Sweden-Number/dlls/wined3d/device.c

5821 lines
202 KiB
C

/*
* Copyright 2002 Lionel Ulmer
* Copyright 2002-2005 Jason Edmeades
* Copyright 2003-2004 Raphael Junqueira
* Copyright 2004 Christian Costa
* Copyright 2005 Oliver Stieber
* Copyright 2006-2008 Stefan Dösinger for CodeWeavers
* Copyright 2006-2008 Henri Verbeet
* Copyright 2007 Andrew Riedi
* Copyright 2009-2011 Henri Verbeet for CodeWeavers
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
*/
#include "config.h"
#include "wine/port.h"
#include <stdio.h>
#ifdef HAVE_FLOAT_H
# include <float.h>
#endif
#include "wined3d_private.h"
WINE_DEFAULT_DEBUG_CHANNEL(d3d);
/* Define the default light parameters as specified by MSDN. */
const struct wined3d_light WINED3D_default_light =
{
WINED3D_LIGHT_DIRECTIONAL, /* Type */
{ 1.0f, 1.0f, 1.0f, 0.0f }, /* Diffuse r,g,b,a */
{ 0.0f, 0.0f, 0.0f, 0.0f }, /* Specular r,g,b,a */
{ 0.0f, 0.0f, 0.0f, 0.0f }, /* Ambient r,g,b,a, */
{ 0.0f, 0.0f, 0.0f }, /* Position x,y,z */
{ 0.0f, 0.0f, 1.0f }, /* Direction x,y,z */
0.0f, /* Range */
0.0f, /* Falloff */
0.0f, 0.0f, 0.0f, /* Attenuation 0,1,2 */
0.0f, /* Theta */
0.0f /* Phi */
};
/**********************************************************
* Global variable / Constants follow
**********************************************************/
const struct wined3d_matrix identity =
{{{
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f,
}}}; /* When needed for comparisons */
/* Note that except for WINED3DPT_POINTLIST and WINED3DPT_LINELIST these
* actually have the same values in GL and D3D. */
static GLenum gl_primitive_type_from_d3d(enum wined3d_primitive_type primitive_type)
{
switch(primitive_type)
{
case WINED3D_PT_POINTLIST:
return GL_POINTS;
case WINED3D_PT_LINELIST:
return GL_LINES;
case WINED3D_PT_LINESTRIP:
return GL_LINE_STRIP;
case WINED3D_PT_TRIANGLELIST:
return GL_TRIANGLES;
case WINED3D_PT_TRIANGLESTRIP:
return GL_TRIANGLE_STRIP;
case WINED3D_PT_TRIANGLEFAN:
return GL_TRIANGLE_FAN;
case WINED3D_PT_LINELIST_ADJ:
return GL_LINES_ADJACENCY_ARB;
case WINED3D_PT_LINESTRIP_ADJ:
return GL_LINE_STRIP_ADJACENCY_ARB;
case WINED3D_PT_TRIANGLELIST_ADJ:
return GL_TRIANGLES_ADJACENCY_ARB;
case WINED3D_PT_TRIANGLESTRIP_ADJ:
return GL_TRIANGLE_STRIP_ADJACENCY_ARB;
default:
FIXME("Unhandled primitive type %s\n", debug_d3dprimitivetype(primitive_type));
return GL_NONE;
}
}
static enum wined3d_primitive_type d3d_primitive_type_from_gl(GLenum primitive_type)
{
switch(primitive_type)
{
case GL_POINTS:
return WINED3D_PT_POINTLIST;
case GL_LINES:
return WINED3D_PT_LINELIST;
case GL_LINE_STRIP:
return WINED3D_PT_LINESTRIP;
case GL_TRIANGLES:
return WINED3D_PT_TRIANGLELIST;
case GL_TRIANGLE_STRIP:
return WINED3D_PT_TRIANGLESTRIP;
case GL_TRIANGLE_FAN:
return WINED3D_PT_TRIANGLEFAN;
case GL_LINES_ADJACENCY_ARB:
return WINED3D_PT_LINELIST_ADJ;
case GL_LINE_STRIP_ADJACENCY_ARB:
return WINED3D_PT_LINESTRIP_ADJ;
case GL_TRIANGLES_ADJACENCY_ARB:
return WINED3D_PT_TRIANGLELIST_ADJ;
case GL_TRIANGLE_STRIP_ADJACENCY_ARB:
return WINED3D_PT_TRIANGLESTRIP_ADJ;
default:
FIXME("Unhandled primitive type %s\n", debug_d3dprimitivetype(primitive_type));
return WINED3D_PT_UNDEFINED;
}
}
static BOOL fixed_get_input(BYTE usage, BYTE usage_idx, unsigned int *regnum)
{
if ((usage == WINED3D_DECL_USAGE_POSITION || usage == WINED3D_DECL_USAGE_POSITIONT) && !usage_idx)
*regnum = WINED3D_FFP_POSITION;
else if (usage == WINED3D_DECL_USAGE_BLEND_WEIGHT && !usage_idx)
*regnum = WINED3D_FFP_BLENDWEIGHT;
else if (usage == WINED3D_DECL_USAGE_BLEND_INDICES && !usage_idx)
*regnum = WINED3D_FFP_BLENDINDICES;
else if (usage == WINED3D_DECL_USAGE_NORMAL && !usage_idx)
*regnum = WINED3D_FFP_NORMAL;
else if (usage == WINED3D_DECL_USAGE_PSIZE && !usage_idx)
*regnum = WINED3D_FFP_PSIZE;
else if (usage == WINED3D_DECL_USAGE_COLOR && !usage_idx)
*regnum = WINED3D_FFP_DIFFUSE;
else if (usage == WINED3D_DECL_USAGE_COLOR && usage_idx == 1)
*regnum = WINED3D_FFP_SPECULAR;
else if (usage == WINED3D_DECL_USAGE_TEXCOORD && usage_idx < WINED3DDP_MAXTEXCOORD)
*regnum = WINED3D_FFP_TEXCOORD0 + usage_idx;
else
{
FIXME("Unsupported input stream [usage=%s, usage_idx=%u]\n", debug_d3ddeclusage(usage), usage_idx);
*regnum = ~0U;
return FALSE;
}
return TRUE;
}
/* Context activation is done by the caller. */
void device_stream_info_from_declaration(struct wined3d_device *device, struct wined3d_stream_info *stream_info)
{
const struct wined3d_state *state = &device->stateBlock->state;
/* We need to deal with frequency data! */
struct wined3d_vertex_declaration *declaration = state->vertex_declaration;
BOOL use_vshader;
unsigned int i;
stream_info->use_map = 0;
stream_info->swizzle_map = 0;
/* Check for transformed vertices, disable vertex shader if present. */
stream_info->position_transformed = declaration->position_transformed;
use_vshader = state->vertex_shader && !declaration->position_transformed;
/* Translate the declaration into strided data. */
for (i = 0; i < declaration->element_count; ++i)
{
const struct wined3d_vertex_declaration_element *element = &declaration->elements[i];
const struct wined3d_stream_state *stream = &state->streams[element->input_slot];
struct wined3d_buffer *buffer = stream->buffer;
struct wined3d_bo_address data;
BOOL stride_used;
unsigned int idx;
DWORD stride;
TRACE("%p Element %p (%u of %u)\n", declaration->elements,
element, i + 1, declaration->element_count);
if (!buffer) continue;
data.buffer_object = 0;
data.addr = NULL;
stride = stream->stride;
if (state->user_stream)
{
TRACE("Stream %u is UP, %p\n", element->input_slot, buffer);
data.buffer_object = 0;
data.addr = (BYTE *)buffer;
}
else
{
TRACE("Stream %u isn't UP, %p\n", element->input_slot, buffer);
buffer_get_memory(buffer, &device->adapter->gl_info, &data);
/* Can't use vbo's if the base vertex index is negative. OpenGL doesn't accept negative offsets
* (or rather offsets bigger than the vbo, because the pointer is unsigned), so use system memory
* sources. In most sane cases the pointer - offset will still be > 0, otherwise it will wrap
* around to some big value. Hope that with the indices, the driver wraps it back internally. If
* not, drawStridedSlow is needed, including a vertex buffer path. */
if (state->load_base_vertex_index < 0)
{
WARN("load_base_vertex_index is < 0 (%d), not using VBOs.\n",
state->load_base_vertex_index);
data.buffer_object = 0;
data.addr = buffer_get_sysmem(buffer, &device->adapter->gl_info);
if ((UINT_PTR)data.addr < -state->load_base_vertex_index * stride)
{
FIXME("System memory vertex data load offset is negative!\n");
}
}
}
data.addr += element->offset;
TRACE("offset %u input_slot %u usage_idx %d\n", element->offset, element->input_slot, element->usage_idx);
if (use_vshader)
{
if (element->output_slot == ~0U)
{
/* TODO: Assuming vertexdeclarations are usually used with the
* same or a similar shader, it might be worth it to store the
* last used output slot and try that one first. */
stride_used = vshader_get_input(state->vertex_shader,
element->usage, element->usage_idx, &idx);
}
else
{
idx = element->output_slot;
stride_used = TRUE;
}
}
else
{
if (!element->ffp_valid)
{
WARN("Skipping unsupported fixed function element of format %s and usage %s\n",
debug_d3dformat(element->format->id), debug_d3ddeclusage(element->usage));
stride_used = FALSE;
}
else
{
stride_used = fixed_get_input(element->usage, element->usage_idx, &idx);
}
}
if (stride_used)
{
TRACE("Load %s array %u [usage %s, usage_idx %u, "
"input_slot %u, offset %u, stride %u, format %s, buffer_object %u]\n",
use_vshader ? "shader": "fixed function", idx,
debug_d3ddeclusage(element->usage), element->usage_idx, element->input_slot,
element->offset, stride, debug_d3dformat(element->format->id), data.buffer_object);
data.addr += stream->offset;
stream_info->elements[idx].format = element->format;
stream_info->elements[idx].data = data;
stream_info->elements[idx].stride = stride;
stream_info->elements[idx].stream_idx = element->input_slot;
if (!device->adapter->gl_info.supported[ARB_VERTEX_ARRAY_BGRA]
&& element->format->id == WINED3DFMT_B8G8R8A8_UNORM)
{
stream_info->swizzle_map |= 1 << idx;
}
stream_info->use_map |= 1 << idx;
}
}
device->num_buffer_queries = 0;
if (!state->user_stream)
{
WORD map = stream_info->use_map;
stream_info->all_vbo = 1;
/* PreLoad all the vertex buffers. */
for (i = 0; map; map >>= 1, ++i)
{
struct wined3d_stream_info_element *element;
struct wined3d_buffer *buffer;
if (!(map & 1)) continue;
element = &stream_info->elements[i];
buffer = state->streams[element->stream_idx].buffer;
wined3d_buffer_preload(buffer);
/* If the preload dropped the buffer object, update the stream info. */
if (buffer->buffer_object != element->data.buffer_object)
{
element->data.buffer_object = 0;
element->data.addr = buffer_get_sysmem(buffer, &device->adapter->gl_info)
+ (ptrdiff_t)element->data.addr;
}
if (!buffer->buffer_object)
stream_info->all_vbo = 0;
if (buffer->query)
device->buffer_queries[device->num_buffer_queries++] = buffer->query;
}
}
else
{
stream_info->all_vbo = 0;
}
}
static void stream_info_element_from_strided(const struct wined3d_gl_info *gl_info,
const struct wined3d_strided_element *strided, struct wined3d_stream_info_element *e)
{
e->data.addr = strided->data;
e->data.buffer_object = 0;
e->format = wined3d_get_format(gl_info, strided->format);
e->stride = strided->stride;
e->stream_idx = 0;
}
static void device_stream_info_from_strided(const struct wined3d_gl_info *gl_info,
const struct wined3d_strided_data *strided, struct wined3d_stream_info *stream_info)
{
unsigned int i;
memset(stream_info, 0, sizeof(*stream_info));
if (strided->position.data)
stream_info_element_from_strided(gl_info, &strided->position, &stream_info->elements[WINED3D_FFP_POSITION]);
if (strided->normal.data)
stream_info_element_from_strided(gl_info, &strided->normal, &stream_info->elements[WINED3D_FFP_NORMAL]);
if (strided->diffuse.data)
stream_info_element_from_strided(gl_info, &strided->diffuse, &stream_info->elements[WINED3D_FFP_DIFFUSE]);
if (strided->specular.data)
stream_info_element_from_strided(gl_info, &strided->specular, &stream_info->elements[WINED3D_FFP_SPECULAR]);
for (i = 0; i < WINED3DDP_MAXTEXCOORD; ++i)
{
if (strided->tex_coords[i].data)
stream_info_element_from_strided(gl_info, &strided->tex_coords[i],
&stream_info->elements[WINED3D_FFP_TEXCOORD0 + i]);
}
stream_info->position_transformed = strided->position_transformed;
for (i = 0; i < sizeof(stream_info->elements) / sizeof(*stream_info->elements); ++i)
{
if (!stream_info->elements[i].format) continue;
if (!gl_info->supported[ARB_VERTEX_ARRAY_BGRA]
&& stream_info->elements[i].format->id == WINED3DFMT_B8G8R8A8_UNORM)
{
stream_info->swizzle_map |= 1 << i;
}
stream_info->use_map |= 1 << i;
}
}
static void device_trace_strided_stream_info(const struct wined3d_stream_info *stream_info)
{
TRACE("Strided Data:\n");
TRACE_STRIDED(stream_info, WINED3D_FFP_POSITION);
TRACE_STRIDED(stream_info, WINED3D_FFP_BLENDWEIGHT);
TRACE_STRIDED(stream_info, WINED3D_FFP_BLENDINDICES);
TRACE_STRIDED(stream_info, WINED3D_FFP_NORMAL);
TRACE_STRIDED(stream_info, WINED3D_FFP_PSIZE);
TRACE_STRIDED(stream_info, WINED3D_FFP_DIFFUSE);
TRACE_STRIDED(stream_info, WINED3D_FFP_SPECULAR);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD0);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD1);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD2);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD3);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD4);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD5);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD6);
TRACE_STRIDED(stream_info, WINED3D_FFP_TEXCOORD7);
}
/* Context activation is done by the caller. */
void device_update_stream_info(struct wined3d_device *device, const struct wined3d_gl_info *gl_info)
{
struct wined3d_stream_info *stream_info = &device->strided_streams;
const struct wined3d_state *state = &device->stateBlock->state;
DWORD prev_all_vbo = stream_info->all_vbo;
if (device->up_strided)
{
/* Note: this is a ddraw fixed-function code path. */
TRACE("=============================== Strided Input ================================\n");
device_stream_info_from_strided(gl_info, device->up_strided, stream_info);
if (TRACE_ON(d3d)) device_trace_strided_stream_info(stream_info);
}
else
{
TRACE("============================= Vertex Declaration =============================\n");
device_stream_info_from_declaration(device, stream_info);
}
if (state->vertex_shader && !stream_info->position_transformed)
{
if (state->vertex_declaration->half_float_conv_needed && !stream_info->all_vbo)
{
TRACE("Using drawStridedSlow with vertex shaders for FLOAT16 conversion.\n");
device->useDrawStridedSlow = TRUE;
}
else
{
device->useDrawStridedSlow = FALSE;
}
}
else
{
WORD slow_mask = (1 << WINED3D_FFP_PSIZE);
slow_mask |= -!gl_info->supported[ARB_VERTEX_ARRAY_BGRA]
& ((1 << WINED3D_FFP_DIFFUSE) | (1 << WINED3D_FFP_SPECULAR));
if ((stream_info->position_transformed || (stream_info->use_map & slow_mask)) && !stream_info->all_vbo)
{
device->useDrawStridedSlow = TRUE;
}
else
{
device->useDrawStridedSlow = FALSE;
}
}
if (prev_all_vbo != stream_info->all_vbo)
device_invalidate_state(device, STATE_INDEXBUFFER);
}
static void device_preload_texture(const struct wined3d_state *state, unsigned int idx)
{
struct wined3d_texture *texture;
enum WINED3DSRGB srgb;
if (!(texture = state->textures[idx])) return;
srgb = state->sampler_states[idx][WINED3D_SAMP_SRGB_TEXTURE] ? SRGB_SRGB : SRGB_RGB;
texture->texture_ops->texture_preload(texture, srgb);
}
void device_preload_textures(const struct wined3d_device *device)
{
const struct wined3d_state *state = &device->stateBlock->state;
unsigned int i;
if (use_vs(state))
{
for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i)
{
if (state->vertex_shader->reg_maps.sampler_type[i])
device_preload_texture(state, MAX_FRAGMENT_SAMPLERS + i);
}
}
if (use_ps(state))
{
for (i = 0; i < MAX_FRAGMENT_SAMPLERS; ++i)
{
if (state->pixel_shader->reg_maps.sampler_type[i])
device_preload_texture(state, i);
}
}
else
{
WORD ffu_map = device->fixed_function_usage_map;
for (i = 0; ffu_map; ffu_map >>= 1, ++i)
{
if (ffu_map & 1)
device_preload_texture(state, i);
}
}
}
BOOL device_context_add(struct wined3d_device *device, struct wined3d_context *context)
{
struct wined3d_context **new_array;
TRACE("Adding context %p.\n", context);
if (!device->contexts) new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_array));
else new_array = HeapReAlloc(GetProcessHeap(), 0, device->contexts,
sizeof(*new_array) * (device->context_count + 1));
if (!new_array)
{
ERR("Failed to grow the context array.\n");
return FALSE;
}
new_array[device->context_count++] = context;
device->contexts = new_array;
return TRUE;
}
void device_context_remove(struct wined3d_device *device, struct wined3d_context *context)
{
struct wined3d_context **new_array;
BOOL found = FALSE;
UINT i;
TRACE("Removing context %p.\n", context);
for (i = 0; i < device->context_count; ++i)
{
if (device->contexts[i] == context)
{
found = TRUE;
break;
}
}
if (!found)
{
ERR("Context %p doesn't exist in context array.\n", context);
return;
}
if (!--device->context_count)
{
HeapFree(GetProcessHeap(), 0, device->contexts);
device->contexts = NULL;
return;
}
memmove(&device->contexts[i], &device->contexts[i + 1], (device->context_count - i) * sizeof(*device->contexts));
new_array = HeapReAlloc(GetProcessHeap(), 0, device->contexts, device->context_count * sizeof(*device->contexts));
if (!new_array)
{
ERR("Failed to shrink context array. Oh well.\n");
return;
}
device->contexts = new_array;
}
/* Do not call while under the GL lock. */
void device_switch_onscreen_ds(struct wined3d_device *device,
struct wined3d_context *context, struct wined3d_surface *depth_stencil)
{
if (device->onscreen_depth_stencil)
{
surface_load_ds_location(device->onscreen_depth_stencil, context, SFLAG_INTEXTURE);
surface_modify_ds_location(device->onscreen_depth_stencil, SFLAG_INTEXTURE,
device->onscreen_depth_stencil->ds_current_size.cx,
device->onscreen_depth_stencil->ds_current_size.cy);
wined3d_surface_decref(device->onscreen_depth_stencil);
}
device->onscreen_depth_stencil = depth_stencil;
wined3d_surface_incref(device->onscreen_depth_stencil);
}
static BOOL is_full_clear(const struct wined3d_surface *target, const RECT *draw_rect, const RECT *clear_rect)
{
/* partial draw rect */
if (draw_rect->left || draw_rect->top
|| draw_rect->right < target->resource.width
|| draw_rect->bottom < target->resource.height)
return FALSE;
/* partial clear rect */
if (clear_rect && (clear_rect->left > 0 || clear_rect->top > 0
|| clear_rect->right < target->resource.width
|| clear_rect->bottom < target->resource.height))
return FALSE;
return TRUE;
}
static void prepare_ds_clear(struct wined3d_surface *ds, struct wined3d_context *context,
DWORD location, const RECT *draw_rect, UINT rect_count, const RECT *clear_rect, RECT *out_rect)
{
RECT current_rect, r;
if (ds->flags & location)
SetRect(&current_rect, 0, 0,
ds->ds_current_size.cx,
ds->ds_current_size.cy);
else
SetRectEmpty(&current_rect);
IntersectRect(&r, draw_rect, &current_rect);
if (EqualRect(&r, draw_rect))
{
/* current_rect ⊇ draw_rect, modify only. */
SetRect(out_rect, 0, 0, ds->ds_current_size.cx, ds->ds_current_size.cy);
return;
}
if (EqualRect(&r, &current_rect))
{
/* draw_rect ⊇ current_rect, test if we're doing a full clear. */
if (!clear_rect)
{
/* Full clear, modify only. */
*out_rect = *draw_rect;
return;
}
IntersectRect(&r, draw_rect, clear_rect);
if (EqualRect(&r, draw_rect))
{
/* clear_rect ⊇ draw_rect, modify only. */
*out_rect = *draw_rect;
return;
}
}
/* Full load. */
surface_load_ds_location(ds, context, location);
SetRect(out_rect, 0, 0, ds->ds_current_size.cx, ds->ds_current_size.cy);
}
/* Do not call while under the GL lock. */
void device_clear_render_targets(struct wined3d_device *device, UINT rt_count, const struct wined3d_fb_state *fb,
UINT rect_count, const RECT *rects, const RECT *draw_rect, DWORD flags, const struct wined3d_color *color,
float depth, DWORD stencil)
{
const RECT *clear_rect = (rect_count > 0 && rects) ? (const RECT *)rects : NULL;
struct wined3d_surface *target = rt_count ? fb->render_targets[0] : NULL;
UINT drawable_width, drawable_height;
struct wined3d_context *context;
GLbitfield clear_mask = 0;
BOOL render_offscreen;
unsigned int i;
RECT ds_rect;
/* When we're clearing parts of the drawable, make sure that the target surface is well up to date in the
* drawable. After the clear we'll mark the drawable up to date, so we have to make sure that this is true
* for the cleared parts, and the untouched parts.
*
* If we're clearing the whole target there is no need to copy it into the drawable, it will be overwritten
* anyway. If we're not clearing the color buffer we don't have to copy either since we're not going to set
* the drawable up to date. We have to check all settings that limit the clear area though. Do not bother
* checking all this if the dest surface is in the drawable anyway. */
if (flags & WINED3DCLEAR_TARGET && !is_full_clear(target, draw_rect, clear_rect))
{
for (i = 0; i < rt_count; ++i)
{
struct wined3d_surface *rt = fb->render_targets[i];
if (rt)
surface_load_location(rt, rt->draw_binding, NULL);
}
}
context = context_acquire(device, target);
if (!context->valid)
{
context_release(context);
WARN("Invalid context, skipping clear.\n");
return;
}
if (target)
{
render_offscreen = context->render_offscreen;
target->get_drawable_size(context, &drawable_width, &drawable_height);
}
else
{
render_offscreen = TRUE;
drawable_width = fb->depth_stencil->pow2Width;
drawable_height = fb->depth_stencil->pow2Height;
}
if (flags & WINED3DCLEAR_ZBUFFER)
{
DWORD location = render_offscreen ? fb->depth_stencil->draw_binding : SFLAG_INDRAWABLE;
if (!render_offscreen && fb->depth_stencil != device->onscreen_depth_stencil)
device_switch_onscreen_ds(device, context, fb->depth_stencil);
prepare_ds_clear(fb->depth_stencil, context, location,
draw_rect, rect_count, clear_rect, &ds_rect);
}
if (!context_apply_clear_state(context, device, rt_count, fb))
{
context_release(context);
WARN("Failed to apply clear state, skipping clear.\n");
return;
}
ENTER_GL();
/* Only set the values up once, as they are not changing. */
if (flags & WINED3DCLEAR_STENCIL)
{
if (context->gl_info->supported[EXT_STENCIL_TWO_SIDE])
{
glDisable(GL_STENCIL_TEST_TWO_SIDE_EXT);
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_TWOSIDEDSTENCILMODE));
}
glStencilMask(~0U);
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_STENCILWRITEMASK));
glClearStencil(stencil);
checkGLcall("glClearStencil");
clear_mask = clear_mask | GL_STENCIL_BUFFER_BIT;
}
if (flags & WINED3DCLEAR_ZBUFFER)
{
DWORD location = render_offscreen ? fb->depth_stencil->draw_binding : SFLAG_INDRAWABLE;
surface_modify_ds_location(fb->depth_stencil, location, ds_rect.right, ds_rect.bottom);
glDepthMask(GL_TRUE);
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_ZWRITEENABLE));
glClearDepth(depth);
checkGLcall("glClearDepth");
clear_mask = clear_mask | GL_DEPTH_BUFFER_BIT;
}
if (flags & WINED3DCLEAR_TARGET)
{
for (i = 0; i < rt_count; ++i)
{
struct wined3d_surface *rt = fb->render_targets[i];
if (rt)
surface_modify_location(rt, rt->draw_binding, TRUE);
}
glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_COLORWRITEENABLE));
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_COLORWRITEENABLE1));
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_COLORWRITEENABLE2));
context_invalidate_state(context, STATE_RENDER(WINED3D_RS_COLORWRITEENABLE3));
glClearColor(color->r, color->g, color->b, color->a);
checkGLcall("glClearColor");
clear_mask = clear_mask | GL_COLOR_BUFFER_BIT;
}
if (!clear_rect)
{
if (render_offscreen)
{
glScissor(draw_rect->left, draw_rect->top,
draw_rect->right - draw_rect->left, draw_rect->bottom - draw_rect->top);
}
else
{
glScissor(draw_rect->left, drawable_height - draw_rect->bottom,
draw_rect->right - draw_rect->left, draw_rect->bottom - draw_rect->top);
}
checkGLcall("glScissor");
glClear(clear_mask);
checkGLcall("glClear");
}
else
{
RECT current_rect;
/* Now process each rect in turn. */
for (i = 0; i < rect_count; ++i)
{
/* Note that GL uses lower left, width/height. */
IntersectRect(&current_rect, draw_rect, &clear_rect[i]);
TRACE("clear_rect[%u] %s, current_rect %s.\n", i,
wine_dbgstr_rect(&clear_rect[i]),
wine_dbgstr_rect(&current_rect));
/* Tests show that rectangles where x1 > x2 or y1 > y2 are ignored silently.
