2 * WINED3D draw functions
4 * Copyright 2002-2004 Jason Edmeades
5 * Copyright 2002-2004 Raphael Junqueira
6 * Copyright 2004 Christian Costa
7 * Copyright 2005 Oliver Stieber
8 * Copyright 2006, 2008 Henri Verbeet
9 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
10 * Copyright 2009 Henri Verbeet for CodeWeavers
12 * This library is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * This library is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with this library; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
28 #include "wined3d_private.h"
30 WINE_DEFAULT_DEBUG_CHANNEL(d3d_draw);
31 #define GLINFO_LOCATION This->adapter->gl_info
36 /* GL locking is done by the caller */
37 static void drawStridedFast(IWineD3DDevice *iface, GLenum primitive_type,
38 UINT count, UINT idx_size, const void *idx_data, UINT start_idx)
42 TRACE("(%p) : glElements(%x, %d, ...)\n", iface, primitive_type, count);
44 glDrawElements(primitive_type, count,
45 idx_size == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
46 (const char *)idx_data + (idx_size * start_idx));
47 checkGLcall("glDrawElements");
51 TRACE("(%p) : glDrawArrays(%#x, %d, %d)\n", iface, primitive_type, start_idx, count);
53 glDrawArrays(primitive_type, start_idx, count);
54 checkGLcall("glDrawArrays");
59 * Actually draw using the supplied information.
60 * Slower GL version which extracts info about each vertex in turn
63 /* GL locking is done by the caller */
64 static void drawStridedSlow(IWineD3DDevice *iface, const struct wined3d_context *context,
65 const struct wined3d_stream_info *si, UINT NumVertexes, GLenum glPrimType,
66 const void *idxData, UINT idxSize, UINT startIdx)
68 unsigned int textureNo = 0;
69 const WORD *pIdxBufS = NULL;
70 const DWORD *pIdxBufL = NULL;
72 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
73 const UINT *streamOffset = This->stateBlock->streamOffset;
74 long SkipnStrides = startIdx + This->stateBlock->loadBaseVertexIndex;
75 BOOL pixelShader = use_ps(This->stateBlock);
76 BOOL specular_fog = FALSE;
77 const BYTE *texCoords[WINED3DDP_MAXTEXCOORD];
78 const BYTE *diffuse = NULL, *specular = NULL, *normal = NULL, *position = NULL;
79 const struct wined3d_gl_info *gl_info = context->gl_info;
80 UINT texture_stages = gl_info->limits.texture_stages;
81 const struct wined3d_stream_info_element *element;
82 UINT num_untracked_materials;
85 TRACE("Using slow vertex array code\n");
87 /* Variable Initialization */
89 /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
90 * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
91 * idxData will be != NULL
94 idxData = buffer_get_sysmem((struct wined3d_buffer *) This->stateBlock->pIndexData);
97 if (idxSize == 2) pIdxBufS = idxData;
98 else pIdxBufL = idxData;
100 ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
104 /* Start drawing in GL */
107 if (si->use_map & (1 << WINED3D_FFP_POSITION))
109 element = &si->elements[WINED3D_FFP_POSITION];
110 position = element->data + streamOffset[element->stream_idx];
113 if (si->use_map & (1 << WINED3D_FFP_NORMAL))
115 element = &si->elements[WINED3D_FFP_NORMAL];
116 normal = element->data + streamOffset[element->stream_idx];
123 num_untracked_materials = context->num_untracked_materials;
124 if (si->use_map & (1 << WINED3D_FFP_DIFFUSE))
126 element = &si->elements[WINED3D_FFP_DIFFUSE];
127 diffuse = element->data + streamOffset[element->stream_idx];
129 if (num_untracked_materials && element->format_desc->format != WINED3DFMT_B8G8R8A8_UNORM)
130 FIXME("Implement diffuse color tracking from %s\n", debug_d3dformat(element->format_desc->format));
134 glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
137 if (si->use_map & (1 << WINED3D_FFP_SPECULAR))
139 element = &si->elements[WINED3D_FFP_SPECULAR];
140 specular = element->data + streamOffset[element->stream_idx];
142 /* special case where the fog density is stored in the specular alpha channel */
143 if (This->stateBlock->renderState[WINED3DRS_FOGENABLE]
144 && (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE
145 || si->elements[WINED3D_FFP_POSITION].format_desc->format == WINED3DFMT_R32G32B32A32_FLOAT)
146 && This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE)
148 if (gl_info->supported[EXT_FOG_COORD])
150 if (element->format_desc->format == WINED3DFMT_B8G8R8A8_UNORM) specular_fog = TRUE;
151 else FIXME("Implement fog coordinates from %s\n", debug_d3dformat(element->format_desc->format));
159 /* TODO: Use the fog table code from old ddraw */
160 FIXME("Implement fog for transformed vertices in software\n");
166 else if (gl_info->supported[EXT_SECONDARY_COLOR])
168 GL_EXTCALL(glSecondaryColor3fEXT)(0, 0, 0);
171 for (textureNo = 0; textureNo < texture_stages; ++textureNo)
173 int coordIdx = This->stateBlock->textureState[textureNo][WINED3DTSS_TEXCOORDINDEX];
174 DWORD texture_idx = This->texUnitMap[textureNo];
176 if (!gl_info->supported[ARB_MULTITEXTURE] && textureNo > 0)
178 FIXME("Program using multiple concurrent textures which this opengl implementation doesn't support\n");
182 if (!pixelShader && !This->stateBlock->textures[textureNo]) continue;
184 if (texture_idx == WINED3D_UNMAPPED_STAGE) continue;
188 TRACE("tex: %d - Skip tex coords, as being system generated\n", textureNo);
191 else if (coordIdx < 0)
193 FIXME("tex: %d - Coord index %d is less than zero, expect a crash.\n", textureNo, coordIdx);
197 if (si->use_map & (1 << (WINED3D_FFP_TEXCOORD0 + coordIdx)))
199 element = &si->elements[WINED3D_FFP_TEXCOORD0 + coordIdx];
200 texCoords[coordIdx] = element->data + streamOffset[element->stream_idx];
201 tex_mask |= (1 << textureNo);
205 TRACE("tex: %d - Skipping tex coords, as no data supplied\n", textureNo);
206 if (gl_info->supported[ARB_MULTITEXTURE])
207 GL_EXTCALL(glMultiTexCoord4fARB(GL_TEXTURE0_ARB + texture_idx, 0, 0, 0, 1));
209 glTexCoord4f(0, 0, 0, 1);
213 /* We shouldn't start this function if any VBO is involved. Should I put a safety check here?
