dlls/wined3d/context.c

   1 /*
   2  * Context and render target management in wined3d
   3  *
   4  * Copyright 2002-2004 Jason Edmeades
   5  * Copyright 2002-2004 Raphael Junqueira
   6  * Copyright 2004 Christian Costa
   7  * Copyright 2005 Oliver Stieber
   8  * Copyright 2006, 2008 Henri Verbeet
   9  * Copyright 2007-2011, 2013 Stefan Dösinger for CodeWeavers
  10  * Copyright 2009-2011 Henri Verbeet for CodeWeavers
  11  *
  12  * This library is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public
  14  * License as published by the Free Software Foundation; either
  15  * version 2.1 of the License, or (at your option) any later version.
  16  *
  17  * This library is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with this library; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
  25  */
  26
  27 #include "wined3d_private.h"
  28
  29 WINE_DEFAULT_DEBUG_CHANNEL(d3d);
  30 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf);
  31
  32 void context_resource_released(const struct wined3d_device *device, struct wined3d_resource *resource)
  33 {
  34     unsigned int i;
  35
  36     for (i = 0; i < device->context_count; ++i)
  37     {
  38         struct wined3d_context *context = device->contexts[i];
  39
  40         if (&context->current_rt.texture->resource == resource)
  41         {
  42             context->current_rt.texture = NULL;
  43             context->current_rt.sub_resource_idx = 0;
  44         }
  45     }
  46 }
  47
  48 void wined3d_context_cleanup(struct wined3d_context *context)
  49 {
  50 }
  51
  52 /* This is used when a context for render target A is active, but a separate context is
  53  * needed to access the WGL framebuffer for render target B. Re-acquire a context for rt
  54  * A to avoid breaking caller code. */
  55 void context_restore(struct wined3d_context *context, struct wined3d_texture *texture, unsigned int sub_resource_idx)
  56 {
  57     if (context->current_rt.texture != texture || context->current_rt.sub_resource_idx != sub_resource_idx)
  58     {
  59         context_release(context);
  60         context = context_acquire(texture->resource.device, texture, sub_resource_idx);
  61     }
  62
  63     context_release(context);
  64 }
  65
  66 void context_invalidate_compute_state(struct wined3d_context *context, DWORD state_id)
  67 {
  68     DWORD representative = context->state_table[state_id].representative - STATE_COMPUTE_OFFSET;
  69     unsigned int index, shift;
  70
  71     index = representative / (sizeof(*context->dirty_compute_states) * CHAR_BIT);
  72     shift = representative & (sizeof(*context->dirty_compute_states) * CHAR_BIT - 1);
  73     context->dirty_compute_states[index] |= (1u << shift);
  74 }
  75
  76 void context_invalidate_state(struct wined3d_context *context, unsigned int state_id)
  77 {
  78     unsigned int representative = context->state_table[state_id].representative;
  79     unsigned int index, shift;
  80
  81     index = representative / (sizeof(*context->dirty_graphics_states) * CHAR_BIT);
  82     shift = representative & ((sizeof(*context->dirty_graphics_states) * CHAR_BIT) - 1);
  83     context->dirty_graphics_states[index] |= (1u << shift);
  84 }
  85
  86 void wined3d_context_init(struct wined3d_context *context, struct wined3d_swapchain *swapchain)
  87 {
  88     struct wined3d_device *device = swapchain->device;
  89     DWORD state;
  90
  91     context->d3d_info = &device->adapter->d3d_info;
  92     context->state_table = device->state_table;
  93
  94     /* Mark all states dirty to force a proper initialization of the states on
  95      * the first use of the context. Compute states do not need initialization. */
  96     for (state = 0; state <= STATE_HIGHEST; ++state)
  97     {
  98         if (context->state_table[state].representative && !STATE_IS_COMPUTE(state))
  99             context_invalidate_state(context, state);
 100     }
 101
 102     context->device = device;
 103     context->swapchain = swapchain;
 104     context->current_rt.texture = swapchain->front_buffer;
 105     context->current_rt.sub_resource_idx = 0;
 106
