src/gallium/auxiliary/vl/vl_compositor_cs.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2019 Advanced Micro Devices, Inc.
   4  * All Rights Reserved.
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the
   8  * "Software"), to deal in the Software without restriction, including
   9  * without limitation the rights to use, copy, modify, merge, publish,
  10  * distribute, sub license, and/or sell copies of the Software, and to
  11  * permit persons to whom the Software is furnished to do so, subject to
  12  * the following conditions:
  13  *
  14  * The above copyright notice and this permission notice (including the
  15  * next paragraph) shall be included in all copies or substantial portions
  16  * of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  19  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  21  * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  22  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  23  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  24  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25  *
  26  * Authors: James Zhu <james.zhu<@amd.com>
  27  *
  28  **************************************************************************/
  29
  30 #include <assert.h>
  31
  32 #include "tgsi/tgsi_text.h"
  33 #include "vl_compositor_cs.h"
  34
  35 struct cs_viewport {
  36    float scale_x;
  37    float scale_y;
  38    struct u_rect area;
  39    int translate_x;
  40    int translate_y;
  41 };
  42
  43 char *compute_shader_video_buffer =
  44       "COMP\n"
  45       "PROPERTY CS_FIXED_BLOCK_WIDTH 8\n"
  46       "PROPERTY CS_FIXED_BLOCK_HEIGHT 8\n"
  47       "PROPERTY CS_FIXED_BLOCK_DEPTH 1\n"
  48
  49       "DCL SV[0], THREAD_ID\n"
  50       "DCL SV[1], BLOCK_ID\n"
  51
  52       "DCL CONST[0..5]\n"
  53       "DCL SVIEW[0..2], RECT, FLOAT\n"
  54       "DCL SAMP[0..2]\n"
  55
  56       "DCL IMAGE[0], 2D, WR\n"
  57       "DCL TEMP[0..7]\n"
  58
  59       "IMM[0] UINT32 { 8, 8, 1, 0}\n"
  60       "IMM[1] FLT32 { 1.0, 2.0, 0.0, 0.0}\n"
  61
  62       "UMAD TEMP[0], SV[1], IMM[0], SV[0]\n"
  63
  64       /* Drawn area check */
  65       "USGE TEMP[1].xy, TEMP[0].xyxy, CONST[4].xyxy\n"
  66       "USLT TEMP[1].zw, TEMP[0].xyxy, CONST[4].zwzw\n"
  67       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].yyyy\n"
  68       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].zzzz\n"
  69       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].wwww\n"
  70
  71       "UIF TEMP[1]\n"
  72          /* Translate */
  73          "UADD TEMP[2].xy, TEMP[0], -CONST[5].xyxy\n"
  74          "U2F TEMP[2], TEMP[2]\n"
  75          "DIV TEMP[3], TEMP[2], IMM[1].yyyy\n"
  76
  77          /* Scale */
  78          "DIV TEMP[2], TEMP[2], CONST[3].zwzw\n"
  79          "DIV TEMP[3], TEMP[3], CONST[3].zwzw\n"
  80
  81          /* Fetch texels */
  82          "TEX_LZ TEMP[4].x, TEMP[2], SAMP[0], RECT\n"
  83          "TEX_LZ TEMP[4].y, TEMP[3], SAMP[1], RECT\n"
  84          "TEX_LZ TEMP[4].z, TEMP[3], SAMP[2], RECT\n"
  85
  86          "MOV TEMP[4].w, IMM[1].xxxx\n"
  87
  88          /* Color Space Conversion */
  89          "DP4 TEMP[7].x, CONST[0], TEMP[4]\n"
  90          "DP4 TEMP[7].y, CONST[1], TEMP[4]\n"
  91          "DP4 TEMP[7].z, CONST[2], TEMP[4]\n"
  92
  93          "MOV TEMP[5].w, TEMP[4].zzzz\n"
  94          "SLE TEMP[6].w, TEMP[5], CONST[3].xxxx\n"
  95          "SGT TEMP[5].w, TEMP[5], CONST[3].yyyy\n"
  96
  97          "MAX TEMP[7].w, TEMP[5], TEMP[6]\n"
  98
  99          "STORE IMAGE[0], TEMP[0], TEMP[7], 2D\n"
