2 * Copyright (c) 2014 Scott Mansell
3 * Copyright © 2014 Broadcom
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
27 #include "util/u_format.h"
28 #include "util/u_pack_color.h"
29 #include "indices/u_primconvert.h"
31 #include "vc4_context.h"
32 #include "vc4_resource.h"
35 * Does the initial bining command list setup for drawing to a given FBO.
38 vc4_start_draw(struct vc4_context
*vc4
)
43 uint32_t width
= vc4
->framebuffer
.width
;
44 uint32_t height
= vc4
->framebuffer
.height
;
45 uint32_t tilew
= align(width
, 64) / 64;
46 uint32_t tileh
= align(height
, 64) / 64;
48 uint32_t tile_alloc_size
= 32 * tilew
* tileh
* 16;
49 uint32_t tile_state_size
= 48 * tilew
* tileh
;
50 if (!vc4
->tile_alloc
|| vc4
->tile_alloc
->size
< tile_alloc_size
) {
51 vc4_bo_unreference(&vc4
->tile_alloc
);
52 vc4
->tile_alloc
= vc4_bo_alloc(vc4
->screen
, tile_alloc_size
,
55 if (!vc4
->tile_state
|| vc4
->tile_state
->size
< tile_state_size
) {
56 vc4_bo_unreference(&vc4
->tile_state
);
57 vc4
->tile_state
= vc4_bo_alloc(vc4
->screen
, tile_state_size
,
61 // Tile state data is 48 bytes per tile, I think it can be thrown away
62 // as soon as binning is finished.
63 cl_start_reloc(&vc4
->bcl
, 2);
64 cl_u8(&vc4
->bcl
, VC4_PACKET_TILE_BINNING_MODE_CONFIG
);
65 cl_reloc(vc4
, &vc4
->bcl
, vc4
->tile_alloc
, 0);
66 cl_u32(&vc4
->bcl
, vc4
->tile_alloc
->size
);
67 cl_reloc(vc4
, &vc4
->bcl
, vc4
->tile_state
, 0);
68 cl_u8(&vc4
->bcl
, tilew
);
69 cl_u8(&vc4
->bcl
, tileh
);
70 cl_u8(&vc4
->bcl
, VC4_BIN_CONFIG_AUTO_INIT_TSDA
);
72 cl_u8(&vc4
->bcl
, VC4_PACKET_START_TILE_BINNING
);
74 cl_u8(&vc4
->bcl
, VC4_PACKET_PRIMITIVE_LIST_FORMAT
);
75 cl_u8(&vc4
->bcl
, 0x12); // 16 bit triangle
77 vc4
->needs_flush
= true;
78 vc4
->draw_call_queued
= true;
82 vc4_draw_vbo(struct pipe_context
*pctx
, const struct pipe_draw_info
*info
)
84 struct vc4_context
*vc4
= vc4_context(pctx
);
86 if (info
->mode
>= PIPE_PRIM_QUADS
) {
87 util_primconvert_save_index_buffer(vc4
->primconvert
, &vc4
->indexbuf
);
88 util_primconvert_save_rasterizer_state(vc4
->primconvert
, &vc4
->rasterizer
->base
);
89 util_primconvert_draw_vbo(vc4
->primconvert
, info
);
94 vc4_update_compiled_shaders(vc4
);
98 /* the actual draw call. */
99 struct vc4_vertex_stateobj
*vtx
= vc4
->vtx
;
100 struct vc4_vertexbuf_stateobj
*vertexbuf
= &vc4
->vertexbuf
;
101 cl_u8(&vc4
->bcl
, VC4_PACKET_GL_SHADER_STATE
);
102 assert(vtx
->num_elements
<= 8);
103 /* Note that number of attributes == 0 in the packet means 8
104 * attributes. This field also contains the offset into shader_rec.
106 cl_u32(&vc4
->bcl
, vtx
->num_elements
& 0x7);
108 /* Note that the primitive type fields match with OpenGL/gallium
109 * definitions, up to but not including QUADS.
112 struct vc4_resource
*rsc
= vc4_resource(vc4
->indexbuf
.buffer
);
114 assert(vc4
->indexbuf
.index_size
== 1 ||
115 vc4
->indexbuf
.index_size
== 2);
117 cl_start_reloc(&vc4
->bcl
, 1);
118 cl_u8(&vc4
->bcl
, VC4_PACKET_GL_INDEXED_PRIMITIVE
);
121 (vc4
->indexbuf
.index_size
== 2 ?