214 * Guess it's not necessary(we crash then anyway) and would only eat CPU time
217 /* For each primitive */
218 for (vx_index = 0; vx_index < NumVertexes; ++vx_index) {
219 UINT texture, tmp_tex_mask;
220 /* Blending data and Point sizes are not supported by this function. They are not supported by the fixed
221 * function pipeline at all. A Fixme for them is printed after decoding the vertex declaration
224 /* For indexed data, we need to go a few more strides in */
225 if (idxData != NULL) {
227 /* Indexed so work out the number of strides to skip */
229 SkipnStrides = pIdxBufS[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
231 SkipnStrides = pIdxBufL[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
234 tmp_tex_mask = tex_mask;
235 for (texture = 0; tmp_tex_mask; tmp_tex_mask >>= 1, ++texture)
241 if (!(tmp_tex_mask & 1)) continue;
243 coord_idx = This->stateBlock->textureState[texture][WINED3DTSS_TEXCOORDINDEX];
244 ptr = texCoords[coord_idx] + (SkipnStrides * si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].stride);
246 texture_idx = This->texUnitMap[texture];
247 multi_texcoord_funcs[si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].format_desc->emit_idx](
248 GL_TEXTURE0_ARB + texture_idx, ptr);
251 /* Diffuse -------------------------------- */
253 const void *ptrToCoords = diffuse + SkipnStrides * si->elements[WINED3D_FFP_DIFFUSE].stride;
255 diffuse_funcs[si->elements[WINED3D_FFP_DIFFUSE].format_desc->emit_idx](ptrToCoords);
256 if (num_untracked_materials)
258 DWORD diffuseColor = ((const DWORD *)ptrToCoords)[0];
262 color[0] = D3DCOLOR_B_R(diffuseColor) / 255.0f;
263 color[1] = D3DCOLOR_B_G(diffuseColor) / 255.0f;
264 color[2] = D3DCOLOR_B_B(diffuseColor) / 255.0f;
265 color[3] = D3DCOLOR_B_A(diffuseColor) / 255.0f;
267 for (i = 0; i < num_untracked_materials; ++i)
269 glMaterialfv(GL_FRONT_AND_BACK, context->untracked_materials[i], color);
274 /* Specular ------------------------------- */
276 const void *ptrToCoords = specular + SkipnStrides * si->elements[WINED3D_FFP_SPECULAR].stride;
278 specular_funcs[si->elements[WINED3D_FFP_SPECULAR].format_desc->emit_idx](ptrToCoords);
282 DWORD specularColor = *(const DWORD *)ptrToCoords;
283 GL_EXTCALL(glFogCoordfEXT(specularColor >> 24));
287 /* Normal -------------------------------- */
288 if (normal != NULL) {
289 const void *ptrToCoords = normal + SkipnStrides * si->elements[WINED3D_FFP_NORMAL].stride;
290 normal_funcs[si->elements[WINED3D_FFP_NORMAL].format_desc->emit_idx](ptrToCoords);
293 /* Position -------------------------------- */
295 const void *ptrToCoords = position + SkipnStrides * si->elements[WINED3D_FFP_POSITION].stride;
296 position_funcs[si->elements[WINED3D_FFP_POSITION].format_desc->emit_idx](ptrToCoords);
299 /* For non indexed mode, step onto next parts */
300 if (idxData == NULL) {
306 checkGLcall("glEnd and previous calls");
309 /* GL locking is done by the caller */
310 static inline void send_attribute(IWineD3DDeviceImpl *This, WINED3DFORMAT format, const UINT index, const void *ptr)
312 const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
316 case WINED3DFMT_R32_FLOAT:
317 GL_EXTCALL(glVertexAttrib1fvARB(index, ptr));
319 case WINED3DFMT_R32G32_FLOAT:
320 GL_EXTCALL(glVertexAttrib2fvARB(index, ptr));
322 case WINED3DFMT_R32G32B32_FLOAT:
323 GL_EXTCALL(glVertexAttrib3fvARB(index, ptr));
325 case WINED3DFMT_R32G32B32A32_FLOAT:
326 GL_EXTCALL(glVertexAttrib4fvARB(index, ptr));
329 case WINED3DFMT_R8G8B8A8_UINT:
330 GL_EXTCALL(glVertexAttrib4ubvARB(index, ptr));
332 case WINED3DFMT_B8G8R8A8_UNORM:
333 if (gl_info->supported[ARB_VERTEX_ARRAY_BGRA])
335 const DWORD *src = ptr;
336 DWORD c = *src & 0xff00ff00;
337 c |= (*src & 0xff0000) >> 16;
338 c |= (*src & 0xff) << 16;
339 GL_EXTCALL(glVertexAttrib4NubvARB(index, (GLubyte *)&c));
342 /* else fallthrough */
343 case WINED3DFMT_R8G8B8A8_UNORM:
344 GL_EXTCALL(glVertexAttrib4NubvARB(index, ptr));
347 case WINED3DFMT_R16G16_SINT:
348 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
350 case WINED3DFMT_R16G16B16A16_SINT:
351 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
