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 "config.h"
  28 #include "wine/port.h"
  29
  30 #include "wined3d_private.h"
  31
  32 WINE_DEFAULT_DEBUG_CHANNEL(d3d);
  33 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf);
  34
  35 void context_resource_released(const struct wined3d_device *device, struct wined3d_resource *resource)
  36 {
  37     unsigned int i;
  38
  39     for (i = 0; i < device->context_count; ++i)
  40     {
  41         struct wined3d_context *context = device->contexts[i];
  42
  43         if (&context->current_rt.texture->resource == resource)
  44         {
  45             context->current_rt.texture = NULL;
  46             context->current_rt.sub_resource_idx = 0;
  47         }
  48     }
  49 }
  50
  51 void wined3d_context_cleanup(struct wined3d_context *context)
  52 {
  53 }
  54
  55 /* This is used when a context for render target A is active, but a separate context is
  56  * needed to access the WGL framebuffer for render target B. Re-acquire a context for rt
  57  * A to avoid breaking caller code. */
  58 void context_restore(struct wined3d_context *context, struct wined3d_texture *texture, unsigned int sub_resource_idx)
  59 {
  60     if (context->current_rt.texture != texture || context->current_rt.sub_resource_idx != sub_resource_idx)
  61     {
  62         context_release(context);
  63         context = context_acquire(texture->resource.device, texture, sub_resource_idx);
  64     }
  65
  66     context_release(context);
  67 }
  68
  69 void context_invalidate_compute_state(struct wined3d_context *context, DWORD state_id)
  70 {
  71     DWORD representative = context->state_table[state_id].representative - STATE_COMPUTE_OFFSET;
  72     unsigned int index, shift;
  73
  74     index = representative / (sizeof(*context->dirty_compute_states) * CHAR_BIT);
  75     shift = representative & (sizeof(*context->dirty_compute_states) * CHAR_BIT - 1);
  76     context->dirty_compute_states[index] |= (1u << shift);
  77 }
  78
  79 void context_invalidate_state(struct wined3d_context *context, unsigned int state_id)
  80 {
  81     unsigned int representative = context->state_table[state_id].representative;
  82     unsigned int index, shift;
  83
  84     index = representative / (sizeof(*context->dirty_graphics_states) * CHAR_BIT);
  85     shift = representative & ((sizeof(*context->dirty_graphics_states) * CHAR_BIT) - 1);
  86     context->dirty_graphics_states[index] |= (1u << shift);
  87 }
  88
  89 void wined3d_context_init(struct wined3d_context *context, struct wined3d_swapchain *swapchain)
  90 {
  91     struct wined3d_device *device = swapchain->device;
  92     DWORD state;
  93
  94     context->d3d_info = &device->adapter->d3d_info;
  95     context->state_table = device->state_table;
  96
  97     /* Mark all states dirty to force a proper initialization of the states on
  98      * the first use of the context. Compute states do not need initialization. */
  99     for (state = 0; state <= STATE_HIGHEST; ++state)
 100     {
 101         if (context->state_table[state].representative && !STATE_IS_COMPUTE(state))
 102             context_invalidate_state(context, state);
 103     }
 104
 105     context->device = device;
 106     context->swapchain = swapchain;
 107     context->current_rt.texture = swapchain->front_buffer;
 108     context->current_rt.sub_resource_idx = 0;
 109
 110     context->shader_update_mask = (1u << WINED3D_SHADER_TYPE_PIXEL)
 111             | (1u << WINED3D_SHADER_TYPE_VERTEX)
 112             | (1u << WINED3D_SHADER_TYPE_GEOMETRY)
 113             | (1u << WINED3D_SHADER_TYPE_HULL)
 114             | (1u << WINED3D_SHADER_TYPE_DOMAIN)
 115             | (1u << WINED3D_SHADER_TYPE_COMPUTE);
 116 }
 117
 118 HRESULT wined3d_context_no3d_init(struct wined3d_context *context_no3d, struct wined3d_swapchain *swapchain)
 119 {
 120     TRACE("context_no3d %p, swapchain %p.\n", context_no3d, swapchain);
 121
 122     wined3d_context_init(context_no3d, swapchain);