* The rectangle is not cleared, no error is returned, but further rectangles are
* still cleared if they are valid. */
if (current_rect.left > current_rect.right || current_rect.top > current_rect.bottom)
{
TRACE("Rectangle with negative dimensions, ignoring.\n");
continue;
}
if (render_offscreen)
{
glScissor(current_rect.left, current_rect.top,
current_rect.right - current_rect.left, current_rect.bottom - current_rect.top);
}
else
{
glScissor(current_rect.left, drawable_height - current_rect.bottom,
current_rect.right - current_rect.left, current_rect.bottom - current_rect.top);
}
checkGLcall("glScissor");
glClear(clear_mask);
checkGLcall("glClear");
}
}
LEAVE_GL();
if (wined3d_settings.strict_draw_ordering || (flags & WINED3DCLEAR_TARGET
&& target->container.type == WINED3D_CONTAINER_SWAPCHAIN
&& target->container.u.swapchain->front_buffer == target))
wglFlush(); /* Flush to ensure ordering across contexts. */
context_release(context);
}
ULONG CDECL wined3d_device_incref(struct wined3d_device *device)
{
ULONG refcount = InterlockedIncrement(&device->ref);
TRACE("%p increasing refcount to %u.\n", device, refcount);
return refcount;
}
ULONG CDECL wined3d_device_decref(struct wined3d_device *device)
{
ULONG refcount = InterlockedDecrement(&device->ref);
TRACE("%p decreasing refcount to %u.\n", device, refcount);
if (!refcount)
{
struct wined3d_stateblock *stateblock;
UINT i;
if (wined3d_stateblock_decref(device->updateStateBlock)
&& device->updateStateBlock != device->stateBlock)
FIXME("Something's still holding the update stateblock.\n");
device->updateStateBlock = NULL;
stateblock = device->stateBlock;
device->stateBlock = NULL;
if (wined3d_stateblock_decref(stateblock))
FIXME("Something's still holding the stateblock.\n");
for (i = 0; i < sizeof(device->multistate_funcs) / sizeof(device->multistate_funcs[0]); ++i)
{
HeapFree(GetProcessHeap(), 0, device->multistate_funcs[i]);
device->multistate_funcs[i] = NULL;
}
if (!list_empty(&device->resources))
{
struct wined3d_resource *resource;
FIXME("Device released with resources still bound, acceptable but unexpected.\n");
LIST_FOR_EACH_ENTRY(resource, &device->resources, struct wined3d_resource, resource_list_entry)
{
FIXME("Leftover resource %p with type %s (%#x).\n",
resource, debug_d3dresourcetype(resource->type), resource->type);
}
}
if (device->contexts)
ERR("Context array not freed!\n");
if (device->hardwareCursor)
DestroyCursor(device->hardwareCursor);
device->hardwareCursor = 0;
wined3d_decref(device->wined3d);
device->wined3d = NULL;
HeapFree(GetProcessHeap(), 0, device);
TRACE("Freed device %p.\n", device);
}
return refcount;
}
UINT CDECL wined3d_device_get_swapchain_count(const struct wined3d_device *device)
{
TRACE("device %p.\n", device);
return device->swapchain_count;
}
HRESULT CDECL wined3d_device_get_swapchain(const struct wined3d_device *device,
UINT swapchain_idx, struct wined3d_swapchain **swapchain)
{
TRACE("device %p, swapchain_idx %u, swapchain %p.\n",
device, swapchain_idx, swapchain);
if (swapchain_idx >= device->swapchain_count)
{
WARN("swapchain_idx %u >= swapchain_count %u.\n",
swapchain_idx, device->swapchain_count);
*swapchain = NULL;
return WINED3DERR_INVALIDCALL;
}
*swapchain = device->swapchains[swapchain_idx];
wined3d_swapchain_incref(*swapchain);
TRACE("Returning %p.\n", *swapchain);
return WINED3D_OK;
}
static void device_load_logo(struct wined3d_device *device, const char *filename)
{
struct wined3d_color_key color_key;
HBITMAP hbm;
BITMAP bm;
HRESULT hr;
HDC dcb = NULL, dcs = NULL;
hbm = LoadImageA(NULL, filename, IMAGE_BITMAP, 0, 0, LR_LOADFROMFILE | LR_CREATEDIBSECTION);
if(hbm)
{
GetObjectA(hbm, sizeof(BITMAP), &bm);
dcb = CreateCompatibleDC(NULL);
if(!dcb) goto out;
SelectObject(dcb, hbm);
}
else
{
/* Create a 32x32 white surface to indicate that wined3d is used, but the specified image
* couldn't be loaded
*/
memset(&bm, 0, sizeof(bm));
bm.bmWidth = 32;
bm.bmHeight = 32;
}
hr = wined3d_surface_create(device, bm.bmWidth, bm.bmHeight, WINED3DFMT_B5G6R5_UNORM, 0, 0,
WINED3D_POOL_SYSTEM_MEM, WINED3D_MULTISAMPLE_NONE, 0, WINED3D_SURFACE_TYPE_OPENGL, WINED3D_SURFACE_MAPPABLE,
NULL, &wined3d_null_parent_ops, &device->logo_surface);
if (FAILED(hr))
{
ERR("Wine logo requested, but failed to create surface, hr %#x.\n", hr);
goto out;
}
if (dcb)
{
if (FAILED(hr = wined3d_surface_getdc(device->logo_surface, &dcs)))
goto out;
BitBlt(dcs, 0, 0, bm.bmWidth, bm.bmHeight, dcb, 0, 0, SRCCOPY);
wined3d_surface_releasedc(device->logo_surface, dcs);
color_key.color_space_low_value = 0;
color_key.color_space_high_value = 0;
wined3d_surface_set_color_key(device->logo_surface, WINEDDCKEY_SRCBLT, &color_key);
}
else
{
const struct wined3d_color c = {1.0f, 1.0f, 1.0f, 1.0f};
/* Fill the surface with a white color to show that wined3d is there */
wined3d_device_color_fill(device, device->logo_surface, NULL, &c);
}
out:
if (dcb) DeleteDC(dcb);
if (hbm) DeleteObject(hbm);
}
/* Context activation is done by the caller. */
static void create_dummy_textures(struct wined3d_device *device, struct wined3d_context *context)
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
unsigned int i, j, count;
/* Under DirectX you can sample even if no texture is bound, whereas
* OpenGL will only allow that when a valid texture is bound.
* We emulate this by creating dummy textures and binding them
* to each texture stage when the currently set D3D texture is NULL. */
ENTER_GL();
if (gl_info->supported[APPLE_CLIENT_STORAGE])
{
/* The dummy texture does not have client storage backing */
glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_FALSE);
checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_FALSE)");
}
count = min(MAX_COMBINED_SAMPLERS, gl_info->limits.combined_samplers);
for (i = 0; i < count; ++i)
{
DWORD color = 0x000000ff;
/* Make appropriate texture active */
context_active_texture(context, gl_info, i);
glGenTextures(1, &device->dummy_texture_2d[i]);
checkGLcall("glGenTextures");
TRACE("Dummy 2D texture %u given name %u.\n", i, device->dummy_texture_2d[i]);
glBindTexture(GL_TEXTURE_2D, device->dummy_texture_2d[i]);
checkGLcall("glBindTexture");
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, 1, 1, 0, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, &color);
checkGLcall("glTexImage2D");
if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
{
glGenTextures(1, &device->dummy_texture_rect[i]);
checkGLcall("glGenTextures");
TRACE("Dummy rectangle texture %u given name %u.\n", i, device->dummy_texture_rect[i]);
glBindTexture(GL_TEXTURE_RECTANGLE_ARB, device->dummy_texture_rect[i]);
checkGLcall("glBindTexture");
glTexImage2D(GL_TEXTURE_RECTANGLE_ARB, 0, GL_RGBA8, 1, 1, 0, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, &color);
checkGLcall("glTexImage2D");
}
if (gl_info->supported[EXT_TEXTURE3D])
{
glGenTextures(1, &device->dummy_texture_3d[i]);
checkGLcall("glGenTextures");
TRACE("Dummy 3D texture %u given name %u.\n", i, device->dummy_texture_3d[i]);
glBindTexture(GL_TEXTURE_3D, device->dummy_texture_3d[i]);
checkGLcall("glBindTexture");
GL_EXTCALL(glTexImage3DEXT(GL_TEXTURE_3D, 0, GL_RGBA8, 1, 1, 1, 0, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, &color));
checkGLcall("glTexImage3D");
}
if (gl_info->supported[ARB_TEXTURE_CUBE_MAP])
{
glGenTextures(1, &device->dummy_texture_cube[i]);
checkGLcall("glGenTextures");
TRACE("Dummy cube texture %u given name %u.\n", i, device->dummy_texture_cube[i]);
glBindTexture(GL_TEXTURE_CUBE_MAP, device->dummy_texture_cube[i]);
checkGLcall("glBindTexture");
for (j = GL_TEXTURE_CUBE_MAP_POSITIVE_X; j <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z; ++j)
{
glTexImage2D(j, 0, GL_RGBA8, 1, 1, 0, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, &color);
checkGLcall("glTexImage2D");
}
}
}
if (gl_info->supported[APPLE_CLIENT_STORAGE])
{
/* Re-enable because if supported it is enabled by default */
glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE);
checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE)");
}
LEAVE_GL();
}
/* Context activation is done by the caller. */
static void destroy_dummy_textures(struct wined3d_device *device, const struct wined3d_gl_info *gl_info)
{
unsigned int count = min(MAX_COMBINED_SAMPLERS, gl_info->limits.combined_samplers);
ENTER_GL();
if (gl_info->supported[ARB_TEXTURE_CUBE_MAP])
{
glDeleteTextures(count, device->dummy_texture_cube);
checkGLcall("glDeleteTextures(count, device->dummy_texture_cube)");
}
if (gl_info->supported[EXT_TEXTURE3D])
{
glDeleteTextures(count, device->dummy_texture_3d);
checkGLcall("glDeleteTextures(count, device->dummy_texture_3d)");
}
if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
{
glDeleteTextures(count, device->dummy_texture_rect);
checkGLcall("glDeleteTextures(count, device->dummy_texture_rect)");
}
glDeleteTextures(count, device->dummy_texture_2d);
checkGLcall("glDeleteTextures(count, device->dummy_texture_2d)");
LEAVE_GL();
memset(device->dummy_texture_cube, 0, gl_info->limits.textures * sizeof(*device->dummy_texture_cube));
memset(device->dummy_texture_3d, 0, gl_info->limits.textures * sizeof(*device->dummy_texture_3d));
memset(device->dummy_texture_rect, 0, gl_info->limits.textures * sizeof(*device->dummy_texture_rect));
memset(device->dummy_texture_2d, 0, gl_info->limits.textures * sizeof(*device->dummy_texture_2d));
}
static LONG fullscreen_style(LONG style)
{
/* Make sure the window is managed, otherwise we won't get keyboard input. */
style |= WS_POPUP | WS_SYSMENU;
style &= ~(WS_CAPTION | WS_THICKFRAME);
return style;
}
static LONG fullscreen_exstyle(LONG exstyle)
{
/* Filter out window decorations. */
exstyle &= ~(WS_EX_WINDOWEDGE | WS_EX_CLIENTEDGE);
return exstyle;
}
void CDECL wined3d_device_setup_fullscreen_window(struct wined3d_device *device, HWND window, UINT w, UINT h)
{
BOOL filter_messages;
LONG style, exstyle;
TRACE("Setting up window %p for fullscreen mode.\n", window);
if (device->style || device->exStyle)
{
ERR("Changing the window style for window %p, but another style (%08x, %08x) is already stored.\n",
window, device->style, device->exStyle);
}
device->style = GetWindowLongW(window, GWL_STYLE);
device->exStyle = GetWindowLongW(window, GWL_EXSTYLE);
style = fullscreen_style(device->style);
exstyle = fullscreen_exstyle(device->exStyle);
TRACE("Old style was %08x, %08x, setting to %08x, %08x.\n",
device->style, device->exStyle, style, exstyle);
filter_messages = device->filter_messages;
device->filter_messages = TRUE;
SetWindowLongW(window, GWL_STYLE, style);
SetWindowLongW(window, GWL_EXSTYLE, exstyle);
SetWindowPos(window, HWND_TOP, 0, 0, w, h, SWP_FRAMECHANGED | SWP_SHOWWINDOW | SWP_NOACTIVATE);
device->filter_messages = filter_messages;
}
void CDECL wined3d_device_restore_fullscreen_window(struct wined3d_device *device, HWND window)
{
BOOL filter_messages;
LONG style, exstyle;
if (!device->style && !device->exStyle) return;
TRACE("Restoring window style of window %p to %08x, %08x.\n",
window, device->style, device->exStyle);
style = GetWindowLongW(window, GWL_STYLE);
exstyle = GetWindowLongW(window, GWL_EXSTYLE);
filter_messages = device->filter_messages;
device->filter_messages = TRUE;
/* Only restore the style if the application didn't modify it during the
* fullscreen phase. Some applications change it before calling Reset()
* when switching between windowed and fullscreen modes (HL2), some
* depend on the original style (Eve Online). */
if (style == fullscreen_style(device->style) && exstyle == fullscreen_exstyle(device->exStyle))
{
SetWindowLongW(window, GWL_STYLE, device->style);
SetWindowLongW(window, GWL_EXSTYLE, device->exStyle);
}
SetWindowPos(window, 0, 0, 0, 0, 0, SWP_FRAMECHANGED | SWP_NOMOVE | SWP_NOSIZE | SWP_NOZORDER | SWP_NOACTIVATE);
device->filter_messages = filter_messages;
/* Delete the old values. */
device->style = 0;
device->exStyle = 0;
}
HRESULT CDECL wined3d_device_acquire_focus_window(struct wined3d_device *device, HWND window)
{
TRACE("device %p, window %p.\n", device, window);
if (!wined3d_register_window(window, device))
{
ERR("Failed to register window %p.\n", window);
return E_FAIL;
}
InterlockedExchangePointer((void **)&device->focus_window, window);
SetWindowPos(window, 0, 0, 0, 0, 0, SWP_NOSIZE | SWP_NOMOVE);
return WINED3D_OK;
}
void CDECL wined3d_device_release_focus_window(struct wined3d_device *device)
{
TRACE("device %p.\n", device);
if (device->focus_window) wined3d_unregister_window(device->focus_window);
InterlockedExchangePointer((void **)&device->focus_window, NULL);
}
HRESULT CDECL wined3d_device_init_3d(struct wined3d_device *device,
struct wined3d_swapchain_desc *swapchain_desc)
{
static const struct wined3d_color black = {0.0f, 0.0f, 0.0f, 0.0f};
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
struct wined3d_swapchain *swapchain = NULL;
struct wined3d_context *context;
HRESULT hr;
DWORD state;
unsigned int i;
TRACE("device %p, swapchain_desc %p.\n", device, swapchain_desc);
if (device->d3d_initialized)
return WINED3DERR_INVALIDCALL;
if (!device->adapter->opengl)
return WINED3DERR_INVALIDCALL;
device->valid_rt_mask = 0;
for (i = 0; i < gl_info->limits.buffers; ++i)
device->valid_rt_mask |= (1 << i);
device->fb.render_targets = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
sizeof(*device->fb.render_targets) * gl_info->limits.buffers);
/* Initialize the texture unit mapping to a 1:1 mapping */
for (state = 0; state < MAX_COMBINED_SAMPLERS; ++state)
{
if (state < gl_info->limits.fragment_samplers)
{
device->texUnitMap[state] = state;
device->rev_tex_unit_map[state] = state;
}
else
{
device->texUnitMap[state] = WINED3D_UNMAPPED_STAGE;
device->rev_tex_unit_map[state] = WINED3D_UNMAPPED_STAGE;
}
}
/* Setup the implicit swapchain. This also initializes a context. */
TRACE("Creating implicit swapchain\n");
hr = device->device_parent->ops->create_swapchain(device->device_parent,
swapchain_desc, &swapchain);
if (FAILED(hr))
{
WARN("Failed to create implicit swapchain\n");
goto err_out;
}
device->swapchain_count = 1;
device->swapchains = HeapAlloc(GetProcessHeap(), 0, device->swapchain_count * sizeof(*device->swapchains));
if (!device->swapchains)
{
ERR("Out of memory!\n");
goto err_out;
}
device->swapchains[0] = swapchain;
if (swapchain->back_buffers && swapchain->back_buffers[0])
{
TRACE("Setting rendertarget to %p.\n", swapchain->back_buffers);
device->fb.render_targets[0] = swapchain->back_buffers[0];
}
else
{
TRACE("Setting rendertarget to %p.\n", swapchain->front_buffer);
device->fb.render_targets[0] = swapchain->front_buffer;
}
wined3d_surface_incref(device->fb.render_targets[0]);
/* Depth Stencil support */
device->fb.depth_stencil = device->auto_depth_stencil;
if (device->fb.depth_stencil)
wined3d_surface_incref(device->fb.depth_stencil);
hr = device->shader_backend->shader_alloc_private(device);
if (FAILED(hr))
{
TRACE("Shader private data couldn't be allocated\n");
goto err_out;
}
hr = device->frag_pipe->alloc_private(device);
if (FAILED(hr))
{
TRACE("Fragment pipeline private data couldn't be allocated\n");
goto err_out;
}
hr = device->blitter->alloc_private(device);
if (FAILED(hr))
{
TRACE("Blitter private data couldn't be allocated\n");
goto err_out;
}
/* Set up some starting GL setup */
/* Setup all the devices defaults */
stateblock_init_default_state(device->stateBlock);
context = context_acquire(device, swapchain->front_buffer);
create_dummy_textures(device, context);
ENTER_GL();
/* Initialize the current view state */
device->view_ident = 1;
device->contexts[0]->last_was_rhw = 0;
switch (wined3d_settings.offscreen_rendering_mode)
{
case ORM_FBO:
device->offscreenBuffer = GL_COLOR_ATTACHMENT0;
break;
case ORM_BACKBUFFER:
{
if (context_get_current()->aux_buffers > 0)
{
TRACE("Using auxiliary buffer for offscreen rendering\n");
device->offscreenBuffer = GL_AUX0;
}
else
{
TRACE("Using back buffer for offscreen rendering\n");
device->offscreenBuffer = GL_BACK;
}
}
}
TRACE("All defaults now set up, leaving 3D init.\n");
LEAVE_GL();
context_release(context);
/* Clear the screen */
wined3d_device_clear(device, 0, NULL, WINED3DCLEAR_TARGET
| (swapchain_desc->enable_auto_depth_stencil ? WINED3DCLEAR_ZBUFFER | WINED3DCLEAR_STENCIL : 0),
&black, 1.0f, 0);
device->d3d_initialized = TRUE;
if (wined3d_settings.logo)
device_load_logo(device, wined3d_settings.logo);
return WINED3D_OK;
err_out:
HeapFree(GetProcessHeap(), 0, device->fb.render_targets);
HeapFree(GetProcessHeap(), 0, device->swapchains);
device->swapchain_count = 0;
if (swapchain)
wined3d_swapchain_decref(swapchain);
if (device->blit_priv)
device->blitter->free_private(device);
if (device->fragment_priv)
device->frag_pipe->free_private(device);
if (device->shader_priv)
device->shader_backend->shader_free_private(device);
return hr;
}
HRESULT CDECL wined3d_device_init_gdi(struct wined3d_device *device,
struct wined3d_swapchain_desc *swapchain_desc)
{
struct wined3d_swapchain *swapchain = NULL;
HRESULT hr;
TRACE("device %p, swapchain_desc %p.\n", device, swapchain_desc);
/* Setup the implicit swapchain */
TRACE("Creating implicit swapchain\n");
hr = device->device_parent->ops->create_swapchain(device->device_parent,
swapchain_desc, &swapchain);
if (FAILED(hr))
{
WARN("Failed to create implicit swapchain\n");
goto err_out;
}
device->swapchain_count = 1;
device->swapchains = HeapAlloc(GetProcessHeap(), 0, device->swapchain_count * sizeof(*device->swapchains));
if (!device->swapchains)
{
ERR("Out of memory!\n");
goto err_out;
}
device->swapchains[0] = swapchain;
return WINED3D_OK;
err_out:
wined3d_swapchain_decref(swapchain);
return hr;
}
HRESULT CDECL wined3d_device_uninit_3d(struct wined3d_device *device)
{
struct wined3d_resource *resource, *cursor;
const struct wined3d_gl_info *gl_info;
struct wined3d_context *context;
struct wined3d_surface *surface;
UINT i;
TRACE("device %p.\n", device);
if (!device->d3d_initialized)
return WINED3DERR_INVALIDCALL;
/* Force making the context current again, to verify it is still valid
* (workaround for broken drivers) */
context_set_current(NULL);
/* I don't think that the interface guarantees that the device is destroyed from the same thread
* it was created. Thus make sure a context is active for the glDelete* calls
*/
context = context_acquire(device, NULL);
gl_info = context->gl_info;
if (device->logo_surface)
wined3d_surface_decref(device->logo_surface);
stateblock_unbind_resources(device->stateBlock);
/* Unload resources */
LIST_FOR_EACH_ENTRY_SAFE(resource, cursor, &device->resources, struct wined3d_resource, resource_list_entry)
{
TRACE("Unloading resource %p.\n", resource);
resource->resource_ops->resource_unload(resource);
}
TRACE("Deleting high order patches\n");
for (i = 0; i < PATCHMAP_SIZE; ++i)
{
struct wined3d_rect_patch *patch;
struct list *e1, *e2;
LIST_FOR_EACH_SAFE(e1, e2, &device->patches[i])
{
patch = LIST_ENTRY(e1, struct wined3d_rect_patch, entry);
wined3d_device_delete_patch(device, patch->Handle);
}
}
/* Delete the mouse cursor texture */
if (device->cursorTexture)
{
ENTER_GL();
glDeleteTextures(1, &device->cursorTexture);
LEAVE_GL();
device->cursorTexture = 0;
}
/* Destroy the depth blt resources, they will be invalid after the reset. Also free shader
* private data, it might contain opengl pointers
*/
if (device->depth_blt_texture)
{
ENTER_GL();
glDeleteTextures(1, &device->depth_blt_texture);
LEAVE_GL();
device->depth_blt_texture = 0;
}
/* Destroy the shader backend. Note that this has to happen after all shaders are destroyed. */
device->blitter->free_private(device);
device->frag_pipe->free_private(device);
device->shader_backend->shader_free_private(device);
/* Release the buffers (with sanity checks)*/
if (device->onscreen_depth_stencil)
{
surface = device->onscreen_depth_stencil;
device->onscreen_depth_stencil = NULL;
wined3d_surface_decref(surface);
}
if (device->fb.depth_stencil)
{
surface = device->fb.depth_stencil;
TRACE("Releasing depth/stencil buffer %p.\n", surface);
device->fb.depth_stencil = NULL;
wined3d_surface_decref(surface);
}
if (device->auto_depth_stencil)
{
surface = device->auto_depth_stencil;
device->auto_depth_stencil = NULL;
if (wined3d_surface_decref(surface))
FIXME("Something's still holding the auto depth stencil buffer (%p).\n", surface);
}
for (i = 1; i < gl_info->limits.buffers; ++i)
{
wined3d_device_set_render_target(device, i, NULL, FALSE);
}
surface = device->fb.render_targets[0];
TRACE("Setting rendertarget 0 to NULL\n");
device->fb.render_targets[0] = NULL;
TRACE("Releasing the render target at %p\n", surface);
wined3d_surface_decref(surface);
context_release(context);
for (i = 0; i < device->swapchain_count; ++i)
{
TRACE("Releasing the implicit swapchain %u.\n", i);
if (wined3d_swapchain_decref(device->swapchains[i]))
FIXME("Something's still holding the implicit swapchain.\n");
}
HeapFree(GetProcessHeap(), 0, device->swapchains);
device->swapchains = NULL;
device->swapchain_count = 0;
HeapFree(GetProcessHeap(), 0, device->fb.render_targets);
device->fb.render_targets = NULL;
device->d3d_initialized = FALSE;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_uninit_gdi(struct wined3d_device *device)
{
unsigned int i;
for (i = 0; i < device->swapchain_count; ++i)
{
TRACE("Releasing the implicit swapchain %u.\n", i);
if (wined3d_swapchain_decref(device->swapchains[i]))
FIXME("Something's still holding the implicit swapchain.\n");
}
HeapFree(GetProcessHeap(), 0, device->swapchains);
device->swapchains = NULL;
device->swapchain_count = 0;
return WINED3D_OK;
}
/* Enables thread safety in the wined3d device and its resources. Called by DirectDraw
* from SetCooperativeLevel if DDSCL_MULTITHREADED is specified, and by d3d8/9 from
* CreateDevice if D3DCREATE_MULTITHREADED is passed.