 107     context->shader_update_mask = (1u << WINED3D_SHADER_TYPE_PIXEL)
 108             | (1u << WINED3D_SHADER_TYPE_VERTEX)
 109             | (1u << WINED3D_SHADER_TYPE_GEOMETRY)
 110             | (1u << WINED3D_SHADER_TYPE_HULL)
 111             | (1u << WINED3D_SHADER_TYPE_DOMAIN)
 112             | (1u << WINED3D_SHADER_TYPE_COMPUTE);
 113 }
 114
 115 HRESULT wined3d_context_no3d_init(struct wined3d_context *context_no3d, struct wined3d_swapchain *swapchain)
 116 {
 117     TRACE("context_no3d %p, swapchain %p.\n", context_no3d, swapchain);
 118
 119     wined3d_context_init(context_no3d, swapchain);
 120
 121     return WINED3D_OK;
 122 }
 123
 124 static BOOL fixed_get_input(BYTE usage, BYTE usage_idx, unsigned int *regnum)
 125 {
 126     if ((usage == WINED3D_DECL_USAGE_POSITION || usage == WINED3D_DECL_USAGE_POSITIONT) && !usage_idx)
 127         *regnum = WINED3D_FFP_POSITION;
 128     else if (usage == WINED3D_DECL_USAGE_BLEND_WEIGHT && !usage_idx)
 129         *regnum = WINED3D_FFP_BLENDWEIGHT;
 130     else if (usage == WINED3D_DECL_USAGE_BLEND_INDICES && !usage_idx)
 131         *regnum = WINED3D_FFP_BLENDINDICES;
 132     else if (usage == WINED3D_DECL_USAGE_NORMAL && !usage_idx)
 133         *regnum = WINED3D_FFP_NORMAL;
 134     else if (usage == WINED3D_DECL_USAGE_PSIZE && !usage_idx)
 135         *regnum = WINED3D_FFP_PSIZE;
 136     else if (usage == WINED3D_DECL_USAGE_COLOR && !usage_idx)
 137         *regnum = WINED3D_FFP_DIFFUSE;
 138     else if (usage == WINED3D_DECL_USAGE_COLOR && usage_idx == 1)
 139         *regnum = WINED3D_FFP_SPECULAR;
 140     else if (usage == WINED3D_DECL_USAGE_TEXCOORD && usage_idx < WINED3DDP_MAXTEXCOORD)
 141         *regnum = WINED3D_FFP_TEXCOORD0 + usage_idx;
 142     else
 143     {
 144         WARN("Unsupported input stream [usage=%s, usage_idx=%u].\n", debug_d3ddeclusage(usage), usage_idx);
 145         *regnum = ~0u;
 146         return FALSE;
 147     }
 148
 149     return TRUE;
 150 }
 151
 152 /* Context activation is done by the caller. */
 153 void wined3d_stream_info_from_declaration(struct wined3d_stream_info *stream_info,
 154         const struct wined3d_state *state, const struct wined3d_d3d_info *d3d_info)
 155 {
 156     /* We need to deal with frequency data! */
 157     struct wined3d_vertex_declaration *declaration = state->vertex_declaration;
 158     BOOL generic_attributes = d3d_info->ffp_generic_attributes;
 159     BOOL use_vshader = use_vs(state);
 160     unsigned int i;
 161
 162     stream_info->use_map = 0;
 163     stream_info->swizzle_map = 0;
 164     stream_info->position_transformed = 0;
 165
 166     if (!declaration)
 167         return;
 168
 169     stream_info->position_transformed = declaration->position_transformed;
 170
 171     /* Translate the declaration into strided data. */
 172     for (i = 0; i < declaration->element_count; ++i)
 173     {
 174         const struct wined3d_vertex_declaration_element *element = &declaration->elements[i];
 175         const struct wined3d_stream_state *stream = &state->streams[element->input_slot];
 176         BOOL stride_used;
 177         unsigned int idx;
 178
 179         TRACE("%p Element %p (%u of %u).\n", declaration->elements,
 180                 element, i + 1, declaration->element_count);
 181
 182         if (!stream->buffer)
 183             continue;
 184
 185         TRACE("offset %u input_slot %u usage_idx %d.\n", element->offset, element->input_slot, element->usage_idx);
 186
 187         if (use_vshader)
 188         {
 189             if (element->output_slot == WINED3D_OUTPUT_SLOT_UNUSED)
 190             {
 191                 stride_used = FALSE;
 192             }
 193             else if (element->output_slot == WINED3D_OUTPUT_SLOT_SEMANTIC)
 194             {
 195                 /* TODO: Assuming vertexdeclarations are usually used with the
 196                  * same or a similar shader, it might be worth it to store the
 197                  * last used output slot and try that one first. */
 198                 stride_used = vshader_get_input(state->shader[WINED3D_SHADER_TYPE_VERTEX],
 199                         element->usage, element->usage_idx, &idx);
 200             }
