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 */
105 VTRACE(("glBegin(%x)\n", glPrimType));
108 if (si->use_map & (1 << WINED3D_FFP_POSITION))
110 element = &si->elements[WINED3D_FFP_POSITION];
111 position = element->data + streamOffset[element->stream_idx];
114 if (si->use_map & (1 << WINED3D_FFP_NORMAL))
116 element = &si->elements[WINED3D_FFP_NORMAL];
117 normal = element->data + streamOffset[element->stream_idx];
124 num_untracked_materials = context->num_untracked_materials;
125 if (si->use_map & (1 << WINED3D_FFP_DIFFUSE))
127 element = &si->elements[WINED3D_FFP_DIFFUSE];
128 diffuse = element->data + streamOffset[element->stream_idx];
130 if (num_untracked_materials && element->format_desc->format != WINED3DFMT_B8G8R8A8_UNORM)
131 FIXME("Implement diffuse color tracking from %s\n", debug_d3dformat(element->format_desc->format));
135 glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
138 if (si->use_map & (1 << WINED3D_FFP_SPECULAR))
140 element = &si->elements[WINED3D_FFP_SPECULAR];
141 specular = element->data + streamOffset[element->stream_idx];
143 /* special case where the fog density is stored in the specular alpha channel */
144 if (This->stateBlock->renderState[WINED3DRS_FOGENABLE]
145 && (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE
146 || si->elements[WINED3D_FFP_POSITION].format_desc->format == WINED3DFMT_R32G32B32A32_FLOAT)
147 && This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE)
149 if (gl_info->supported[EXT_FOG_COORD])
151 if (element->format_desc->format == WINED3DFMT_B8G8R8A8_UNORM) specular_fog = TRUE;
152 else FIXME("Implement fog coordinates from %s\n", debug_d3dformat(element->format_desc->format));
160 /* TODO: Use the fog table code from old ddraw */
161 FIXME("Implement fog for transformed vertices in software\n");
167 else if (gl_info->supported[EXT_SECONDARY_COLOR])
169 GL_EXTCALL(glSecondaryColor3fEXT)(0, 0, 0);
172 for (textureNo = 0; textureNo < texture_stages; ++textureNo)
174 int coordIdx = This->stateBlock->textureState[textureNo][WINED3DTSS_TEXCOORDINDEX];
175 DWORD texture_idx = This->texUnitMap[textureNo];
177 if (!gl_info->supported[ARB_MULTITEXTURE] && textureNo > 0)
179 FIXME("Program using multiple concurrent textures which this opengl implementation doesn't support\n");
183 if (!pixelShader && !This->stateBlock->textures[textureNo]) continue;
185 if (texture_idx == WINED3D_UNMAPPED_STAGE) continue;
189 TRACE("tex: %d - Skip tex coords, as being system generated\n", textureNo);
192 else if (coordIdx < 0)
194 FIXME("tex: %d - Coord index %d is less than zero, expect a crash.\n", textureNo, coordIdx);
198 if (si->use_map & (1 << (WINED3D_FFP_TEXCOORD0 + coordIdx)))
200 element = &si->elements[WINED3D_FFP_TEXCOORD0 + coordIdx];
201 texCoords[coordIdx] = element->data + streamOffset[element->stream_idx];
202 tex_mask |= (1 << textureNo);
206 TRACE("tex: %d - Skipping tex coords, as no data supplied\n", textureNo);
207 if (gl_info->supported[ARB_MULTITEXTURE])
208 GL_EXTCALL(glMultiTexCoord4fARB(GL_TEXTURE0_ARB + texture_idx, 0, 0, 0, 1));
210 glTexCoord4f(0, 0, 0, 1);
214 /* We shouldn't start this function if any VBO is involved. Should I put a safety check here?
