stencil
[0].enabled
) {
1272 depthStencilState
.stencilWriteEnable
= 1;
1273 depthStencilState
.stencilTestEnable
= 1;
1274 depthStencilState
.stencilTestFunc
=
1275 swr_convert_depth_func(stencil
[0].func
);
1277 depthStencilState
.stencilPassDepthPassOp
=
1278 swr_convert_stencil_op(stencil
[0].zpass_op
);
1279 depthStencilState
.stencilPassDepthFailOp
=
1280 swr_convert_stencil_op(stencil
[0].zfail_op
);
1281 depthStencilState
.stencilFailOp
=
1282 swr_convert_stencil_op(stencil
[0].fail_op
);
1283 depthStencilState
.stencilWriteMask
= stencil
[0].writemask
;
1284 depthStencilState
.stencilTestMask
= stencil
[0].valuemask
;
1285 depthStencilState
.stencilRefValue
= ctx
->stencil_ref
.ref_value
[0];
1287 if (stencil
[1].enabled
) {
1288 depthStencilState
.doubleSidedStencilTestEnable
= 1;
1290 depthStencilState
.backfaceStencilTestFunc
=
1291 swr_convert_depth_func(stencil
[1].func
);
1293 depthStencilState
.backfaceStencilPassDepthPassOp
=
1294 swr_convert_stencil_op(stencil
[1].zpass_op
);
1295 depthStencilState
.backfaceStencilPassDepthFailOp
=
1296 swr_convert_stencil_op(stencil
[1].zfail_op
);
1297 depthStencilState
.backfaceStencilFailOp
=
1298 swr_convert_stencil_op(stencil
[1].fail_op
);
1299 depthStencilState
.backfaceStencilWriteMask
= stencil
[1].writemask
;
1300 depthStencilState
.backfaceStencilTestMask
= stencil
[1].valuemask
;
1302 depthStencilState
.backfaceStencilRefValue
=
1303 ctx
->stencil_ref
.ref_value
[1];
1306 depthStencilState
.depthTestEnable
= depth
->enabled
;
1307 depthStencilState
.depthTestFunc
= swr_convert_depth_func(depth
->func
);
1308 depthStencilState
.depthWriteEnable
= depth
->writemask
;
1309 SwrSetDepthStencilState(ctx
->swrContext
, &depthStencilState
);
1311 depthBoundsState
.depthBoundsTestEnable
= depth
->bounds_test
;
1312 depthBoundsState
.depthBoundsTestMinValue
= depth
->bounds_min
;
1313 depthBoundsState
.depthBoundsTestMaxValue
= depth
->bounds_max
;
1314 SwrSetDepthBoundsState(ctx
->swrContext
, &depthBoundsState
);
1318 if (ctx
->dirty
& (SWR_NEW_BLEND
|
1319 SWR_NEW_FRAMEBUFFER
|
1320 SWR_NEW_DEPTH_STENCIL_ALPHA
)) {
1321 struct pipe_framebuffer_state
*fb
= &ctx
->framebuffer
;
1323 SWR_BLEND_STATE blendState
;
1324 memcpy(&blendState
, &ctx
->blend
->blendState
, sizeof(blendState
));
1325 blendState
.constantColor
[0] = ctx
->blend_color
.color
[0];
1326 blendState
.constantColor
[1] = ctx
->blend_color
.color
[1];
1327 blendState
.constantColor
[2] = ctx
->blend_color
.color
[2];
1328 blendState
.constantColor
[3] = ctx
->blend_color
.color
[3];
1329 blendState
.alphaTestReference
=
1330 *((uint32_t*)&ctx
->depth_stencil
->alpha
.ref_value
);
1333 blendState
.sampleMask
= 0;
1334 blendState
.sampleCount
= SWR_MULTISAMPLE_1X
;
1336 /* If there are no color buffers bound, disable writes on RT0
1338 if (fb
->nr_cbufs
== 0) {
1339 blendState
.renderTarget
