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 #include "swr_screen.h"
25 #include "swr_context.h"
26 #include "swr_resource.h"
27 #include "swr_fence.h"
28 #include "swr_query.h"
31 #include "util/u_draw.h"
32 #include "util/u_prim.h"
35 * Draw vertex arrays, with optional indexing, optional instancing.
38 swr_draw_vbo(struct pipe_context
*pipe
, const struct pipe_draw_info
*info
)
40 struct swr_context
*ctx
= swr_context(pipe
);
42 if (!info
->count_from_stream_output
&& !info
->indirect
&&
43 !info
->primitive_restart
&&
44 !u_trim_pipe_prim(info
->mode
, (unsigned*)&info
->count
))
47 if (!swr_check_render_cond(pipe
))
51 util_draw_indirect(pipe
, info
);
55 /* Update derived state, pass draw info to update function */
56 swr_update_derived(pipe
, info
);
58 swr_update_draw_context(ctx
);
60 if (ctx
->vs
->pipe
.stream_output
.num_outputs
) {
61 if (!ctx
->vs
->soFunc
[info
->mode
]) {
62 STREAMOUT_COMPILE_STATE state
= {0};
63 struct pipe_stream_output_info
*so
= &ctx
->vs
->pipe
.stream_output
;
65 state
.numVertsPerPrim
= u_vertices_per_prim(info
->mode
);
67 uint32_t offsets
[MAX_SO_STREAMS
] = {0};
70 for (uint32_t i
= 0; i
< so
->num_outputs
; i
++) {
71 assert(so
->output
[i
].stream
== 0); // @todo
72 uint32_t output_buffer
= so
->output
[i
].output_buffer
;
73 if (so
->output
[i
].dst_offset
!= offsets
[output_buffer
]) {
74 // hole - need to fill
75 state
.stream
.decl
[num
].bufferIndex
= output_buffer
;
76 state
.stream
.decl
[num
].hole
= true;
77 state
.stream
.decl
[num
].componentMask
=
78 (1 << (so
->output
[i
].dst_offset
- offsets
[output_buffer
]))
81 offsets
[output_buffer
] = so
->output
[i
].dst_offset
;
84 unsigned attrib_slot
= so
->output
[i
].register_index
;
85 attrib_slot
= swr_so_adjust_attrib(attrib_slot
, ctx
->vs
);
87 state
.stream
.decl
[num
].bufferIndex
= output_buffer
;
88 state
.stream
.decl
[num
].attribSlot
= attrib_slot
;
89 state
.stream
.decl
[num
].componentMask
=
90 ((1 << so
->output
[i
].num_components
) - 1)
91 << so
->output
[i
].start_component
;
92 state
.stream
.decl
[num
].hole
= false;
95 offsets
[output_buffer
] += so
->output
[i
].num_components
;
98 state
.stream
.numDecls
= num
;
100 HANDLE hJitMgr
= swr_screen(pipe
->screen
)->hJitMgr
;
101 ctx
->vs
->soFunc
[info
->mode
] = JitCompileStreamout(hJitMgr
, state
);
102 debug_printf("so shader %p\n", ctx
->vs
->soFunc
[info
->mode
]);
103 assert(ctx
->vs
->soFunc
[info
->mode
] && "Error: SoShader = NULL");
106 ctx
->api
.pfnSwrSetSoFunc(ctx
->swrContext
, ctx
->vs
->soFunc
[info
->mode
], 0);
109 struct swr_vertex_element_state
*velems
= ctx
->velems
;
110 if (info
->primitive_restart
)
111 velems
->fsState
.cutIndex
= info
->restart_index
;
113 velems
->fsState
.cutIndex
= 0;
114 velems
->fsState
.bEnableCutIndex
= info
->primitive_restart
;
115 velems
->fsState
.bPartialVertexBuffer
= (info
->min_index
> 0);
117 swr_jit_fetch_key key
;
118 swr_generate_fetch_key(key
, velems
);
119 auto search
= velems
->map
.find(key
);
120 if (search
!= velems
->map
.end()) {
121 velems
->fsFunc
= search
->second
;
123 HANDLE hJitMgr