215 * Guess it's not necessary(we crash then anyway) and would only eat CPU time
218 /* For each primitive */
219 for (vx_index = 0; vx_index < NumVertexes; ++vx_index) {
220 UINT texture, tmp_tex_mask;
221 /* Blending data and Point sizes are not supported by this function. They are not supported by the fixed
222 * function pipeline at all. A Fixme for them is printed after decoding the vertex declaration
225 /* For indexed data, we need to go a few more strides in */
226 if (idxData != NULL) {
228 /* Indexed so work out the number of strides to skip */
230 VTRACE(("Idx for vertex %u = %u\n", vx_index, pIdxBufS[startIdx+vx_index]));
231 SkipnStrides = pIdxBufS[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
233 VTRACE(("Idx for vertex %u = %u\n", vx_index, pIdxBufL[startIdx+vx_index]));
234 SkipnStrides = pIdxBufL[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
238 tmp_tex_mask = tex_mask;
239 for (texture = 0; tmp_tex_mask; tmp_tex_mask >>= 1, ++texture)
245 if (!(tmp_tex_mask & 1)) continue;
247 coord_idx = This->stateBlock->textureState[texture][WINED3DTSS_TEXCOORDINDEX];
248 ptr = texCoords[coord_idx] + (SkipnStrides * si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].stride);
250 texture_idx = This->texUnitMap[texture];
251 multi_texcoord_funcs[si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].format_desc->emit_idx](
252 GL_TEXTURE0_ARB + texture_idx, ptr);
255 /* Diffuse -------------------------------- */
257 const void *ptrToCoords = diffuse + SkipnStrides * si->elements[WINED3D_FFP_DIFFUSE].stride;
259 diffuse_funcs[si->elements[WINED3D_FFP_DIFFUSE].format_desc->emit_idx](ptrToCoords);
260 if (num_untracked_materials)
262 DWORD diffuseColor = ((const DWORD *)ptrToCoords)[0];
266 color[0] = D3DCOLOR_B_R(diffuseColor) / 255.0f;
267 color[1] = D3DCOLOR_B_G(diffuseColor) / 255.0f;
268 color[2] = D3DCOLOR_B_B(diffuseColor) / 255.0f;
269 color[3] = D3DCOLOR_B_A(diffuseColor) / 255.0f;
271 for (i = 0; i < num_untracked_materials; ++i)
273 glMaterialfv(GL_FRONT_AND_BACK, context->untracked_materials[i], color);
278 /* Specular ------------------------------- */
280 const void *ptrToCoords = specular + SkipnStrides * si->elements[WINED3D_FFP_SPECULAR].stride;
282 specular_funcs[si->elements[WINED3D_FFP_SPECULAR].format_desc->emit_idx](ptrToCoords);
286 DWORD specularColor = *(const DWORD *)ptrToCoords;
287 GL_EXTCALL(glFogCoordfEXT(specularColor >> 24));
291 /* Normal -------------------------------- */
292 if (normal != NULL) {
293 const void *ptrToCoords = normal + SkipnStrides * si->elements[WINED3D_FFP_NORMAL].stride;
294 normal_funcs[si->elements[WINED3D_FFP_NORMAL].format_desc->emit_idx](ptrToCoords);
297 /* Position -------------------------------- */
299 const void *ptrToCoords = position + SkipnStrides * si->elements[WINED3D_FFP_POSITION].stride;
300 position_funcs[si->elements[WINED3D_FFP_POSITION].format_desc->emit_idx](ptrToCoords);
303 /* For non indexed mode, step onto next parts */
304 if (idxData == NULL) {
310 checkGLcall("glEnd and previous calls");
313 /* GL locking is done by the caller */
314 static inline void send_attribute(IWineD3DDeviceImpl *This, WINED3DFORMAT format, const UINT index, const void *ptr)
316 const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
320 case WINED3DFMT_R32_FLOAT:
321 GL_EXTCALL(glVertexAttrib1fvARB(index, ptr));
323 case WINED3DFMT_R32G32_FLOAT:
324 GL_EXTCALL(glVertexAttrib2fvARB(index, ptr));
326 case WINED3DFMT_R32G32B32_FLOAT:
327 GL_EXTCALL(glVertexAttrib3fvARB(index, ptr));
329 case WINED3DFMT_R32G32B32A32_FLOAT:
330 GL_EXTCALL(glVertexAttrib4fvARB(index, ptr));
333 case WINED3DFMT_R8G8B8A8_UINT:
334 GL_EXTCALL(glVertexAttrib4ubvARB(index, ptr));
336 case WINED3DFMT_B8G8R8A8_UNORM:
337 if (gl_info->supported[ARB_VERTEX_ARRAY_BGRA])
339 const DWORD *src = ptr;
340 DWORD c = *src & 0xff00ff00;
341 c |= (*src & 0xff0000) >> 16;
342 c |= (*src & 0xff) << 16;
343 GL_EXTCALL(glVertexAttrib4NubvARB(index, (GLubyte *)&c));
346 /* else fallthrough */
347 case WINED3DFMT_R8G8B8A8_UNORM:
348 GL_EXTCALL(glVertexAttrib4NubvARB(index, ptr));
351 case WINED3DFMT_R16G16_SINT:
352 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
354 case WINED3DFMT_R16G16B16A16_SINT:
355 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
358 case WINED3DFMT_R16G16_SNORM:
360 GLshort s[4] = {((const GLshort *)ptr)[0], ((const GLshort *)ptr)[1], 0, 1};
361 GL_EXTCALL(glVertexAttrib4NsvARB(index, s));
364 case WINED3DFMT_R16G16_UNORM:
366 GLushort s[4] = {((const GLushort *)ptr)[0], ((const GLushort *)ptr)[1], 0, 1};
367 GL_EXTCALL(glVertexAttrib4NusvARB(index, s));
370 case WINED3DFMT_R16G16B16A16_SNORM:
371 GL_EXTCALL(glVertexAttrib4NsvARB(index, ptr));
373 case WINED3DFMT_R16G16B16A16_UNORM:
374 GL_EXTCALL(glVertexAttrib4NusvARB(index, ptr));
377 case WINED3DFMT_R10G10B10A2_UINT:
378 FIXME("Unsure about WINED3DDECLTYPE_UDEC3\n");
379 /*glVertexAttrib3usvARB(instancedData[j], (GLushort *) ptr); Does not exist */
381 case WINED3DFMT_R10G10B10A2_SNORM:
382 FIXME("Unsure about WINED3DDECLTYPE_DEC3N\n");
383 /*glVertexAttrib3NusvARB(instancedData[j], (GLushort *) ptr); Does not exist */
386 case WINED3DFMT_R16G16_FLOAT:
387 /* Are those 16 bit floats. C doesn't have a 16 bit float type. I could read the single bits and calculate a 4
388 * byte float according to the IEEE standard
390 if (gl_info->supported[NV_HALF_FLOAT])
392 /* Not supported by GL_ARB_half_float_vertex */
393 GL_EXTCALL(glVertexAttrib2hvNV(index, ptr));
397 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
398 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
399 GL_EXTCALL(glVertexAttrib2fARB(index, x, y));
402 case WINED3DFMT_R16G16B16A16_FLOAT:
403 if (gl_info->supported[NV_HALF_FLOAT])
405 /* Not supported by GL_ARB_half_float_vertex */
406 GL_EXTCALL(glVertexAttrib4hvNV(index, ptr));
410 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
411 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
412 float z = float_16_to_32(((const unsigned short *)ptr) + 2);
413 float w = float_16_to_32(((const unsigned short *)ptr) + 3);
414 GL_EXTCALL(glVertexAttrib4fARB(index, x, y, z, w));
419 ERR("Unexpected attribute format: %s\n", debug_d3dformat(format));
424 /* GL locking is done by the caller */
425 static void drawStridedSlowVs(IWineD3DDevice *iface, const struct wined3d_stream_info *si, UINT numberOfVertices,
426 GLenum glPrimitiveType, const void *idxData, UINT idxSize, UINT startIdx)
428 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
429 long SkipnStrides = startIdx + This->stateBlock->loadBaseVertexIndex;
430 const WORD *pIdxBufS = NULL;
431 const DWORD *pIdxBufL = NULL;
434 IWineD3DStateBlockImpl *stateblock = This->stateBlock;
438 /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
439 * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
440 * idxData will be != NULL
442 if(idxData == NULL) {
443 idxData = buffer_get_sysmem((struct wined3d_buffer *) This->stateBlock->pIndexData);
446 if (idxSize == 2) pIdxBufS = idxData;
447 else pIdxBufL = idxData;
448 } else if (idxData) {
449 ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
453 /* Start drawing in GL */
454 VTRACE(("glBegin(%x)\n", glPrimitiveType));
455 glBegin(glPrimitiveType);
457 for (vx_index = 0; vx_index < numberOfVertices; ++vx_index) {
458 if (idxData != NULL) {
460 /* Indexed so work out the number of strides to skip */
462 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufS[startIdx+vx_index]));
463 SkipnStrides = pIdxBufS[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
465 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufL[startIdx+vx_index]));
466 SkipnStrides = pIdxBufL[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
470 for (i = MAX_ATTRIBS - 1; i >= 0; i--)
472 if (!(si->use_map & (1 << i))) continue;
474 ptr = si->elements[i].data +
475 si->elements[i].stride * SkipnStrides +
476 stateblock->streamOffset[si->elements[i].stream_idx];
478 send_attribute(This, si->elements[i].format_desc->format, i, ptr);
486 /* GL locking is done by the caller */
487 static inline void drawStridedInstanced(IWineD3DDevice *iface, const struct wined3d_stream_info *si,
488 UINT numberOfVertices, GLenum glPrimitiveType, const void *idxData, UINT idxSize,
491 UINT numInstances = 0, i;
492 int numInstancedAttribs = 0, j;
493 UINT instancedData[sizeof(si->elements) / sizeof(*si->elements) /* 16 */];
494 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
495 IWineD3DStateBlockImpl *stateblock = This->stateBlock;
498 /* This is a nasty thing. MSDN says no hardware supports that and apps have to use software vertex processing.