[0].writeDisableRed
= 1;
1340 blendState
.renderTarget
[0].writeDisableGreen
= 1;
1341 blendState
.renderTarget
[0].writeDisableBlue
= 1;
1342 blendState
.renderTarget
[0].writeDisableAlpha
= 1;
1343 SwrSetBlendFunc(ctx
->swrContext
, 0, NULL
);
1346 for (int target
= 0;
1347 target
< std::min(SWR_NUM_RENDERTARGETS
,
1348 PIPE_MAX_COLOR_BUFS
);
1350 if (!fb
->cbufs
[target
])
1353 struct swr_resource
*colorBuffer
=
1354 swr_resource(fb
->cbufs
[target
]->texture
);
1356 BLEND_COMPILE_STATE compileState
;
1357 memset(&compileState
, 0, sizeof(compileState
));
1358 compileState
.format
= colorBuffer
->swr
.format
;
1359 memcpy(&compileState
.blendState
,
1360 &ctx
->blend
->compileState
[target
],
1361 sizeof(compileState
.blendState
));
1363 const SWR_FORMAT_INFO
& info
= GetFormatInfo(compileState
.format
);
1364 if (compileState
.blendState
.logicOpEnable
&&
1365 ((info
.type
[0] == SWR_TYPE_FLOAT
) || info
.isSRGB
)) {
1366 compileState
.blendState
.logicOpEnable
= false;
1369 if (info
.type
[0] == SWR_TYPE_SINT
|| info
.type
[0] == SWR_TYPE_UINT
)
1370 compileState
.blendState
.blendEnable
= false;
1372 if (compileState
.blendState
.blendEnable
== false &&
1373 compileState
.blendState
.logicOpEnable
== false &&
1374 ctx
->depth_stencil
->alpha
.enabled
== 0) {
1375 SwrSetBlendFunc(ctx
->swrContext
, target
, NULL
);
1379 compileState
.desc
.alphaTestEnable
=
1380 ctx
->depth_stencil
->alpha
.enabled
;
1381 compileState
.desc
.independentAlphaBlendEnable
=
1382 (compileState
.blendState
.sourceBlendFactor
!=
1383 compileState
.blendState
.sourceAlphaBlendFactor
) ||
1384 (compileState
.blendState
.destBlendFactor
!=
1385 compileState
.blendState
.destAlphaBlendFactor
) ||
1386 (compileState
.blendState
.colorBlendFunc
!=
1387 compileState
.blendState
.alphaBlendFunc
);
1388 compileState
.desc
.alphaToCoverageEnable
=
1389 ctx
->blend
->pipe
.alpha_to_coverage
;
1390 compileState
.desc
.sampleMaskEnable
= 0; // XXX
1391 compileState
.desc
.numSamples
= 1; // XXX
1393 compileState
.alphaTestFunction
=
1394 swr_convert_depth_func(ctx
->depth_stencil
->alpha
.func
);
1395 compileState
.alphaTestFormat
= ALPHA_TEST_FLOAT32
; // xxx
1397 compileState
.Canonicalize();
1399 PFN_BLEND_JIT_FUNC func
= NULL
;
1400 auto search
= ctx
->blendJIT
->find(compileState
);
1401 if (search
!= ctx
->blendJIT
->end()) {
1402 func
= search
->second
;
1404 HANDLE hJitMgr
= screen
->hJitMgr
;
1405 func
= JitCompileBlend(hJitMgr
, compileState
);
1406 debug_printf("BLEND shader %p\n", func
);
1407 assert(func
&& "Error: BlendShader = NULL");
1409 ctx
->blendJIT
->insert(std::make_pair(compileState
, func
));
1411 SwrSetBlendFunc(ctx
->swrContext
, target
, func
);
1414 SwrSetBlendState(ctx
->swrContext
, &blendState
);
1417 if (ctx
->dirty
& SWR_NEW_STIPPLE
) {
1418 /* XXX What to do with this one??? SWR doesn't stipple */
1421 if (ctx