 201             else
 202             {
 203                 idx = element->output_slot;
 204                 stride_used = TRUE;
 205             }
 206         }
 207         else
 208         {
 209             if (!generic_attributes && !element->ffp_valid)
 210             {
 211                 WARN("Skipping unsupported fixed function element of format %s and usage %s.\n",
 212                         debug_d3dformat(element->format->id), debug_d3ddeclusage(element->usage));
 213                 stride_used = FALSE;
 214             }
 215             else
 216             {
 217                 stride_used = fixed_get_input(element->usage, element->usage_idx, &idx);
 218             }
 219         }
 220
 221         if (stride_used)
 222         {
 223             TRACE("Load %s array %u [usage %s, usage_idx %u, "
 224                     "input_slot %u, offset %u, stride %u, format %s, class %s, step_rate %u].\n",
 225                     use_vshader ? "shader": "fixed function", idx,
 226                     debug_d3ddeclusage(element->usage), element->usage_idx, element->input_slot,
 227                     element->offset, stream->stride, debug_d3dformat(element->format->id),
 228                     debug_d3dinput_classification(element->input_slot_class), element->instance_data_step_rate);
 229
 230             stream_info->elements[idx].format = element->format;
 231             stream_info->elements[idx].data.buffer_object = 0;
 232             stream_info->elements[idx].data.addr = (BYTE *)NULL + stream->offset + element->offset;
 233             stream_info->elements[idx].stride = stream->stride;
 234             stream_info->elements[idx].stream_idx = element->input_slot;
 235             if (stream->flags & WINED3DSTREAMSOURCE_INSTANCEDATA)
 236             {
 237                 stream_info->elements[idx].divisor = 1;
 238                 stream_info->elements[idx].instanced = true;
 239             }
 240             else if (element->input_slot_class == WINED3D_INPUT_PER_INSTANCE_DATA)
 241             {
 242                 stream_info->elements[idx].divisor = element->instance_data_step_rate;
 243                 stream_info->elements[idx].instanced = true;
 244             }
 245             else
 246             {
 247                 stream_info->elements[idx].divisor = 0;
 248                 stream_info->elements[idx].instanced = false;
 249             }
 250
 251             if (!d3d_info->vertex_bgra && element->format->id == WINED3DFMT_B8G8R8A8_UNORM)
 252             {
 253                 stream_info->swizzle_map |= 1u << idx;
 254             }
 255             stream_info->use_map |= 1u << idx;
 256         }
 257     }
 258 }
 259
 260 /* Context activation is done by the caller. */
 261 void context_update_stream_info(struct wined3d_context *context, const struct wined3d_state *state)
 262 {
 263     struct wined3d_stream_info *stream_info = &context->stream_info;
 264     const struct wined3d_d3d_info *d3d_info = context->d3d_info;
 265     DWORD prev_all_vbo = stream_info->all_vbo;
 266     unsigned int i;
 267     uint32_t map;
 268
 269     wined3d_stream_info_from_declaration(stream_info, state, d3d_info);
 270
 271     stream_info->all_vbo = 1;
 272     map = stream_info->use_map;
 273     while (map)
 274     {
 275         struct wined3d_stream_info_element *element;
 276         struct wined3d_bo_address data;
 277         struct wined3d_buffer *buffer;
 278
 279         i = wined3d_bit_scan(&map);
 280         element = &stream_info->elements[i];
 281         buffer = state->streams[element->stream_idx].buffer;
 282
 283         /* We can't use VBOs if the base vertex index is negative. OpenGL
 284          * doesn't accept negative offsets (or rather offsets bigger than the
 285          * VBO, because the pointer is unsigned), so use system memory
 286          * sources. In most sane cases the pointer - offset will still be > 0,
 287          * otherwise it will wrap around to some big value. Hope that with the
 288          * indices the driver wraps it back internally. If not,
 289          * draw_primitive_immediate_mode() is needed, including a vertex buffer
