src/gallium/drivers/vc4/vc4_draw.c

   1 /*
   2  * Copyright (c) 2014 Scott Mansell
   3  * Copyright © 2014 Broadcom
   4  *
   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:
  11  *
  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
  14  * Software.
  15  *
  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
  22  * IN THE SOFTWARE.
  23  */
  24
  25 #include <stdio.h>
  26
  27 #include "vc4_context.h"
  28 #include "vc4_resource.h"
  29
  30 static struct vc4_bo *
  31 get_vbo(struct vc4_context *vc4, uint32_t width, uint32_t height)
  32 {
  33         struct {
  34                 uint16_t x, y;
  35                 float z, rhw, r, g, b;
  36         } verts[] = {
  37                 {
  38                         // Vertex: Top, red
  39                         (-(int)width / 3) << 4, // X in 12.4 fixed point
  40                         (-(int)height / 3) << 4, // Y in 12.4 fixed point
  41                         1.0f, // Z
  42                         1.0f, // 1/W
  43                         1.0f, // Varying 0 (Red)
  44                         0.0f, // Varying 1 (Green)
  45                         0.0f, // Varying 2 (Blue)
  46                 },
  47                 {
  48                         // Vertex: bottom left, Green
  49                         (width / 3) << 4, // X in 12.4 fixed point
  50                         (-(int)height / 3) << 4, // Y in 12.4 fixed point
  51                         1.0f, // Z
  52                         1.0f, // 1/W
  53                         0.0f, // Varying 0 (Red)
  54                         1.0f, // Varying 1 (Green)
  55                         0.0f, // Varying 2 (Blue)
  56                 },
  57
  58                 {
  59                         // Vertex: bottom right, Blue
  60                         (width / 3) << 4, // X in 12.4 fixed point
  61                         (height / 3) << 4, // Y in 12.4 fixed point
  62                         1.0f, // Z
  63                         1.0f, // 1/W
  64                         0.0f, // Varying 0 (Red)
  65                         0.0f, // Varying 1 (Green)
  66                         1.0f, // Varying 2 (Blue)
  67                 },
  68         };
  69
  70         return vc4_bo_alloc_mem(vc4->screen, verts, sizeof(verts), "verts");
  71 }
  72 static struct vc4_bo *
  73 get_ibo(struct vc4_context *vc4)
  74 {
  75         static const uint8_t indices[] = { 0, 1, 2 };
  76
  77         return vc4_bo_alloc_mem(vc4->screen, indices, sizeof(indices), "indices");
  78 }
  79
  80 static void
  81 vc4_rcl_tile_calls(struct vc4_context *vc4,
  82                    uint32_t xtiles, uint32_t ytiles,
  83                    struct vc4_bo *tile_alloc)
  84 {
  85         for (int x = 0; x < xtiles; x++) {
  86                 for (int y = 0; y < ytiles; y++) {
  87                         cl_u8(&vc4->rcl, VC4_PACKET_TILE_COORDINATES);
  88                         cl_u8(&vc4->rcl, x);
  89                         cl_u8(&vc4->rcl, y);
  90
  91                         cl_start_reloc(&vc4->rcl, 1);
  92                         cl_u8(&vc4->rcl, VC4_PACKET_BRANCH_TO_SUB_LIST);
  93                         cl_reloc(vc4, &vc4->rcl, tile_alloc,
  94                                  (y * xtiles + x) * 32);
  95
  96                         if (x == xtiles - 1 && y == ytiles - 1) {
  97                                 cl_u8(&vc4->rcl,
  98                                       VC4_PACKET_STORE_MS_TILE_BUFFER_AND_EOF);
  99                         } else {
 100                                 cl_u8(&vc4->rcl,
 101                                       VC4_PACKET_STORE_MS_TILE_BUFFER);
 102                         }
 103                 }
 104         }
 105 }
 106
 107 static void
 108 vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
 109 {
 110         struct vc4_context *vc4 = vc4_context(pctx);
 111         uint32_t width = vc4->framebuffer.width;
 112         uint32_t height = vc4->framebuffer.height;
 113         uint32_t tilew = align(width, 64) / 64;
 114         uint32_t tileh = align(height, 64) / 64;
 115         struct vc4_bo *tile_alloc = vc4_bo_alloc(vc4->screen,
 116                                                  32 * tilew * tileh, "tilea");
 117         struct vc4_bo *tile_state = vc4_bo_alloc(vc4->screen,
 118                                                  48 * tilew * tileh, "tilestate");
 119         struct vc4_bo *ibo = get_ibo(vc4);
 120
 121         struct vc4_bo *fs_uniform = vc4_bo_alloc(vc4->screen, 0x1000, "fsu");
 122         struct vc4_bo *vbo = get_vbo(vc4, width, height);
 123
 124         vc4->needs_flush = true;
 125
 126         //   Tile state data is 48 bytes per tile, I think it can be thrown away
 127         //   as soon as binning is finished.