499 * We don't support this for now
501 * Shouldn't be too hard to support with opengl, in theory just call glDrawArrays instead of drawElements.
502 * But the StreamSourceFreq value has a different meaning in that situation.
504 FIXME("Non-indexed instanced drawing is not supported\n");
508 TRACE("(%p) : glElements(%x, %d, ...)\n", This, glPrimitiveType, numberOfVertices);
510 /* First, figure out how many instances we have to draw */
511 for(i = 0; i < MAX_STREAMS; i++) {
512 /* Look at the streams and take the first one which matches */
513 if(((stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INSTANCEDATA) || (stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INDEXEDDATA)) && stateblock->streamSource[i]) {
514 /* D3D9 could set streamFreq 0 with (INSTANCEDATA or INDEXEDDATA) and then it is handled as 1. See d3d9/tests/visual.c-> stream_test() */
515 if(stateblock->streamFreq[i] == 0){
518 numInstances = stateblock->streamFreq[i]; /* use the specified number of instances from the first matched stream. See d3d9/tests/visual.c-> stream_test() */
520 break; /* break, because only the first suitable value is interesting */
524 for (i = 0; i < sizeof(si->elements) / sizeof(*si->elements); ++i)
526 if (!(si->use_map & (1 << i))) continue;
528 if (stateblock->streamFlags[si->elements[i].stream_idx] & WINED3DSTREAMSOURCE_INSTANCEDATA)
530 instancedData[numInstancedAttribs] = i;
531 numInstancedAttribs++;
535 /* now draw numInstances instances :-) */
536 for(i = 0; i < numInstances; i++) {
537 /* Specify the instanced attributes using immediate mode calls */
538 for(j = 0; j < numInstancedAttribs; j++) {
539 const BYTE *ptr = si->elements[instancedData[j]].data +
540 si->elements[instancedData[j]].stride * i +
541 stateblock->streamOffset[si->elements[instancedData[j]].stream_idx];
542 if (si->elements[instancedData[j]].buffer_object)
544 struct wined3d_buffer *vb =
545 (struct wined3d_buffer *)stateblock->streamSource[si->elements[instancedData[j]].stream_idx];
546 ptr += (long) buffer_get_sysmem(vb);
549 send_attribute(This, si->elements[instancedData[j]].format_desc->format, instancedData[j], ptr);
552 glDrawElements(glPrimitiveType, numberOfVertices, idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
553 (const char *)idxData+(idxSize * startIdx));
554 checkGLcall("glDrawElements");
558 static inline void remove_vbos(IWineD3DDeviceImpl *This, struct wined3d_stream_info *s)
562 for (i = 0; i < (sizeof(s->elements) / sizeof(*s->elements)); ++i)
564 struct wined3d_stream_info_element *e;
566 if (!(s->use_map & (1 << i))) continue;
569 if (e->buffer_object)
571 struct wined3d_buffer *vb = (struct wined3d_buffer *)This->stateBlock->streamSource[e->stream_idx];
572 e->buffer_object = 0;
573 e->data = (BYTE *)((unsigned long)e->data + (unsigned long)buffer_get_sysmem(vb));
578 /* Routine common to the draw primitive and draw indexed primitive routines */
579 void drawPrimitive(IWineD3DDevice *iface, UINT index_count, UINT StartIdx, UINT idxSize, const void *idxData)
582 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
583 IWineD3DSurfaceImpl *target;
584 struct wined3d_context *context;
587 if (!index_count) return;
589 if (This->stateBlock->renderState[WINED3DRS_COLORWRITEENABLE])
591 /* Invalidate the back buffer memory so LockRect will read it the next time */
592 for (i = 0; i < This->adapter->gl_info.limits.buffers; ++i)
594 target = (IWineD3DSurfaceImpl *)This->render_targets[i];
597 IWineD3DSurface_LoadLocation((IWineD3DSurface *)target, SFLAG_INDRAWABLE, NULL);
598 IWineD3DSurface_ModifyLocation((IWineD3DSurface *)target, SFLAG_INDRAWABLE, TRUE);