354 case WINED3DFMT_R16G16_SNORM:
356 GLshort s[4] = {((const GLshort *)ptr)[0], ((const GLshort *)ptr)[1], 0, 1};
357 GL_EXTCALL(glVertexAttrib4NsvARB(index, s));
360 case WINED3DFMT_R16G16_UNORM:
362 GLushort s[4] = {((const GLushort *)ptr)[0], ((const GLushort *)ptr)[1], 0, 1};
363 GL_EXTCALL(glVertexAttrib4NusvARB(index, s));
366 case WINED3DFMT_R16G16B16A16_SNORM:
367 GL_EXTCALL(glVertexAttrib4NsvARB(index, ptr));
369 case WINED3DFMT_R16G16B16A16_UNORM:
370 GL_EXTCALL(glVertexAttrib4NusvARB(index, ptr));
373 case WINED3DFMT_R10G10B10A2_UINT:
374 FIXME("Unsure about WINED3DDECLTYPE_UDEC3\n");
375 /*glVertexAttrib3usvARB(instancedData[j], (GLushort *) ptr); Does not exist */
377 case WINED3DFMT_R10G10B10A2_SNORM:
378 FIXME("Unsure about WINED3DDECLTYPE_DEC3N\n");
379 /*glVertexAttrib3NusvARB(instancedData[j], (GLushort *) ptr); Does not exist */
382 case WINED3DFMT_R16G16_FLOAT:
383 /* Are those 16 bit floats. C doesn't have a 16 bit float type. I could read the single bits and calculate a 4
384 * byte float according to the IEEE standard
386 if (gl_info->supported[NV_HALF_FLOAT])
388 /* Not supported by GL_ARB_half_float_vertex */
389 GL_EXTCALL(glVertexAttrib2hvNV(index, ptr));
393 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
394 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
395 GL_EXTCALL(glVertexAttrib2fARB(index, x, y));
398 case WINED3DFMT_R16G16B16A16_FLOAT:
399 if (gl_info->supported[NV_HALF_FLOAT])
401 /* Not supported by GL_ARB_half_float_vertex */
402 GL_EXTCALL(glVertexAttrib4hvNV(index, ptr));
406 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
407 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
408 float z = float_16_to_32(((const unsigned short *)ptr) + 2);
409 float w = float_16_to_32(((const unsigned short *)ptr) + 3);
410 GL_EXTCALL(glVertexAttrib4fARB(index, x, y, z, w));
415 ERR("Unexpected attribute format: %s\n", debug_d3dformat(format));
420 /* GL locking is done by the caller */
421 static void drawStridedSlowVs(IWineD3DDevice *iface, const struct wined3d_stream_info *si, UINT numberOfVertices,
422 GLenum glPrimitiveType, const void *idxData, UINT idxSize, UINT startIdx)
424 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
425 long SkipnStrides = startIdx + This->stateBlock->loadBaseVertexIndex;
426 const WORD *pIdxBufS = NULL;
427 const DWORD *pIdxBufL = NULL;
430 IWineD3DStateBlockImpl *stateblock = This->stateBlock;
434 /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
435 * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
436 * idxData will be != NULL
438 if(idxData == NULL) {
439 idxData = buffer_get_sysmem((struct wined3d_buffer *) This->stateBlock->pIndexData);
442 if (idxSize == 2) pIdxBufS = idxData;
443 else pIdxBufL = idxData;
444 } else if (idxData) {
445 ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
449 /* Start drawing in GL */
450 glBegin(glPrimitiveType);
452 for (vx_index = 0; vx_index < numberOfVertices; ++vx_index) {
453 if (idxData != NULL) {
455 /* Indexed so work out the number of strides to skip */
457 SkipnStrides = pIdxBufS[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
459 SkipnStrides = pIdxBufL[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
462 for (i = MAX_ATTRIBS - 1; i >= 0; i--)
464 if (!(si->use_map & (1 << i))) continue;
466 ptr = si->elements[i].data +
467 si->elements[i].stride * SkipnStrides +
468 stateblock->streamOffset[si->elements[i].stream_idx];
470 send_attribute(This, si->elements[i].format_desc->format, i, ptr);
478 /* GL locking is done by the caller */
479 static inline void drawStridedInstanced(IWineD3DDevice *iface, const struct wined3d_stream_info *si,
480 UINT numberOfVertices, GLenum glPrimitiveType, const void *idxData, UINT idxSize,
483 UINT numInstances = 0, i;
484 int numInstancedAttribs = 0, j;
485 UINT instancedData[sizeof(si->elements) / sizeof(*si->elements) /* 16 */];
486 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
487 IWineD3DStateBlockImpl *stateblock = This->stateBlock;
490 /* This is a nasty thing. MSDN says no hardware supports that and apps have to use software vertex processing.