*
* There is no way to deactivate thread safety once it is enabled.
*/
void CDECL wined3d_device_set_multithreaded(struct wined3d_device *device)
{
TRACE("device %p.\n", device);
/* For now just store the flag (needed in case of ddraw). */
device->create_parms.flags |= WINED3DCREATE_MULTITHREADED;
}
HRESULT CDECL wined3d_device_get_wined3d(const struct wined3d_device *device, struct wined3d **wined3d)
{
TRACE("device %p, wined3d %p.\n", device, wined3d);
*wined3d = device->wined3d;
wined3d_incref(*wined3d);
TRACE("Returning %p.\n", *wined3d);
return WINED3D_OK;
}
UINT CDECL wined3d_device_get_available_texture_mem(const struct wined3d_device *device)
{
TRACE("device %p.\n", device);
TRACE("Emulating %d MB, returning %d MB left.\n",
device->adapter->TextureRam / (1024 * 1024),
(device->adapter->TextureRam - device->adapter->UsedTextureRam) / (1024 * 1024));
return device->adapter->TextureRam - device->adapter->UsedTextureRam;
}
HRESULT CDECL wined3d_device_set_stream_source(struct wined3d_device *device, UINT stream_idx,
struct wined3d_buffer *buffer, UINT offset, UINT stride)
{
struct wined3d_stream_state *stream;
struct wined3d_buffer *prev_buffer;
TRACE("device %p, stream_idx %u, buffer %p, offset %u, stride %u.\n",
device, stream_idx, buffer, offset, stride);
if (stream_idx >= MAX_STREAMS)
{
WARN("Stream index %u out of range.\n", stream_idx);
return WINED3DERR_INVALIDCALL;
}
else if (offset & 0x3)
{
WARN("Offset %u is not 4 byte aligned.\n", offset);
return WINED3DERR_INVALIDCALL;
}
stream = &device->updateStateBlock->state.streams[stream_idx];
prev_buffer = stream->buffer;
device->updateStateBlock->changed.streamSource |= 1 << stream_idx;
if (prev_buffer == buffer
&& stream->stride == stride
&& stream->offset == offset)
{
TRACE("Application is setting the old values over, nothing to do.\n");
return WINED3D_OK;
}
stream->buffer = buffer;
if (buffer)
{
stream->stride = stride;
stream->offset = offset;
}
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
if (buffer)
wined3d_buffer_incref(buffer);
if (prev_buffer)
wined3d_buffer_decref(prev_buffer);
return WINED3D_OK;
}
if (buffer)
{
InterlockedIncrement(&buffer->resource.bind_count);
wined3d_buffer_incref(buffer);
}
if (prev_buffer)
{
InterlockedDecrement(&prev_buffer->resource.bind_count);
wined3d_buffer_decref(prev_buffer);
}
device_invalidate_state(device, STATE_STREAMSRC);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_stream_source(const struct wined3d_device *device,
UINT stream_idx, struct wined3d_buffer **buffer, UINT *offset, UINT *stride)
{
struct wined3d_stream_state *stream;
TRACE("device %p, stream_idx %u, buffer %p, offset %p, stride %p.\n",
device, stream_idx, buffer, offset, stride);
if (stream_idx >= MAX_STREAMS)
{
WARN("Stream index %u out of range.\n", stream_idx);
return WINED3DERR_INVALIDCALL;
}
stream = &device->stateBlock->state.streams[stream_idx];
*buffer = stream->buffer;
if (*buffer)
wined3d_buffer_incref(*buffer);
if (offset)
*offset = stream->offset;
*stride = stream->stride;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_stream_source_freq(struct wined3d_device *device, UINT stream_idx, UINT divider)
{
struct wined3d_stream_state *stream;
UINT old_flags, old_freq;
TRACE("device %p, stream_idx %u, divider %#x.\n", device, stream_idx, divider);
/* Verify input. At least in d3d9 this is invalid. */
if ((divider & WINED3DSTREAMSOURCE_INSTANCEDATA) && (divider & WINED3DSTREAMSOURCE_INDEXEDDATA))
{
WARN("INSTANCEDATA and INDEXEDDATA were set, returning D3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
if ((divider & WINED3DSTREAMSOURCE_INSTANCEDATA) && !stream_idx)
{
WARN("INSTANCEDATA used on stream 0, returning D3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
if (!divider)
{
WARN("Divider is 0, returning D3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
stream = &device->updateStateBlock->state.streams[stream_idx];
old_flags = stream->flags;
old_freq = stream->frequency;
stream->flags = divider & (WINED3DSTREAMSOURCE_INSTANCEDATA | WINED3DSTREAMSOURCE_INDEXEDDATA);
stream->frequency = divider & 0x7fffff;
device->updateStateBlock->changed.streamFreq |= 1 << stream_idx;
if (stream->frequency != old_freq || stream->flags != old_flags)
device_invalidate_state(device, STATE_STREAMSRC);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_stream_source_freq(const struct wined3d_device *device,
UINT stream_idx, UINT *divider)
{
struct wined3d_stream_state *stream;
TRACE("device %p, stream_idx %u, divider %p.\n", device, stream_idx, divider);
stream = &device->updateStateBlock->state.streams[stream_idx];
*divider = stream->flags | stream->frequency;
TRACE("Returning %#x.\n", *divider);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_transform(struct wined3d_device *device,
enum wined3d_transform_state d3dts, const struct wined3d_matrix *matrix)
{
TRACE("device %p, state %s, matrix %p.\n",
device, debug_d3dtstype(d3dts), matrix);
TRACE("%.8e %.8e %.8e %.8e\n", matrix->u.s._11, matrix->u.s._12, matrix->u.s._13, matrix->u.s._14);
TRACE("%.8e %.8e %.8e %.8e\n", matrix->u.s._21, matrix->u.s._22, matrix->u.s._23, matrix->u.s._24);
TRACE("%.8e %.8e %.8e %.8e\n", matrix->u.s._31, matrix->u.s._32, matrix->u.s._33, matrix->u.s._34);
TRACE("%.8e %.8e %.8e %.8e\n", matrix->u.s._41, matrix->u.s._42, matrix->u.s._43, matrix->u.s._44);
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
device->updateStateBlock->changed.transform[d3dts >> 5] |= 1 << (d3dts & 0x1f);
device->updateStateBlock->state.transforms[d3dts] = *matrix;
return WINED3D_OK;
}
/* If the new matrix is the same as the current one,
* we cut off any further processing. this seems to be a reasonable
* optimization because as was noticed, some apps (warcraft3 for example)
* tend towards setting the same matrix repeatedly for some reason.
*
* From here on we assume that the new matrix is different, wherever it matters. */
if (!memcmp(&device->stateBlock->state.transforms[d3dts].u.m[0][0], matrix, sizeof(*matrix)))
{
TRACE("The application is setting the same matrix over again.\n");
return WINED3D_OK;
}
device->stateBlock->state.transforms[d3dts] = *matrix;
if (d3dts == WINED3D_TS_VIEW)
device->view_ident = !memcmp(matrix, &identity, sizeof(identity));
if (d3dts < WINED3D_TS_WORLD_MATRIX(device->adapter->gl_info.limits.blends))
device_invalidate_state(device, STATE_TRANSFORM(d3dts));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_transform(const struct wined3d_device *device,
enum wined3d_transform_state state, struct wined3d_matrix *matrix)
{
TRACE("device %p, state %s, matrix %p.\n", device, debug_d3dtstype(state), matrix);
*matrix = device->stateBlock->state.transforms[state];
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_multiply_transform(struct wined3d_device *device,
enum wined3d_transform_state state, const struct wined3d_matrix *matrix)
{
const struct wined3d_matrix *mat = NULL;
struct wined3d_matrix temp;
TRACE("device %p, state %s, matrix %p.\n", device, debug_d3dtstype(state), matrix);
/* Note: Using 'updateStateBlock' rather than 'stateblock' in the code
* below means it will be recorded in a state block change, but it
* works regardless where it is recorded.
* If this is found to be wrong, change to StateBlock. */
if (state > HIGHEST_TRANSFORMSTATE)
{
WARN("Unhandled transform state %#x.\n", state);
return WINED3D_OK;
}
mat = &device->updateStateBlock->state.transforms[state];
multiply_matrix(&temp, mat, matrix);
/* Apply change via set transform - will reapply to eg. lights this way. */
return wined3d_device_set_transform(device, state, &temp);
}
/* Note lights are real special cases. Although the device caps state only
* e.g. 8 are supported, you can reference any indexes you want as long as
* that number max are enabled at any one point in time. Therefore since the
* indices can be anything, we need a hashmap of them. However, this causes
* stateblock problems. When capturing the state block, I duplicate the
* hashmap, but when recording, just build a chain pretty much of commands to
* be replayed. */
HRESULT CDECL wined3d_device_set_light(struct wined3d_device *device,
UINT light_idx, const struct wined3d_light *light)
{
UINT hash_idx = LIGHTMAP_HASHFUNC(light_idx);
struct wined3d_light_info *object = NULL;
struct list *e;
float rho;
TRACE("device %p, light_idx %u, light %p.\n", device, light_idx, light);
/* Check the parameter range. Need for speed most wanted sets junk lights
* which confuse the GL driver. */
if (!light)
return WINED3DERR_INVALIDCALL;
switch (light->type)
{
case WINED3D_LIGHT_POINT:
case WINED3D_LIGHT_SPOT:
case WINED3D_LIGHT_PARALLELPOINT:
case WINED3D_LIGHT_GLSPOT:
/* Incorrect attenuation values can cause the gl driver to crash.
* Happens with Need for speed most wanted. */
if (light->attenuation0 < 0.0f || light->attenuation1 < 0.0f || light->attenuation2 < 0.0f)
{
WARN("Attenuation is negative, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
break;
case WINED3D_LIGHT_DIRECTIONAL:
/* Ignores attenuation */
break;
default:
WARN("Light type out of range, returning WINED3DERR_INVALIDCALL\n");
return WINED3DERR_INVALIDCALL;
}
LIST_FOR_EACH(e, &device->updateStateBlock->state.light_map[hash_idx])
{
object = LIST_ENTRY(e, struct wined3d_light_info, entry);
if (object->OriginalIndex == light_idx)
break;
object = NULL;
}
if (!object)
{
TRACE("Adding new light\n");
object = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*object));
if (!object)
{
ERR("Out of memory error when allocating a light\n");
return E_OUTOFMEMORY;
}
list_add_head(&device->updateStateBlock->state.light_map[hash_idx], &object->entry);
object->glIndex = -1;
object->OriginalIndex = light_idx;
}
/* Initialize the object. */
TRACE("Light %d setting to type %d, Diffuse(%f,%f,%f,%f), Specular(%f,%f,%f,%f), Ambient(%f,%f,%f,%f)\n",
light_idx, light->type,
light->diffuse.r, light->diffuse.g, light->diffuse.b, light->diffuse.a,
light->specular.r, light->specular.g, light->specular.b, light->specular.a,
light->ambient.r, light->ambient.g, light->ambient.b, light->ambient.a);
TRACE("... Pos(%f,%f,%f), Dir(%f,%f,%f)\n", light->position.x, light->position.y, light->position.z,
light->direction.x, light->direction.y, light->direction.z);
TRACE("... Range(%f), Falloff(%f), Theta(%f), Phi(%f)\n",
light->range, light->falloff, light->theta, light->phi);
/* Save away the information. */
object->OriginalParms = *light;
switch (light->type)
{
case WINED3D_LIGHT_POINT:
/* Position */
object->lightPosn[0] = light->position.x;
object->lightPosn[1] = light->position.y;
object->lightPosn[2] = light->position.z;
object->lightPosn[3] = 1.0f;
object->cutoff = 180.0f;
/* FIXME: Range */
break;
case WINED3D_LIGHT_DIRECTIONAL:
/* Direction */
object->lightPosn[0] = -light->direction.x;
object->lightPosn[1] = -light->direction.y;
object->lightPosn[2] = -light->direction.z;
object->lightPosn[3] = 0.0f;
object->exponent = 0.0f;
object->cutoff = 180.0f;
break;
case WINED3D_LIGHT_SPOT:
/* Position */
object->lightPosn[0] = light->position.x;
object->lightPosn[1] = light->position.y;
object->lightPosn[2] = light->position.z;
object->lightPosn[3] = 1.0f;
/* Direction */
object->lightDirn[0] = light->direction.x;
object->lightDirn[1] = light->direction.y;
object->lightDirn[2] = light->direction.z;
object->lightDirn[3] = 1.0f;
/* opengl-ish and d3d-ish spot lights use too different models
* for the light "intensity" as a function of the angle towards
* the main light direction, so we only can approximate very
* roughly. However, spot lights are rather rarely used in games
* (if ever used at all). Furthermore if still used, probably
* nobody pays attention to such details. */
if (!light->falloff)
{
/* Falloff = 0 is easy, because d3d's and opengl's spot light
* equations have the falloff resp. exponent parameter as an
* exponent, so the spot light lighting will always be 1.0 for
* both of them, and we don't have to care for the rest of the
* rather complex calculation. */
object->exponent = 0.0f;
}
else
{
rho = light->theta + (light->phi - light->theta) / (2 * light->falloff);
if (rho < 0.0001f)
rho = 0.0001f;
object->exponent = -0.3f / logf(cosf(rho / 2));
}
if (object->exponent > 128.0f)
object->exponent = 128.0f;
object->cutoff = (float)(light->phi * 90 / M_PI);
/* FIXME: Range */
break;
default:
FIXME("Unrecognized light type %#x.\n", light->type);
}
/* Update the live definitions if the light is currently assigned a glIndex. */
if (object->glIndex != -1 && !device->isRecordingState)
device_invalidate_state(device, STATE_ACTIVELIGHT(object->glIndex));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_light(const struct wined3d_device *device,
UINT light_idx, struct wined3d_light *light)
{
UINT hash_idx = LIGHTMAP_HASHFUNC(light_idx);
struct wined3d_light_info *light_info = NULL;
struct list *e;
TRACE("device %p, light_idx %u, light %p.\n", device, light_idx, light);
LIST_FOR_EACH(e, &device->stateBlock->state.light_map[hash_idx])
{
light_info = LIST_ENTRY(e, struct wined3d_light_info, entry);
if (light_info->OriginalIndex == light_idx)
break;
light_info = NULL;
}
if (!light_info)
{
TRACE("Light information requested but light not defined\n");
return WINED3DERR_INVALIDCALL;
}
*light = light_info->OriginalParms;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_light_enable(struct wined3d_device *device, UINT light_idx, BOOL enable)
{
UINT hash_idx = LIGHTMAP_HASHFUNC(light_idx);
struct wined3d_light_info *light_info = NULL;
struct list *e;
TRACE("device %p, light_idx %u, enable %#x.\n", device, light_idx, enable);
LIST_FOR_EACH(e, &device->updateStateBlock->state.light_map[hash_idx])
{
light_info = LIST_ENTRY(e, struct wined3d_light_info, entry);
if (light_info->OriginalIndex == light_idx)
break;
light_info = NULL;
}
TRACE("Found light %p.\n", light_info);
/* Special case - enabling an undefined light creates one with a strict set of parameters. */
if (!light_info)
{
TRACE("Light enabled requested but light not defined, so defining one!\n");
wined3d_device_set_light(device, light_idx, &WINED3D_default_light);
/* Search for it again! Should be fairly quick as near head of list. */
LIST_FOR_EACH(e, &device->updateStateBlock->state.light_map[hash_idx])
{
light_info = LIST_ENTRY(e, struct wined3d_light_info, entry);
if (light_info->OriginalIndex == light_idx)
break;
light_info = NULL;
}
if (!light_info)
{
FIXME("Adding default lights has failed dismally\n");
return WINED3DERR_INVALIDCALL;
}
}
if (!enable)
{
if (light_info->glIndex != -1)
{
if (!device->isRecordingState)
device_invalidate_state(device, STATE_ACTIVELIGHT(light_info->glIndex));
device->updateStateBlock->state.lights[light_info->glIndex] = NULL;
light_info->glIndex = -1;
}
else
{
TRACE("Light already disabled, nothing to do\n");
}
light_info->enabled = FALSE;
}
else
{
light_info->enabled = TRUE;
if (light_info->glIndex != -1)
{
TRACE("Nothing to do as light was enabled\n");
}
else
{
unsigned int i;
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
/* Find a free GL light. */
for (i = 0; i < gl_info->limits.lights; ++i)
{
if (!device->updateStateBlock->state.lights[i])
{
device->updateStateBlock->state.lights[i] = light_info;
light_info->glIndex = i;
break;
}
}
if (light_info->glIndex == -1)
{
/* Our tests show that Windows returns D3D_OK in this situation, even with
* D3DCREATE_HARDWARE_VERTEXPROCESSING | D3DCREATE_PUREDEVICE devices. This
* is consistent among ddraw, d3d8 and d3d9. GetLightEnable returns TRUE
* as well for those lights.