= swr_screen(ctx
->pipe
.screen
)->hJitMgr
;
124 velems
->fsFunc
= JitCompileFetch(hJitMgr
, velems
->fsState
);
126 debug_printf("fetch shader %p\n", velems
->fsFunc
);
127 assert(velems
->fsFunc
&& "Error: FetchShader = NULL");
129 velems
->map
.insert(std::make_pair(key
, velems
->fsFunc
));
132 ctx
->api
.pfnSwrSetFetchFunc(ctx
->swrContext
, velems
->fsFunc
);
134 /* Set up frontend state
135 * XXX setup provokingVertex & topologyProvokingVertex */
136 SWR_FRONTEND_STATE feState
= {0};
138 // feState.vsVertexSize seeds the PA size that is used as an interface
139 // between all the shader stages, so it has to be large enough to
140 // incorporate all interfaces between stages
142 // max of gs and vs num_outputs
143 feState
.vsVertexSize
= ctx
->vs
->info
.base
.num_outputs
;
145 ctx
->gs
->info
.base
.num_outputs
> feState
.vsVertexSize
) {
146 feState
.vsVertexSize
= ctx
->gs
->info
.base
.num_outputs
;
149 if (ctx
->vs
->info
.base
.num_outputs
) {
150 // gs does not adjust for position in SGV slot at input from vs
152 feState
.vsVertexSize
--;
155 // other (non-SGV) slots start at VERTEX_ATTRIB_START_SLOT
156 feState
.vsVertexSize
+= VERTEX_ATTRIB_START_SLOT
;
158 // The PA in the clipper does not handle BE vertex sizes
159 // different from FE. Increase vertexsize only for the cases that needed it
161 // primid needs a slot
162 if (ctx
->fs
->info
.base
.uses_primid
)
163 feState
.vsVertexSize
++;
164 // sprite coord enable
165 if (ctx
->rasterizer
->sprite_coord_enable
)
166 feState
.vsVertexSize
++;
169 if (ctx
->rasterizer
->flatshade_first
) {
170 feState
.provokingVertex
= {1, 0, 0};
172 feState
.provokingVertex
= {2, 1, 2};
175 enum pipe_prim_type topology
;
177 topology
= (pipe_prim_type
)ctx
->gs
->info
.base
.properties
[TGSI_PROPERTY_GS_OUTPUT_PRIM
];
179 topology
= info
->mode
;
182 case PIPE_PRIM_TRIANGLE_FAN
:
183 feState
.topologyProvokingVertex
= feState
.provokingVertex
.triFan
;
185 case PIPE_PRIM_TRIANGLE_STRIP
:
186 case PIPE_PRIM_TRIANGLES
:
187 feState
.topologyProvokingVertex
= feState
.provokingVertex
.triStripList
;
189 case PIPE_PRIM_QUAD_STRIP
:
190 case PIPE_PRIM_QUADS
:
191 if (ctx
->rasterizer
->flatshade_first
)
192 feState
.topologyProvokingVertex
= 0;
194 feState
.topologyProvokingVertex
= 3;
196 case PIPE_PRIM_LINES
:
197 case PIPE_PRIM_LINE_LOOP
:
198 case PIPE_PRIM_LINE_STRIP
:
199 feState
.topologyProvokingVertex
= feState
.provokingVertex
.lineStripList
;
202 feState
.topologyProvokingVertex
= 0;
205 feState
.bEnableCutIndex
= info
->primitive_restart
;
206 ctx
->api
.pfnSwrSetFrontendState(ctx
->swrContext
, &feState
);
208 if (info
->index_size
)
209 ctx
->api
.pfnSwrDrawIndexedInstanced(ctx
->swrContext
,
210 swr_convert_prim_topology(info
->mode
),
212 info
->instance_count
,
215 info
->start_instance
);
217 ctx
->api
.pfnSwrDrawInstanced(ctx
->swrContext
,
218 swr_convert_prim_topology(info
->mode
),
220 info
->instance_count
,
222 info
->start_instance
);
224 /* On large client-buffer draw, we used client buffer directly, without
225 * copy. Block until draw is finished.