491 * We don't support this for now
493 * Shouldn't be too hard to support with opengl, in theory just call glDrawArrays instead of drawElements.
494 * But the StreamSourceFreq value has a different meaning in that situation.
496 FIXME("Non-indexed instanced drawing is not supported\n");
500 TRACE("(%p) : glElements(%x, %d, ...)\n", This, glPrimitiveType, numberOfVertices);
502 /* First, figure out how many instances we have to draw */
503 for(i = 0; i < MAX_STREAMS; i++) {
504 /* Look at the streams and take the first one which matches */
505 if(((stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INSTANCEDATA) || (stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INDEXEDDATA)) && stateblock->streamSource[i]) {
506 /* D3D9 could set streamFreq 0 with (INSTANCEDATA or INDEXEDDATA) and then it is handled as 1. See d3d9/tests/visual.c-> stream_test() */
507 if(stateblock->streamFreq[i] == 0){
510 numInstances = stateblock->streamFreq[i]; /* use the specified number of instances from the first matched stream. See d3d9/tests/visual.c-> stream_test() */
512 break; /* break, because only the first suitable value is interesting */
516 for (i = 0; i < sizeof(si->elements) / sizeof(*si->elements); ++i)
518 if (!(si->use_map & (1 << i))) continue;
520 if (stateblock->streamFlags[si->elements[i].stream_idx] & WINED3DSTREAMSOURCE_INSTANCEDATA)
522 instancedData[numInstancedAttribs] = i;
523 numInstancedAttribs++;
527 /* now draw numInstances instances :-) */
528 for(i = 0; i < numInstances; i++) {
529 /* Specify the instanced attributes using immediate mode calls */
530 for(j = 0; j < numInstancedAttribs; j++) {
531 const BYTE *ptr = si->elements[instancedData[j]].data +
532 si->elements[instancedData[j]].stride * i +
533 stateblock->streamOffset[si->elements[instancedData[j]].stream_idx];
534 if (si->elements[instancedData[j]].buffer_object)
536 struct wined3d_buffer *vb =
537 (struct wined3d_buffer *)stateblock->streamSource[si->elements[instancedData[j]].stream_idx];
538 ptr += (long) buffer_get_sysmem(vb);
541 send_attribute(This, si->elements[instancedData[j]].format_desc->format, instancedData[j], ptr);
544 glDrawElements(glPrimitiveType, numberOfVertices, idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
545 (const char *)idxData+(idxSize * startIdx));
546 checkGLcall("glDrawElements");
550 static inline void remove_vbos(IWineD3DDeviceImpl *This, struct wined3d_stream_info *s)
554 for (i = 0; i < (sizeof(s->elements) / sizeof(*s->elements)); ++i)
556 struct wined3d_stream_info_element *e;
558 if (!(s->use_map & (1 << i))) continue;
561 if (e->buffer_object)
563 struct wined3d_buffer *vb = (struct wined3d_buffer *)This->stateBlock->streamSource[e->stream_idx];
564 e->buffer_object = 0;
565 e->data = (BYTE *)((unsigned long)e->data + (unsigned long)buffer_get_sysmem(vb));
570 /* Routine common to the draw primitive and draw indexed primitive routines */
571 void drawPrimitive(IWineD3DDevice *iface, UINT index_count, UINT StartIdx, UINT idxSize, const void *idxData)
574 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
575 struct wined3d_context *context;
578 if (!index_count) return;
580 if (This->stateBlock->renderState[WINED3DRS_COLORWRITEENABLE])
582 /* Invalidate the back buffer memory so LockRect will read it the next time */
583 for (i = 0; i < This->adapter->gl_info.limits.buffers; ++i)
585 IWineD3DSurface *target = (IWineD3DSurface *)This->render_targets[i];
588 IWineD3DSurface_LoadLocation(target, SFLAG_INDRAWABLE, NULL);
589 IWineD3DSurface_ModifyLocation(target, SFLAG_INDRAWABLE, TRUE);
594 /* Signals other modules that a drawing is in progress and the stateblock finalized */
595 This->isInDraw = TRUE;
597 context = context_acquire(This, This->render_targets[0]);
600 context_release(context);
601 WARN("Invalid context, skipping draw.\n");
605 context_apply_draw_state(context, This);
607 if (This->depth_stencil)
609 /* Note that this depends on the context_acquire() call above to set
610 * This->render_offscreen properly. We don't currently take the
611 * Z-compare function into account, but we could skip loading the
612 * depthstencil for D3DCMP_NEVER and D3DCMP_ALWAYS as well. Also note
613 * that we never copy the stencil data.*/
614 DWORD location = context->render_offscreen ? SFLAG_DS_OFFSCREEN : SFLAG_DS_ONSCREEN;
615 if (This->stateBlock->renderState[WINED3DRS_ZWRITEENABLE]
616 || This->stateBlock->renderState[WINED3DRS_ZENABLE])
618 RECT current_rect, draw_rect, r;
620 if (location == SFLAG_DS_ONSCREEN && This->depth_stencil != This->onscreen_depth_stencil)
621 device_switch_onscreen_ds(This, context, This->depth_stencil);
623 if (This->depth_stencil->Flags & location)
624 SetRect(¤t_rect, 0, 0,
625 This->depth_stencil->ds_current_size.cx,
626 This->depth_stencil->ds_current_size.cy);
628 SetRectEmpty(¤t_rect);
630 device_get_draw_rect(This, &draw_rect);
632 IntersectRect(&r, &draw_rect, ¤t_rect);
633 if (!EqualRect(&r, &draw_rect))
