src/mesa/drivers/dri/intel/intel_span.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
   4  * Copyright 2011 Intel Corporation
   5  * All Rights Reserved.
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a
   8  * copy of this software and associated documentation files (the
   9  * "Software"), to deal in the Software without restriction, including
  10  * without limitation the rights to use, copy, modify, merge, publish,
  11  * distribute, sub license, and/or sell copies of the Software, and to
  12  * permit persons to whom the Software is furnished to do so, subject to
  13  * the following conditions:
  14  *
  15  * The above copyright notice and this permission notice (including the
  16  * next paragraph) shall be included in all copies or substantial portions
  17  * of the Software.
  18  *
  19  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  20  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  21  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  22  * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
  23  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  24  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  25  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  26  *
  27  * Authors:
  28  *     Chad Versace <chad@chad-versace.us>
  29  *
  30  **************************************************************************/
  31
  32 #include <stdbool.h>
  33 #include <stdint.h>
  34 #include "main/glheader.h"
  35 #include "main/macros.h"
  36 #include "main/mtypes.h"
  37 #include "main/colormac.h"
  38 #include "main/renderbuffer.h"
  39
  40 #include "intel_buffers.h"
  41 #include "intel_fbo.h"
  42 #include "intel_mipmap_tree.h"
  43 #include "intel_screen.h"
  44 #include "intel_span.h"
  45 #include "intel_regions.h"
  46 #include "intel_tex.h"
  47
  48 #include "swrast/swrast.h"
  49 #include "swrast/s_renderbuffer.h"
  50
  51 /**
  52  * \brief Get pointer offset into stencil buffer.
  53  *
  54  * The stencil buffer is W tiled. Since the GTT is incapable of W fencing, we
  55  * must decode the tile's layout in software.
  56  *
  57  * See
  58  *   - PRM, 2011 Sandy Bridge, Volume 1, Part 2, Section 4.5.2.1 W-Major Tile
  59  *     Format.
  60  *   - PRM, 2011 Sandy Bridge, Volume 1, Part 2, Section 4.5.3 Tiling Algorithm
  61  *
  62  * Even though the returned offset is always positive, the return type is
  63  * signed due to
  64  *    commit e8b1c6d6f55f5be3bef25084fdd8b6127517e137
  65  *    mesa: Fix return type of  _mesa_get_format_bytes() (#37351)
  66  */
  67 intptr_t
  68 intel_offset_S8(uint32_t stride, uint32_t x, uint32_t y)
  69 {
  70    uint32_t tile_size = 4096;
  71    uint32_t tile_width = 64;
  72    uint32_t tile_height = 64;
  73    uint32_t row_size = 64 * stride;
  74
  75    uint32_t tile_x = x / tile_width;
  76    uint32_t tile_y = y / tile_height;
  77
  78    /* The byte's address relative to the tile's base addres. */
  79    uint32_t byte_x = x % tile_width;
  80    uint32_t byte_y = y % tile_height;
  81
  82    uintptr_t u = tile_y * row_size
  83                + tile_x * tile_size
  84                + 512 * (byte_x / 8)
  85                +  64 * (byte_y / 8)
  86                +  32 * ((byte_y / 4) % 2)
  87                +  16 * ((byte_x / 4) % 2)
  88                +   8 * ((byte_y / 2) % 2)
  89                +   4 * ((byte_x / 2) % 2)
  90                +   2 * (byte_y % 2)
  91                +   1 * (byte_x % 2);
  92
  93    /*
  94     * Errata for Gen5:
  95     *
  96     * An additional offset is needed which is not documented in the PRM.