 128         cl_start_reloc(&vc4->bcl, 2);
 129         cl_u8(&vc4->bcl, VC4_PACKET_TILE_BINNING_MODE_CONFIG);
 130         cl_reloc(vc4, &vc4->bcl, tile_alloc, 0);
 131         cl_u32(&vc4->bcl, 0x8000); /* tile allocation memory size */
 132         cl_reloc(vc4, &vc4->bcl, tile_state, 0);
 133         cl_u8(&vc4->bcl, tilew);
 134         cl_u8(&vc4->bcl, tileh);
 135         cl_u8(&vc4->bcl, VC4_BIN_CONFIG_AUTO_INIT_TSDA);
 136
 137         cl_u8(&vc4->bcl, VC4_PACKET_START_TILE_BINNING);
 138
 139         cl_u8(&vc4->bcl, VC4_PACKET_PRIMITIVE_LIST_FORMAT);
 140         cl_u8(&vc4->bcl, 0x32); // 16 bit triangle
 141
 142         vc4_emit_state(pctx);
 143
 144         /* the actual draw call. */
 145         cl_u8(&vc4->bcl, VC4_PACKET_NV_SHADER_STATE);
 146 #ifndef USE_VC4_SIMULATOR
 147         cl_u32(&vc4->bcl, 0); /* offset into shader_rec */
 148 #else
 149         cl_u32(&vc4->bcl, simpenrose_hw_addr(vc4->shader_rec.next));
 150 #endif
 151
 152         cl_start_reloc(&vc4->bcl, 1);
 153         cl_u8(&vc4->bcl, VC4_PACKET_GL_INDEXED_PRIMITIVE);
 154         cl_u8(&vc4->bcl, 0x04); // 8bit index, trinagles
 155         cl_u32(&vc4->bcl, 3); // Length
 156         cl_reloc(vc4, &vc4->bcl, ibo, 0);
 157         cl_u32(&vc4->bcl, 2); // Maximum index
 158
 159         cl_u8(&vc4->bcl, VC4_PACKET_FLUSH_ALL);
 160         cl_u8(&vc4->bcl, VC4_PACKET_NOP);
 161         cl_u8(&vc4->bcl, VC4_PACKET_HALT);
 162
 163 // Shader Record
 164         cl_start_shader_reloc(&vc4->shader_rec, 3);
 165
 166         cl_u8(&vc4->shader_rec, 0);
 167         cl_u8(&vc4->shader_rec, 6*4); // stride
 168         cl_u8(&vc4->shader_rec, 0xcc); // num uniforms (not used)
 169         cl_u8(&vc4->shader_rec, 3); // num varyings
 170         cl_reloc(vc4, &vc4->shader_rec, vc4->prog.fs->bo, 0);
 171         cl_reloc(vc4, &vc4->shader_rec, fs_uniform, 0);
 172         cl_reloc(vc4, &vc4->shader_rec, vbo, 0);
 173
 174         vc4->shader_rec_count++;
 175
 176         cl_u8(&vc4->rcl, VC4_PACKET_CLEAR_COLORS);
 177         cl_u32(&vc4->rcl, 0xff000000); // Opaque Black
 178         cl_u32(&vc4->rcl, 0xff000000); // 32 bit clear colours need to be repeated twice
 179         cl_u32(&vc4->rcl, 0);
 180         cl_u8(&vc4->rcl, 0);
 181
 182         struct vc4_surface *csurf = vc4_surface(vc4->framebuffer.cbufs[0]);
 183         struct vc4_resource *ctex = vc4_resource(csurf->base.texture);
 184
 185         cl_start_reloc(&vc4->rcl, 1);
 186         cl_u8(&vc4->rcl, VC4_PACKET_TILE_RENDERING_MODE_CONFIG);
 187         cl_reloc(vc4, &vc4->rcl, ctex->bo, csurf->offset);
 188         cl_u16(&vc4->rcl, width);
 189         cl_u16(&vc4->rcl, height);
 190         cl_u8(&vc4->rcl, (VC4_RENDER_CONFIG_MEMORY_FORMAT_LINEAR |
 191                           VC4_RENDER_CONFIG_FORMAT_RGBA8888));
 192         cl_u8(&vc4->rcl, 0);
 193
 194         // Do a store of the first tile to force the tile buffer to be cleared
 195         /* XXX: I think these two packets may be unnecessary. */
 196         cl_u8(&vc4->rcl, VC4_PACKET_TILE_COORDINATES);
 197         cl_u8(&vc4->rcl, 0);
 198         cl_u8(&vc4->rcl, 0);
 199
 200         cl_u8(&vc4->rcl, VC4_PACKET_STORE_TILE_BUFFER_GENERAL);
 201         cl_u16(&vc4->rcl, 0); // Store nothing (just clear)
 202         cl_u32(&vc4->rcl, 0); // no address is needed
 203
 204         vc4_rcl_tile_calls(vc4, tilew, tileh, tile_alloc);
 205
 206         vc4_flush(pctx);
 207 }
 208
 209 static void
 210 vc4_clear(struct pipe_context *pctx, unsigned buffers,
 211           const union pipe_color_union *color, double depth, unsigned stencil)
 212 {
 213         struct vc4_context *vc4 = vc4_context(pctx);
 214
 215         vc4->needs_flush = true;
 216 }
 217
 218 static void
 219 vc4_clear_render_target(struct pipe_context *pctx, struct pipe_surface *ps,
 220                         const union pipe_color_union *color,
 221                         unsigned x, unsigned y, unsigned w, unsigned h)
 222 {
 223         fprintf(stderr, "unimpl: clear RT\n");
 224 }
 225
 226 static void
 227 vc4_clear_depth_stencil(struct pipe_context *pctx, struct pipe_surface *ps,
 228                         unsigned buffers, double depth, unsigned stencil,
 229                         unsigned x, unsigned y, unsigned w, unsigned h)
 230 {
 231         fprintf(stderr, "unimpl: clear DS\n");
 232 }
 233
 234 void
 235 vc4_draw_init(struct pipe_context *pctx)
 236 {
 237         pctx->draw_vbo = vc4_draw_vbo;
 238         pctx->clear = vc4_clear;
 239         pctx->clear_render_target = vc4_clear_render_target;
 240         pctx->clear_depth_stencil = vc4_clear_depth_stencil;
 241 }