*
* TODO: Test how this affects rendering. */
WARN("Too many concurrently active lights\n");
return WINED3D_OK;
}
/* i == light_info->glIndex */
if (!device->isRecordingState)
device_invalidate_state(device, STATE_ACTIVELIGHT(i));
}
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_light_enable(const struct wined3d_device *device, UINT light_idx, BOOL *enable)
{
UINT hash_idx = LIGHTMAP_HASHFUNC(light_idx);
struct wined3d_light_info *light_info = NULL;
struct list *e;
TRACE("device %p, light_idx %u, enable %p.\n", device, light_idx, enable);
LIST_FOR_EACH(e, &device->stateBlock->state.light_map[hash_idx])
{
light_info = LIST_ENTRY(e, struct wined3d_light_info, entry);
if (light_info->OriginalIndex == light_idx)
break;
light_info = NULL;
}
if (!light_info)
{
TRACE("Light enabled state requested but light not defined.\n");
return WINED3DERR_INVALIDCALL;
}
/* true is 128 according to SetLightEnable */
*enable = light_info->enabled ? 128 : 0;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_clip_plane(struct wined3d_device *device,
UINT plane_idx, const struct wined3d_vec4 *plane)
{
TRACE("device %p, plane_idx %u, plane %p.\n", device, plane_idx, plane);
/* Validate plane_idx. */
if (plane_idx >= device->adapter->gl_info.limits.clipplanes)
{
TRACE("Application has requested clipplane this device doesn't support.\n");
return WINED3DERR_INVALIDCALL;
}
device->updateStateBlock->changed.clipplane |= 1 << plane_idx;
if (!memcmp(&device->updateStateBlock->state.clip_planes[plane_idx], plane, sizeof(*plane)))
{
TRACE("Application is setting old values over, nothing to do.\n");
return WINED3D_OK;
}
device->updateStateBlock->state.clip_planes[plane_idx] = *plane;
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_CLIPPLANE(plane_idx));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_clip_plane(const struct wined3d_device *device,
UINT plane_idx, struct wined3d_vec4 *plane)
{
TRACE("device %p, plane_idx %u, plane %p.\n", device, plane_idx, plane);
/* Validate plane_idx. */
if (plane_idx >= device->adapter->gl_info.limits.clipplanes)
{
TRACE("Application has requested clipplane this device doesn't support.\n");
return WINED3DERR_INVALIDCALL;
}
*plane = device->stateBlock->state.clip_planes[plane_idx];
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_clip_status(struct wined3d_device *device,
const struct wined3d_clip_status *clip_status)
{
FIXME("device %p, clip_status %p stub!\n", device, clip_status);
if (!clip_status)
return WINED3DERR_INVALIDCALL;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_clip_status(const struct wined3d_device *device,
struct wined3d_clip_status *clip_status)
{
FIXME("device %p, clip_status %p stub!\n", device, clip_status);
if (!clip_status)
return WINED3DERR_INVALIDCALL;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_material(struct wined3d_device *device, const struct wined3d_material *material)
{
TRACE("device %p, material %p.\n", device, material);
device->updateStateBlock->changed.material = TRUE;
device->updateStateBlock->state.material = *material;
/* Handle recording of state blocks */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_MATERIAL);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_material(const struct wined3d_device *device, struct wined3d_material *material)
{
TRACE("device %p, material %p.\n", device, material);
*material = device->updateStateBlock->state.material;
TRACE("diffuse {%.8e, %.8e, %.8e, %.8e}\n",
material->diffuse.r, material->diffuse.g,
material->diffuse.b, material->diffuse.a);
TRACE("ambient {%.8e, %.8e, %.8e, %.8e}\n",
material->ambient.r, material->ambient.g,
material->ambient.b, material->ambient.a);
TRACE("specular {%.8e, %.8e, %.8e, %.8e}\n",
material->specular.r, material->specular.g,
material->specular.b, material->specular.a);
TRACE("emissive {%.8e, %.8e, %.8e, %.8e}\n",
material->emissive.r, material->emissive.g,
material->emissive.b, material->emissive.a);
TRACE("power %.8e.\n", material->power);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_index_buffer(struct wined3d_device *device,
struct wined3d_buffer *buffer, enum wined3d_format_id format_id)
{
struct wined3d_buffer *prev_buffer;
TRACE("device %p, buffer %p, format %s.\n",
device, buffer, debug_d3dformat(format_id));
prev_buffer = device->updateStateBlock->state.index_buffer;
device->updateStateBlock->changed.indices = TRUE;
device->updateStateBlock->state.index_buffer = buffer;
device->updateStateBlock->state.index_format = format_id;
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
if (buffer)
wined3d_buffer_incref(buffer);
if (prev_buffer)
wined3d_buffer_decref(prev_buffer);
return WINED3D_OK;
}
if (prev_buffer != buffer)
{
device_invalidate_state(device, STATE_INDEXBUFFER);
if (buffer)
{
InterlockedIncrement(&buffer->resource.bind_count);
wined3d_buffer_incref(buffer);
}
if (prev_buffer)
{
InterlockedDecrement(&prev_buffer->resource.bind_count);
wined3d_buffer_decref(prev_buffer);
}
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_index_buffer(const struct wined3d_device *device, struct wined3d_buffer **buffer)
{
TRACE("device %p, buffer %p.\n", device, buffer);
*buffer = device->stateBlock->state.index_buffer;
if (*buffer)
wined3d_buffer_incref(*buffer);
TRACE("Returning %p.\n", *buffer);
return WINED3D_OK;
}
/* Method to offer d3d9 a simple way to set the base vertex index without messing with the index buffer */
HRESULT CDECL wined3d_device_set_base_vertex_index(struct wined3d_device *device, INT base_index)
{
TRACE("device %p, base_index %d.\n", device, base_index);
if (device->updateStateBlock->state.base_vertex_index == base_index)
{
TRACE("Application is setting the old value over, nothing to do\n");
return WINED3D_OK;
}
device->updateStateBlock->state.base_vertex_index = base_index;
if (device->isRecordingState)
{
TRACE("Recording... not performing anything\n");
return WINED3D_OK;
}
return WINED3D_OK;
}
INT CDECL wined3d_device_get_base_vertex_index(const struct wined3d_device *device)
{
TRACE("device %p.\n", device);
return device->stateBlock->state.base_vertex_index;
}
HRESULT CDECL wined3d_device_set_viewport(struct wined3d_device *device, const struct wined3d_viewport *viewport)
{
TRACE("device %p, viewport %p.\n", device, viewport);
TRACE("x %u, y %u, w %u, h %u, min_z %.8e, max_z %.8e.\n",
viewport->x, viewport->y, viewport->width, viewport->height, viewport->min_z, viewport->max_z);
device->updateStateBlock->changed.viewport = TRUE;
device->updateStateBlock->state.viewport = *viewport;
/* Handle recording of state blocks */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_VIEWPORT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_viewport(const struct wined3d_device *device, struct wined3d_viewport *viewport)
{
TRACE("device %p, viewport %p.\n", device, viewport);
*viewport = device->stateBlock->state.viewport;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_render_state(struct wined3d_device *device,
enum wined3d_render_state state, DWORD value)
{
DWORD old_value = device->stateBlock->state.render_states[state];
TRACE("device %p, state %s (%#x), value %#x.\n", device, debug_d3drenderstate(state), state, value);
device->updateStateBlock->changed.renderState[state >> 5] |= 1 << (state & 0x1f);
device->updateStateBlock->state.render_states[state] = value;
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
/* Compared here and not before the assignment to allow proper stateblock recording. */
if (value == old_value)
TRACE("Application is setting the old value over, nothing to do.\n");
else
device_invalidate_state(device, STATE_RENDER(state));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_render_state(const struct wined3d_device *device,
enum wined3d_render_state state, DWORD *value)
{
TRACE("device %p, state %s (%#x), value %p.\n", device, debug_d3drenderstate(state), state, value);
*value = device->stateBlock->state.render_states[state];
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_sampler_state(struct wined3d_device *device,
UINT sampler_idx, enum wined3d_sampler_state state, DWORD value)
{
DWORD old_value;
TRACE("device %p, sampler_idx %u, state %s, value %#x.\n",
device, sampler_idx, debug_d3dsamplerstate(state), value);
if (sampler_idx >= WINED3DVERTEXTEXTURESAMPLER0 && sampler_idx <= WINED3DVERTEXTEXTURESAMPLER3)
sampler_idx -= (WINED3DVERTEXTEXTURESAMPLER0 - MAX_FRAGMENT_SAMPLERS);
if (sampler_idx >= sizeof(device->stateBlock->state.sampler_states)
/ sizeof(*device->stateBlock->state.sampler_states))
{
WARN("Invalid sampler %u.\n", sampler_idx);
return WINED3D_OK; /* Windows accepts overflowing this array ... we do not. */
}
old_value = device->stateBlock->state.sampler_states[sampler_idx][state];
device->updateStateBlock->state.sampler_states[sampler_idx][state] = value;
device->updateStateBlock->changed.samplerState[sampler_idx] |= 1 << state;
/* Handle recording of state blocks. */
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
if (old_value == value)
{
TRACE("Application is setting the old value over, nothing to do.\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_SAMPLER(sampler_idx));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_sampler_state(const struct wined3d_device *device,
UINT sampler_idx, enum wined3d_sampler_state state, DWORD *value)
{
TRACE("device %p, sampler_idx %u, state %s, value %p.\n",
device, sampler_idx, debug_d3dsamplerstate(state), value);
if (sampler_idx >= WINED3DVERTEXTEXTURESAMPLER0 && sampler_idx <= WINED3DVERTEXTEXTURESAMPLER3)
sampler_idx -= (WINED3DVERTEXTEXTURESAMPLER0 - MAX_FRAGMENT_SAMPLERS);
if (sampler_idx >= sizeof(device->stateBlock->state.sampler_states)
/ sizeof(*device->stateBlock->state.sampler_states))
{
WARN("Invalid sampler %u.\n", sampler_idx);
return WINED3D_OK; /* Windows accepts overflowing this array ... we do not. */
}
*value = device->stateBlock->state.sampler_states[sampler_idx][state];
TRACE("Returning %#x.\n", *value);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_scissor_rect(struct wined3d_device *device, const RECT *rect)
{
TRACE("device %p, rect %s.\n", device, wine_dbgstr_rect(rect));
device->updateStateBlock->changed.scissorRect = TRUE;
if (EqualRect(&device->updateStateBlock->state.scissor_rect, rect))
{
TRACE("App is setting the old scissor rectangle over, nothing to do.\n");
return WINED3D_OK;
}
CopyRect(&device->updateStateBlock->state.scissor_rect, rect);
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_SCISSORRECT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_scissor_rect(const struct wined3d_device *device, RECT *rect)
{
TRACE("device %p, rect %p.\n", device, rect);
*rect = device->updateStateBlock->state.scissor_rect;
TRACE("Returning rect %s.\n", wine_dbgstr_rect(rect));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_vertex_declaration(struct wined3d_device *device,
struct wined3d_vertex_declaration *declaration)
{
struct wined3d_vertex_declaration *prev = device->updateStateBlock->state.vertex_declaration;
TRACE("device %p, declaration %p.\n", device, declaration);
if (declaration)
wined3d_vertex_declaration_incref(declaration);
if (prev)
wined3d_vertex_declaration_decref(prev);
device->updateStateBlock->state.vertex_declaration = declaration;
device->updateStateBlock->changed.vertexDecl = TRUE;
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
else if (declaration == prev)
{
/* Checked after the assignment to allow proper stateblock recording. */
TRACE("Application is setting the old declaration over, nothing to do.\n");
return WINED3D_OK;
}
device_invalidate_state(device, STATE_VDECL);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_vertex_declaration(const struct wined3d_device *device,
struct wined3d_vertex_declaration **declaration)
{
TRACE("device %p, declaration %p.\n", device, declaration);
*declaration = device->stateBlock->state.vertex_declaration;
if (*declaration)
wined3d_vertex_declaration_incref(*declaration);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_vertex_shader(struct wined3d_device *device, struct wined3d_shader *shader)
{
struct wined3d_shader *prev = device->updateStateBlock->state.vertex_shader;
TRACE("device %p, shader %p.\n", device, shader);
device->updateStateBlock->state.vertex_shader = shader;
device->updateStateBlock->changed.vertexShader = TRUE;
if (device->isRecordingState)
{
if (shader)
wined3d_shader_incref(shader);
if (prev)
wined3d_shader_decref(prev);
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
if (shader == prev)
{
TRACE("Application is setting the old shader over, nothing to do.\n");
return WINED3D_OK;
}
if (shader)
wined3d_shader_incref(shader);
if (prev)
wined3d_shader_decref(prev);
device_invalidate_state(device, STATE_VSHADER);
return WINED3D_OK;
}
struct wined3d_shader * CDECL wined3d_device_get_vertex_shader(const struct wined3d_device *device)
{
struct wined3d_shader *shader;
TRACE("device %p.\n", device);
shader = device->stateBlock->state.vertex_shader;
if (shader)
wined3d_shader_incref(shader);
TRACE("Returning %p.\n", shader);
return shader;
}
HRESULT CDECL wined3d_device_set_vs_consts_b(struct wined3d_device *device,
UINT start_register, const BOOL *constants, UINT bool_count)
{
UINT count = min(bool_count, MAX_CONST_B - start_register);
UINT i;
TRACE("device %p, start_register %u, constants %p, bool_count %u.\n",
device, start_register, constants, bool_count);
if (!constants || start_register >= MAX_CONST_B)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.vs_consts_b[start_register], constants, count * sizeof(BOOL));
for (i = 0; i < count; ++i)
TRACE("Set BOOL constant %u to %s.\n", start_register + i, constants[i] ? "true" : "false");
for (i = start_register; i < count + start_register; ++i)
device->updateStateBlock->changed.vertexShaderConstantsB |= (1 << i);
if (!device->isRecordingState)
device_invalidate_state(device, STATE_VERTEXSHADERCONSTANT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_vs_consts_b(const struct wined3d_device *device,
UINT start_register, BOOL *constants, UINT bool_count)
{
UINT count = min(bool_count, MAX_CONST_B - start_register);
TRACE("device %p, start_register %u, constants %p, bool_count %u.\n",
device, start_register, constants, bool_count);
if (!constants || start_register >= MAX_CONST_B)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.vs_consts_b[start_register], count * sizeof(BOOL));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_vs_consts_i(struct wined3d_device *device,
UINT start_register, const int *constants, UINT vector4i_count)
{
UINT count = min(vector4i_count, MAX_CONST_I - start_register);
UINT i;
TRACE("device %p, start_register %u, constants %p, vector4i_count %u.\n",
device, start_register, constants, vector4i_count);
if (!constants || start_register >= MAX_CONST_I)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.vs_consts_i[start_register * 4], constants, count * sizeof(int) * 4);
for (i = 0; i < count; ++i)
TRACE("Set INT constant %u to {%d, %d, %d, %d}.\n", start_register + i,
constants[i * 4], constants[i * 4 + 1],
constants[i * 4 + 2], constants[i * 4 + 3]);
for (i = start_register; i < count + start_register; ++i)
device->updateStateBlock->changed.vertexShaderConstantsI |= (1 << i);
if (!device->isRecordingState)
device_invalidate_state(device, STATE_VERTEXSHADERCONSTANT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_vs_consts_i(const struct wined3d_device *device,
UINT start_register, int *constants, UINT vector4i_count)
{
UINT count = min(vector4i_count, MAX_CONST_I - start_register);
TRACE("device %p, start_register %u, constants %p, vector4i_count %u.\n",
device, start_register, constants, vector4i_count);
if (!constants || start_register >= MAX_CONST_I)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.vs_consts_i[start_register * 4], count * sizeof(int) * 4);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_vs_consts_f(struct wined3d_device *device,
UINT start_register, const float *constants, UINT vector4f_count)
{
UINT i;
TRACE("device %p, start_register %u, constants %p, vector4f_count %u.\n",
device, start_register, constants, vector4f_count);
/* Specifically test start_register > limit to catch MAX_UINT overflows
* when adding start_register + vector4f_count. */
if (!constants
|| start_register + vector4f_count > device->d3d_vshader_constantF
|| start_register > device->d3d_vshader_constantF)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.vs_consts_f[start_register * 4],
constants, vector4f_count * sizeof(float) * 4);
if (TRACE_ON(d3d))
{
for (i = 0; i < vector4f_count; ++i)
TRACE("Set FLOAT constant %u to {%.8e, %.8e, %.8e, %.8e}.\n", start_register + i,
constants[i * 4], constants[i * 4 + 1],
constants[i * 4 + 2], constants[i * 4 + 3]);
}
if (!device->isRecordingState)
{
device->shader_backend->shader_update_float_vertex_constants(device, start_register, vector4f_count);
device_invalidate_state(device, STATE_VERTEXSHADERCONSTANT);
}
memset(device->updateStateBlock->changed.vertexShaderConstantsF + start_register, 1,
sizeof(*device->updateStateBlock->changed.vertexShaderConstantsF) * vector4f_count);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_vs_consts_f(const struct wined3d_device *device,
UINT start_register, float *constants, UINT vector4f_count)
{
int count = min(vector4f_count, device->d3d_vshader_constantF - start_register);
TRACE("device %p, start_register %u, constants %p, vector4f_count %u.\n",
device, start_register, constants, vector4f_count);
if (!constants || count < 0)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.vs_consts_f[start_register * 4], count * sizeof(float) * 4);
return WINED3D_OK;
}
static void device_invalidate_texture_stage(const struct wined3d_device *device, DWORD stage)
{
DWORD i;
for (i = 0; i <= WINED3D_HIGHEST_TEXTURE_STATE; ++i)
{
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, i));
}
}
static void device_map_stage(struct wined3d_device *device, DWORD stage, DWORD unit)
{
DWORD i = device->rev_tex_unit_map[unit];
DWORD j = device->texUnitMap[stage];
device->texUnitMap[stage] = unit;
if (i != WINED3D_UNMAPPED_STAGE && i != stage)
device->texUnitMap[i] = WINED3D_UNMAPPED_STAGE;
device->rev_tex_unit_map[unit] = stage;
if (j != WINED3D_UNMAPPED_STAGE && j != unit)
device->rev_tex_unit_map[j] = WINED3D_UNMAPPED_STAGE;
}
static void device_update_fixed_function_usage_map(struct wined3d_device *device)
{
UINT i;
device->fixed_function_usage_map = 0;
for (i = 0; i < MAX_TEXTURES; ++i)
{
const struct wined3d_state *state = &device->stateBlock->state;
enum wined3d_texture_op color_op = state->texture_states[i][WINED3D_TSS_COLOR_OP];
enum wined3d_texture_op alpha_op = state->texture_states[i][WINED3D_TSS_ALPHA_OP];
DWORD color_arg1 = state->texture_states[i][WINED3D_TSS_COLOR_ARG1] & WINED3DTA_SELECTMASK;
DWORD color_arg2 = state->texture_states[i][WINED3D_TSS_COLOR_ARG2] & WINED3DTA_SELECTMASK;
DWORD color_arg3 = state->texture_states[i][WINED3D_TSS_COLOR_ARG0] & WINED3DTA_SELECTMASK;
DWORD alpha_arg1 = state->texture_states[i][WINED3D_TSS_ALPHA_ARG1] & WINED3DTA_SELECTMASK;
DWORD alpha_arg2 = state->texture_states[i][WINED3D_TSS_ALPHA_ARG2] & WINED3DTA_SELECTMASK;
DWORD alpha_arg3 = state->texture_states[i][WINED3D_TSS_ALPHA_ARG0] & WINED3DTA_SELECTMASK;
/* Not used, and disable higher stages. */
if (color_op == WINED3D_TOP_DISABLE)
break;
if (((color_arg1 == WINED3DTA_TEXTURE) && color_op != WINED3D_TOP_SELECT_ARG2)
|| ((color_arg2 == WINED3DTA_TEXTURE) && color_op != WINED3D_TOP_SELECT_ARG1)
|| ((color_arg3 == WINED3DTA_TEXTURE)
&& (color_op == WINED3D_TOP_MULTIPLY_ADD || color_op == WINED3D_TOP_LERP))
|| ((alpha_arg1 == WINED3DTA_TEXTURE) && alpha_op != WINED3D_TOP_SELECT_ARG2)
|| ((alpha_arg2 == WINED3DTA_TEXTURE) && alpha_op != WINED3D_TOP_SELECT_ARG1)
|| ((alpha_arg3 == WINED3DTA_TEXTURE)
&& (alpha_op == WINED3D_TOP_MULTIPLY_ADD || alpha_op == WINED3D_TOP_LERP)))
device->fixed_function_usage_map |= (1 << i);
if ((color_op == WINED3D_TOP_BUMPENVMAP || color_op == WINED3D_TOP_BUMPENVMAP_LUMINANCE)
&& i < MAX_TEXTURES - 1)
device->fixed_function_usage_map |= (1 << (i + 1));
}
}
static void device_map_fixed_function_samplers(struct wined3d_device *device, const struct wined3d_gl_info *gl_info)
{
unsigned int i, tex;
WORD ffu_map;
device_update_fixed_function_usage_map(device);
ffu_map = device->fixed_function_usage_map;
if (device->max_ffp_textures == gl_info->limits.texture_stages
|| device->stateBlock->state.lowest_disabled_stage <= device->max_ffp_textures)
{
for (i = 0; ffu_map; ffu_map >>= 1, ++i)
{
if (!(ffu_map & 1)) continue;
if (device->texUnitMap[i] != i)
{
device_map_stage(device, i, i);
device_invalidate_state(device, STATE_SAMPLER(i));
device_invalidate_texture_stage(device, i);
}
}
return;
}
/* Now work out the mapping */
tex = 0;
for (i = 0; ffu_map; ffu_map >>= 1, ++i)
{
if (!(ffu_map & 1)) continue;
if (device->texUnitMap[i] != tex)
{
device_map_stage(device, i, tex);
device_invalidate_state(device, STATE_SAMPLER(i));
device_invalidate_texture_stage(device, i);
}
++tex;
}
}
static void device_map_psamplers(struct wined3d_device *device, const struct wined3d_gl_info *gl_info)
{
const enum wined3d_sampler_texture_type *sampler_type =
device->stateBlock->state.pixel_shader->reg_maps.sampler_type;
unsigned int i;
for (i = 0; i < MAX_FRAGMENT_SAMPLERS; ++i)
{
if (sampler_type[i] && device->texUnitMap[i] != i)
{
device_map_stage(device, i, i);
device_invalidate_state(device, STATE_SAMPLER(i));
if (i < gl_info->limits.texture_stages)
device_invalidate_texture_stage(device, i);
}
}
}
static BOOL device_unit_free_for_vs(const struct wined3d_device *device,
const enum wined3d_sampler_texture_type *pshader_sampler_tokens,
const enum wined3d_sampler_texture_type *vshader_sampler_tokens, DWORD unit)
{
DWORD current_mapping = device->rev_tex_unit_map[unit];
/* Not currently used */
if (current_mapping == WINED3D_UNMAPPED_STAGE) return TRUE;
if (current_mapping < MAX_FRAGMENT_SAMPLERS) {
/* Used by a fragment sampler */
if (!pshader_sampler_tokens) {
/* No pixel shader, check fixed function */
return current_mapping >= MAX_TEXTURES || !(device->fixed_function_usage_map & (1 << current_mapping));
}
/* Pixel shader, check the shader's sampler map */
return !pshader_sampler_tokens[current_mapping];
}
/* Used by a vertex sampler */
return !vshader_sampler_tokens[current_mapping - MAX_FRAGMENT_SAMPLERS];
}
static void device_map_vsamplers(struct wined3d_device *device, BOOL ps, const struct wined3d_gl_info *gl_info)
{
const enum wined3d_sampler_texture_type *vshader_sampler_type =
device->stateBlock->state.vertex_shader->reg_maps.sampler_type;
const enum wined3d_sampler_texture_type *pshader_sampler_type = NULL;
int start = min(MAX_COMBINED_SAMPLERS, gl_info->limits.combined_samplers) - 1;
int i;
if (ps)
{
/* Note that we only care if a sampler is sampled or not, not the sampler's specific type.
* Otherwise we'd need to call shader_update_samplers() here for 1.x pixelshaders. */
pshader_sampler_type = device->stateBlock->state.pixel_shader->reg_maps.sampler_type;
}
for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i) {
DWORD vsampler_idx = i + MAX_FRAGMENT_SAMPLERS;
if (vshader_sampler_type[i])
{
if (device->texUnitMap[vsampler_idx] != WINED3D_UNMAPPED_STAGE)
{
/* Already mapped somewhere */
continue;
}
while (start >= 0)
{
if (device_unit_free_for_vs(device, pshader_sampler_type, vshader_sampler_type, start))
{
device_map_stage(device, vsampler_idx, start);
device_invalidate_state(device, STATE_SAMPLER(vsampler_idx));
--start;
break;
}
--start;
}
}
}
}
void device_update_tex_unit_map(struct wined3d_device *device)
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
const struct wined3d_state *state = &device->stateBlock->state;
BOOL vs = use_vs(state);
BOOL ps = use_ps(state);
/*
* Rules are:
* -> Pixel shaders need a 1:1 map. In theory the shader input could be mapped too, but
* that would be really messy and require shader recompilation
* -> When the mapping of a stage is changed, sampler and ALL texture stage states have
* to be reset. Because of that try to work with a 1:1 mapping as much as possible
*/
if (ps)
device_map_psamplers(device, gl_info);
else
device_map_fixed_function_samplers(device, gl_info);
if (vs)
device_map_vsamplers(device, ps, gl_info);
}
HRESULT CDECL wined3d_device_set_pixel_shader(struct wined3d_device *device, struct wined3d_shader *shader)
{
struct wined3d_shader *prev = device->updateStateBlock->state.pixel_shader;
TRACE("device %p, shader %p.\n", device, shader);
device->updateStateBlock->state.pixel_shader = shader;
device->updateStateBlock->changed.pixelShader = TRUE;
if (device->isRecordingState)
{
if (shader)
wined3d_shader_incref(shader);
if (prev)
wined3d_shader_decref(prev);
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
if (shader == prev)
{
TRACE("Application is setting the old shader over, nothing to do.\n");
return WINED3D_OK;
}
if (shader)
wined3d_shader_incref(shader);
if (prev)
wined3d_shader_decref(prev);
device_invalidate_state(device, STATE_PIXELSHADER);
return WINED3D_OK;
}
struct wined3d_shader * CDECL wined3d_device_get_pixel_shader(const struct wined3d_device *device)
{
struct wined3d_shader *shader;
TRACE("device %p.\n", device);
shader = device->stateBlock->state.pixel_shader;
if (shader)
wined3d_shader_incref(shader);
TRACE("Returning %p.\n", shader);
return shader;
}
HRESULT CDECL wined3d_device_set_ps_consts_b(struct wined3d_device *device,
UINT start_register, const BOOL *constants, UINT bool_count)
{
UINT count = min(bool_count, MAX_CONST_B - start_register);
UINT i;
TRACE("device %p, start_register %u, constants %p, bool_count %u.\n",
device, start_register, constants, bool_count);
if (!constants || start_register >= MAX_CONST_B)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.ps_consts_b[start_register], constants, count * sizeof(BOOL));
for (i = 0; i < count; ++i)
TRACE("Set BOOL constant %u to %s.\n", start_register + i, constants[i] ? "true" : "false");
for (i = start_register; i < count + start_register; ++i)
device->updateStateBlock->changed.pixelShaderConstantsB |= (1 << i);
if (!device->isRecordingState)
device_invalidate_state(device, STATE_PIXELSHADERCONSTANT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_ps_consts_b(const struct wined3d_device *device,
UINT start_register, BOOL *constants, UINT bool_count)
{
UINT count = min(bool_count, MAX_CONST_B - start_register);
TRACE("device %p, start_register %u, constants %p, bool_count %u.\n",
device, start_register, constants, bool_count);
if (!constants || start_register >= MAX_CONST_B)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.ps_consts_b[start_register], count * sizeof(BOOL));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_ps_consts_i(struct wined3d_device *device,
UINT start_register, const int *constants, UINT vector4i_count)
{
UINT count = min(vector4i_count, MAX_CONST_I - start_register);
UINT i;
TRACE("device %p, start_register %u, constants %p, vector4i_count %u.\n",
device, start_register, constants, vector4i_count);
if (!constants || start_register >= MAX_CONST_I)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.ps_consts_i[start_register * 4], constants, count * sizeof(int) * 4);
for (i = 0; i < count; ++i)
TRACE("Set INT constant %u to {%d, %d, %d, %d}.\n", start_register + i,
constants[i * 4], constants[i * 4 + 1],
constants[i * 4 + 2], constants[i * 4 + 3]);
for (i = start_register; i < count + start_register; ++i)
device->updateStateBlock->changed.pixelShaderConstantsI |= (1 << i);
if (!device->isRecordingState)
device_invalidate_state(device, STATE_PIXELSHADERCONSTANT);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_ps_consts_i(const struct wined3d_device *device,
UINT start_register, int *constants, UINT vector4i_count)
{
UINT count = min(vector4i_count, MAX_CONST_I - start_register);
TRACE("device %p, start_register %u, constants %p, vector4i_count %u.\n",
device, start_register, constants, vector4i_count);
if (!constants || start_register >= MAX_CONST_I)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.ps_consts_i[start_register * 4], count * sizeof(int) * 4);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_ps_consts_f(struct wined3d_device *device,
UINT start_register, const float *constants, UINT vector4f_count)
{
UINT i;
TRACE("device %p, start_register %u, constants %p, vector4f_count %u.\n",
device, start_register, constants, vector4f_count);
/* Specifically test start_register > limit to catch MAX_UINT overflows
* when adding start_register + vector4f_count. */
if (!constants
|| start_register + vector4f_count > device->d3d_pshader_constantF
|| start_register > device->d3d_pshader_constantF)
return WINED3DERR_INVALIDCALL;
memcpy(&device->updateStateBlock->state.ps_consts_f[start_register * 4],
constants, vector4f_count * sizeof(float) * 4);
if (TRACE_ON(d3d))
{
for (i = 0; i < vector4f_count; ++i)
TRACE("Set FLOAT constant %u to {%.8e, %.8e, %.8e, %.8e}.\n", start_register + i,
constants[i * 4], constants[i * 4 + 1],
constants[i * 4 + 2], constants[i * 4 + 3]);
}
if (!device->isRecordingState)
{
device->shader_backend->shader_update_float_pixel_constants(device, start_register, vector4f_count);
device_invalidate_state(device, STATE_PIXELSHADERCONSTANT);
}
memset(device->updateStateBlock->changed.pixelShaderConstantsF + start_register, 1,
sizeof(*device->updateStateBlock->changed.pixelShaderConstantsF) * vector4f_count);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_ps_consts_f(const struct wined3d_device *device,
UINT start_register, float *constants, UINT vector4f_count)
{
int count = min(vector4f_count, device->d3d_pshader_constantF - start_register);
TRACE("device %p, start_register %u, constants %p, vector4f_count %u.\n",
device, start_register, constants, vector4f_count);
if (!constants || count < 0)
return WINED3DERR_INVALIDCALL;
memcpy(constants, &device->stateBlock->state.ps_consts_f[start_register * 4], count * sizeof(float) * 4);
return WINED3D_OK;
}
/* Context activation is done by the caller. */
/* Do not call while under the GL lock. */
#define copy_and_next(dest, src, size) memcpy(dest, src, size); dest += (size)
static HRESULT process_vertices_strided(const struct wined3d_device *device, DWORD dwDestIndex, DWORD dwCount,
const struct wined3d_stream_info *stream_info, struct wined3d_buffer *dest, DWORD flags,
DWORD DestFVF)
{
struct wined3d_matrix mat, proj_mat, view_mat, world_mat;
struct wined3d_viewport vp;
UINT vertex_size;
unsigned int i;
BYTE *dest_ptr;
BOOL doClip;
DWORD numTextures;
HRESULT hr;
if (stream_info->use_map & (1 << WINED3D_FFP_NORMAL))
{
WARN(" lighting state not saved yet... Some strange stuff may happen !\n");
}
if (!(stream_info->use_map & (1 << WINED3D_FFP_POSITION)))
{
ERR("Source has no position mask\n");
return WINED3DERR_INVALIDCALL;
}
if (device->stateBlock->state.render_states[WINED3D_RS_CLIPPING])
{
static BOOL warned = FALSE;
/*
* The clipping code is not quite correct. Some things need
* to be checked against IDirect3DDevice3 (!), d3d8 and d3d9,
* so disable clipping for now.