 100       "ENDIF\n"
 101
 102       "END\n";
 103
 104 char *compute_shader_weave =
 105       "COMP\n"
 106       "PROPERTY CS_FIXED_BLOCK_WIDTH 8\n"
 107       "PROPERTY CS_FIXED_BLOCK_HEIGHT 8\n"
 108       "PROPERTY CS_FIXED_BLOCK_DEPTH 1\n"
 109
 110       "DCL SV[0], THREAD_ID\n"
 111       "DCL SV[1], BLOCK_ID\n"
 112
 113       "DCL CONST[0..5]\n"
 114       "DCL SVIEW[0..2], RECT, FLOAT\n"
 115       "DCL SAMP[0..2]\n"
 116
 117       "DCL IMAGE[0], 2D, WR\n"
 118       "DCL TEMP[0..9]\n"
 119
 120       "IMM[0] UINT32 { 8, 8, 1, 0}\n"
 121       "IMM[1] FLT32 { 1.0, 2.0, 0.0, 0.0}\n"
 122       "IMM[2] UINT32 { 1, 2, 4, 0}\n"
 123
 124       "UMAD TEMP[0], SV[1], IMM[0], SV[0]\n"
 125
 126       /* Drawn area check */
 127       "USGE TEMP[1].xy, TEMP[0].xyxy, CONST[4].xyxy\n"
 128       "USLT TEMP[1].zw, TEMP[0].xyxy, CONST[4].zwzw\n"
 129       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].yyyy\n"
 130       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].zzzz\n"
 131       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].wwww\n"
 132
 133       "UIF TEMP[1]\n"
 134          "MOV TEMP[2], TEMP[0]\n"
 135          /* Translate */
 136          "UADD TEMP[2].xy, TEMP[2], -CONST[5].xyxy\n"
 137
 138          /* Texture layer */
 139          "UMOD TEMP[2].z, TEMP[2].yyyy, IMM[2].yyyy\n"
 140          "UMOD TEMP[3].z, TEMP[2].yyyy, IMM[2].zzzz\n"
 141          "USHR TEMP[3].z, TEMP[3].zzzz, IMM[2].xxxx\n"
 142
 143          "USHR TEMP[2].y, TEMP[2], IMM[2].xxxx\n"
 144          "USHR TEMP[3].xy, TEMP[2], IMM[2].xxxx\n"
 145
 146          "U2F TEMP[4], TEMP[2]\n"
 147          "U2F TEMP[5], TEMP[3]\n"
 148
 149          /* Scale */
 150          "DIV TEMP[4], TEMP[4], CONST[3].zwzw\n"
 151          "DIV TEMP[5], TEMP[5], CONST[3].zwzw\n"
 152
 153          /* Fetch texels */
 154          "TEX_LZ TEMP[6].x, TEMP[4], SAMP[0], RECT\n"
 155          "TEX_LZ TEMP[6].y, TEMP[5], SAMP[1], RECT\n"
 156          "TEX_LZ TEMP[6].z, TEMP[5], SAMP[2], RECT\n"
 157
 158          "MOV TEMP[6].w, IMM[1].xxxx\n"
 159
 160          /* Color Space Conversion */
 161          "DP4 TEMP[9].x, CONST[0], TEMP[6]\n"
 162          "DP4 TEMP[9].y, CONST[1], TEMP[6]\n"
 163          "DP4 TEMP[9].z, CONST[2], TEMP[6]\n"
 164
 165          "MOV TEMP[7].w, TEMP[6].zzzz\n"
 166          "SLE TEMP[8].w, TEMP[7], CONST[3].xxxx\n"
 167          "SGT TEMP[7].w, TEMP[7], CONST[3].yyyy\n"
 168
 169          "MAX TEMP[9].w, TEMP[7], TEMP[8]\n"
 170
 171          "STORE IMAGE[0], TEMP[0], TEMP[9], 2D\n"
 172       "ENDIF\n"
 173
 174       "END\n";
 175
 176 char *compute_shader_rgba =
 177       "COMP\n"
 178       "PROPERTY CS_FIXED_BLOCK_WIDTH 8\n"
 179       "PROPERTY CS_FIXED_BLOCK_HEIGHT 8\n"
 180       "PROPERTY CS_FIXED_BLOCK_DEPTH 1\n"
 181
 182       "DCL SV[0], THREAD_ID\n"
 183       "DCL SV[1], BLOCK_ID\n"
 184
 185       "DCL CONST[0..5]\n"
 186       "DCL SVIEW[0], RECT, FLOAT\n"
 187       "DCL SAMP[0]\n"
 188
 189       "DCL IMAGE[0], 2D, WR\n"
 190       "DCL TEMP[0..3]\n"
 191
 192       "IMM[0] UINT32 { 8, 8, 1, 0}\n"
 193       "IMM[1] FLT32 { 1.0, 2.0, 0.0, 0.0}\n"
 194
 195       "UMAD TEMP[0], SV[1], IMM[0], SV[0]\n"
 196
 197       /* Drawn area check */
 198       "USGE TEMP[1].xy, TEMP[0].xyxy, CONST[4].xyxy\n"
 199       "USLT TEMP[1].zw, TEMP[0].xyxy, CONST[4].zwzw\n"
 200       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].yyyy\n"
 201       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].zzzz\n"
 202       "AND TEMP[1].x, TEMP[1].xxxx, TEMP[1].wwww\n"
 203
 204       "UIF TEMP[1]\n"