->dirty
& (SWR_NEW_VS
| SWR_NEW_SO
| SWR_NEW_RASTERIZER
)) {
1422 ctx
->vs
->soState
.rasterizerDisable
=
1423 ctx
->rasterizer
->rasterizer_discard
;
1424 SwrSetSoState(ctx
->swrContext
, &ctx
->vs
->soState
);
1426 pipe_stream_output_info
*stream_output
= &ctx
->vs
->pipe
.stream_output
;
1428 for (uint32_t i
= 0; i
< ctx
->num_so_targets
; i
++) {
1429 SWR_STREAMOUT_BUFFER buffer
= {0};
1430 if (!ctx
->so_targets
[i
])
1432 buffer
.enable
= true;
1434 (uint32_t *)swr_resource_data(ctx
->so_targets
[i
]->buffer
);
1435 buffer
.bufferSize
= ctx
->so_targets
[i
]->buffer_size
>> 2;
1436 buffer
.pitch
= stream_output
->stride
[i
];
1437 buffer
.streamOffset
= ctx
->so_targets
[i
]->buffer_offset
>> 2;
1439 SwrSetSoBuffers(ctx
->swrContext
, &buffer
, i
);
1443 if (ctx
->dirty
& SWR_NEW_CLIP
) {
1444 // shader exporting clip distances overrides all user clip planes
1445 if (ctx
->rasterizer
->clip_plane_enable
&&
1446 !ctx
->vs
->info
.base
.num_written_clipdistance
)
1448 swr_draw_context
*pDC
= &ctx
->swrDC
;
1449 memcpy(pDC
->userClipPlanes
,
1451 sizeof(pDC
->userClipPlanes
));
1455 // set up backend state
1456 SWR_BACKEND_STATE backendState
= {0};
1457 backendState
.numAttributes
=
1458 ctx
->vs
->info
.base
.num_outputs
- 1 +
1459 (ctx
->rasterizer
->sprite_coord_enable
? 1 : 0);
1460 for (unsigned i
= 0; i
< backendState
.numAttributes
; i
++)
1461 backendState
.numComponents
[i
] = 4;
1462 backendState
.constantInterpolationMask
=
1463 ctx
->rasterizer
->flatshade
?
1464 ctx
->fs
->flatConstantMask
:
1465 ctx
->fs
->constantMask
;
1466 backendState
.pointSpriteTexCoordMask
= ctx
->fs
->pointSpriteMask
;
1468 SwrSetBackendState(ctx
->swrContext
, &backendState
);
1470 /* Ensure that any in-progress attachment change StoreTiles finish */
1471 if (swr_is_fence_pending(screen
->flush_fence
))
1472 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
1474 /* Finally, update the in-use status of all resources involved in draw */
1475 swr_update_resource_status(pipe
, p_draw_info
);
1477 ctx
->dirty
= post_update_dirty_flags
;
1481 static struct pipe_stream_output_target
*
1482 swr_create_so_target(struct pipe_context
*pipe
,
1483 struct pipe_resource
*buffer
,
1484 unsigned buffer_offset
,
1485 unsigned buffer_size
)
1487 struct pipe_stream_output_target
*target
;
1489 target
= CALLOC_STRUCT(pipe_stream_output_target
);
1493 target
->context
= pipe
;
1494 target
->reference
.count
= 1;
1495 pipe_resource_reference(&target
->buffer
, buffer
);
1496 target
->buffer_offset
= buffer_offset
;
1497 target
->buffer_size
= buffer_size
;
1502 swr_destroy_so_target(struct pipe_context
*pipe
,
1503 struct pipe_stream_output_target
*target
)
1505 pipe_resource_reference(&target
->buffer
, NULL
);
1510 swr_set_so_targets(struct pipe_context
*pipe
,
1511 unsigned num_targets
,
1512 struct pipe_stream_output_target
**targets
,