* (The graphics in Half-Life are broken, and my processvertices
* test crashes with IDirect3DDevice3)
doClip = TRUE;
*/
doClip = FALSE;
if(!warned) {
warned = TRUE;
FIXME("Clipping is broken and disabled for now\n");
}
}
else
doClip = FALSE;
vertex_size = get_flexible_vertex_size(DestFVF);
if (FAILED(hr = wined3d_buffer_map(dest, dwDestIndex * vertex_size, dwCount * vertex_size, &dest_ptr, 0)))
{
WARN("Failed to map buffer, hr %#x.\n", hr);
return hr;
}
wined3d_device_get_transform(device, WINED3D_TS_VIEW, &view_mat);
wined3d_device_get_transform(device, WINED3D_TS_PROJECTION, &proj_mat);
wined3d_device_get_transform(device, WINED3D_TS_WORLD_MATRIX(0), &world_mat);
TRACE("View mat:\n");
TRACE("%f %f %f %f\n", view_mat.u.s._11, view_mat.u.s._12, view_mat.u.s._13, view_mat.u.s._14);
TRACE("%f %f %f %f\n", view_mat.u.s._21, view_mat.u.s._22, view_mat.u.s._23, view_mat.u.s._24);
TRACE("%f %f %f %f\n", view_mat.u.s._31, view_mat.u.s._32, view_mat.u.s._33, view_mat.u.s._34);
TRACE("%f %f %f %f\n", view_mat.u.s._41, view_mat.u.s._42, view_mat.u.s._43, view_mat.u.s._44);
TRACE("Proj mat:\n");
TRACE("%f %f %f %f\n", proj_mat.u.s._11, proj_mat.u.s._12, proj_mat.u.s._13, proj_mat.u.s._14);
TRACE("%f %f %f %f\n", proj_mat.u.s._21, proj_mat.u.s._22, proj_mat.u.s._23, proj_mat.u.s._24);
TRACE("%f %f %f %f\n", proj_mat.u.s._31, proj_mat.u.s._32, proj_mat.u.s._33, proj_mat.u.s._34);
TRACE("%f %f %f %f\n", proj_mat.u.s._41, proj_mat.u.s._42, proj_mat.u.s._43, proj_mat.u.s._44);
TRACE("World mat:\n");
TRACE("%f %f %f %f\n", world_mat.u.s._11, world_mat.u.s._12, world_mat.u.s._13, world_mat.u.s._14);
TRACE("%f %f %f %f\n", world_mat.u.s._21, world_mat.u.s._22, world_mat.u.s._23, world_mat.u.s._24);
TRACE("%f %f %f %f\n", world_mat.u.s._31, world_mat.u.s._32, world_mat.u.s._33, world_mat.u.s._34);
TRACE("%f %f %f %f\n", world_mat.u.s._41, world_mat.u.s._42, world_mat.u.s._43, world_mat.u.s._44);
/* Get the viewport */
wined3d_device_get_viewport(device, &vp);
TRACE("viewport x %u, y %u, width %u, height %u, min_z %.8e, max_z %.8e.\n",
vp.x, vp.y, vp.width, vp.height, vp.min_z, vp.max_z);
multiply_matrix(&mat,&view_mat,&world_mat);
multiply_matrix(&mat,&proj_mat,&mat);
numTextures = (DestFVF & WINED3DFVF_TEXCOUNT_MASK) >> WINED3DFVF_TEXCOUNT_SHIFT;
for (i = 0; i < dwCount; i+= 1) {
unsigned int tex_index;
if ( ((DestFVF & WINED3DFVF_POSITION_MASK) == WINED3DFVF_XYZ ) ||
((DestFVF & WINED3DFVF_POSITION_MASK) == WINED3DFVF_XYZRHW ) ) {
/* The position first */
const struct wined3d_stream_info_element *element = &stream_info->elements[WINED3D_FFP_POSITION];
const float *p = (const float *)(element->data.addr + i * element->stride);
float x, y, z, rhw;
TRACE("In: ( %06.2f %06.2f %06.2f )\n", p[0], p[1], p[2]);
/* Multiplication with world, view and projection matrix */
x = (p[0] * mat.u.s._11) + (p[1] * mat.u.s._21) + (p[2] * mat.u.s._31) + (1.0f * mat.u.s._41);
y = (p[0] * mat.u.s._12) + (p[1] * mat.u.s._22) + (p[2] * mat.u.s._32) + (1.0f * mat.u.s._42);
z = (p[0] * mat.u.s._13) + (p[1] * mat.u.s._23) + (p[2] * mat.u.s._33) + (1.0f * mat.u.s._43);
rhw = (p[0] * mat.u.s._14) + (p[1] * mat.u.s._24) + (p[2] * mat.u.s._34) + (1.0f * mat.u.s._44);
TRACE("x=%f y=%f z=%f rhw=%f\n", x, y, z, rhw);
/* WARNING: The following things are taken from d3d7 and were not yet checked
* against d3d8 or d3d9!
*/
/* Clipping conditions: From msdn
*
* A vertex is clipped if it does not match the following requirements
* -rhw < x <= rhw
* -rhw < y <= rhw
* 0 < z <= rhw
* 0 < rhw ( Not in d3d7, but tested in d3d7)
*
* If clipping is on is determined by the D3DVOP_CLIP flag in D3D7, and
* by the D3DRS_CLIPPING in D3D9(according to the msdn, not checked)
*
*/
if( !doClip ||
( (-rhw -eps < x) && (-rhw -eps < y) && ( -eps < z) &&
(x <= rhw + eps) && (y <= rhw + eps ) && (z <= rhw + eps) &&
( rhw > eps ) ) ) {
/* "Normal" viewport transformation (not clipped)
* 1) The values are divided by rhw
* 2) The y axis is negative, so multiply it with -1
* 3) Screen coordinates go from -(Width/2) to +(Width/2) and
* -(Height/2) to +(Height/2). The z range is MinZ to MaxZ
* 4) Multiply x with Width/2 and add Width/2
* 5) The same for the height
* 6) Add the viewpoint X and Y to the 2D coordinates and
* The minimum Z value to z
* 7) rhw = 1 / rhw Reciprocal of Homogeneous W....
*
* Well, basically it's simply a linear transformation into viewport
* coordinates
*/
x /= rhw;
y /= rhw;
z /= rhw;
y *= -1;
x *= vp.width / 2;
y *= vp.height / 2;
z *= vp.max_z - vp.min_z;
x += vp.width / 2 + vp.x;
y += vp.height / 2 + vp.y;
z += vp.min_z;
rhw = 1 / rhw;
} else {
/* That vertex got clipped
* Contrary to OpenGL it is not dropped completely, it just
* undergoes a different calculation.
*/
TRACE("Vertex got clipped\n");
x += rhw;
y += rhw;
x /= 2;
y /= 2;
/* Msdn mentions that Direct3D9 keeps a list of clipped vertices
* outside of the main vertex buffer memory. That needs some more
* investigation...
*/
}
TRACE("Writing (%f %f %f) %f\n", x, y, z, rhw);
( (float *) dest_ptr)[0] = x;
( (float *) dest_ptr)[1] = y;
( (float *) dest_ptr)[2] = z;
( (float *) dest_ptr)[3] = rhw; /* SIC, see ddraw test! */
dest_ptr += 3 * sizeof(float);
if ((DestFVF & WINED3DFVF_POSITION_MASK) == WINED3DFVF_XYZRHW)
dest_ptr += sizeof(float);
}
if (DestFVF & WINED3DFVF_PSIZE)
dest_ptr += sizeof(DWORD);
if (DestFVF & WINED3DFVF_NORMAL)
{
const struct wined3d_stream_info_element *element = &stream_info->elements[WINED3D_FFP_NORMAL];
const float *normal = (const float *)(element->data.addr + i * element->stride);
/* AFAIK this should go into the lighting information */
FIXME("Didn't expect the destination to have a normal\n");
copy_and_next(dest_ptr, normal, 3 * sizeof(float));
}
if (DestFVF & WINED3DFVF_DIFFUSE)
{
const struct wined3d_stream_info_element *element = &stream_info->elements[WINED3D_FFP_DIFFUSE];
const DWORD *color_d = (const DWORD *)(element->data.addr + i * element->stride);
if (!(stream_info->use_map & (1 << WINED3D_FFP_DIFFUSE)))
{
static BOOL warned = FALSE;
if(!warned) {
ERR("No diffuse color in source, but destination has one\n");
warned = TRUE;
}
*( (DWORD *) dest_ptr) = 0xffffffff;
dest_ptr += sizeof(DWORD);
}
else
{
copy_and_next(dest_ptr, color_d, sizeof(DWORD));
}
}
if (DestFVF & WINED3DFVF_SPECULAR)
{
/* What's the color value in the feedback buffer? */
const struct wined3d_stream_info_element *element = &stream_info->elements[WINED3D_FFP_SPECULAR];
const DWORD *color_s = (const DWORD *)(element->data.addr + i * element->stride);
if (!(stream_info->use_map & (1 << WINED3D_FFP_SPECULAR)))
{
static BOOL warned = FALSE;
if(!warned) {
ERR("No specular color in source, but destination has one\n");
warned = TRUE;
}
*( (DWORD *) dest_ptr) = 0xFF000000;
dest_ptr += sizeof(DWORD);
}
else
{
copy_and_next(dest_ptr, color_s, sizeof(DWORD));
}
}
for (tex_index = 0; tex_index < numTextures; ++tex_index)
{
const struct wined3d_stream_info_element *element = &stream_info->elements[WINED3D_FFP_TEXCOORD0 + tex_index];
const float *tex_coord = (const float *)(element->data.addr + i * element->stride);
if (!(stream_info->use_map & (1 << (WINED3D_FFP_TEXCOORD0 + tex_index))))
{
ERR("No source texture, but destination requests one\n");
dest_ptr += GET_TEXCOORD_SIZE_FROM_FVF(DestFVF, tex_index) * sizeof(float);
}
else
{
copy_and_next(dest_ptr, tex_coord, GET_TEXCOORD_SIZE_FROM_FVF(DestFVF, tex_index) * sizeof(float));
}
}
}
wined3d_buffer_unmap(dest);
return WINED3D_OK;
}
#undef copy_and_next
/* Do not call while under the GL lock. */
HRESULT CDECL wined3d_device_process_vertices(struct wined3d_device *device,
UINT src_start_idx, UINT dst_idx, UINT vertex_count, struct wined3d_buffer *dst_buffer,
const struct wined3d_vertex_declaration *declaration, DWORD flags, DWORD dst_fvf)
{
struct wined3d_state *state = &device->stateBlock->state;
struct wined3d_stream_info stream_info;
const struct wined3d_gl_info *gl_info;
BOOL streamWasUP = state->user_stream;
struct wined3d_context *context;
struct wined3d_shader *vs;
unsigned int i;
HRESULT hr;
TRACE("device %p, src_start_idx %u, dst_idx %u, vertex_count %u, "
"dst_buffer %p, declaration %p, flags %#x, dst_fvf %#x.\n",
device, src_start_idx, dst_idx, vertex_count,
dst_buffer, declaration, flags, dst_fvf);
if (declaration)
FIXME("Output vertex declaration not implemented yet.\n");
/* Need any context to write to the vbo. */
context = context_acquire(device, NULL);
gl_info = context->gl_info;
/* ProcessVertices reads from vertex buffers, which have to be assigned.
* DrawPrimitive and DrawPrimitiveUP control the streamIsUP flag, thus
* restore it afterwards. */
vs = state->vertex_shader;
state->vertex_shader = NULL;
state->user_stream = FALSE;
device_stream_info_from_declaration(device, &stream_info);
state->user_stream = streamWasUP;
state->vertex_shader = vs;
/* We can't convert FROM a VBO, and vertex buffers used to source into
* process_vertices() are unlikely to ever be used for drawing. Release
* VBOs in those buffers and fix up the stream_info structure.
*
* Also apply the start index. */
for (i = 0; i < (sizeof(stream_info.elements) / sizeof(*stream_info.elements)); ++i)
{
struct wined3d_stream_info_element *e;
if (!(stream_info.use_map & (1 << i)))
continue;
e = &stream_info.elements[i];
if (e->data.buffer_object)
{
struct wined3d_buffer *vb = state->streams[e->stream_idx].buffer;
e->data.buffer_object = 0;
e->data.addr = (BYTE *)((ULONG_PTR)e->data.addr + (ULONG_PTR)buffer_get_sysmem(vb, gl_info));
ENTER_GL();
GL_EXTCALL(glDeleteBuffersARB(1, &vb->buffer_object));
vb->buffer_object = 0;
LEAVE_GL();
}
if (e->data.addr)
e->data.addr += e->stride * src_start_idx;
}
hr = process_vertices_strided(device, dst_idx, vertex_count,
&stream_info, dst_buffer, flags, dst_fvf);
context_release(context);
return hr;
}
HRESULT CDECL wined3d_device_set_texture_stage_state(struct wined3d_device *device,
UINT stage, enum wined3d_texture_stage_state state, DWORD value)
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
DWORD old_value;
TRACE("device %p, stage %u, state %s, value %#x.\n",
device, stage, debug_d3dtexturestate(state), value);
if (state > WINED3D_HIGHEST_TEXTURE_STATE)
{
WARN("Invalid state %#x passed.\n", state);
return WINED3D_OK;
}
if (stage >= gl_info->limits.texture_stages)
{
WARN("Attempting to set stage %u which is higher than the max stage %u, ignoring.\n",
stage, gl_info->limits.texture_stages - 1);
return WINED3D_OK;
}
old_value = device->updateStateBlock->state.texture_states[stage][state];
device->updateStateBlock->changed.textureState[stage] |= 1 << state;
device->updateStateBlock->state.texture_states[stage][state] = value;
if (device->isRecordingState)
{
TRACE("Recording... not performing anything.\n");
return WINED3D_OK;
}
/* Checked after the assignments to allow proper stateblock recording. */
if (old_value == value)
{
TRACE("Application is setting the old value over, nothing to do.\n");
return WINED3D_OK;
}
if (stage > device->stateBlock->state.lowest_disabled_stage
&& device->StateTable[STATE_TEXTURESTAGE(0, state)].representative
== STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP))
{
/* Colorop change above lowest disabled stage? That won't change
* anything in the GL setup. Changes in other states are important on
* disabled stages too. */
return WINED3D_OK;
}
if (state == WINED3D_TSS_COLOR_OP)
{
unsigned int i;
if (value == WINED3D_TOP_DISABLE && old_value != WINED3D_TOP_DISABLE)
{
/* Previously enabled stage disabled now. Make sure to dirtify
* all enabled stages above stage, they have to be disabled.
*
* The current stage is dirtified below. */
for (i = stage + 1; i < device->stateBlock->state.lowest_disabled_stage; ++i)
{
TRACE("Additionally dirtifying stage %u.\n", i);
device_invalidate_state(device, STATE_TEXTURESTAGE(i, WINED3D_TSS_COLOR_OP));
}
device->stateBlock->state.lowest_disabled_stage = stage;
TRACE("New lowest disabled: %u.\n", stage);
}
else if (value != WINED3D_TOP_DISABLE && old_value == WINED3D_TOP_DISABLE)
{
/* Previously disabled stage enabled. Stages above it may need
* enabling. Stage must be lowest_disabled_stage here, if it's
* bigger success is returned above, and stages below the lowest
* disabled stage can't be enabled (because they are enabled
* already).