634 surface_load_ds_location(This->depth_stencil, context, location);
636 if (This->stateBlock->renderState[WINED3DRS_ZWRITEENABLE])
637 surface_modify_ds_location(This->depth_stencil, location,
638 This->depth_stencil->ds_current_size.cx,
639 This->depth_stencil->ds_current_size.cy);
643 /* Ok, we will be updating the screen from here onwards so grab the lock */
646 GLenum glPrimType = This->stateBlock->gl_primitive_type;
647 BOOL emulation = FALSE;
648 const struct wined3d_stream_info *stream_info = &This->strided_streams;
649 struct wined3d_stream_info stridedlcl;
651 if (!use_vs(This->stateBlock))
653 if (!This->strided_streams.position_transformed && context->num_untracked_materials
654 && This->stateBlock->renderState[WINED3DRS_LIGHTING])
658 FIXME("Using software emulation because not all material properties could be tracked\n");
661 TRACE("Using software emulation because not all material properties could be tracked\n");
665 else if (context->fog_coord && This->stateBlock->renderState[WINED3DRS_FOGENABLE])
667 /* Either write a pipeline replacement shader or convert the specular alpha from unsigned byte
668 * to a float in the vertex buffer
672 FIXME("Using software emulation because manual fog coordinates are provided\n");
675 TRACE("Using software emulation because manual fog coordinates are provided\n");
681 stream_info = &stridedlcl;
682 memcpy(&stridedlcl, &This->strided_streams, sizeof(stridedlcl));
683 remove_vbos(This, &stridedlcl);
687 if (This->useDrawStridedSlow || emulation) {
688 /* Immediate mode drawing */
689 if (use_vs(This->stateBlock))
693 FIXME("Using immediate mode with vertex shaders for half float emulation\n");
696 TRACE("Using immediate mode with vertex shaders for half float emulation\n");
698 drawStridedSlowVs(iface, stream_info, index_count, glPrimType, idxData, idxSize, StartIdx);
700 drawStridedSlow(iface, context, stream_info, index_count,
701 glPrimType, idxData, idxSize, StartIdx);
703 } else if(This->instancedDraw) {
704 /* Instancing emulation with mixing immediate mode and arrays */
705 drawStridedInstanced(iface, &This->strided_streams, index_count,
706 glPrimType, idxData, idxSize, StartIdx);
708 drawStridedFast(iface, glPrimType, index_count, idxSize, idxData, StartIdx);
712 /* Finished updating the screen, restore lock */
715 for(i = 0; i < This->num_buffer_queries; i++)
717 wined3d_event_query_issue(This->buffer_queries[i], This);
720 if (wined3d_settings.strict_draw_ordering) wglFlush(); /* Flush to ensure ordering across contexts. */
722 context_release(context);
724 TRACE("Done all gl drawing\n");
727 #ifdef SHOW_FRAME_MAKEUP
729 static long int primCounter = 0;
730 /* NOTE: set primCounter to the value reported by drawprim
731 before you want to to write frame makeup to /tmp */
732 if (primCounter >= 0) {
733 WINED3DLOCKED_RECT r;
735 IWineD3DSurface_LockRect(This->render_targets[0], &r, NULL, WINED3DLOCK_READONLY);
736 sprintf(buffer, "/tmp/backbuffer_%ld.tga", primCounter);
737 TRACE("Saving screenshot %s\n", buffer);
738 IWineD3DSurface_SaveSnapshot(This->render_targets[0], buffer);
739 IWineD3DSurface_UnlockRect(This->render_targets[0]);
741 #ifdef SHOW_TEXTURE_MAKEUP
743 IWineD3DSurface *pSur;
745 for (textureNo = 0; textureNo < MAX_COMBINED_SAMPLERS; ++textureNo) {
746 if (This->stateBlock->textures[textureNo] != NULL) {
747 sprintf(buffer, "/tmp/texture_%p_%ld_%d.tga", This->stateBlock->textures[textureNo], primCounter, textureNo);
748 TRACE("Saving texture %s\n", buffer);
749 if (IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]) == WINED3DRTYPE_TEXTURE) {
750 IWineD3DTexture_GetSurfaceLevel(This->stateBlock->textures[textureNo], 0, &pSur);
751 IWineD3DSurface_SaveSnapshot(pSur, buffer);
752 IWineD3DSurface_Release(pSur);
754 FIXME("base Texture isn't of type texture %d\n", IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]));
761 TRACE("drawprim #%ld\n", primCounter);
766 /* Control goes back to the device, stateblock values may change again */
767 This->isInDraw = FALSE;
770 static void normalize_normal(float *n) {
771 float length = n[0] * n[0] + n[1] * n[1] + n[2] * n[2];
772 if (length == 0.0f) return;
773 length = sqrt(length);
774 n[0] = n[0] / length;
775 n[1] = n[1] / length;
776 n[2] = n[2] / length;
779 /* Tesselates a high order rectangular patch into single triangles using gl evaluators
781 * The problem is that OpenGL does not offer a direct way to return the tesselated primitives,
782 * and they can't be sent off for rendering directly either. Tesselating is slow, so we want
783 * to cache the patches in a vertex buffer. But more importantly, gl can't bind generated
784 * attributes to numbered shader attributes, so we have to store them and rebind them as needed
787 * To read back, the opengl feedback mode is used. This creates a problem because we want
788 * untransformed, unlit vertices, but feedback runs everything through transform and lighting.