 123
 124     return WINED3D_OK;
 125 }
 126
 127 static BOOL fixed_get_input(BYTE usage, BYTE usage_idx, unsigned int *regnum)
 128 {
 129     if ((usage == WINED3D_DECL_USAGE_POSITION || usage == WINED3D_DECL_USAGE_POSITIONT) && !usage_idx)
 130         *regnum = WINED3D_FFP_POSITION;
 131     else if (usage == WINED3D_DECL_USAGE_BLEND_WEIGHT && !usage_idx)
 132         *regnum = WINED3D_FFP_BLENDWEIGHT;
 133     else if (usage == WINED3D_DECL_USAGE_BLEND_INDICES && !usage_idx)
 134         *regnum = WINED3D_FFP_BLENDINDICES;
 135     else if (usage == WINED3D_DECL_USAGE_NORMAL && !usage_idx)
 136         *regnum = WINED3D_FFP_NORMAL;
 137     else if (usage == WINED3D_DECL_USAGE_PSIZE && !usage_idx)
 138         *regnum = WINED3D_FFP_PSIZE;
 139     else if (usage == WINED3D_DECL_USAGE_COLOR && !usage_idx)
 140         *regnum = WINED3D_FFP_DIFFUSE;
 141     else if (usage == WINED3D_DECL_USAGE_COLOR && usage_idx == 1)
 142         *regnum = WINED3D_FFP_SPECULAR;
 143     else if (usage == WINED3D_DECL_USAGE_TEXCOORD && usage_idx < WINED3DDP_MAXTEXCOORD)
 144         *regnum = WINED3D_FFP_TEXCOORD0 + usage_idx;
 145     else
 146     {
 147         WARN("Unsupported input stream [usage=%s, usage_idx=%u].\n", debug_d3ddeclusage(usage), usage_idx);
 148         *regnum = ~0u;
 149         return FALSE;
 150     }
 151
 152     return TRUE;
 153 }
 154
 155 /* Context activation is done by the caller. */
 156 void wined3d_stream_info_from_declaration(struct wined3d_stream_info *stream_info,
 157         const struct wined3d_state *state, const struct wined3d_d3d_info *d3d_info)
 158 {
 159     /* We need to deal with frequency data! */
 160     struct wined3d_vertex_declaration *declaration = state->vertex_declaration;
 161     BOOL generic_attributes = d3d_info->ffp_generic_attributes;
 162     BOOL use_vshader = use_vs(state);
 163     unsigned int i;
 164
 165     stream_info->use_map = 0;
 166     stream_info->swizzle_map = 0;
 167     stream_info->position_transformed = 0;
 168
 169     if (!declaration)
 170         return;
 171
 172     stream_info->position_transformed = declaration->position_transformed;
 173
 174     /* Translate the declaration into strided data. */
 175     for (i = 0; i < declaration->element_count; ++i)
 176     {
 177         const struct wined3d_vertex_declaration_element *element = &declaration->elements[i];
 178         const struct wined3d_stream_state *stream = &state->streams[element->input_slot];
 179         BOOL stride_used;
 180         unsigned int idx;
 181
 182         TRACE("%p Element %p (%u of %u).\n", declaration->elements,
 183                 element, i + 1, declaration->element_count);
 184
 185         if (!stream->buffer)
 186             continue;
 187
 188         TRACE("offset %u input_slot %u usage_idx %d.\n", element->offset, element->input_slot, element->usage_idx);
 189
 190         if (use_vshader)
 191         {
 192             if (element->output_slot == WINED3D_OUTPUT_SLOT_UNUSED)
 193             {
 194                 stride_used = FALSE;
 195             }
 196             else if (element->output_slot == WINED3D_OUTPUT_SLOT_SEMANTIC)
 197             {
 198                 /* TODO: Assuming vertexdeclarations are usually used with the
 199                  * same or a similar shader, it might be worth it to store the
 200                  * last used output slot and try that one first. */
 201                 stride_used = vshader_get_input(state->shader[WINED3D_SHADER_TYPE_VERTEX],
 202                         element->usage, element->usage_idx, &idx);
 203             }
 204             else
 205             {
 206                 idx = element->output_slot;
 207                 stride_used = TRUE;
 208             }
 209         }
 210         else
 211         {
 212             if (!generic_attributes && !element->ffp_valid)
 213             {
 214                 WARN("Skipping unsupported fixed function element of format %s and usage %s.\n",