*
* Again stage stage doesn't need to be dirtified here, it is
* handled below. */
for (i = stage + 1; i < gl_info->limits.texture_stages; ++i)
{
if (device->updateStateBlock->state.texture_states[i][WINED3D_TSS_COLOR_OP] == WINED3D_TOP_DISABLE)
break;
TRACE("Additionally dirtifying stage %u due to enable.\n", i);
device_invalidate_state(device, STATE_TEXTURESTAGE(i, WINED3D_TSS_COLOR_OP));
}
device->stateBlock->state.lowest_disabled_stage = i;
TRACE("New lowest disabled: %u.\n", i);
}
}
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, state));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_texture_stage_state(const struct wined3d_device *device,
UINT stage, enum wined3d_texture_stage_state state, DWORD *value)
{
TRACE("device %p, stage %u, state %s, value %p.\n",
device, stage, debug_d3dtexturestate(state), value);
if (state > WINED3D_HIGHEST_TEXTURE_STATE)
{
WARN("Invalid state %#x passed.\n", state);
return WINED3D_OK;
}
*value = device->updateStateBlock->state.texture_states[stage][state];
TRACE("Returning %#x.\n", *value);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_texture(struct wined3d_device *device,
UINT stage, struct wined3d_texture *texture)
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
struct wined3d_texture *prev;
TRACE("device %p, stage %u, texture %p.\n", device, stage, texture);
if (stage >= WINED3DVERTEXTEXTURESAMPLER0 && stage <= WINED3DVERTEXTEXTURESAMPLER3)
stage -= (WINED3DVERTEXTEXTURESAMPLER0 - MAX_FRAGMENT_SAMPLERS);
/* Windows accepts overflowing this array... we do not. */
if (stage >= sizeof(device->stateBlock->state.textures) / sizeof(*device->stateBlock->state.textures))
{
WARN("Ignoring invalid stage %u.\n", stage);
return WINED3D_OK;
}
if (texture && texture->resource.pool == WINED3D_POOL_SCRATCH)
{
WARN("Rejecting attempt to set scratch texture.\n");
return WINED3DERR_INVALIDCALL;
}
device->updateStateBlock->changed.textures |= 1 << stage;
prev = device->updateStateBlock->state.textures[stage];
TRACE("Previous texture %p.\n", prev);
if (texture == prev)
{
TRACE("App is setting the same texture again, nothing to do.\n");
return WINED3D_OK;
}
TRACE("Setting new texture to %p.\n", texture);
device->updateStateBlock->state.textures[stage] = texture;
if (device->isRecordingState)
{
TRACE("Recording... not performing anything\n");
if (texture) wined3d_texture_incref(texture);
if (prev) wined3d_texture_decref(prev);
return WINED3D_OK;
}
if (texture)
{
LONG bind_count = InterlockedIncrement(&texture->resource.bind_count);
wined3d_texture_incref(texture);
if (!prev || texture->target != prev->target)
device_invalidate_state(device, STATE_PIXELSHADER);
if (!prev && stage < gl_info->limits.texture_stages)
{
/* The source arguments for color and alpha ops have different
* meanings when a NULL texture is bound, so the COLOR_OP and
* ALPHA_OP have to be dirtified. */
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, WINED3D_TSS_COLOR_OP));
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, WINED3D_TSS_ALPHA_OP));
}
if (bind_count == 1)
texture->sampler = stage;
}
if (prev)
{
LONG bind_count = InterlockedDecrement(&prev->resource.bind_count);
wined3d_texture_decref(prev);
if (!texture && stage < gl_info->limits.texture_stages)
{
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, WINED3D_TSS_COLOR_OP));
device_invalidate_state(device, STATE_TEXTURESTAGE(stage, WINED3D_TSS_ALPHA_OP));
}
if (bind_count && prev->sampler == stage)
{
unsigned int i;
/* Search for other stages the texture is bound to. Shouldn't
* happen if applications bind textures to a single stage only. */
TRACE("Searching for other stages the texture is bound to.\n");
for (i = 0; i < MAX_COMBINED_SAMPLERS; ++i)
{
if (device->updateStateBlock->state.textures[i] == prev)
{
TRACE("Texture is also bound to stage %u.\n", i);
prev->sampler = i;
break;
}
}
}
}
device_invalidate_state(device, STATE_SAMPLER(stage));
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_texture(const struct wined3d_device *device,
UINT stage, struct wined3d_texture **texture)
{
TRACE("device %p, stage %u, texture %p.\n", device, stage, texture);
if (stage >= WINED3DVERTEXTEXTURESAMPLER0 && stage <= WINED3DVERTEXTEXTURESAMPLER3)
stage -= (WINED3DVERTEXTEXTURESAMPLER0 - MAX_FRAGMENT_SAMPLERS);
if (stage >= sizeof(device->stateBlock->state.textures) / sizeof(*device->stateBlock->state.textures))
{
WARN("Ignoring invalid stage %u.\n", stage);
return WINED3D_OK; /* Windows accepts overflowing this array ... we do not. */
}
*texture = device->stateBlock->state.textures[stage];
if (*texture)
wined3d_texture_incref(*texture);
TRACE("Returning %p.\n", *texture);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_back_buffer(const struct wined3d_device *device, UINT swapchain_idx,
UINT backbuffer_idx, enum wined3d_backbuffer_type backbuffer_type, struct wined3d_surface **backbuffer)
{
struct wined3d_swapchain *swapchain;
HRESULT hr;
TRACE("device %p, swapchain_idx %u, backbuffer_idx %u, backbuffer_type %#x, backbuffer %p.\n",
device, swapchain_idx, backbuffer_idx, backbuffer_type, backbuffer);
hr = wined3d_device_get_swapchain(device, swapchain_idx, &swapchain);
if (FAILED(hr))
{
WARN("Failed to get swapchain %u, hr %#x.\n", swapchain_idx, hr);
return hr;
}
hr = wined3d_swapchain_get_back_buffer(swapchain, backbuffer_idx, backbuffer_type, backbuffer);
wined3d_swapchain_decref(swapchain);
if (FAILED(hr))
{
WARN("Failed to get backbuffer %u, hr %#x.\n", backbuffer_idx, hr);
return hr;
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_device_caps(const struct wined3d_device *device, WINED3DCAPS *caps)
{
TRACE("device %p, caps %p.\n", device, caps);
return wined3d_get_device_caps(device->wined3d, device->adapter->ordinal,
device->create_parms.device_type, caps);
}
HRESULT CDECL wined3d_device_get_display_mode(const struct wined3d_device *device,
UINT swapchain_idx, struct wined3d_display_mode *mode)
{
struct wined3d_swapchain *swapchain;
HRESULT hr;
TRACE("device %p, swapchain_idx %u, mode %p.\n", device, swapchain_idx, mode);
if (SUCCEEDED(hr = wined3d_device_get_swapchain(device, swapchain_idx, &swapchain)))
{
hr = wined3d_swapchain_get_display_mode(swapchain, mode);
wined3d_swapchain_decref(swapchain);
}
return hr;
}
HRESULT CDECL wined3d_device_begin_stateblock(struct wined3d_device *device)
{
struct wined3d_stateblock *stateblock;
HRESULT hr;
TRACE("device %p.\n", device);
if (device->isRecordingState)
return WINED3DERR_INVALIDCALL;
hr = wined3d_stateblock_create(device, WINED3D_SBT_RECORDED, &stateblock);
if (FAILED(hr))
return hr;
wined3d_stateblock_decref(device->updateStateBlock);
device->updateStateBlock = stateblock;
device->isRecordingState = TRUE;
TRACE("Recording stateblock %p.\n", stateblock);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_end_stateblock(struct wined3d_device *device,
struct wined3d_stateblock **stateblock)
{
struct wined3d_stateblock *object = device->updateStateBlock;
TRACE("device %p, stateblock %p.\n", device, stateblock);
if (!device->isRecordingState)
{
WARN("Not recording.\n");
*stateblock = NULL;
return WINED3DERR_INVALIDCALL;
}
stateblock_init_contained_states(object);
*stateblock = object;
device->isRecordingState = FALSE;
device->updateStateBlock = device->stateBlock;
wined3d_stateblock_incref(device->updateStateBlock);
TRACE("Returning stateblock %p.\n", *stateblock);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_begin_scene(struct wined3d_device *device)
{
/* At the moment we have no need for any functionality at the beginning
* of a scene. */
TRACE("device %p.\n", device);
if (device->inScene)
{
WARN("Already in scene, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
device->inScene = TRUE;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_end_scene(struct wined3d_device *device)
{
struct wined3d_context *context;
TRACE("device %p.\n", device);
if (!device->inScene)
{
WARN("Not in scene, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
context = context_acquire(device, NULL);
/* We only have to do this if we need to read the, swapbuffers performs a flush for us */
wglFlush();
/* No checkGLcall here to avoid locking the lock just for checking a call that hardly ever
* fails. */
context_release(context);
device->inScene = FALSE;
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_present(const struct wined3d_device *device, const RECT *src_rect,
const RECT *dst_rect, HWND dst_window_override, const RGNDATA *dirty_region)
{
UINT i;
TRACE("device %p, src_rect %s, dst_rect %s, dst_window_override %p, dirty_region %p.\n",
device, wine_dbgstr_rect(src_rect), wine_dbgstr_rect(dst_rect),
dst_window_override, dirty_region);
for (i = 0; i < device->swapchain_count; ++i)
{
wined3d_swapchain_present(device->swapchains[i], src_rect,
dst_rect, dst_window_override, dirty_region, 0);
}
return WINED3D_OK;
}
/* Do not call while under the GL lock. */
HRESULT CDECL wined3d_device_clear(struct wined3d_device *device, DWORD rect_count,
const RECT *rects, DWORD flags, const struct wined3d_color *color, float depth, DWORD stencil)
{
RECT draw_rect;
TRACE("device %p, rect_count %u, rects %p, flags %#x, color {%.8e, %.8e, %.8e, %.8e}, depth %.8e, stencil %u.\n",
device, rect_count, rects, flags, color->r, color->g, color->b, color->a, depth, stencil);
if (flags & (WINED3DCLEAR_ZBUFFER | WINED3DCLEAR_STENCIL))
{
struct wined3d_surface *ds = device->fb.depth_stencil;
if (!ds)
{
WARN("Clearing depth and/or stencil without a depth stencil buffer attached, returning WINED3DERR_INVALIDCALL\n");
/* TODO: What about depth stencil buffers without stencil bits? */
return WINED3DERR_INVALIDCALL;
}
else if (flags & WINED3DCLEAR_TARGET)
{
if (ds->resource.width < device->fb.render_targets[0]->resource.width
|| ds->resource.height < device->fb.render_targets[0]->resource.height)
{
WARN("Silently ignoring depth and target clear with mismatching sizes\n");
return WINED3D_OK;
}
}
}
wined3d_get_draw_rect(&device->stateBlock->state, &draw_rect);
device_clear_render_targets(device, device->adapter->gl_info.limits.buffers,
&device->fb, rect_count, rects, &draw_rect, flags, color, depth, stencil);
return WINED3D_OK;
}
void CDECL wined3d_device_set_primitive_type(struct wined3d_device *device,
enum wined3d_primitive_type primitive_type)
{
TRACE("device %p, primitive_type %s\n", device, debug_d3dprimitivetype(primitive_type));
device->updateStateBlock->changed.primitive_type = TRUE;
device->updateStateBlock->state.gl_primitive_type = gl_primitive_type_from_d3d(primitive_type);
}
void CDECL wined3d_device_get_primitive_type(const struct wined3d_device *device,
enum wined3d_primitive_type *primitive_type)
{
TRACE("device %p, primitive_type %p\n", device, primitive_type);
*primitive_type = d3d_primitive_type_from_gl(device->stateBlock->state.gl_primitive_type);
TRACE("Returning %s\n", debug_d3dprimitivetype(*primitive_type));
}
HRESULT CDECL wined3d_device_draw_primitive(struct wined3d_device *device, UINT start_vertex, UINT vertex_count)
{
TRACE("device %p, start_vertex %u, vertex_count %u.\n", device, start_vertex, vertex_count);
if (!device->stateBlock->state.vertex_declaration)
{
WARN("Called without a valid vertex declaration set.\n");
return WINED3DERR_INVALIDCALL;
}
/* The index buffer is not needed here, but restore it, otherwise it is hell to keep track of */
if (device->stateBlock->state.user_stream)
{
device_invalidate_state(device, STATE_INDEXBUFFER);
device->stateBlock->state.user_stream = FALSE;
}
if (device->stateBlock->state.load_base_vertex_index)
{
device->stateBlock->state.load_base_vertex_index = 0;
device_invalidate_state(device, STATE_BASEVERTEXINDEX);
}
/* Account for the loading offset due to index buffers. Instead of
* reloading all sources correct it with the startvertex parameter. */
drawPrimitive(device, vertex_count, start_vertex, FALSE, NULL);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_indexed_primitive(struct wined3d_device *device, UINT start_idx, UINT index_count)
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
TRACE("device %p, start_idx %u, index_count %u.\n", device, start_idx, index_count);
if (!device->stateBlock->state.index_buffer)
{
/* D3D9 returns D3DERR_INVALIDCALL when DrawIndexedPrimitive is called
* without an index buffer set. (The first time at least...)
* D3D8 simply dies, but I doubt it can do much harm to return
* D3DERR_INVALIDCALL there as well. */
WARN("Called without a valid index buffer set, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
if (!device->stateBlock->state.vertex_declaration)
{
WARN("Called without a valid vertex declaration set.\n");
return WINED3DERR_INVALIDCALL;
}
if (device->stateBlock->state.user_stream)
{
device_invalidate_state(device, STATE_INDEXBUFFER);
device->stateBlock->state.user_stream = FALSE;
}
if (!gl_info->supported[ARB_DRAW_ELEMENTS_BASE_VERTEX] &&
device->stateBlock->state.load_base_vertex_index != device->stateBlock->state.base_vertex_index)
{
device->stateBlock->state.load_base_vertex_index = device->stateBlock->state.base_vertex_index;
device_invalidate_state(device, STATE_BASEVERTEXINDEX);
}
drawPrimitive(device, index_count, start_idx, TRUE, NULL);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_primitive_up(struct wined3d_device *device, UINT vertex_count,
const void *stream_data, UINT stream_stride)
{
struct wined3d_stream_state *stream;
struct wined3d_buffer *vb;
TRACE("device %p, vertex count %u, stream_data %p, stream_stride %u.\n",
device, vertex_count, stream_data, stream_stride);
if (!device->stateBlock->state.vertex_declaration)
{
WARN("Called without a valid vertex declaration set.\n");
return WINED3DERR_INVALIDCALL;
}
/* Note in the following, it's not this type, but that's the purpose of streamIsUP */
stream = &device->stateBlock->state.streams[0];
vb = stream->buffer;
stream->buffer = (struct wined3d_buffer *)stream_data;
if (vb)
wined3d_buffer_decref(vb);
stream->offset = 0;
stream->stride = stream_stride;
device->stateBlock->state.user_stream = TRUE;
if (device->stateBlock->state.load_base_vertex_index)
{
device->stateBlock->state.load_base_vertex_index = 0;
device_invalidate_state(device, STATE_BASEVERTEXINDEX);
}
/* TODO: Only mark dirty if drawing from a different UP address */
device_invalidate_state(device, STATE_STREAMSRC);
drawPrimitive(device, vertex_count, 0, FALSE, NULL);
/* MSDN specifies stream zero settings must be set to NULL */
stream->buffer = NULL;
stream->stride = 0;
/* stream zero settings set to null at end, as per the msdn. No need to
* mark dirty here, the app has to set the new stream sources or use UP
* drawing again. */
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_indexed_primitive_up(struct wined3d_device *device,
UINT index_count, const void *index_data, enum wined3d_format_id index_data_format_id,
const void *stream_data, UINT stream_stride)
{
struct wined3d_stream_state *stream;
struct wined3d_buffer *vb, *ib;
TRACE("device %p, index_count %u, index_data %p, index_data_format %s, stream_data %p, stream_stride %u.\n",
device, index_count, index_data, debug_d3dformat(index_data_format_id), stream_data, stream_stride);
if (!device->stateBlock->state.vertex_declaration)
{
WARN("(%p) : Called without a valid vertex declaration set\n", device);
return WINED3DERR_INVALIDCALL;
}
stream = &device->stateBlock->state.streams[0];
vb = stream->buffer;
stream->buffer = (struct wined3d_buffer *)stream_data;
if (vb)
wined3d_buffer_decref(vb);
stream->offset = 0;
stream->stride = stream_stride;
device->stateBlock->state.user_stream = TRUE;
device->stateBlock->state.index_format = index_data_format_id;
/* Set to 0 as per msdn. Do it now due to the stream source loading during drawPrimitive */
device->stateBlock->state.base_vertex_index = 0;
if (device->stateBlock->state.load_base_vertex_index)
{
device->stateBlock->state.load_base_vertex_index = 0;
device_invalidate_state(device, STATE_BASEVERTEXINDEX);
}
/* Invalidate the state until we have nicer tracking of the stream source pointers */
device_invalidate_state(device, STATE_STREAMSRC);
device_invalidate_state(device, STATE_INDEXBUFFER);
drawPrimitive(device, index_count, 0, TRUE, index_data);
/* MSDN specifies stream zero settings and index buffer must be set to NULL */
stream->buffer = NULL;
stream->stride = 0;
ib = device->stateBlock->state.index_buffer;
if (ib)
{
wined3d_buffer_decref(ib);
device->stateBlock->state.index_buffer = NULL;
}
/* No need to mark the stream source state dirty here. Either the app calls UP drawing again, or it has to call
* SetStreamSource to specify a vertex buffer
*/
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_primitive_strided(struct wined3d_device *device,
UINT vertex_count, const struct wined3d_strided_data *strided_data)
{
/* Mark the state dirty until we have nicer tracking. It's fine to change
* baseVertexIndex because that call is only called by ddraw which does
* not need that value. */
device_invalidate_state(device, STATE_VDECL);
device_invalidate_state(device, STATE_STREAMSRC);
device_invalidate_state(device, STATE_INDEXBUFFER);
device->stateBlock->state.base_vertex_index = 0;
device->up_strided = strided_data;
drawPrimitive(device, vertex_count, 0, FALSE, NULL);
device->up_strided = NULL;
/* Invalidate the states again to make sure the values from the stateblock
* are properly applied in the next regular draw. Note that the application-
* provided strided data has ovwritten pretty much the entire vertex and
* and index stream related states */
device_invalidate_state(device, STATE_VDECL);
device_invalidate_state(device, STATE_STREAMSRC);
device_invalidate_state(device, STATE_INDEXBUFFER);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_indexed_primitive_strided(struct wined3d_device *device,
UINT index_count, const struct wined3d_strided_data *strided_data,
UINT vertex_count, const void *index_data, enum wined3d_format_id index_data_format_id)
{
enum wined3d_format_id prev_idx_format;
/* Mark the state dirty until we have nicer tracking
* its fine to change baseVertexIndex because that call is only called by ddraw which does not need
* that value.
*/
device_invalidate_state(device, STATE_VDECL);
device_invalidate_state(device, STATE_STREAMSRC);
device_invalidate_state(device, STATE_INDEXBUFFER);
prev_idx_format = device->stateBlock->state.index_format;
device->stateBlock->state.index_format = index_data_format_id;
device->stateBlock->state.user_stream = TRUE;
device->stateBlock->state.base_vertex_index = 0;
device->up_strided = strided_data;
drawPrimitive(device, index_count, 0, TRUE, index_data);
device->up_strided = NULL;
device->stateBlock->state.index_format = prev_idx_format;
device_invalidate_state(device, STATE_VDECL);
device_invalidate_state(device, STATE_STREAMSRC);
device_invalidate_state(device, STATE_INDEXBUFFER);
return WINED3D_OK;
}
/* This is a helper function for UpdateTexture, there is no UpdateVolume method in D3D. */
static HRESULT device_update_volume(struct wined3d_device *device,
struct wined3d_volume *src_volume, struct wined3d_volume *dst_volume)
{
struct wined3d_map_desc src;
struct wined3d_map_desc dst;
HRESULT hr;
TRACE("device %p, src_volume %p, dst_volume %p.\n",
device, src_volume, dst_volume);
/* TODO: Implement direct loading into the gl volume instead of using
* memcpy and dirtification to improve loading performance. */
if (FAILED(hr = wined3d_volume_map(src_volume, &src, NULL, WINED3D_MAP_READONLY)))
return hr;
if (FAILED(hr = wined3d_volume_map(dst_volume, &dst, NULL, WINED3D_MAP_DISCARD)))
{
wined3d_volume_unmap(src_volume);
return hr;
}
memcpy(dst.data, src.data, dst_volume->resource.size);
hr = wined3d_volume_unmap(dst_volume);
if (FAILED(hr))
wined3d_volume_unmap(src_volume);
else
hr = wined3d_volume_unmap(src_volume);
return hr;
}
HRESULT CDECL wined3d_device_update_texture(struct wined3d_device *device,
struct wined3d_texture *src_texture, struct wined3d_texture *dst_texture)
{
enum wined3d_resource_type type;
unsigned int level_count, i;
HRESULT hr;
TRACE("device %p, src_texture %p, dst_texture %p.\n", device, src_texture, dst_texture);
/* Verify that the source and destination textures are non-NULL. */
if (!src_texture || !dst_texture)
{
WARN("Source and destination textures must be non-NULL, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
if (src_texture == dst_texture)
{
WARN("Source and destination are the same object, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
/* Verify that the source and destination textures are the same type. */
type = src_texture->resource.type;
if (dst_texture->resource.type != type)
{
WARN("Source and destination have different types, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
/* Check that both textures have the identical numbers of levels. */
level_count = wined3d_texture_get_level_count(src_texture);
if (wined3d_texture_get_level_count(dst_texture) != level_count)
{
WARN("Source and destination have different level counts, returning WINED3DERR_INVALIDCALL.\n");
return WINED3DERR_INVALIDCALL;
}
/* Make sure that the destination texture is loaded. */
dst_texture->texture_ops->texture_preload(dst_texture, SRGB_RGB);
/* Update every surface level of the texture. */
switch (type)
{
case WINED3D_RTYPE_TEXTURE:
{
struct wined3d_surface *src_surface;
struct wined3d_surface *dst_surface;
for (i = 0; i < level_count; ++i)
{
src_surface = surface_from_resource(wined3d_texture_get_sub_resource(src_texture, i));
dst_surface = surface_from_resource(wined3d_texture_get_sub_resource(dst_texture, i));
hr = wined3d_device_update_surface(device, src_surface, NULL, dst_surface, NULL);
if (FAILED(hr))
{
WARN("Failed to update surface, hr %#x.\n", hr);
return hr;
}
}
break;
}
case WINED3D_RTYPE_CUBE_TEXTURE:
{
struct wined3d_surface *src_surface;
struct wined3d_surface *dst_surface;
for (i = 0; i < level_count * 6; ++i)
{
src_surface = surface_from_resource(wined3d_texture_get_sub_resource(src_texture, i));
dst_surface = surface_from_resource(wined3d_texture_get_sub_resource(dst_texture, i));
hr = wined3d_device_update_surface(device, src_surface, NULL, dst_surface, NULL);
if (FAILED(hr))
{
WARN("Failed to update surface, hr %#x.\n", hr);
return hr;
}
}
break;
}
case WINED3D_RTYPE_VOLUME_TEXTURE:
{
for (i = 0; i < level_count; ++i)
{
hr = device_update_volume(device,
volume_from_resource(wined3d_texture_get_sub_resource(src_texture, i)),
volume_from_resource(wined3d_texture_get_sub_resource(dst_texture, i)));
if (FAILED(hr))
{
WARN("Failed to update volume, hr %#x.\n", hr);
return hr;
}
}
break;
}
default:
FIXME("Unsupported texture type %#x.\n", type);
return WINED3DERR_INVALIDCALL;
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_front_buffer_data(const struct wined3d_device *device,
UINT swapchain_idx, struct wined3d_surface *dst_surface)
{
struct wined3d_swapchain *swapchain;
HRESULT hr;
TRACE("device %p, swapchain_idx %u, dst_surface %p.\n", device, swapchain_idx, dst_surface);
hr = wined3d_device_get_swapchain(device, swapchain_idx, &swapchain);
if (FAILED(hr)) return hr;
hr = wined3d_swapchain_get_front_buffer_data(swapchain, dst_surface);
wined3d_swapchain_decref(swapchain);
return hr;
}
HRESULT CDECL wined3d_device_validate_device(const struct wined3d_device *device, DWORD *num_passes)
{
const struct wined3d_state *state = &device->stateBlock->state;
struct wined3d_texture *texture;
DWORD i;
TRACE("device %p, num_passes %p.\n", device, num_passes);
for (i = 0; i < MAX_COMBINED_SAMPLERS; ++i)
{
if (state->sampler_states[i][WINED3D_SAMP_MIN_FILTER] == WINED3D_TEXF_NONE)
{
WARN("Sampler state %u has minfilter D3DTEXF_NONE, returning D3DERR_UNSUPPORTEDTEXTUREFILTER\n", i);
return WINED3DERR_UNSUPPORTEDTEXTUREFILTER;
}
if (state->sampler_states[i][WINED3D_SAMP_MAG_FILTER] == WINED3D_TEXF_NONE)
{
WARN("Sampler state %u has magfilter D3DTEXF_NONE, returning D3DERR_UNSUPPORTEDTEXTUREFILTER\n", i);
return WINED3DERR_UNSUPPORTEDTEXTUREFILTER;
}
texture = state->textures[i];
if (!texture || texture->resource.format->flags & WINED3DFMT_FLAG_FILTERING) continue;
if (state->sampler_states[i][WINED3D_SAMP_MAG_FILTER] != WINED3D_TEXF_POINT)
{
WARN("Non-filterable texture and mag filter enabled on samper %u, returning E_FAIL\n", i);
return E_FAIL;
}
if (state->sampler_states[i][WINED3D_SAMP_MIN_FILTER] != WINED3D_TEXF_POINT)
{
WARN("Non-filterable texture and min filter enabled on samper %u, returning E_FAIL\n", i);
return E_FAIL;
}
if (state->sampler_states[i][WINED3D_SAMP_MIP_FILTER] != WINED3D_TEXF_NONE
&& state->sampler_states[i][WINED3D_SAMP_MIP_FILTER] != WINED3D_TEXF_POINT)
{
WARN("Non-filterable texture and mip filter enabled on samper %u, returning E_FAIL\n", i);
return E_FAIL;
}
}
if (state->render_states[WINED3D_RS_ZENABLE] || state->render_states[WINED3D_RS_ZWRITEENABLE]
|| state->render_states[WINED3D_RS_STENCILENABLE])
{
struct wined3d_surface *ds = device->fb.depth_stencil;
struct wined3d_surface *target = device->fb.render_targets[0];
if(ds && target
&& (ds->resource.width < target->resource.width || ds->resource.height < target->resource.height))
{
WARN("Depth stencil is smaller than the color buffer, returning D3DERR_CONFLICTINGRENDERSTATE\n");
return WINED3DERR_CONFLICTINGRENDERSTATE;
}
}
/* return a sensible default */
*num_passes = 1;
TRACE("returning D3D_OK\n");
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_software_vertex_processing(struct wined3d_device *device, BOOL software)
{
static BOOL warned;
TRACE("device %p, software %#x.\n", device, software);
if (!warned)
{
FIXME("device %p, software %#x stub!\n", device, software);
warned = TRUE;
}
device->softwareVertexProcessing = software;
return WINED3D_OK;
}
BOOL CDECL wined3d_device_get_software_vertex_processing(const struct wined3d_device *device)
{
static BOOL warned;
TRACE("device %p.\n", device);
if (!warned)
{
TRACE("device %p stub!\n", device);
warned = TRUE;
}
return device->softwareVertexProcessing;
}
HRESULT CDECL wined3d_device_get_raster_status(const struct wined3d_device *device,
UINT swapchain_idx, struct wined3d_raster_status *raster_status)
{
struct wined3d_swapchain *swapchain;
HRESULT hr;
TRACE("device %p, swapchain_idx %u, raster_status %p.\n",
device, swapchain_idx, raster_status);
hr = wined3d_device_get_swapchain(device, swapchain_idx, &swapchain);
if (FAILED(hr))
{
WARN("Failed to get swapchain %u, hr %#x.\n", swapchain_idx, hr);
return hr;
}
hr = wined3d_swapchain_get_raster_status(swapchain, raster_status);
wined3d_swapchain_decref(swapchain);
if (FAILED(hr))
{
WARN("Failed to get raster status, hr %#x.\n", hr);
return hr;
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_npatch_mode(struct wined3d_device *device, float segments)
{
static BOOL warned;
TRACE("device %p, segments %.8e.\n", device, segments);
if (segments != 0.0f)
{
if (!warned)
{
FIXME("device %p, segments %.8e stub!\n", device, segments);
warned = TRUE;
}
}
return WINED3D_OK;
}
float CDECL wined3d_device_get_npatch_mode(const struct wined3d_device *device)
{
static BOOL warned;
TRACE("device %p.\n", device);
if (!warned)
{
FIXME("device %p stub!\n", device);
warned = TRUE;
}
return 0.0f;
}
HRESULT CDECL wined3d_device_update_surface(struct wined3d_device *device,
struct wined3d_surface *src_surface, const RECT *src_rect,
struct wined3d_surface *dst_surface, const POINT *dst_point)
{
TRACE("device %p, src_surface %p, src_rect %s, dst_surface %p, dst_point %s.\n",
device, src_surface, wine_dbgstr_rect(src_rect),
dst_surface, wine_dbgstr_point(dst_point));
if (src_surface->resource.pool != WINED3D_POOL_SYSTEM_MEM || dst_surface->resource.pool != WINED3D_POOL_DEFAULT)
{
WARN("source %p must be SYSTEMMEM and dest %p must be DEFAULT, returning WINED3DERR_INVALIDCALL\n",
src_surface, dst_surface);
return WINED3DERR_INVALIDCALL;
}
return surface_upload_from_surface(dst_surface, dst_point, src_surface, src_rect);
}
HRESULT CDECL wined3d_device_draw_rect_patch(struct wined3d_device *device, UINT handle,
const float *num_segs, const struct wined3d_rect_patch_info *rect_patch_info)
{
struct wined3d_rect_patch *patch;
GLenum old_primitive_type;
unsigned int i;
struct list *e;
BOOL found;
TRACE("device %p, handle %#x, num_segs %p, rect_patch_info %p.\n",
device, handle, num_segs, rect_patch_info);
if (!(handle || rect_patch_info))
{
/* TODO: Write a test for the return value, thus the FIXME */
FIXME("Both handle and rect_patch_info are NULL.\n");
return WINED3DERR_INVALIDCALL;
}
if (handle)
{
i = PATCHMAP_HASHFUNC(handle);
found = FALSE;
LIST_FOR_EACH(e, &device->patches[i])
{
patch = LIST_ENTRY(e, struct wined3d_rect_patch, entry);
if (patch->Handle == handle)
{
found = TRUE;
break;
}
}
if (!found)
{
TRACE("Patch does not exist. Creating a new one\n");
patch = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*patch));
patch->Handle = handle;
list_add_head(&device->patches[i], &patch->entry);
} else {
TRACE("Found existing patch %p\n", patch);
}
}
else
{
/* Since opengl does not load tesselated vertex attributes into numbered vertex
* attributes we have to tesselate, read back, and draw. This needs a patch
* management structure instance. Create one.