789 * Thus disable lighting and set identity matrices to get unmodified colors and positions.
790 * To overcome clipping find the biggest x, y and z values of the vertices in the patch and scale
791 * them to [-1.0;+1.0] and set the viewport up to scale them back.
793 * Normals are more tricky: Draw white vertices with 3 directional lights, and calculate the
794 * resulting colors back to the normals.
796 * NOTE: This function activates a context for blitting, modifies matrices & viewport, but
797 * does not restore it because normally a draw follows immediately afterwards. The caller is
798 * responsible of taking care that either the gl states are restored, or the context activated
799 * for drawing to reset the lastWasBlit flag.
801 HRESULT tesselate_rectpatch(IWineD3DDeviceImpl *This,
802 struct WineD3DRectPatch *patch) {
803 unsigned int i, j, num_quads, out_vertex_size, buffer_size, d3d_out_vertex_size;
804 float max_x = 0.0f, max_y = 0.0f, max_z = 0.0f, neg_z = 0.0f;
805 struct wined3d_stream_info stream_info;
806 struct wined3d_stream_info_element *e;
807 struct wined3d_context *context;
809 const WINED3DRECTPATCH_INFO *info = &patch->RectPatchInfo;
811 GLenum feedback_type;
814 /* Simply activate the context for blitting. This disables all the things we don't want and
815 * takes care of dirtifying. Dirtifying is preferred over pushing / popping, since drawing the
816 * patch (as opposed to normal draws) will most likely need different changes anyway. */
817 context = context_acquire(This, NULL);
818 context_apply_blit_state(context, This);
820 /* First, locate the position data. This is provided in a vertex buffer in the stateblock.
823 device_stream_info_from_declaration(This, FALSE, &stream_info, NULL);
825 e = &stream_info.elements[WINED3D_FFP_POSITION];
826 if (e->buffer_object)
828 struct wined3d_buffer *vb;
829 vb = (struct wined3d_buffer *)This->stateBlock->streamSource[e->stream_idx];
830 e->data = (BYTE *)((unsigned long)e->data + (unsigned long)buffer_get_sysmem(vb));
832 vtxStride = e->stride;
834 vtxStride * info->Stride * info->StartVertexOffsetHeight +
835 vtxStride * info->StartVertexOffsetWidth;
837 /* Not entirely sure about what happens with transformed vertices */
838 if (stream_info.position_transformed) FIXME("Transformed position in rectpatch generation\n");
840 if(vtxStride % sizeof(GLfloat)) {
841 /* glMap2f reads vertex sizes in GLfloats, the d3d stride is in bytes.
842 * I don't see how the stride could not be a multiple of 4, but make sure
845 ERR("Vertex stride is not a multiple of sizeof(GLfloat)\n");
847 if(info->Basis != WINED3DBASIS_BEZIER) {
848 FIXME("Basis is %s, how to handle this?\n", debug_d3dbasis(info->Basis));
850 if(info->Degree != WINED3DDEGREE_CUBIC) {
851 FIXME("Degree is %s, how to handle this?\n", debug_d3ddegree(info->Degree));
854 /* First, get the boundary cube of the input data */
855 for(j = 0; j < info->Height; j++) {
856 for(i = 0; i < info->Width; i++) {
857 const float *v = (const float *)(data + vtxStride * i + vtxStride * info->Stride * j);
858 if(fabs(v[0]) > max_x) max_x = fabs(v[0]);
859 if(fabs(v[1]) > max_y) max_y = fabs(v[1]);
860 if(fabs(v[2]) > max_z) max_z = fabs(v[2]);
861 if(v[2] < neg_z) neg_z = v[2];
865 /* This needs some improvements in the vertex decl code */
866 FIXME("Cannot find data to generate. Only generating position and normals\n");
867 patch->has_normals = TRUE;
868 patch->has_texcoords = FALSE;
872 glMatrixMode(GL_PROJECTION);
873 checkGLcall("glMatrixMode(GL_PROJECTION)");
875 checkGLcall("glLoadIndentity()");
876 glScalef(1.0f / (max_x), 1.0f / (max_y), max_z == 0.0f ? 1.0f : 1.0f / (2.0f * max_z));
877 glTranslatef(0.0f, 0.0f, 0.5f);
878 checkGLcall("glScalef");
879 glViewport(-max_x, -max_y, 2 * (max_x), 2 * (max_y));
880 checkGLcall("glViewport");
882 /* Some states to take care of. If we're in wireframe opengl will produce lines, and confuse
883 * our feedback buffer parser
885 glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
886 checkGLcall("glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)");
887 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_FILLMODE));
888 if(patch->has_normals) {
889 static const GLfloat black[] = {0.0f, 0.0f, 0.0f, 0.0f};
890 static const GLfloat red[] = {1.0f, 0.0f, 0.0f, 0.0f};
891 static const GLfloat green[] = {0.0f, 1.0f, 0.0f, 0.0f};
892 static const GLfloat blue[] = {0.0f, 0.0f, 1.0f, 0.0f};
893 static const GLfloat white[] = {1.0f, 1.0f, 1.0f, 1.0f};
894 glEnable(GL_LIGHTING);
895 checkGLcall("glEnable(GL_LIGHTING)");
896 glLightModelfv(GL_LIGHT_MODEL_AMBIENT, black);
897 checkGLcall("glLightModel for MODEL_AMBIENT");
898 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_AMBIENT));
900 for (i = 3; i < context->gl_info->limits.lights; ++i)