 215                         debug_d3dformat(element->format->id), debug_d3ddeclusage(element->usage));
 216                 stride_used = FALSE;
 217             }
 218             else
 219             {
 220                 stride_used = fixed_get_input(element->usage, element->usage_idx, &idx);
 221             }
 222         }
 223
 224         if (stride_used)
 225         {
 226             TRACE("Load %s array %u [usage %s, usage_idx %u, "
 227                     "input_slot %u, offset %u, stride %u, format %s, class %s, step_rate %u].\n",
 228                     use_vshader ? "shader": "fixed function", idx,
 229                     debug_d3ddeclusage(element->usage), element->usage_idx, element->input_slot,
 230                     element->offset, stream->stride, debug_d3dformat(element->format->id),
 231                     debug_d3dinput_classification(element->input_slot_class), element->instance_data_step_rate);
 232
 233             stream_info->elements[idx].format = element->format;
 234             stream_info->elements[idx].data.buffer_object = 0;
 235             stream_info->elements[idx].data.addr = (BYTE *)NULL + stream->offset + element->offset;
 236             stream_info->elements[idx].stride = stream->stride;
 237             stream_info->elements[idx].stream_idx = element->input_slot;
 238             if (stream->flags & WINED3DSTREAMSOURCE_INSTANCEDATA)
 239             {
 240                 stream_info->elements[idx].divisor = 1;
 241                 stream_info->elements[idx].instanced = true;
 242             }
 243             else if (element->input_slot_class == WINED3D_INPUT_PER_INSTANCE_DATA)
 244             {
 245                 stream_info->elements[idx].divisor = element->instance_data_step_rate;
 246                 stream_info->elements[idx].instanced = true;
 247             }
 248             else
 249             {
 250                 stream_info->elements[idx].divisor = 0;
 251                 stream_info->elements[idx].instanced = false;
 252             }
 253
 254             if (!d3d_info->vertex_bgra && element->format->id == WINED3DFMT_B8G8R8A8_UNORM)
 255             {
 256                 stream_info->swizzle_map |= 1u << idx;
 257             }
 258             stream_info->use_map |= 1u << idx;
 259         }
 260     }
 261 }
 262
 263 /* Context activation is done by the caller. */
 264 void context_update_stream_info(struct wined3d_context *context, const struct wined3d_state *state)
 265 {
 266     struct wined3d_stream_info *stream_info = &context->stream_info;
 267     const struct wined3d_d3d_info *d3d_info = context->d3d_info;
 268     DWORD prev_all_vbo = stream_info->all_vbo;
 269     unsigned int i;
 270     WORD map;
 271
 272     wined3d_stream_info_from_declaration(stream_info, state, d3d_info);
 273
 274     stream_info->all_vbo = 1;
 275     for (i = 0, map = stream_info->use_map; map; map >>= 1, ++i)
 276     {
 277         struct wined3d_stream_info_element *element;
 278         struct wined3d_bo_address data;
 279         struct wined3d_buffer *buffer;
 280
 281         if (!(map & 1))
 282             continue;
 283
 284         element = &stream_info->elements[i];
 285         buffer = state->streams[element->stream_idx].buffer;
 286
 287         /* We can't use VBOs if the base vertex index is negative. OpenGL
 288          * doesn't accept negative offsets (or rather offsets bigger than the
 289          * VBO, because the pointer is unsigned), so use system memory
 290          * sources. In most sane cases the pointer - offset will still be > 0,
 291          * otherwise it will wrap around to some big value. Hope that with the
 292          * indices the driver wraps it back internally. If not,
 293          * draw_primitive_immediate_mode() is needed, including a vertex buffer
 294          * path. */
 295         if (state->load_base_vertex_index < 0)
 296         {
 297             WARN_(d3d_perf)("load_base_vertex_index is < 0 (%d), not using VBOs.\n",