226 * VMD is an example application that benefits from this. */
227 if (ctx
->dirty
& SWR_LARGE_CLIENT_DRAW
) {
228 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
229 swr_fence_submit(ctx
, screen
->flush_fence
);
230 swr_fence_finish(pipe
->screen
, NULL
, screen
->flush_fence
, 0);
236 swr_flush(struct pipe_context
*pipe
,
237 struct pipe_fence_handle
**fence
,
240 struct swr_context
*ctx
= swr_context(pipe
);
241 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
242 struct pipe_surface
*cb
= ctx
->framebuffer
.cbufs
[0];
244 /* If the current renderTarget is the display surface, store tiles back to
245 * the surface, in preparation for present (swr_flush_frontbuffer).
246 * Other renderTargets get stored back when attachment changes or
247 * swr_surface_destroy */
248 if (cb
&& swr_resource(cb
->texture
)->display_target
)
249 swr_store_dirty_resource(pipe
, cb
->texture
, SWR_TILE_RESOLVED
);
252 swr_fence_reference(pipe
->screen
, fence
, screen
->flush_fence
);
256 swr_finish(struct pipe_context
*pipe
)
258 struct pipe_fence_handle
*fence
= nullptr;
260 swr_flush(pipe
, &fence
, 0);
261 swr_fence_finish(pipe
->screen
, NULL
, fence
, 0);
262 swr_fence_reference(pipe
->screen
, &fence
, NULL
);
266 * Invalidate tiles so they can be reloaded back when needed
269 swr_invalidate_render_target(struct pipe_context
*pipe
,
271 uint16_t width
, uint16_t height
)
273 struct swr_context
*ctx
= swr_context(pipe
);
275 /* grab the rect from the passed in arguments */
276 swr_update_draw_context(ctx
);
278 {0, 0, (int32_t)width
, (int32_t)height
};
279 ctx
->api
.pfnSwrInvalidateTiles(ctx
->swrContext
,
286 * Store SWR HotTiles back to renderTarget surface.
289 swr_store_render_target(struct pipe_context
*pipe
,
291 enum SWR_TILE_STATE post_tile_state
)
293 struct swr_context
*ctx
= swr_context(pipe
);
294 struct swr_draw_context
*pDC
= &ctx
->swrDC
;
295 struct SWR_SURFACE_STATE
*renderTarget
= &pDC
->renderTargets
[attachment
];
297 /* Only proceed if there's a valid surface to store to */
298 if (renderTarget
->xpBaseAddress
) {
299 swr_update_draw_context(ctx
);
302 (int32_t)u_minify(renderTarget
->width
, renderTarget
->lod
),
303 (int32_t)u_minify(renderTarget
->height
, renderTarget
->lod
)};
304 ctx
->api
.pfnSwrStoreTiles(ctx
->swrContext
,
312 swr_store_dirty_resource(struct pipe_context
*pipe
,
313 struct pipe_resource
*resource
,
314 enum SWR_TILE_STATE post_tile_state
)
316 /* Only store resource if it has been written to */
317 if (swr_resource(resource
)->status
& SWR_RESOURCE_WRITE
) {
318 struct swr_context
*ctx
= swr_context(pipe
);
319 struct swr_screen
*screen
= swr_screen(pipe
->screen
);
320 struct swr_resource
*spr
= swr_resource(resource
);
322 swr_draw_context
*pDC
= &ctx
->swrDC
;
323 SWR_SURFACE_STATE
*renderTargets
= pDC
->renderTargets
;
324 for (uint32_t i
= 0; i
< SWR_NUM_ATTACHMENTS
; i
++)
325 if (renderTargets
[i
].xpBaseAddress
== spr
->swr
.xpBaseAddress
||
326 (spr
->secondary
.xpBaseAddress
&&
327 renderTargets
[i
].xpBaseAddress
== spr
->secondary
.xpBaseAddress
)) {
328 swr_store_render_target(pipe
, i
, post_tile_state
);
330 /* Mesa thinks depth/stencil are fused, so we'll never get an
331 * explicit resource for stencil. So, if checking depth, then
332 * also check for stencil. */
333 if (spr
->has_stencil
&& (i
== SWR_ATTACHMENT_DEPTH
)) {
334 swr_store_render_target(
335 pipe
, SWR_ATTACHMENT_STENCIL
, post_tile_state
);
338 /* This fence signals StoreTiles completion */
339 swr_fence_submit(ctx
, screen
->flush_fence
);
347 swr_draw_init(struct pipe_context
*pipe
)
349 pipe
->draw_vbo
= swr_draw_vbo
;
350 pipe
->flush
= swr_flush
;