*
* A possible improvement is to check if a vertex shader is used, and if not directly
* draw the patch.
*/
FIXME("Drawing an uncached patch. This is slow\n");
patch = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*patch));
}
if (num_segs[0] != patch->numSegs[0] || num_segs[1] != patch->numSegs[1]
|| num_segs[2] != patch->numSegs[2] || num_segs[3] != patch->numSegs[3]
|| (rect_patch_info && memcmp(rect_patch_info, &patch->rect_patch_info, sizeof(*rect_patch_info))))
{
HRESULT hr;
TRACE("Tesselation density or patch info changed, retesselating\n");
if (rect_patch_info)
patch->rect_patch_info = *rect_patch_info;
patch->numSegs[0] = num_segs[0];
patch->numSegs[1] = num_segs[1];
patch->numSegs[2] = num_segs[2];
patch->numSegs[3] = num_segs[3];
hr = tesselate_rectpatch(device, patch);
if (FAILED(hr))
{
WARN("Patch tesselation failed.\n");
/* Do not release the handle to store the params of the patch */
if (!handle)
HeapFree(GetProcessHeap(), 0, patch);
return hr;
}
}
old_primitive_type = device->stateBlock->state.gl_primitive_type;
device->stateBlock->state.gl_primitive_type = GL_TRIANGLES;
wined3d_device_draw_primitive_strided(device, patch->numSegs[0] * patch->numSegs[1] * 2 * 3, &patch->strided);
device->stateBlock->state.gl_primitive_type = old_primitive_type;
/* Destroy uncached patches */
if (!handle)
{
HeapFree(GetProcessHeap(), 0, patch->mem);
HeapFree(GetProcessHeap(), 0, patch);
}
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_draw_tri_patch(struct wined3d_device *device, UINT handle,
const float *segment_count, const struct wined3d_tri_patch_info *patch_info)
{
FIXME("device %p, handle %#x, segment_count %p, patch_info %p stub!\n",
device, handle, segment_count, patch_info);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_delete_patch(struct wined3d_device *device, UINT handle)
{
struct wined3d_rect_patch *patch;
struct list *e;
int i;
TRACE("device %p, handle %#x.\n", device, handle);
i = PATCHMAP_HASHFUNC(handle);
LIST_FOR_EACH(e, &device->patches[i])
{
patch = LIST_ENTRY(e, struct wined3d_rect_patch, entry);
if (patch->Handle == handle)
{
TRACE("Deleting patch %p\n", patch);
list_remove(&patch->entry);
HeapFree(GetProcessHeap(), 0, patch->mem);
HeapFree(GetProcessHeap(), 0, patch);
return WINED3D_OK;
}
}
/* TODO: Write a test for the return value */
FIXME("Attempt to destroy nonexistent patch\n");
return WINED3DERR_INVALIDCALL;
}
/* Do not call while under the GL lock. */
HRESULT CDECL wined3d_device_color_fill(struct wined3d_device *device,
struct wined3d_surface *surface, const RECT *rect, const struct wined3d_color *color)
{
RECT r;
TRACE("device %p, surface %p, rect %s, color {%.8e, %.8e, %.8e, %.8e}.\n",
device, surface, wine_dbgstr_rect(rect),
color->r, color->g, color->b, color->a);
if (surface->resource.pool != WINED3D_POOL_DEFAULT && surface->resource.pool != WINED3D_POOL_SYSTEM_MEM)
{
WARN("Color-fill not allowed on %s surfaces.\n", debug_d3dpool(surface->resource.pool));
return WINED3DERR_INVALIDCALL;
}
if (!rect)
{
SetRect(&r, 0, 0, surface->resource.width, surface->resource.height);
rect = &r;
}
return surface_color_fill(surface, rect, color);
}
/* Do not call while under the GL lock. */
void CDECL wined3d_device_clear_rendertarget_view(struct wined3d_device *device,
struct wined3d_rendertarget_view *rendertarget_view, const struct wined3d_color *color)
{
struct wined3d_resource *resource;
HRESULT hr;
RECT rect;
resource = rendertarget_view->resource;
if (resource->type != WINED3D_RTYPE_SURFACE)
{
FIXME("Only supported on surface resources\n");
return;
}
SetRect(&rect, 0, 0, resource->width, resource->height);
hr = surface_color_fill(surface_from_resource(resource), &rect, color);
if (FAILED(hr)) ERR("Color fill failed, hr %#x.\n", hr);
}
HRESULT CDECL wined3d_device_get_render_target(const struct wined3d_device *device,
UINT render_target_idx, struct wined3d_surface **render_target)
{
TRACE("device %p, render_target_idx %u, render_target %p.\n",
device, render_target_idx, render_target);
if (render_target_idx >= device->adapter->gl_info.limits.buffers)
{
WARN("Only %u render targets are supported.\n", device->adapter->gl_info.limits.buffers);
return WINED3DERR_INVALIDCALL;
}
*render_target = device->fb.render_targets[render_target_idx];
TRACE("Returning render target %p.\n", *render_target);
if (!*render_target)
return WINED3DERR_NOTFOUND;
wined3d_surface_incref(*render_target);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_depth_stencil(const struct wined3d_device *device,
struct wined3d_surface **depth_stencil)
{
TRACE("device %p, depth_stencil %p.\n", device, depth_stencil);
*depth_stencil = device->fb.depth_stencil;
TRACE("Returning depth/stencil surface %p.\n", *depth_stencil);
if (!*depth_stencil)
return WINED3DERR_NOTFOUND;
wined3d_surface_incref(*depth_stencil);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_render_target(struct wined3d_device *device,
UINT render_target_idx, struct wined3d_surface *render_target, BOOL set_viewport)
{
struct wined3d_surface *prev;
TRACE("device %p, render_target_idx %u, render_target %p, set_viewport %#x.\n",
device, render_target_idx, render_target, set_viewport);
if (render_target_idx >= device->adapter->gl_info.limits.buffers)
{
WARN("Only %u render targets are supported.\n", device->adapter->gl_info.limits.buffers);
return WINED3DERR_INVALIDCALL;
}
prev = device->fb.render_targets[render_target_idx];
if (render_target == prev)
{
TRACE("Trying to do a NOP SetRenderTarget operation.\n");
return WINED3D_OK;
}
/* Render target 0 can't be set to NULL. */
if (!render_target && !render_target_idx)
{
WARN("Trying to set render target 0 to NULL.\n");
return WINED3DERR_INVALIDCALL;
}
if (render_target && !(render_target->resource.usage & WINED3DUSAGE_RENDERTARGET))
{
FIXME("Surface %p doesn't have render target usage.\n", render_target);
return WINED3DERR_INVALIDCALL;
}
if (render_target)
wined3d_surface_incref(render_target);
device->fb.render_targets[render_target_idx] = render_target;
/* Release after the assignment, to prevent device_resource_released()
* from seeing the surface as still in use. */
if (prev)
wined3d_surface_decref(prev);
/* Render target 0 is special. */
if (!render_target_idx && set_viewport)
{
/* Set the viewport and scissor rectangles, if requested. Tests show
* that stateblock recording is ignored, the change goes directly
* into the primary stateblock. */
device->stateBlock->state.viewport.height = device->fb.render_targets[0]->resource.height;
device->stateBlock->state.viewport.width = device->fb.render_targets[0]->resource.width;
device->stateBlock->state.viewport.x = 0;
device->stateBlock->state.viewport.y = 0;
device->stateBlock->state.viewport.max_z = 1.0f;
device->stateBlock->state.viewport.min_z = 0.0f;
device_invalidate_state(device, STATE_VIEWPORT);
device->stateBlock->state.scissor_rect.top = 0;
device->stateBlock->state.scissor_rect.left = 0;
device->stateBlock->state.scissor_rect.right = device->stateBlock->state.viewport.width;
device->stateBlock->state.scissor_rect.bottom = device->stateBlock->state.viewport.height;
device_invalidate_state(device, STATE_SCISSORRECT);
}
device_invalidate_state(device, STATE_FRAMEBUFFER);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_depth_stencil(struct wined3d_device *device, struct wined3d_surface *depth_stencil)
{
struct wined3d_surface *prev = device->fb.depth_stencil;
TRACE("device %p, depth_stencil %p, old depth_stencil %p.\n",
device, depth_stencil, prev);
if (prev == depth_stencil)
{
TRACE("Trying to do a NOP SetRenderTarget operation.\n");
return WINED3D_OK;
}
if (prev)
{
if (device->swapchains[0]->desc.flags & WINED3DPRESENTFLAG_DISCARD_DEPTHSTENCIL
|| prev->flags & SFLAG_DISCARD)
{
surface_modify_ds_location(prev, SFLAG_DISCARDED,
prev->resource.width, prev->resource.height);
if (prev == device->onscreen_depth_stencil)
{
wined3d_surface_decref(device->onscreen_depth_stencil);
device->onscreen_depth_stencil = NULL;
}
}
}
device->fb.depth_stencil = depth_stencil;
if (depth_stencil)
wined3d_surface_incref(depth_stencil);
if (!prev != !depth_stencil)
{
/* Swapping NULL / non NULL depth stencil affects the depth and tests */
device_invalidate_state(device, STATE_RENDER(WINED3D_RS_ZENABLE));
device_invalidate_state(device, STATE_RENDER(WINED3D_RS_STENCILENABLE));
device_invalidate_state(device, STATE_RENDER(WINED3D_RS_STENCILWRITEMASK));
device_invalidate_state(device, STATE_RENDER(WINED3D_RS_DEPTHBIAS));
}
else if (prev && prev->resource.format->depth_size != depth_stencil->resource.format->depth_size)
{
device_invalidate_state(device, STATE_RENDER(WINED3D_RS_DEPTHBIAS));
}
if (prev)
wined3d_surface_decref(prev);
device_invalidate_state(device, STATE_FRAMEBUFFER);
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_set_cursor_properties(struct wined3d_device *device,
UINT x_hotspot, UINT y_hotspot, struct wined3d_surface *cursor_image)
{
TRACE("device %p, x_hotspot %u, y_hotspot %u, cursor_image %p.\n",
device, x_hotspot, y_hotspot, cursor_image);
/* some basic validation checks */
if (device->cursorTexture)
{
struct wined3d_context *context = context_acquire(device, NULL);
ENTER_GL();
glDeleteTextures(1, &device->cursorTexture);
LEAVE_GL();
context_release(context);
device->cursorTexture = 0;
}
if (cursor_image)
{
struct wined3d_display_mode mode;
struct wined3d_map_desc map_desc;
HRESULT hr;
/* MSDN: Cursor must be A8R8G8B8 */
if (cursor_image->resource.format->id != WINED3DFMT_B8G8R8A8_UNORM)
{
WARN("surface %p has an invalid format.\n", cursor_image);
return WINED3DERR_INVALIDCALL;
}
if (FAILED(hr = wined3d_get_adapter_display_mode(device->wined3d, device->adapter->ordinal, &mode)))
{
ERR("Failed to get display mode, hr %#x.\n", hr);
return WINED3DERR_INVALIDCALL;
}
/* MSDN: Cursor must be smaller than the display mode */
if (cursor_image->resource.width > mode.width || cursor_image->resource.height > mode.height)
{
WARN("Surface %p dimensions are %ux%u, but screen dimensions are %ux%u.\n",
cursor_image, cursor_image->resource.width, cursor_image->resource.height,
mode.width, mode.height);
return WINED3DERR_INVALIDCALL;
}
/* TODO: MSDN: Cursor sizes must be a power of 2 */
/* Do not store the surface's pointer because the application may
* release it after setting the cursor image. Windows doesn't
* addref the set surface, so we can't do this either without
* creating circular refcount dependencies. Copy out the gl texture
* instead. */
device->cursorWidth = cursor_image->resource.width;
device->cursorHeight = cursor_image->resource.height;
if (SUCCEEDED(wined3d_surface_map(cursor_image, &map_desc, NULL, WINED3D_MAP_READONLY)))
{
const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
const struct wined3d_format *format = wined3d_get_format(gl_info, WINED3DFMT_B8G8R8A8_UNORM);
struct wined3d_context *context;
char *mem, *bits = map_desc.data;
GLint intfmt = format->glInternal;
GLint gl_format = format->glFormat;
GLint type = format->glType;
INT height = device->cursorHeight;
INT width = device->cursorWidth;
INT bpp = format->byte_count;
INT i;
/* Reformat the texture memory (pitch and width can be
* different) */
mem = HeapAlloc(GetProcessHeap(), 0, width * height * bpp);
for (i = 0; i < height; ++i)
memcpy(&mem[width * bpp * i], &bits[map_desc.row_pitch * i], width * bpp);
wined3d_surface_unmap(cursor_image);
context = context_acquire(device, NULL);
ENTER_GL();
if (gl_info->supported[APPLE_CLIENT_STORAGE])
{
glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_FALSE);
checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_FALSE)");
}
invalidate_active_texture(device, context);
/* Create a new cursor texture */
glGenTextures(1, &device->cursorTexture);
checkGLcall("glGenTextures");
context_bind_texture(context, GL_TEXTURE_2D, device->cursorTexture);
/* Copy the bitmap memory into the cursor texture */
glTexImage2D(GL_TEXTURE_2D, 0, intfmt, width, height, 0, gl_format, type, mem);
checkGLcall("glTexImage2D");
HeapFree(GetProcessHeap(), 0, mem);
if (gl_info->supported[APPLE_CLIENT_STORAGE])
{
glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE);
checkGLcall("glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE)");
}
LEAVE_GL();
context_release(context);
}
else
{
FIXME("A cursor texture was not returned.\n");
device->cursorTexture = 0;
}
if (cursor_image->resource.width == 32 && cursor_image->resource.height == 32)
{
UINT mask_size = cursor_image->resource.width * cursor_image->resource.height / 8;
ICONINFO cursorInfo;
DWORD *maskBits;
HCURSOR cursor;
/* 32-bit user32 cursors ignore the alpha channel if it's all
* zeroes, and use the mask instead. Fill the mask with all ones
* to ensure we still get a fully transparent cursor. */
maskBits = HeapAlloc(GetProcessHeap(), 0, mask_size);
memset(maskBits, 0xff, mask_size);
wined3d_surface_map(cursor_image, &map_desc, NULL,
WINED3D_MAP_NO_DIRTY_UPDATE | WINED3D_MAP_READONLY);
TRACE("width: %u height: %u.\n", cursor_image->resource.width, cursor_image->resource.height);
cursorInfo.fIcon = FALSE;
cursorInfo.xHotspot = x_hotspot;
cursorInfo.yHotspot = y_hotspot;
cursorInfo.hbmMask = CreateBitmap(cursor_image->resource.width, cursor_image->resource.height,
1, 1, maskBits);
cursorInfo.hbmColor = CreateBitmap(cursor_image->resource.width, cursor_image->resource.height,
1, 32, map_desc.data);
wined3d_surface_unmap(cursor_image);
/* Create our cursor and clean up. */
cursor = CreateIconIndirect(&cursorInfo);
if (cursorInfo.hbmMask) DeleteObject(cursorInfo.hbmMask);
if (cursorInfo.hbmColor) DeleteObject(cursorInfo.hbmColor);
if (device->hardwareCursor) DestroyCursor(device->hardwareCursor);
device->hardwareCursor = cursor;
if (device->bCursorVisible) SetCursor( cursor );
HeapFree(GetProcessHeap(), 0, maskBits);
}
}
device->xHotSpot = x_hotspot;
device->yHotSpot = y_hotspot;
return WINED3D_OK;
}
void CDECL wined3d_device_set_cursor_position(struct wined3d_device *device,
int x_screen_space, int y_screen_space, DWORD flags)
{
TRACE("device %p, x %d, y %d, flags %#x.\n",
device, x_screen_space, y_screen_space, flags);
device->xScreenSpace = x_screen_space;
device->yScreenSpace = y_screen_space;
if (device->hardwareCursor)
{
POINT pt;
GetCursorPos( &pt );
if (x_screen_space == pt.x && y_screen_space == pt.y)
return;
SetCursorPos( x_screen_space, y_screen_space );
/* Switch to the software cursor if position diverges from the hardware one. */
GetCursorPos( &pt );
if (x_screen_space != pt.x || y_screen_space != pt.y)
{
if (device->bCursorVisible) SetCursor( NULL );
DestroyCursor( device->hardwareCursor );
device->hardwareCursor = 0;
}
}
}
BOOL CDECL wined3d_device_show_cursor(struct wined3d_device *device, BOOL show)
{
BOOL oldVisible = device->bCursorVisible;
TRACE("device %p, show %#x.\n", device, show);
/*
* When ShowCursor is first called it should make the cursor appear at the OS's last
* known cursor position.
*/
if (show && !oldVisible)
{
POINT pt;
GetCursorPos(&pt);
device->xScreenSpace = pt.x;
device->yScreenSpace = pt.y;
}
if (device->hardwareCursor)
{
device->bCursorVisible = show;
if (show)
SetCursor(device->hardwareCursor);
else
SetCursor(NULL);
}
else
{
if (device->cursorTexture)
device->bCursorVisible = show;
}
return oldVisible;
}
void CDECL wined3d_device_evict_managed_resources(struct wined3d_device *device)
{
struct wined3d_resource *resource, *cursor;
TRACE("device %p.\n", device);
LIST_FOR_EACH_ENTRY_SAFE(resource, cursor, &device->resources, struct wined3d_resource, resource_list_entry)
{
TRACE("Checking resource %p for eviction.\n", resource);
if (resource->pool == WINED3D_POOL_MANAGED && !resource->map_count)
{
TRACE("Evicting %p.\n", resource);
resource->resource_ops->resource_unload(resource);
}
}
/* Invalidate stream sources, the buffer(s) may have been evicted. */
device_invalidate_state(device, STATE_STREAMSRC);
}
static BOOL is_display_mode_supported(const struct wined3d_device *device,
const struct wined3d_swapchain_desc *swapchain_desc)
{
struct wined3d_display_mode m;
UINT i, count;
HRESULT hr;
/* All Windowed modes are supported, as is leaving the current mode */
if (swapchain_desc->windowed)
return TRUE;
if (!swapchain_desc->backbuffer_width)
return TRUE;
if (!swapchain_desc->backbuffer_height)
return TRUE;
count = wined3d_get_adapter_mode_count(device->wined3d, device->adapter->ordinal, WINED3DFMT_UNKNOWN);
for (i = 0; i < count; ++i)
{
memset(&m, 0, sizeof(m));
hr = wined3d_enum_adapter_modes(device->wined3d, device->adapter->ordinal, WINED3DFMT_UNKNOWN, i, &m);
if (FAILED(hr))
ERR("Failed to enumerate adapter mode.\n");
if (m.width == swapchain_desc->backbuffer_width && m.height == swapchain_desc->backbuffer_height)
/* Mode found, it is supported. */
return TRUE;
}
/* Mode not found -> not supported */
return FALSE;
}
/* Do not call while under the GL lock. */
static void delete_opengl_contexts(struct wined3d_device *device, struct wined3d_swapchain *swapchain)
{
struct wined3d_resource *resource, *cursor;
const struct wined3d_gl_info *gl_info;
struct wined3d_context *context;
struct wined3d_shader *shader;
context = context_acquire(device, NULL);
gl_info = context->gl_info;
LIST_FOR_EACH_ENTRY_SAFE(resource, cursor, &device->resources, struct wined3d_resource, resource_list_entry)
{
TRACE("Unloading resource %p.\n", resource);
resource->resource_ops->resource_unload(resource);
}
LIST_FOR_EACH_ENTRY(shader, &device->shaders, struct wined3d_shader, shader_list_entry)
{
device->shader_backend->shader_destroy(shader);
}
ENTER_GL();
if (device->depth_blt_texture)
{
glDeleteTextures(1, &device->depth_blt_texture);
device->depth_blt_texture = 0;
}
if (device->cursorTexture)
{
glDeleteTextures(1, &device->cursorTexture);
device->cursorTexture = 0;
}
LEAVE_GL();
device->blitter->free_private(device);
device->frag_pipe->free_private(device);
device->shader_backend->shader_free_private(device);
destroy_dummy_textures(device, gl_info);
context_release(context);
while (device->context_count)
{
swapchain_destroy_contexts(device->contexts[0]->swapchain);
}
HeapFree(GetProcessHeap(), 0, swapchain->context);
swapchain->context = NULL;
}
/* Do not call while under the GL lock. */
static HRESULT create_primary_opengl_context(struct wined3d_device *device, struct wined3d_swapchain *swapchain)
{
struct wined3d_context *context;
struct wined3d_surface *target;
HRESULT hr;
/* Recreate the primary swapchain's context */
swapchain->context = HeapAlloc(GetProcessHeap(), 0, sizeof(*swapchain->context));
if (!swapchain->context)
{
ERR("Failed to allocate memory for swapchain context array.\n");
return E_OUTOFMEMORY;
}
target = swapchain->back_buffers ? swapchain->back_buffers[0] : swapchain->front_buffer;
if (!(context = context_create(swapchain, target, swapchain->ds_format)))
{
WARN("Failed to create context.\n");
HeapFree(GetProcessHeap(), 0, swapchain->context);
return E_FAIL;
}
swapchain->context[0] = context;
swapchain->num_contexts = 1;
create_dummy_textures(device, context);
context_release(context);
hr = device->shader_backend->shader_alloc_private(device);
if (FAILED(hr))
{
ERR("Failed to allocate shader private data, hr %#x.\n", hr);
goto err;
}
hr = device->frag_pipe->alloc_private(device);
if (FAILED(hr))
{
ERR("Failed to allocate fragment pipe private data, hr %#x.\n", hr);
device->shader_backend->shader_free_private(device);
goto err;
}
hr = device->blitter->alloc_private(device);
if (FAILED(hr))
{
ERR("Failed to allocate blitter private data, hr %#x.\n", hr);
device->frag_pipe->free_private(device);
device->shader_backend->shader_free_private(device);
goto err;
}
return WINED3D_OK;
err:
context_acquire(device, NULL);
destroy_dummy_textures(device, context->gl_info);
context_release(context);
context_destroy(device, context);
HeapFree(GetProcessHeap(), 0, swapchain->context);
swapchain->num_contexts = 0;
return hr;
}
/* Do not call while under the GL lock. */
HRESULT CDECL wined3d_device_reset(struct wined3d_device *device,
const struct wined3d_swapchain_desc *swapchain_desc,
wined3d_device_reset_cb callback)
{
struct wined3d_resource *resource, *cursor;
struct wined3d_swapchain *swapchain;
struct wined3d_display_mode mode;
BOOL DisplayModeChanged = FALSE;
BOOL update_desc = FALSE;
unsigned int i;
HRESULT hr;
TRACE("device %p, swapchain_desc %p.\n", device, swapchain_desc);
if (FAILED(hr = wined3d_device_get_swapchain(device, 0, &swapchain)))
{
ERR("Failed to get the first implicit swapchain.\n");
return hr;
}
stateblock_unbind_resources(device->stateBlock);
if (swapchain->back_buffers && swapchain->back_buffers[0])
wined3d_device_set_render_target(device, 0, swapchain->back_buffers[0], FALSE);
else
wined3d_device_set_render_target(device, 0, swapchain->front_buffer, FALSE);
for (i = 1; i < device->adapter->gl_info.limits.buffers; ++i)
{
wined3d_device_set_render_target(device, i, NULL, FALSE);
}
wined3d_device_set_depth_stencil(device, NULL);
if (device->onscreen_depth_stencil)
{
wined3d_surface_decref(device->onscreen_depth_stencil);
device->onscreen_depth_stencil = NULL;
}
LIST_FOR_EACH_ENTRY_SAFE(resource, cursor, &device->resources, struct wined3d_resource, resource_list_entry)
{
TRACE("Enumerating resource %p.\n", resource);
if (FAILED(hr = callback(resource)))
{
wined3d_swapchain_decref(swapchain);
return hr;
}
}
if (!is_display_mode_supported(device, swapchain_desc))
{
WARN("Rejecting reset() call because the requested display mode is not supported.\n");
WARN("Requested mode: %ux%u.\n",
swapchain_desc->backbuffer_width,
swapchain_desc->backbuffer_height);
wined3d_swapchain_decref(swapchain);
return WINED3DERR_INVALIDCALL;
}
/* Is it necessary to recreate the gl context? Actually every setting can be changed
* on an existing gl context, so there's no real need for recreation.