902 glDisable(GL_LIGHT0 + i);
903 checkGLcall("glDisable(GL_LIGHT0 + i)");
904 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(i));
907 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(0));
908 glLightfv(GL_LIGHT0, GL_DIFFUSE, red);
909 glLightfv(GL_LIGHT0, GL_SPECULAR, black);
910 glLightfv(GL_LIGHT0, GL_AMBIENT, black);
911 glLightfv(GL_LIGHT0, GL_POSITION, red);
913 checkGLcall("Setting up light 1");
914 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(1));
915 glLightfv(GL_LIGHT1, GL_DIFFUSE, green);
916 glLightfv(GL_LIGHT1, GL_SPECULAR, black);
917 glLightfv(GL_LIGHT1, GL_AMBIENT, black);
918 glLightfv(GL_LIGHT1, GL_POSITION, green);
920 checkGLcall("Setting up light 2");
921 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(2));
922 glLightfv(GL_LIGHT2, GL_DIFFUSE, blue);
923 glLightfv(GL_LIGHT2, GL_SPECULAR, black);
924 glLightfv(GL_LIGHT2, GL_AMBIENT, black);
925 glLightfv(GL_LIGHT2, GL_POSITION, blue);
927 checkGLcall("Setting up light 3");
929 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_MATERIAL);
930 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_COLORVERTEX));
931 glDisable(GL_COLOR_MATERIAL);
932 glMaterialfv(GL_FRONT_AND_BACK, GL_EMISSION, black);
933 glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR, black);
934 glMaterialfv(GL_FRONT_AND_BACK, GL_DIFFUSE, white);
935 checkGLcall("Setting up materials");
938 /* Enable the needed maps.
939 * GL_MAP2_VERTEX_3 is needed for positional data.
940 * GL_AUTO_NORMAL to generate normals from the position. Do not use GL_MAP2_NORMAL.
941 * GL_MAP2_TEXTURE_COORD_4 for texture coords
943 num_quads = ceilf(patch->numSegs[0]) * ceilf(patch->numSegs[1]);
944 out_vertex_size = 3 /* position */;
945 d3d_out_vertex_size = 3;
946 glEnable(GL_MAP2_VERTEX_3);
947 if(patch->has_normals && patch->has_texcoords) {
948 FIXME("Texcoords not handled yet\n");
949 feedback_type = GL_3D_COLOR_TEXTURE;
950 out_vertex_size += 8;
951 d3d_out_vertex_size += 7;
952 glEnable(GL_AUTO_NORMAL);
953 glEnable(GL_MAP2_TEXTURE_COORD_4);
954 } else if(patch->has_texcoords) {
955 FIXME("Texcoords not handled yet\n");
956 feedback_type = GL_3D_COLOR_TEXTURE;
957 out_vertex_size += 7;
958 d3d_out_vertex_size += 4;
959 glEnable(GL_MAP2_TEXTURE_COORD_4);
960 } else if(patch->has_normals) {
961 feedback_type = GL_3D_COLOR;
962 out_vertex_size += 4;
963 d3d_out_vertex_size += 3;
964 glEnable(GL_AUTO_NORMAL);
966 feedback_type = GL_3D;
968 checkGLcall("glEnable vertex attrib generation");
970 buffer_size = num_quads * out_vertex_size * 2 /* triangle list */ * 3 /* verts per tri */
971 + 4 * num_quads /* 2 triangle markers per quad + num verts in tri */;
972 feedbuffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, buffer_size * sizeof(float) * 8);
974 glMap2f(GL_MAP2_VERTEX_3,
975 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
976 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
977 (const GLfloat *)data);
978 checkGLcall("glMap2f");
979 if(patch->has_texcoords) {
980 glMap2f(GL_MAP2_TEXTURE_COORD_4,
981 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
982 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
983 (const GLfloat *)data);
984 checkGLcall("glMap2f");
986 glMapGrid2f(ceilf(patch->numSegs[0]), 0.0f, 1.0f, ceilf(patch->numSegs[1]), 0.0f, 1.0f);
987 checkGLcall("glMapGrid2f");
989 glFeedbackBuffer(buffer_size * 2, feedback_type, feedbuffer);
990 checkGLcall("glFeedbackBuffer");
991 glRenderMode(GL_FEEDBACK);
993 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
994 checkGLcall("glEvalMesh2");
996 i = glRenderMode(GL_RENDER);
999 ERR("Feedback failed. Expected %d elements back\n", buffer_size);
1000 HeapFree(GetProcessHeap(), 0, feedbuffer);
1001 context_release(context);
1002 return WINED3DERR_DRIVERINTERNALERROR;
1003 } else if(i != buffer_size) {
1005 ERR("Unexpected amount of elements returned. Expected %d, got %d\n", buffer_size, i);
1006 HeapFree(GetProcessHeap(), 0, feedbuffer);
1007 context_release(context);
1008 return WINED3DERR_DRIVERINTERNALERROR;
1010 TRACE("Got %d elements as expected\n", i);
1013 HeapFree(GetProcessHeap(), 0, patch->mem);
1014 patch->mem = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, num_quads * 6 * d3d_out_vertex_size * sizeof(float) * 8);
1016 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1017 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1018 ERR("Unexpected token: %f\n", feedbuffer[j]);
1021 if(feedbuffer[j + 1] != 3) {
1022 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1025 /* Somehow there are different ideas about back / front facing, so fix up the
1028 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 2 + 2]; /* x, triangle 2 */
1029 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 2 + 3]; /* y, triangle 2 */