 298                     state->load_base_vertex_index);
 299             element->data.buffer_object = 0;
 300             element->data.addr += (ULONG_PTR)wined3d_buffer_load_sysmem(buffer, context);
 301             if ((UINT_PTR)element->data.addr < -state->load_base_vertex_index * element->stride)
 302                 FIXME("System memory vertex data load offset is negative!\n");
 303         }
 304         else
 305         {
 306             wined3d_buffer_load(buffer, context, state);
 307             wined3d_buffer_get_memory(buffer, &data, buffer->locations);
 308             element->data.buffer_object = data.buffer_object;
 309             element->data.addr += (ULONG_PTR)data.addr;
 310         }
 311
 312         if (!element->data.buffer_object)
 313             stream_info->all_vbo = 0;
 314
 315         TRACE("Load array %u %s.\n", i, debug_bo_address(&element->data));
 316     }
 317
 318     if (prev_all_vbo != stream_info->all_vbo)
 319         context_invalidate_state(context, STATE_INDEXBUFFER);
 320
 321     context->use_immediate_mode_draw = FALSE;
 322
 323     if (stream_info->all_vbo)
 324         return;
 325
 326     if (!use_vs(state))
 327     {
 328         WORD slow_mask = -!d3d_info->ffp_generic_attributes & (1u << WINED3D_FFP_PSIZE);
 329         slow_mask |= -(!d3d_info->vertex_bgra && !d3d_info->ffp_generic_attributes)
 330                 & ((1u << WINED3D_FFP_DIFFUSE) | (1u << WINED3D_FFP_SPECULAR) | (1u << WINED3D_FFP_BLENDWEIGHT));
 331
 332         if ((stream_info->position_transformed && !d3d_info->xyzrhw)
 333                 || (stream_info->use_map & slow_mask))
 334             context->use_immediate_mode_draw = TRUE;
 335     }
 336 }
 337
 338 static bool is_resource_rtv_bound(const struct wined3d_state *state,
 339         const struct wined3d_resource *resource)
 340 {
 341     unsigned int i;
 342
 343     if (!resource->rtv_bind_count_device)
 344         return false;
 345
 346     for (i = 0; i < ARRAY_SIZE(state->fb.render_targets); ++i)
 347     {
 348         if (state->fb.render_targets[i] && state->fb.render_targets[i]->resource == resource)
 349             return true;
 350     }
 351
 352     return false;
 353 }
 354
 355 /* Context activation is done by the caller. */
 356 static void context_preload_texture(struct wined3d_context *context,
 357         const struct wined3d_state *state, unsigned int idx)
 358 {
 359     struct wined3d_texture *texture;
 360
 361     if (!(texture = state->textures[idx]))
 362         return;
 363
 364     if (is_resource_rtv_bound(state, &texture->resource)
 365             || (state->fb.depth_stencil && state->fb.depth_stencil->resource == &texture->resource))
 366         context->uses_fbo_attached_resources = 1;
 367
 368     wined3d_texture_load(texture, context, is_srgb_enabled(state->sampler_states[idx]));
 369 }
 370
 371 /* Context activation is done by the caller. */
 372 void context_preload_textures(struct wined3d_context *context, const struct wined3d_state *state)
 373 {
 374     unsigned int i;
 375
 376     if (use_vs(state))
 377     {
 378         for (i = 0; i < WINED3D_MAX_VERTEX_SAMPLERS; ++i)
 379         {
 380             if (state->shader[WINED3D_SHADER_TYPE_VERTEX]->reg_maps.resource_info[i].type)
 381                 context_preload_texture(context, state, WINED3D_MAX_FRAGMENT_SAMPLERS + i);
 382         }
 383     }
 384
 385     if (use_ps(state))
 386     {
 387         for (i = 0; i < WINED3D_MAX_FRAGMENT_SAMPLERS; ++i)
 388         {
 389             if (state->shader[WINED3D_SHADER_TYPE_PIXEL]->reg_maps.resource_info[i].type)
 390                 context_preload_texture(context, state, i);
 391         }
 392     }
 393     else
 394     {
 395         WORD ffu_map = context->fixed_function_usage_map;
 396
 397         for (i = 0; ffu_map; ffu_map >>= 1, ++i)
 398         {
 399             if (ffu_map & 1)
 400                 context_preload_texture(context, state, i);
 401         }
 402     }
 403 }