*
* TODO: Figure out how Reset influences resources in D3DPOOL_DEFAULT, D3DPOOL_SYSTEMMEMORY and D3DPOOL_MANAGED
*
* TODO: Figure out what happens to explicit swapchains, or if we have more than one implicit swapchain
*/
TRACE("New params:\n");
TRACE("backbuffer_width %u\n", swapchain_desc->backbuffer_width);
TRACE("backbuffer_height %u\n", swapchain_desc->backbuffer_height);
TRACE("backbuffer_format %s\n", debug_d3dformat(swapchain_desc->backbuffer_format));
TRACE("backbuffer_count %u\n", swapchain_desc->backbuffer_count);
TRACE("multisample_type %#x\n", swapchain_desc->multisample_type);
TRACE("multisample_quality %u\n", swapchain_desc->multisample_quality);
TRACE("swap_effect %#x\n", swapchain_desc->swap_effect);
TRACE("device_window %p\n", swapchain_desc->device_window);
TRACE("windowed %#x\n", swapchain_desc->windowed);
TRACE("enable_auto_depth_stencil %#x\n", swapchain_desc->enable_auto_depth_stencil);
if (swapchain_desc->enable_auto_depth_stencil)
TRACE("auto_depth_stencil_format %s\n", debug_d3dformat(swapchain_desc->auto_depth_stencil_format));
TRACE("flags %#x\n", swapchain_desc->flags);
TRACE("refresh_rate %u\n", swapchain_desc->refresh_rate);
TRACE("swap_interval %u\n", swapchain_desc->swap_interval);
TRACE("auto_restore_display_mode %#x\n", swapchain_desc->auto_restore_display_mode);
/* No special treatment of these parameters. Just store them */
swapchain->desc.swap_effect = swapchain_desc->swap_effect;
swapchain->desc.flags = swapchain_desc->flags;
swapchain->desc.swap_interval = swapchain_desc->swap_interval;
swapchain->desc.refresh_rate = swapchain_desc->refresh_rate;
/* What to do about these? */
if (swapchain_desc->backbuffer_count
&& swapchain_desc->backbuffer_count != swapchain->desc.backbuffer_count)
FIXME("Cannot change the back buffer count yet.\n");
if (swapchain_desc->device_window
&& swapchain_desc->device_window != swapchain->desc.device_window)
{
TRACE("Changing the device window from %p to %p.\n",
swapchain->desc.device_window, swapchain_desc->device_window);
swapchain->desc.device_window = swapchain_desc->device_window;
swapchain->device_window = swapchain_desc->device_window;
wined3d_swapchain_set_window(swapchain, NULL);
}
if (swapchain_desc->enable_auto_depth_stencil && !device->auto_depth_stencil)
{
HRESULT hrc;
TRACE("Creating the depth stencil buffer\n");
hrc = device->device_parent->ops->create_depth_stencil(device->device_parent,
swapchain_desc->backbuffer_width,
swapchain_desc->backbuffer_height,
swapchain_desc->auto_depth_stencil_format,
swapchain_desc->multisample_type,
swapchain_desc->multisample_quality,
FALSE,
&device->auto_depth_stencil);
if (FAILED(hrc))
{
ERR("Failed to create the depth stencil buffer.\n");
wined3d_swapchain_decref(swapchain);
return WINED3DERR_INVALIDCALL;
}
}
if (device->onscreen_depth_stencil)
{
wined3d_surface_decref(device->onscreen_depth_stencil);
device->onscreen_depth_stencil = NULL;
}
/* Reset the depth stencil */
if (swapchain_desc->enable_auto_depth_stencil)
wined3d_device_set_depth_stencil(device, device->auto_depth_stencil);
TRACE("Resetting stateblock\n");
wined3d_stateblock_decref(device->updateStateBlock);
wined3d_stateblock_decref(device->stateBlock);
if (swapchain_desc->windowed)
{
mode.width = swapchain->orig_width;
mode.height = swapchain->orig_height;
mode.refresh_rate = 0;
mode.format_id = swapchain->desc.backbuffer_format;
}
else
{
mode.width = swapchain_desc->backbuffer_width;
mode.height = swapchain_desc->backbuffer_height;
mode.refresh_rate = swapchain_desc->refresh_rate;
mode.format_id = swapchain_desc->backbuffer_format;
}
/* Should Width == 800 && Height == 0 set 800x600? */
if (swapchain_desc->backbuffer_width && swapchain_desc->backbuffer_height
&& (swapchain_desc->backbuffer_width != swapchain->desc.backbuffer_width
|| swapchain_desc->backbuffer_height != swapchain->desc.backbuffer_height))
{
if (!swapchain_desc->windowed)
DisplayModeChanged = TRUE;
swapchain->desc.backbuffer_width = swapchain_desc->backbuffer_width;
swapchain->desc.backbuffer_height = swapchain_desc->backbuffer_height;
update_desc = TRUE;
}
if (swapchain_desc->backbuffer_format != WINED3DFMT_UNKNOWN
&& swapchain_desc->backbuffer_format != swapchain->desc.backbuffer_format)
{
swapchain->desc.backbuffer_format = swapchain_desc->backbuffer_format;
update_desc = TRUE;
}
if (swapchain_desc->multisample_type != swapchain->desc.multisample_type
|| swapchain_desc->multisample_quality != swapchain->desc.multisample_quality)
{
swapchain->desc.multisample_type = swapchain_desc->multisample_type;
swapchain->desc.multisample_quality = swapchain_desc->multisample_quality;
update_desc = TRUE;
}
if (update_desc)
{
UINT i;
hr = wined3d_surface_update_desc(swapchain->front_buffer, swapchain->desc.backbuffer_width,
swapchain->desc.backbuffer_height, swapchain->desc.backbuffer_format,
swapchain->desc.multisample_type, swapchain->desc.multisample_quality);
if (FAILED(hr))
{
wined3d_swapchain_decref(swapchain);
return hr;
}
for (i = 0; i < swapchain->desc.backbuffer_count; ++i)
{
hr = wined3d_surface_update_desc(swapchain->back_buffers[i], swapchain->desc.backbuffer_width,
swapchain->desc.backbuffer_height, swapchain->desc.backbuffer_format,
swapchain->desc.multisample_type, swapchain->desc.multisample_quality);
if (FAILED(hr))
{
wined3d_swapchain_decref(swapchain);
return hr;
}
}
if (device->auto_depth_stencil)
{
hr = wined3d_surface_update_desc(device->auto_depth_stencil, swapchain->desc.backbuffer_width,
swapchain->desc.backbuffer_height, device->auto_depth_stencil->resource.format->id,
swapchain->desc.multisample_type, swapchain->desc.multisample_quality);
if (FAILED(hr))
{
wined3d_swapchain_decref(swapchain);
return hr;
}
}
}
if (device->d3d_initialized)
delete_opengl_contexts(device, swapchain);
if (!swapchain_desc->windowed != !swapchain->desc.windowed
|| DisplayModeChanged)
{
if (FAILED(hr = wined3d_set_adapter_display_mode(device->wined3d, device->adapter->ordinal, &mode)))
{
WARN("Failed to set display mode, hr %#x.\n", hr);
wined3d_swapchain_decref(swapchain);
return hr;
}
if (!swapchain_desc->windowed)
{
if (swapchain->desc.windowed)
{
HWND focus_window = device->create_parms.focus_window;
if (!focus_window)
focus_window = swapchain_desc->device_window;
if (FAILED(hr = wined3d_device_acquire_focus_window(device, focus_window)))
{
ERR("Failed to acquire focus window, hr %#x.\n", hr);
wined3d_swapchain_decref(swapchain);
return hr;
}
/* switch from windowed to fs */
wined3d_device_setup_fullscreen_window(device, swapchain->device_window,
swapchain_desc->backbuffer_width,
swapchain_desc->backbuffer_height);
}
else
{
/* Fullscreen -> fullscreen mode change */
MoveWindow(swapchain->device_window, 0, 0,
swapchain_desc->backbuffer_width,
swapchain_desc->backbuffer_height,
TRUE);
}
}
else if (!swapchain->desc.windowed)
{
/* Fullscreen -> windowed switch */
wined3d_device_restore_fullscreen_window(device, swapchain->device_window);
wined3d_device_release_focus_window(device);
}
swapchain->desc.windowed = swapchain_desc->windowed;
}
else if (!swapchain_desc->windowed)
{
DWORD style = device->style;
DWORD exStyle = device->exStyle;
/* If we're in fullscreen, and the mode wasn't changed, we have to get the window back into
* the right position. Some applications(Battlefield 2, Guild Wars) move it and then call
* Reset to clear up their mess. Guild Wars also loses the device during that.
*/
device->style = 0;
device->exStyle = 0;
wined3d_device_setup_fullscreen_window(device, swapchain->device_window,
swapchain_desc->backbuffer_width,
swapchain_desc->backbuffer_height);
device->style = style;
device->exStyle = exStyle;
}
/* Note: No parent needed for initial internal stateblock */
hr = wined3d_stateblock_create(device, WINED3D_SBT_INIT, &device->stateBlock);
if (FAILED(hr))
ERR("Resetting the stateblock failed with error %#x.\n", hr);
else
TRACE("Created stateblock %p.\n", device->stateBlock);
device->updateStateBlock = device->stateBlock;
wined3d_stateblock_incref(device->updateStateBlock);
stateblock_init_default_state(device->stateBlock);
swapchain_update_render_to_fbo(swapchain);
swapchain_update_draw_bindings(swapchain);
if (device->d3d_initialized)
hr = create_primary_opengl_context(device, swapchain);
wined3d_swapchain_decref(swapchain);
/* All done. There is no need to reload resources or shaders, this will happen automatically on the
* first use
*/
return hr;
}
HRESULT CDECL wined3d_device_set_dialog_box_mode(struct wined3d_device *device, BOOL enable_dialogs)
{
TRACE("device %p, enable_dialogs %#x.\n", device, enable_dialogs);
if (!enable_dialogs) FIXME("Dialogs cannot be disabled yet.\n");
return WINED3D_OK;
}
HRESULT CDECL wined3d_device_get_creation_parameters(const struct wined3d_device *device,
struct wined3d_device_creation_parameters *parameters)
{
TRACE("device %p, parameters %p.\n", device, parameters);
*parameters = device->create_parms;
return WINED3D_OK;
}
void CDECL wined3d_device_set_gamma_ramp(const struct wined3d_device *device,
UINT swapchain_idx, DWORD flags, const struct wined3d_gamma_ramp *ramp)
{
struct wined3d_swapchain *swapchain;
TRACE("device %p, swapchain_idx %u, flags %#x, ramp %p.\n",
device, swapchain_idx, flags, ramp);
if (SUCCEEDED(wined3d_device_get_swapchain(device, swapchain_idx, &swapchain)))
{
wined3d_swapchain_set_gamma_ramp(swapchain, flags, ramp);
wined3d_swapchain_decref(swapchain);
}
}
void CDECL wined3d_device_get_gamma_ramp(const struct wined3d_device *device,
UINT swapchain_idx, struct wined3d_gamma_ramp *ramp)
{
struct wined3d_swapchain *swapchain;
TRACE("device %p, swapchain_idx %u, ramp %p.\n",
device, swapchain_idx, ramp);
if (SUCCEEDED(wined3d_device_get_swapchain(device, swapchain_idx, &swapchain)))
{
wined3d_swapchain_get_gamma_ramp(swapchain, ramp);
wined3d_swapchain_decref(swapchain);
}
}
void device_resource_add(struct wined3d_device *device, struct wined3d_resource *resource)
{
TRACE("device %p, resource %p.\n", device, resource);
list_add_head(&device->resources, &resource->resource_list_entry);
}
static void device_resource_remove(struct wined3d_device *device, struct wined3d_resource *resource)
{
TRACE("device %p, resource %p.\n", device, resource);
list_remove(&resource->resource_list_entry);
}
void device_resource_released(struct wined3d_device *device, struct wined3d_resource *resource)
{
enum wined3d_resource_type type = resource->type;
unsigned int i;
TRACE("device %p, resource %p, type %s.\n", device, resource, debug_d3dresourcetype(type));
context_resource_released(device, resource, type);
switch (type)
{
case WINED3D_RTYPE_SURFACE:
{
struct wined3d_surface *surface = surface_from_resource(resource);
if (!device->d3d_initialized) break;
for (i = 0; i < device->adapter->gl_info.limits.buffers; ++i)
{
if (device->fb.render_targets[i] == surface)
{
ERR("Surface %p is still in use as render target %u.\n", surface, i);
device->fb.render_targets[i] = NULL;
}
}
if (device->fb.depth_stencil == surface)
{
ERR("Surface %p is still in use as depth/stencil buffer.\n", surface);
device->fb.depth_stencil = NULL;
}
}
break;
case WINED3D_RTYPE_TEXTURE:
case WINED3D_RTYPE_CUBE_TEXTURE:
case WINED3D_RTYPE_VOLUME_TEXTURE:
for (i = 0; i < MAX_COMBINED_SAMPLERS; ++i)
{
struct wined3d_texture *texture = wined3d_texture_from_resource(resource);
if (device->stateBlock && device->stateBlock->state.textures[i] == texture)
{
ERR("Texture %p is still in use by stateblock %p, stage %u.\n",
texture, device->stateBlock, i);
device->stateBlock->state.textures[i] = NULL;
}
if (device->updateStateBlock != device->stateBlock
&& device->updateStateBlock->state.textures[i] == texture)
{
ERR("Texture %p is still in use by stateblock %p, stage %u.\n",
texture, device->updateStateBlock, i);
device->updateStateBlock->state.textures[i] = NULL;
}
}
break;
case WINED3D_RTYPE_BUFFER:
{
struct wined3d_buffer *buffer = buffer_from_resource(resource);
for (i = 0; i < MAX_STREAMS; ++i)
{
if (device->stateBlock && device->stateBlock->state.streams[i].buffer == buffer)
{
ERR("Buffer %p is still in use by stateblock %p, stream %u.\n",
buffer, device->stateBlock, i);
device->stateBlock->state.streams[i].buffer = NULL;
}
if (device->updateStateBlock != device->stateBlock
&& device->updateStateBlock->state.streams[i].buffer == buffer)
{
ERR("Buffer %p is still in use by stateblock %p, stream %u.\n",
buffer, device->updateStateBlock, i);
device->updateStateBlock->state.streams[i].buffer = NULL;
}
}
if (device->stateBlock && device->stateBlock->state.index_buffer == buffer)
{
ERR("Buffer %p is still in use by stateblock %p as index buffer.\n",
buffer, device->stateBlock);
device->stateBlock->state.index_buffer = NULL;
}
if (device->updateStateBlock != device->stateBlock
&& device->updateStateBlock->state.index_buffer == buffer)
{
ERR("Buffer %p is still in use by stateblock %p as index buffer.\n",
buffer, device->updateStateBlock);
device->updateStateBlock->state.index_buffer = NULL;
}
}
break;
default:
break;
}
/* Remove the resource from the resourceStore */
device_resource_remove(device, resource);
TRACE("Resource released.\n");
}
HRESULT CDECL wined3d_device_get_surface_from_dc(const struct wined3d_device *device,
HDC dc, struct wined3d_surface **surface)
{
struct wined3d_resource *resource;
TRACE("device %p, dc %p, surface %p.\n", device, dc, surface);
if (!dc)
return WINED3DERR_INVALIDCALL;
LIST_FOR_EACH_ENTRY(resource, &device->resources, struct wined3d_resource, resource_list_entry)
{
if (resource->type == WINED3D_RTYPE_SURFACE)
{
struct wined3d_surface *s = surface_from_resource(resource);
if (s->hDC == dc)
{
TRACE("Found surface %p for dc %p.\n", s, dc);
*surface = s;
return WINED3D_OK;
}
}
}
return WINED3DERR_INVALIDCALL;
}
HRESULT device_init(struct wined3d_device *device, struct wined3d *wined3d,
UINT adapter_idx, enum wined3d_device_type device_type, HWND focus_window, DWORD flags,
BYTE surface_alignment, struct wined3d_device_parent *device_parent)
{
struct wined3d_adapter *adapter = &wined3d->adapters[adapter_idx];
const struct fragment_pipeline *fragment_pipeline;
struct shader_caps shader_caps;
struct fragment_caps ffp_caps;
unsigned int i;
HRESULT hr;
device->ref = 1;
device->wined3d = wined3d;
wined3d_incref(device->wined3d);
device->adapter = wined3d->adapter_count ? adapter : NULL;
device->device_parent = device_parent;
list_init(&device->resources);
list_init(&device->shaders);
device->surface_alignment = surface_alignment;
/* Save the creation parameters. */
device->create_parms.adapter_idx = adapter_idx;
device->create_parms.device_type = device_type;
device->create_parms.focus_window = focus_window;
device->create_parms.flags = flags;
for (i = 0; i < PATCHMAP_SIZE; ++i) list_init(&device->patches[i]);
select_shader_mode(&adapter->gl_info, &device->ps_selected_mode, &device->vs_selected_mode);
device->shader_backend = adapter->shader_backend;
if (device->shader_backend)
{
device->shader_backend->shader_get_caps(&adapter->gl_info, &shader_caps);
device->vshader_version = shader_caps.VertexShaderVersion;
device->pshader_version = shader_caps.PixelShaderVersion;
device->d3d_vshader_constantF = shader_caps.MaxVertexShaderConst;
device->d3d_pshader_constantF = shader_caps.MaxPixelShaderConst;
device->vs_clipping = shader_caps.VSClipping;
}
fragment_pipeline = adapter->fragment_pipe;
device->frag_pipe = fragment_pipeline;
if (fragment_pipeline)
{
fragment_pipeline->get_caps(&adapter->gl_info, &ffp_caps);
device->max_ffp_textures = ffp_caps.MaxSimultaneousTextures;
hr = compile_state_table(device->StateTable, device->multistate_funcs, &adapter->gl_info,
ffp_vertexstate_template, fragment_pipeline, misc_state_template);
if (FAILED(hr))
{
ERR("Failed to compile state table, hr %#x.\n", hr);
wined3d_decref(device->wined3d);
return hr;
}
}
device->blitter = adapter->blitter;
hr = wined3d_stateblock_create(device, WINED3D_SBT_INIT, &device->stateBlock);
if (FAILED(hr))
{
WARN("Failed to create stateblock.\n");
for (i = 0; i < sizeof(device->multistate_funcs) / sizeof(device->multistate_funcs[0]); ++i)
{
HeapFree(GetProcessHeap(), 0, device->multistate_funcs[i]);
}
wined3d_decref(device->wined3d);
return hr;
}
TRACE("Created stateblock %p.\n", device->stateBlock);
device->updateStateBlock = device->stateBlock;
wined3d_stateblock_incref(device->updateStateBlock);
return WINED3D_OK;
}
void device_invalidate_state(const struct wined3d_device *device, DWORD state)
{
DWORD rep = device->StateTable[state].representative;
struct wined3d_context *context;
DWORD idx;
BYTE shift;
UINT i;
for (i = 0; i < device->context_count; ++i)
{
context = device->contexts[i];
if(isStateDirty(context, rep)) continue;
context->dirtyArray[context->numDirtyEntries++] = rep;
idx = rep / (sizeof(*context->isStateDirty) * CHAR_BIT);
shift = rep & ((sizeof(*context->isStateDirty) * CHAR_BIT) - 1);
context->isStateDirty[idx] |= (1 << shift);
}
}
void get_drawable_size_fbo(const struct wined3d_context *context, UINT *width, UINT *height)
{
/* The drawable size of a fbo target is the opengl texture size, which is the power of two size. */
*width = context->current_rt->pow2Width;
*height = context->current_rt->pow2Height;
}
void get_drawable_size_backbuffer(const struct wined3d_context *context, UINT *width, UINT *height)
{
const struct wined3d_swapchain *swapchain = context->swapchain;
/* The drawable size of a backbuffer / aux buffer offscreen target is the size of the
* current context's drawable, which is the size of the back buffer of the swapchain
* the active context belongs to. */
*width = swapchain->desc.backbuffer_width;
*height = swapchain->desc.backbuffer_height;
}
LRESULT device_process_message(struct wined3d_device *device, HWND window, BOOL unicode,
UINT message, WPARAM wparam, LPARAM lparam, WNDPROC proc)
{
if (device->filter_messages)
{
TRACE("Filtering message: window %p, message %#x, wparam %#lx, lparam %#lx.\n",
window, message, wparam, lparam);
if (unicode)
return DefWindowProcW(window, message, wparam, lparam);
else
return DefWindowProcA(window, message, wparam, lparam);
}
if (message == WM_DESTROY)
{
TRACE("unregister window %p.\n", window);
wined3d_unregister_window(window);
if (InterlockedCompareExchangePointer((void **)&device->focus_window, NULL, window) != window)
ERR("Window %p is not the focus window for device %p.\n", window, device);
}
else if (message == WM_DISPLAYCHANGE)
{
device->device_parent->ops->mode_changed(device->device_parent);
}
if (unicode)
return CallWindowProcW(proc, window, message, wparam, lparam);
else
return CallWindowProcA(proc, window, message, wparam, lparam);
}