1 /*
2 * Copyright © 2015 Broadcom
3 *
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:
10 *
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
13 * Software.
14 *
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
21 * IN THE SOFTWARE.
22 */
23
24 #include "util/u_format.h"
25 #include "util/u_surface.h"
26 #include "util/u_blitter.h"
27 #include "vc4_context.h"
28
29 static struct pipe_surface *
vc4_get_blit_surface(struct pipe_context * pctx,struct pipe_resource * prsc,unsigned level)30 vc4_get_blit_surface(struct pipe_context *pctx,
31 struct pipe_resource *prsc, unsigned level)
32 {
33 struct pipe_surface tmpl;
34
35 memset(&tmpl, 0, sizeof(tmpl));
36 tmpl.format = prsc->format;
37 tmpl.u.tex.level = level;
38 tmpl.u.tex.first_layer = 0;
39 tmpl.u.tex.last_layer = 0;
40
41 return pctx->create_surface(pctx, prsc, &tmpl);
42 }
43
44 static bool
is_tile_unaligned(unsigned size,unsigned tile_size)45 is_tile_unaligned(unsigned size, unsigned tile_size)
46 {
47 return size & (tile_size - 1);
48 }
49
50 static bool
vc4_tile_blit(struct pipe_context * pctx,const struct pipe_blit_info * info)51 vc4_tile_blit(struct pipe_context *pctx, const struct pipe_blit_info *info)
52 {
53 struct vc4_context *vc4 = vc4_context(pctx);
54 bool msaa = (info->src.resource->nr_samples > 1 ||
55 info->dst.resource->nr_samples > 1);
56 int tile_width = msaa ? 32 : 64;
57 int tile_height = msaa ? 32 : 64;
58
59 if (util_format_is_depth_or_stencil(info->dst.resource->format))
60 return false;
61
62 if (info->scissor_enable)
63 return false;
64
65 if ((info->mask & PIPE_MASK_RGBA) == 0)
66 return false;
67
68 if (info->dst.box.x != info->src.box.x ||
69 info->dst.box.y != info->src.box.y ||
70 info->dst.box.width != info->src.box.width ||
71 info->dst.box.height != info->src.box.height) {
72 return false;
73 }
74
75 int dst_surface_width = u_minify(info->dst.resource->width0,
76 info->dst.level);
77 int dst_surface_height = u_minify(info->dst.resource->height0,
78 info->dst.level);
79 if (is_tile_unaligned(info->dst.box.x, tile_width) ||
80 is_tile_unaligned(info->dst.box.y, tile_height) ||
81 (is_tile_unaligned(info->dst.box.width, tile_width) &&
82 info->dst.box.x + info->dst.box.width != dst_surface_width) ||
83 (is_tile_unaligned(info->dst.box.height, tile_height) &&
84 info->dst.box.y + info->dst.box.height != dst_surface_height)) {
85 return false;
86 }
87
88 /* VC4_PACKET_LOAD_TILE_BUFFER_GENERAL uses the
89 * VC4_PACKET_TILE_RENDERING_MODE_CONFIG's width (determined by our
90 * destination surface) to determine the stride. This may be wrong
91 * when reading from texture miplevels > 0, which are stored in
92 * POT-sized areas. For MSAA, the tile addresses are computed
93 * explicitly by the RCL, but still use the destination width to
94 * determine the stride (which could be fixed by explicitly supplying
95 * it in the ABI).
96 */
97 struct vc4_resource *rsc = vc4_resource(info->src.resource);
98
99 uint32_t stride;
100
101 if (info->src.resource->nr_samples > 1)
102 stride = align(dst_surface_width, 32) * 4 * rsc->cpp;
103 else if (rsc->slices[info->src.level].tiling == VC4_TILING_FORMAT_T)
104 stride = align(dst_surface_width * rsc->cpp, 128);
105 else
106 stride = align(dst_surface_width * rsc->cpp, 16);
107
108 if (stride != rsc->slices[info->src.level].stride)
109 return false;
110
111 if (info->dst.resource->format != info->src.resource->format)
112 return false;
113
114 if (false) {
115 fprintf(stderr, "RCL blit from %d,%d to %d,%d (%d,%d)\n",
116 info->src.box.x,
117 info->src.box.y,
118 info->dst.box.x,
119 info->dst.box.y,
120 info->dst.box.width,
121 info->dst.box.height);
122 }
123
124 struct pipe_surface *dst_surf =
125 vc4_get_blit_surface(pctx, info->dst.resource, info->dst.level);
126 struct pipe_surface *src_surf =
127 vc4_get_blit_surface(pctx, info->src.resource, info->src.level);
128
129 vc4_flush_jobs_reading_resource(vc4, info->src.resource);
130
131 struct vc4_job *job = vc4_get_job(vc4, dst_surf, NULL);
132 pipe_surface_reference(&job->color_read, src_surf);
133
134 /* If we're resolving from MSAA to single sample, we still need to run
135 * the engine in MSAA mode for the load.