122 VC4_INDEX_BUFFER_U16
:
123 VC4_INDEX_BUFFER_U8
));
124 cl_u32(&vc4
->bcl
, info
->count
);
125 cl_reloc(vc4
, &vc4
->bcl
, rsc
->bo
, vc4
->indexbuf
.offset
);
126 cl_u32(&vc4
->bcl
, info
->max_index
);
128 cl_u8(&vc4
->bcl
, VC4_PACKET_GL_ARRAY_PRIMITIVE
);
129 cl_u8(&vc4
->bcl
, info
->mode
);
130 cl_u32(&vc4
->bcl
, info
->count
);
131 cl_u32(&vc4
->bcl
, info
->start
);
136 vc4_write_uniforms(vc4
, vc4
->prog
.fs
,
137 &vc4
->constbuf
[PIPE_SHADER_FRAGMENT
],
140 vc4_write_uniforms(vc4
, vc4
->prog
.vs
,
141 &vc4
->constbuf
[PIPE_SHADER_VERTEX
],
144 vc4_write_uniforms(vc4
, vc4
->prog
.vs
,
145 &vc4
->constbuf
[PIPE_SHADER_VERTEX
],
149 cl_start_shader_reloc(&vc4
->shader_rec
, 3 + vtx
->num_elements
);
150 cl_u16(&vc4
->shader_rec
, VC4_SHADER_FLAG_ENABLE_CLIPPING
);
151 cl_u8(&vc4
->shader_rec
, 0); /* fs num uniforms (unused) */
152 cl_u8(&vc4
->shader_rec
, vc4
->prog
.fs
->num_inputs
);
153 cl_reloc(vc4
, &vc4
->shader_rec
, vc4
->prog
.fs
->bo
, 0);
154 cl_u32(&vc4
->shader_rec
, 0); /* UBO offset written by kernel */
156 cl_u16(&vc4
->shader_rec
, 0); /* vs num uniforms */
157 cl_u8(&vc4
->shader_rec
, (1 << vtx
->num_elements
) - 1); /* vs attribute array bitfield */
158 cl_u8(&vc4
->shader_rec
, 16 * vtx
->num_elements
); /* vs total attribute size */
159 cl_reloc(vc4
, &vc4
->shader_rec
, vc4
->prog
.vs
->bo
, 0);
160 cl_u32(&vc4
->shader_rec
, 0); /* UBO offset written by kernel */
162 cl_u16(&vc4
->shader_rec
, 0); /* cs num uniforms */
163 cl_u8(&vc4
->shader_rec
, (1 << vtx
->num_elements
) - 1); /* cs attribute array bitfield */
164 cl_u8(&vc4
->shader_rec
, 16 * vtx
->num_elements
); /* vs total attribute size */
165 cl_reloc(vc4
, &vc4
->shader_rec
, vc4
->prog
.vs
->bo
,
166 vc4
->prog
.vs
->coord_shader_offset
);
167 cl_u32(&vc4
->shader_rec
, 0); /* UBO offset written by kernel */
169 for (int i
= 0; i
< vtx
->num_elements
; i
++) {
170 struct pipe_vertex_element
*elem
= &vtx
->pipe
[i
];
171 struct pipe_vertex_buffer
*vb
=
172 &vertexbuf
->vb
[elem
->vertex_buffer_index
];
173 struct vc4_resource
*rsc
= vc4_resource(vb
->buffer
);
175 cl_reloc(vc4
, &vc4
->shader_rec
, rsc
->bo
,
176 vb
->buffer_offset
+ elem
->src_offset
);
177 cl_u8(&vc4
->shader_rec
,
178 util_format_get_blocksize(elem
->src_format
) - 1);
179 cl_u8(&vc4
->shader_rec
, vb
->stride
);
180 cl_u8(&vc4
->shader_rec
, i
* 16); /* VS VPM offset */
181 cl_u8(&vc4
->shader_rec
, i
* 16); /* CS VPM offset */
184 if (vc4
->zsa
&& vc4
->zsa
->depth
.enabled
) {
185 vc4
->resolve
|= PIPE_CLEAR_DEPTH
;
187 vc4
->resolve
|= PIPE_CLEAR_COLOR0
;
189 vc4
->shader_rec_count
++;
195 pack_rgba(enum pipe_format format
, const float *rgba
)
198 util_pack_color(rgba
, format
, &uc
);
203 vc4_clear(struct pipe_context
*pctx
, unsigned buffers
,
204 const union pipe_color_union
*color
, double depth
, unsigned stencil
)
206 struct vc4_context
*vc4
= vc4_context(pctx
);
208 /* We can't flag new buffers for clearing once we've queued draws. We
209 * could avoid this by using the 3d engine to clear.
211 if (vc4
->draw_call_queued
)
214 if (buffers
& PIPE_CLEAR_COLOR0
) {
215 vc4
->clear_color
[0] = vc4
->clear_color
[1] =
216 pack_rgba(vc4
->framebuffer
.cbufs
[0]->format
,
220 vc4
->cleared
|= buffers
;
221 vc4
->resolve
|= buffers
;
227 vc4_clear_render_target(struct pipe_context
*pctx
, struct pipe_surface
*ps
,
228 const union pipe_color_union
*color
,
229 unsigned x
, unsigned y
, unsigned w
, unsigned h
)
231 fprintf(stderr
, "unimpl: clear RT\n");
235 vc4_clear_depth_stencil(struct pipe_context
*pctx
, struct pipe_surface
*ps
,
236 unsigned buffers
, double depth
, unsigned stencil
,
237 unsigned x
, unsigned y
, unsigned w
, unsigned h
)
239 fprintf(stderr
, "unimpl: clear DS\n");
243 vc4_draw_init(struct pipe_context
*pctx
)
245 pctx
->draw_vbo
= vc4_draw_vbo
;
246 pctx
->clear
= vc4_clear
;
247 pctx
->clear_render_target
= vc4_clear_render_target
;
248 pctx
->clear_depth_stencil
= vc4_clear_depth_stencil
;