  97     *
  98     * if ((byte_x / 8) % 2 == 1) {
  99     *    if ((byte_y / 8) % 2) == 0) {
 100     *       u += 64;
 101     *    } else {
 102     *       u -= 64;
 103     *    }
 104     * }
 105     *
 106     * The offset is expressed more tersely as
 107     * u += ((int) x & 0x8) * (8 - (((int) y & 0x8) << 1));
 108     */
 109
 110    return u;
 111 }
 112
 113 static void
 114 intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb)
 115 {
 116    struct gl_context *ctx = &intel->ctx;
 117    struct intel_renderbuffer *irb = intel_renderbuffer(rb);
 118    GLubyte *map;
 119    int stride;
 120
 121    if (!irb)
 122       return;
 123
 124    if (irb->Base.Map) {
 125       /* Renderbuffer is already mapped. This usually happens when a single
 126        * buffer is attached to the framebuffer's depth and stencil attachment
 127        * points.
 128        */
 129       return;
 130    }
 131
 132    ctx->Driver.MapRenderbuffer(ctx, rb, 0, 0, rb->Width, rb->Height,
 133                                GL_MAP_READ_BIT | GL_MAP_WRITE_BIT,
 134                                &map, &stride);
 135    irb->Base.Map = map;
 136    irb->Base.RowStride = stride;
 137 }
 138
 139 static void
 140 intel_renderbuffer_unmap(struct intel_context *intel,
 141                          struct gl_renderbuffer *rb)
 142 {
 143    struct gl_context *ctx = &intel->ctx;
 144    struct intel_renderbuffer *irb = intel_renderbuffer(rb);
 145
 146    if (!irb)
 147       return;
 148
 149    if (!irb->Base.Map) {
 150       /* Renderbuffer is already unmapped. This usually happens when a single
 151        * buffer is attached to the framebuffer's depth and stencil attachment
 152        * points.
 153        */
 154       return;
 155    }
 156
 157    ctx->Driver.UnmapRenderbuffer(ctx, rb);
 158
 159    irb->Base.Map = NULL;
 160    irb->Base.RowStride = 0;
 161 }
 162
 163 static void
 164 intel_framebuffer_map(struct intel_context *intel, struct gl_framebuffer *fb)
 165 {
 166    int i;
 167
 168    for (i = 0; i < BUFFER_COUNT; i++) {
 169       intel_renderbuffer_map(intel, fb->Attachment[i].Renderbuffer);
 170    }
 171
 172    intel_check_front_buffer_rendering(intel);
 173 }
 174
 175 static void
 176 intel_framebuffer_unmap(struct intel_context *intel, struct gl_framebuffer *fb)
 177 {
 178    int i;
 179
 180    for (i = 0; i < BUFFER_COUNT; i++) {
 181       intel_renderbuffer_unmap(intel, fb->Attachment[i].Renderbuffer);
 182    }
 183 }
 184
 185 /**
 186  * Resolve all buffers that will be mapped by intelSpanRenderStart().
 187  *
 188  * Resolve the depth buffer of each enabled texture and of the read and draw
 189  * buffers.
 190  *
 191  * (Note: In the future this will also perform MSAA resolves.)
 192  */
 193 static void
 194 intel_span_resolve_buffers(struct intel_context *intel)
 195 {
 196    struct gl_context *ctx = &intel->ctx;
 197    struct intel_renderbuffer *draw_irb;
 198    struct intel_renderbuffer *read_irb;
 199    struct intel_texture_object *tex_obj;
 200
 201    /* Resolve depth buffer of each enabled texture. */
 202    for (int i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
 203       if (!ctx->Texture.Unit[i]._ReallyEnabled)
 204          continue;
 205       tex_obj = intel_texture_object(ctx->Texture.Unit[i]._Current);
 206       intel_finalize_mipmap_tree(intel, i);
 207       if (!tex_obj || !tex_obj->mt)
 208          continue;
 209       intel_miptree_all_slices_resolve_depth(intel, tex_obj->mt);
 210    }
 211
 212    /* Resolve each attached depth buffer. */
 213    draw_irb = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_DEPTH);
 214    read_irb = intel_get_renderbuffer(ctx->ReadBuffer, BUFFER_DEPTH);
 215    if (draw_irb)
 216       intel_renderbuffer_resolve_depth(intel, draw_irb);
 217    if (read_irb != draw_irb && read_irb)
 218       intel_renderbuffer_resolve_depth(intel, read_irb);
 219 }
 220
 221 /**
 222  * Map the regions needed by intelSpanRenderStart().