1030 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 2 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 3 */
1031 if(patch->has_normals) {
1032 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 2 + 5];
1033 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 2 + 6];
1034 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 2 + 7];
1036 i += d3d_out_vertex_size;
1038 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 1 + 2]; /* x, triangle 2 */
1039 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 1 + 3]; /* y, triangle 2 */
1040 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 1 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 2 */
1041 if(patch->has_normals) {
1042 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 1 + 5];
1043 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 1 + 6];
1044 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 1 + 7];
1046 i += d3d_out_vertex_size;
1048 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 0 + 2]; /* x, triangle 1 */
1049 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 0 + 3]; /* y, triangle 1 */
1050 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 0 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 1 */
1051 if(patch->has_normals) {
1052 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 0 + 5];
1053 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 0 + 6];
1054 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 0 + 7];
1056 i += d3d_out_vertex_size;
1059 if(patch->has_normals) {
1060 /* Now do the same with reverse light directions */
1061 static const GLfloat x[] = {-1.0f, 0.0f, 0.0f, 0.0f};
1062 static const GLfloat y[] = { 0.0f, -1.0f, 0.0f, 0.0f};
1063 static const GLfloat z[] = { 0.0f, 0.0f, -1.0f, 0.0f};
1064 glLightfv(GL_LIGHT0, GL_POSITION, x);
1065 glLightfv(GL_LIGHT1, GL_POSITION, y);
1066 glLightfv(GL_LIGHT2, GL_POSITION, z);
1067 checkGLcall("Setting up reverse light directions");
1069 glRenderMode(GL_FEEDBACK);
1070 checkGLcall("glRenderMode(GL_FEEDBACK)");
1071 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1072 checkGLcall("glEvalMesh2");
1073 i = glRenderMode(GL_RENDER);
1074 checkGLcall("glRenderMode(GL_RENDER)");
1077 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1078 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1079 ERR("Unexpected token: %f\n", feedbuffer[j]);
1082 if(feedbuffer[j + 1] != 3) {
1083 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1086 if(patch->mem[i + 3] == 0.0f)
1087 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 2 + 5];
1088 if(patch->mem[i + 4] == 0.0f)
1089 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 2 + 6];
1090 if(patch->mem[i + 5] == 0.0f)
1091 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 2 + 7];
1092 normalize_normal(patch->mem + i + 3);
1093 i += d3d_out_vertex_size;
1095 if(patch->mem[i + 3] == 0.0f)
1096 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 1 + 5];
1097 if(patch->mem[i + 4] == 0.0f)
1098 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 1 + 6];
1099 if(patch->mem[i + 5] == 0.0f)
1100 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 1 + 7];
1101 normalize_normal(patch->mem + i + 3);
1102 i += d3d_out_vertex_size;
1104 if(patch->mem[i + 3] == 0.0f)
1105 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 0 + 5];
1106 if(patch->mem[i + 4] == 0.0f)
1107 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 0 + 6];
1108 if(patch->mem[i + 5] == 0.0f)
1109 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 0 + 7];
1110 normalize_normal(patch->mem + i + 3);
1111 i += d3d_out_vertex_size;
1115 glDisable(GL_MAP2_VERTEX_3);
1116 glDisable(GL_AUTO_NORMAL);
1117 glDisable(GL_MAP2_NORMAL);
1118 glDisable(GL_MAP2_TEXTURE_COORD_4);
1119 checkGLcall("glDisable vertex attrib generation");
1122 context_release(context);
1124 HeapFree(GetProcessHeap(), 0, feedbuffer);
1126 vtxStride = 3 * sizeof(float);
1127 if(patch->has_normals) {
1128 vtxStride += 3 * sizeof(float);
1130 if(patch->has_texcoords) {
1131 vtxStride += 4 * sizeof(float);
1133 memset(&patch->strided, 0, sizeof(patch->strided));
1134 patch->strided.position.format = WINED3DFMT_R32G32B32_FLOAT;
1135 patch->strided.position.lpData = (BYTE *) patch->mem;
1136 patch->strided.position.dwStride = vtxStride;
1138 if(patch->has_normals) {
1139 patch->strided.normal.format = WINED3DFMT_R32G32B32_FLOAT;
1140 patch->strided.normal.lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1141 patch->strided.normal.dwStride = vtxStride;
1143 if(patch->has_texcoords) {
1144 patch->strided.texCoords[0].format = WINED3DFMT_R32G32B32A32_FLOAT;
1145 patch->strided.texCoords[0].lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1146 if(patch->has_normals) {
1147 patch->strided.texCoords[0].lpData += 3 * sizeof(float);
1149 patch->strided.texCoords[0].dwStride = vtxStride;