 205          /* Translate */
 206          "UADD TEMP[2].xy, TEMP[0], -CONST[5].xyxy\n"
 207          "U2F TEMP[2], TEMP[2]\n"
 208
 209          /* Scale */
 210          "DIV TEMP[2], TEMP[2], CONST[3].zwzw\n"
 211
 212          /* Fetch texels */
 213          "TEX_LZ TEMP[3], TEMP[2], SAMP[0], RECT\n"
 214
 215          "STORE IMAGE[0], TEMP[0], TEMP[3], 2D\n"
 216       "ENDIF\n"
 217
 218       "END\n";
 219
 220 static void
 221 cs_launch(struct vl_compositor *c,
 222           void                 *cs)
 223 {
 224    struct pipe_context *ctx = c->pipe;
 225
 226    /* Bind the image */
 227    struct pipe_image_view image = {};
 228    image.resource = c->fb_state.cbufs[0]->texture;
 229    image.shader_access = image.access = PIPE_IMAGE_ACCESS_READ_WRITE;
 230    image.format = c->fb_state.cbufs[0]->texture->format;
 231
 232    ctx->set_shader_images(c->pipe, PIPE_SHADER_COMPUTE, 0, 1, &image);
 233
 234    /* Bind compute shader */
 235    ctx->bind_compute_state(ctx, cs);
 236
 237    /* Dispatch compute */
 238    struct pipe_grid_info info = {};
 239    info.block[0] = 8;
 240    info.block[1] = 8;
 241    info.block[2] = 1;
 242    info.grid[0] = DIV_ROUND_UP(c->fb_state.width, info.block[0]);
 243    info.grid[1] = DIV_ROUND_UP(c->fb_state.height, info.block[1]);
 244    info.grid[2] = 1;
 245
 246    ctx->launch_grid(ctx, &info);
 247 }
 248
 249 static inline struct u_rect
 250 calc_drawn_area(struct vl_compositor_state *s,
 251                 struct vl_compositor_layer *layer)
 252 {
 253    struct vertex2f tl, br;
 254    struct u_rect result;
 255
 256    assert(s && layer);
 257
 258    tl = layer->dst.tl;
 259    br = layer->dst.br;
 260
 261    /* Scale */
 262    result.x0 = tl.x * layer->viewport.scale[0] + layer->viewport.translate[0];
 263    result.y0 = tl.y * layer->viewport.scale[1] + layer->viewport.translate[1];
 264    result.x1 = br.x * layer->viewport.scale[0] + layer->viewport.translate[0];
 265    result.y1 = br.y * layer->viewport.scale[1] + layer->viewport.translate[1];
 266
 267    /* Clip */
 268    result.x0 = MAX2(result.x0, s->scissor.minx);
 269    result.y0 = MAX2(result.y0, s->scissor.miny);
 270    result.x1 = MIN2(result.x1, s->scissor.maxx);
 271    result.y1 = MIN2(result.y1, s->scissor.maxy);
 272    return result;
 273 }
 274
 275 static bool
 276 set_viewport(struct vl_compositor_state *s,
 277              struct cs_viewport         *drawn)
 278 {
 279    struct pipe_transfer *buf_transfer;
 280
 281    assert(s && drawn);
 282
 283    void *ptr = pipe_buffer_map(s->pipe, s->shader_params,
 284                                PIPE_TRANSFER_READ | PIPE_TRANSFER_WRITE,
 285                                &buf_transfer);
 286
 287    if (!ptr)
 288      return false;
 289
 290    float *ptr_float = (float *)ptr;
 291    ptr_float += sizeof(vl_csc_matrix)/sizeof(float) + 2;
 292    *ptr_float++ = drawn->scale_x;
 293    *ptr_float++ = drawn->scale_y;
 294
 295    int *ptr_int = (int *)ptr_float;
 296    *ptr_int++ = drawn->area.x0;
 297    *ptr_int++ = drawn->area.y0;
 298    *ptr_int++ = drawn->area.x1;
 299    *ptr_int++ = drawn->area.y1;
 300    *ptr_int++ = drawn->translate_x;
 301    *ptr_int = drawn->translate_y;
 302
 303    pipe_buffer_unmap(s->pipe, buf_transfer);
 304
 305    return true;
 306 }
 307
 308 static void
 309 draw_layers(struct vl_compositor       *c,
 310             struct vl_compositor_state *s,
 311             struct u_rect              *dirty)
 312 {
 313    unsigned i;
 314    static struct cs_viewport old_drawn;
 315
 316    assert(c);