 223  */
 224 static void
 225 intel_span_map_buffers(struct intel_context *intel)
 226 {
 227    struct gl_context *ctx = &intel->ctx;
 228    struct intel_texture_object *tex_obj;
 229
 230    for (int i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
 231       if (!ctx->Texture.Unit[i]._ReallyEnabled)
 232          continue;
 233       tex_obj = intel_texture_object(ctx->Texture.Unit[i]._Current);
 234       intel_finalize_mipmap_tree(intel, i);
 235       intel_tex_map_images(intel, tex_obj,
 236                            GL_MAP_READ_BIT | GL_MAP_WRITE_BIT);
 237    }
 238
 239    intel_framebuffer_map(intel, ctx->DrawBuffer);
 240    if (ctx->ReadBuffer != ctx->DrawBuffer) {
 241       intel_framebuffer_map(intel, ctx->ReadBuffer);
 242    }
 243 }
 244
 245 /**
 246  * Prepare for software rendering.  Map current read/draw framebuffers'
 247  * renderbuffes and all currently bound texture objects.
 248  *
 249  * Old note: Moved locking out to get reasonable span performance.
 250  */
 251 void
 252 intelSpanRenderStart(struct gl_context * ctx)
 253 {
 254    struct intel_context *intel = intel_context(ctx);
 255
 256    intel_flush(ctx);
 257    intel_prepare_render(intel);
 258    intel_span_resolve_buffers(intel);
 259    intel_flush(ctx);
 260    intel_span_map_buffers(intel);
 261 }
 262
 263 /**
 264  * Called when done software rendering.  Unmap the buffers we mapped in
 265  * the above function.
 266  */
 267 void
 268 intelSpanRenderFinish(struct gl_context * ctx)
 269 {
 270    struct intel_context *intel = intel_context(ctx);
 271    GLuint i;
 272
 273    _swrast_flush(ctx);
 274
 275    for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
 276       if (ctx->Texture.Unit[i]._ReallyEnabled) {
 277          struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
 278          intel_tex_unmap_images(intel, intel_texture_object(texObj));
 279       }
 280    }
 281
 282    intel_framebuffer_unmap(intel, ctx->DrawBuffer);
 283    if (ctx->ReadBuffer != ctx->DrawBuffer) {
 284       intel_framebuffer_unmap(intel, ctx->ReadBuffer);
 285    }
 286 }
 287
 288
 289 void
 290 intelInitSpanFuncs(struct gl_context * ctx)
 291 {
 292    struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
 293    swdd->SpanRenderStart = intelSpanRenderStart;
 294    swdd->SpanRenderFinish = intelSpanRenderFinish;
 295 }
 296
 297 void
 298 intel_map_vertex_shader_textures(struct gl_context *ctx)
 299 {
 300    struct intel_context *intel = intel_context(ctx);
 301    int i;
 302
 303    if (ctx->VertexProgram._Current == NULL)
 304       return;
 305
 306    for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
 307       if (ctx->Texture.Unit[i]._ReallyEnabled &&
 308           ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) {
 309          struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
 310
 311          intel_tex_map_images(intel, intel_texture_object(texObj),
 312                               GL_MAP_READ_BIT | GL_MAP_WRITE_BIT);
 313       }
 314    }
 315 }
 316
 317 void
 318 intel_unmap_vertex_shader_textures(struct gl_context *ctx)
 319 {
 320    struct intel_context *intel = intel_context(ctx);
 321    int i;
 322
 323    if (ctx->VertexProgram._Current == NULL)
 324       return;
 325
 326    for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
 327       if (ctx->Texture.Unit[i]._ReallyEnabled &&
 328           ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) {
 329          struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
 330
 331          intel_tex_unmap_images(intel, intel_texture_object(texObj));
 332       }
 333    }
 334 }