 290          * path. */
 291         if (state->load_base_vertex_index < 0)
 292         {
 293             WARN_(d3d_perf)("load_base_vertex_index is < 0 (%d), not using VBOs.\n",
 294                     state->load_base_vertex_index);
 295             element->data.buffer_object = 0;
 296             element->data.addr += (ULONG_PTR)wined3d_buffer_load_sysmem(buffer, context);
 297             if ((UINT_PTR)element->data.addr < -state->load_base_vertex_index * element->stride)
 298                 FIXME("System memory vertex data load offset is negative!\n");
 299         }
 300         else
 301         {
 302             wined3d_buffer_load(buffer, context, state);
 303             wined3d_buffer_get_memory(buffer, context, &data);
 304             element->data.buffer_object = data.buffer_object;
 305             element->data.addr += (ULONG_PTR)data.addr;
 306         }
 307
 308         if (!element->data.buffer_object)
 309             stream_info->all_vbo = 0;
 310
 311         TRACE("Load array %u %s.\n", i, debug_bo_address(&element->data));
 312     }
 313
 314     if (prev_all_vbo != stream_info->all_vbo)
 315         context_invalidate_state(context, STATE_INDEXBUFFER);
 316
 317     context->use_immediate_mode_draw = FALSE;
 318
 319     if (stream_info->all_vbo)
 320         return;
 321
 322     if (!use_vs(state))
 323     {
 324         WORD slow_mask = -!d3d_info->ffp_generic_attributes & (1u << WINED3D_FFP_PSIZE);
 325         slow_mask |= -(!d3d_info->vertex_bgra && !d3d_info->ffp_generic_attributes)
 326                 & ((1u << WINED3D_FFP_DIFFUSE) | (1u << WINED3D_FFP_SPECULAR) | (1u << WINED3D_FFP_BLENDWEIGHT));
 327
 328         if ((stream_info->position_transformed && !d3d_info->xyzrhw)
 329                 || (stream_info->use_map & slow_mask))
 330             context->use_immediate_mode_draw = TRUE;
 331     }
 332 }
 333
 334 static bool is_resource_rtv_bound(const struct wined3d_state *state,
 335         const struct wined3d_resource *resource)
 336 {
 337     unsigned int i;
 338
 339     if (!resource->rtv_bind_count_device)
 340         return false;
 341
 342     for (i = 0; i < ARRAY_SIZE(state->fb.render_targets); ++i)
 343     {
 344         if (state->fb.render_targets[i] && state->fb.render_targets[i]->resource == resource)
 345             return true;
 346     }
 347
 348     return false;
 349 }
 350
 351 /* Context activation is done by the caller. */
 352 static void context_preload_texture(struct wined3d_context *context,
 353         const struct wined3d_state *state, unsigned int idx)
 354 {
 355     struct wined3d_texture *texture;
 356
 357     if (!(texture = state->textures[idx]))
 358         return;
 359
 360     if (is_resource_rtv_bound(state, &texture->resource)
 361             || (state->fb.depth_stencil && state->fb.depth_stencil->resource == &texture->resource))
 362         context->uses_fbo_attached_resources = 1;
 363
 364     wined3d_texture_load(texture, context, is_srgb_enabled(state->sampler_states[idx]));
 365 }
 366
 367 /* Context activation is done by the caller. */
 368 void context_preload_textures(struct wined3d_context *context, const struct wined3d_state *state)
 369 {
 370     unsigned int i;
 371
 372     if (use_vs(state))
 373     {
 374         for (i = 0; i < WINED3D_MAX_VERTEX_SAMPLERS; ++i)
 375         {
 376             if (state->shader[WINED3D_SHADER_TYPE_VERTEX]->reg_maps.resource_info[i].type)
 377                 context_preload_texture(context, state, WINED3D_MAX_FRAGMENT_SAMPLERS + i);
 378         }
 379     }
 380
 381     if (use_ps(state))
 382     {
 383         for (i = 0; i < WINED3D_MAX_FRAGMENT_SAMPLERS; ++i)
 384         {
 385             if (state->shader[WINED3D_SHADER_TYPE_PIXEL]->reg_maps.resource_info[i].type)
 386                 context_preload_texture(context, state, i);
 387         }
 388     }
 389     else
 390     {
 391         uint32_t ffu_map = context->fixed_function_usage_map;
 392
 393         while (ffu_map)
 394         {
 395             i = wined3d_bit_scan(&ffu_map);
 396             context_preload_texture(context, state, i);
 397         }
 398     }
 399 }