 317
 318    for (i = 0; i < VL_COMPOSITOR_MAX_LAYERS; ++i) {
 319       if (s->used_layers & (1 << i)) {
 320          struct vl_compositor_layer *layer = &s->layers[i];
 321          struct pipe_sampler_view **samplers = &layer->sampler_views[0];
 322          unsigned num_sampler_views = !samplers[1] ? 1 : !samplers[2] ? 2 : 3;
 323          struct cs_viewport drawn;
 324
 325          drawn.area = calc_drawn_area(s, layer);
 326          drawn.scale_x = layer->viewport.scale[0] /
 327                   (float)layer->sampler_views[0]->texture->width0;
 328          drawn.scale_y = drawn.scale_x;
 329          drawn.translate_x = (int)layer->viewport.translate[0];
 330          drawn.translate_y = (int)layer->viewport.translate[1];
 331
 332          if (memcmp(&drawn, &old_drawn, sizeof(struct cs_viewport))) {
 333             set_viewport(s, &drawn);
 334             old_drawn = drawn;
 335          }
 336
 337          c->pipe->bind_sampler_states(c->pipe, PIPE_SHADER_COMPUTE, 0,
 338                         num_sampler_views, layer->samplers);
 339          c->pipe->set_sampler_views(c->pipe, PIPE_SHADER_COMPUTE, 0,
 340                         num_sampler_views, samplers);
 341
 342          cs_launch(c, layer->cs);
 343
 344          if (dirty) {
 345             struct u_rect drawn = calc_drawn_area(s, layer);
 346             dirty->x0 = MIN2(drawn.x0, dirty->x0);
 347             dirty->y0 = MIN2(drawn.y0, dirty->y0);
 348             dirty->x1 = MAX2(drawn.x1, dirty->x1);
 349             dirty->y1 = MAX2(drawn.y1, dirty->y1);
 350          }
 351       }
 352    }
 353 }
 354
 355 void *
 356 vl_compositor_cs_create_shader(struct vl_compositor *c,
 357                                const char           *compute_shader_text)
 358 {
 359    assert(c && compute_shader_text);
 360
 361    struct tgsi_token tokens[1024];
 362    if (!tgsi_text_translate(compute_shader_text, tokens, ARRAY_SIZE(tokens))) {
 363       assert(0);
 364       return NULL;
 365    }
 366
 367    struct pipe_compute_state state = {};
 368    state.ir_type = PIPE_SHADER_IR_TGSI;
 369    state.prog = tokens;
 370
 371    /* create compute shader */
 372    return c->pipe->create_compute_state(c->pipe, &state);
 373 }
 374
 375 void
 376 vl_compositor_cs_render(struct vl_compositor_state *s,
 377                         struct vl_compositor       *c,
 378                         struct pipe_surface        *dst_surface,
 379                         struct u_rect              *dirty_area,
 380                         bool                        clear_dirty)
 381 {
 382    assert(c && s);
 383    assert(dst_surface);
 384
 385    c->fb_state.width = dst_surface->width;
 386    c->fb_state.height = dst_surface->height;
 387    c->fb_state.cbufs[0] = dst_surface;
 388
 389    if (!s->scissor_valid) {
 390       s->scissor.minx = 0;
 391       s->scissor.miny = 0;
 392       s->scissor.maxx = dst_surface->width;
 393       s->scissor.maxy = dst_surface->height;
 394    }
 395
 396    if (clear_dirty && dirty_area &&
 397        (dirty_area->x0 < dirty_area->x1 || dirty_area->y0 < dirty_area->y1)) {
 398
 399       c->pipe->clear_render_target(c->pipe, dst_surface, &s->clear_color,
 400                        0, 0, dst_surface->width, dst_surface->height, false);
 401       dirty_area->x0 = dirty_area->y0 = VL_COMPOSITOR_MAX_DIRTY;
 402       dirty_area->x1 = dirty_area->y1 = VL_COMPOSITOR_MIN_DIRTY;
 403    }
 404
 405    pipe_set_constant_buffer(c->pipe, PIPE_SHADER_COMPUTE, 0, s->shader_params);
 406
 407    draw_layers(c, s, dirty_area);
 408 }