136 */
137 if (!job->msaa && info->src.resource->nr_samples > 1) {
138 job->msaa = true;
139 job->tile_width = 32;
140 job->tile_height = 32;
141 }
142
143 job->draw_min_x = info->dst.box.x;
144 job->draw_min_y = info->dst.box.y;
145 job->draw_max_x = info->dst.box.x + info->dst.box.width;
146 job->draw_max_y = info->dst.box.y + info->dst.box.height;
147 job->draw_width = dst_surf->width;
148 job->draw_height = dst_surf->height;
149
150 job->tile_width = tile_width;
151 job->tile_height = tile_height;
152 job->msaa = msaa;
153 job->needs_flush = true;
154 job->resolve |= PIPE_CLEAR_COLOR;
155
156 vc4_job_submit(vc4, job);
157
158 pipe_surface_reference(&dst_surf, NULL);
159 pipe_surface_reference(&src_surf, NULL);
160
161 return true;
162 }
163
164 void
vc4_blitter_save(struct vc4_context * vc4)165 vc4_blitter_save(struct vc4_context *vc4)
166 {
167 util_blitter_save_vertex_buffer_slot(vc4->blitter, vc4->vertexbuf.vb);
168 util_blitter_save_vertex_elements(vc4->blitter, vc4->vtx);
169 util_blitter_save_vertex_shader(vc4->blitter, vc4->prog.bind_vs);
170 util_blitter_save_rasterizer(vc4->blitter, vc4->rasterizer);
171 util_blitter_save_viewport(vc4->blitter, &vc4->viewport);
172 util_blitter_save_scissor(vc4->blitter, &vc4->scissor);
173 util_blitter_save_fragment_shader(vc4->blitter, vc4->prog.bind_fs);
174 util_blitter_save_blend(vc4->blitter, vc4->blend);
175 util_blitter_save_depth_stencil_alpha(vc4->blitter, vc4->zsa);
176 util_blitter_save_stencil_ref(vc4->blitter, &vc4->stencil_ref);
177 util_blitter_save_sample_mask(vc4->blitter, vc4->sample_mask);
178 util_blitter_save_framebuffer(vc4->blitter, &vc4->framebuffer);
179 util_blitter_save_fragment_sampler_states(vc4->blitter,
180 vc4->fragtex.num_samplers,
181 (void **)vc4->fragtex.samplers);
182 util_blitter_save_fragment_sampler_views(vc4->blitter,
183 vc4->fragtex.num_textures, vc4->fragtex.textures);
184 }
185
186 static bool
vc4_render_blit(struct pipe_context * ctx,struct pipe_blit_info * info)187 vc4_render_blit(struct pipe_context *ctx, struct pipe_blit_info *info)
188 {
189 struct vc4_context *vc4 = vc4_context(ctx);
190
191 if (!util_blitter_is_blit_supported(vc4->blitter, info)) {
192 fprintf(stderr, "blit unsupported %s -> %s\n",
193 util_format_short_name(info->src.resource->format),
194 util_format_short_name(info->dst.resource->format));
195 return false;
196 }
197
198 vc4_blitter_save(vc4);
199 util_blitter_blit(vc4->blitter, info);
200
201 return true;
202 }
203
204 /* Optimal hardware path for blitting pixels.
205 * Scaling, format conversion, up- and downsampling (resolve) are allowed.
206 */
207 void
vc4_blit(struct pipe_context * pctx,const struct pipe_blit_info * blit_info)208 vc4_blit(struct pipe_context *pctx, const struct pipe_blit_info *blit_info)
209 {
210 struct pipe_blit_info info = *blit_info;
211
212 if (vc4_tile_blit(pctx, blit_info))
213 return;
214
215 if (util_try_blit_via_copy_region(pctx, &info)) {
216 return; /* done */
217 }
218
219 if (info.mask & PIPE_MASK_S) {
220 fprintf(stderr, "cannot blit stencil, skipping\n");
221 info.mask &= ~PIPE_MASK_S;
222 }
223
224 vc4_render_blit(pctx, &info);
225 }
226