1513 const unsigned *offsets
)
1515 struct swr_context
*swr
= swr_context(pipe
);
1518 assert(num_targets
< MAX_SO_STREAMS
);
1520 for (i
= 0; i
< num_targets
; i
++) {
1521 pipe_so_target_reference(
1522 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
],
1526 for (/* fall-through */; i
< swr
->num_so_targets
; i
++) {
1527 pipe_so_target_reference(
1528 (struct pipe_stream_output_target
**)&swr
->so_targets
[i
], NULL
);
1531 swr
->num_so_targets
= num_targets
;
1533 swr
->dirty
= SWR_NEW_SO
;
1538 swr_state_init(struct pipe_context
*pipe
)
1540 pipe
->create_blend_state
= swr_create_blend_state
;
1541 pipe
->bind_blend_state
= swr_bind_blend_state
;
1542 pipe
->delete_blend_state
= swr_delete_blend_state
;
1544 pipe
->create_depth_stencil_alpha_state
= swr_create_depth_stencil_state
;
1545 pipe
->bind_depth_stencil_alpha_state
= swr_bind_depth_stencil_state
;
1546 pipe
->delete_depth_stencil_alpha_state
= swr_delete_depth_stencil_state
;
1548 pipe
->create_rasterizer_state
= swr_create_rasterizer_state
;
1549 pipe
->bind_rasterizer_state
= swr_bind_rasterizer_state
;
1550 pipe
->delete_rasterizer_state
= swr_delete_rasterizer_state
;
1552 pipe
->create_sampler_state
= swr_create_sampler_state
;
1553 pipe
->bind_sampler_states
= swr_bind_sampler_states
;
1554 pipe
->delete_sampler_state
= swr_delete_sampler_state
;
1556 pipe
->create_sampler_view
= swr_create_sampler_view
;
1557 pipe
->set_sampler_views
= swr_set_sampler_views
;
1558 pipe
->sampler_view_destroy
= swr_sampler_view_destroy
;
1560 pipe
->create_vs_state
= swr_create_vs_state
;
1561 pipe
->bind_vs_state
= swr_bind_vs_state
;
1562 pipe
->delete_vs_state
= swr_delete_vs_state
;
1564 pipe
->create_fs_state
= swr_create_fs_state
;
1565 pipe
->bind_fs_state
= swr_bind_fs_state
;
1566 pipe
->delete_fs_state
= swr_delete_fs_state
;
1568 pipe
->set_constant_buffer
= swr_set_constant_buffer
;
1570 pipe
->create_vertex_elements_state
= swr_create_vertex_elements_state
;
1571 pipe
->bind_vertex_elements_state
= swr_bind_vertex_elements_state
;
1572 pipe
->delete_vertex_elements_state
= swr_delete_vertex_elements_state
;
1574 pipe
->set_vertex_buffers
= swr_set_vertex_buffers
;
1575 pipe
->set_index_buffer
= swr_set_index_buffer
;
1577 pipe
->set_polygon_stipple
= swr_set_polygon_stipple
;
1578 pipe
->set_clip_state
= swr_set_clip_state
;
1579 pipe
->set_scissor_states
= swr_set_scissor_states
;
1580 pipe
->set_viewport_states
= swr_set_viewport_states
;
1582 pipe
->set_framebuffer_state
= swr_set_framebuffer_state
;
1584 pipe
->set_blend_color
= swr_set_blend_color
;
1585 pipe
->set_stencil_ref
= swr_set_stencil_ref
;
1587 pipe
->set_sample_mask
= swr_set_sample_mask
;
1589 pipe
->create_stream_output_target
= swr_create_so_target
;
1590 pipe
->stream_output_target_destroy
= swr_destroy_so_target
;
1591 pipe
->set_stream_output_targets
= swr_set_so_targets
;