603 /* Signals other modules that a drawing is in progress and the stateblock finalized */
604 This->isInDraw = TRUE;
606 context = context_acquire(This, This->render_targets[0], CTXUSAGE_DRAWPRIM);
608 if (This->stencilBufferTarget) {
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])
617 surface_load_ds_location(This->stencilBufferTarget, context, location);
618 if (This->stateBlock->renderState[WINED3DRS_ZWRITEENABLE])
619 surface_modify_ds_location(This->stencilBufferTarget, location);
622 /* Ok, we will be updating the screen from here onwards so grab the lock */
625 GLenum glPrimType = This->stateBlock->gl_primitive_type;
626 BOOL emulation = FALSE;
627 const struct wined3d_stream_info *stream_info = &This->strided_streams;
628 struct wined3d_stream_info stridedlcl;
630 if (!use_vs(This->stateBlock))
632 if (!This->strided_streams.position_transformed && context->num_untracked_materials
633 && This->stateBlock->renderState[WINED3DRS_LIGHTING])
637 FIXME("Using software emulation because not all material properties could be tracked\n");
640 TRACE("Using software emulation because not all material properties could be tracked\n");
644 else if (context->fog_coord && This->stateBlock->renderState[WINED3DRS_FOGENABLE])
646 /* Either write a pipeline replacement shader or convert the specular alpha from unsigned byte
647 * to a float in the vertex buffer
651 FIXME("Using software emulation because manual fog coordinates are provided\n");
654 TRACE("Using software emulation because manual fog coordinates are provided\n");
660 stream_info = &stridedlcl;
661 memcpy(&stridedlcl, &This->strided_streams, sizeof(stridedlcl));
662 remove_vbos(This, &stridedlcl);
666 if (This->useDrawStridedSlow || emulation) {
667 /* Immediate mode drawing */
668 if (use_vs(This->stateBlock))
672 FIXME("Using immediate mode with vertex shaders for half float emulation\n");
675 TRACE("Using immediate mode with vertex shaders for half float emulation\n");
677 drawStridedSlowVs(iface, stream_info, index_count, glPrimType, idxData, idxSize, StartIdx);
679 drawStridedSlow(iface, context, stream_info, index_count,
680 glPrimType, idxData, idxSize, StartIdx);
682 } else if(This->instancedDraw) {
683 /* Instancing emulation with mixing immediate mode and arrays */
684 drawStridedInstanced(iface, &This->strided_streams, index_count,
685 glPrimType, idxData, idxSize, StartIdx);
687 drawStridedFast(iface, glPrimType, index_count, idxSize, idxData, StartIdx);
691 /* Finished updating the screen, restore lock */
694 wglFlush(); /* Flush to ensure ordering across contexts. */
696 context_release(context);
698 TRACE("Done all gl drawing\n");
701 #ifdef SHOW_FRAME_MAKEUP
703 static long int primCounter = 0;
704 /* NOTE: set primCounter to the value reported by drawprim
705 before you want to to write frame makeup to /tmp */
706 if (primCounter >= 0) {
707 WINED3DLOCKED_RECT r;
709 IWineD3DSurface_LockRect(This->render_targets[0], &r, NULL, WINED3DLOCK_READONLY);
710 sprintf(buffer, "/tmp/backbuffer_%ld.tga", primCounter);
711 TRACE("Saving screenshot %s\n", buffer);
712 IWineD3DSurface_SaveSnapshot(This->render_targets[0], buffer);
713 IWineD3DSurface_UnlockRect(This->render_targets[0]);
715 #ifdef SHOW_TEXTURE_MAKEUP
717 IWineD3DSurface *pSur;
719 for (textureNo = 0; textureNo < MAX_COMBINED_SAMPLERS; ++textureNo) {
720 if (This->stateBlock->textures[textureNo] != NULL) {
721 sprintf(buffer, "/tmp/texture_%p_%ld_%d.tga", This->stateBlock->textures[textureNo], primCounter, textureNo);
722 TRACE("Saving texture %s\n", buffer);
723 if (IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]) == WINED3DRTYPE_TEXTURE) {
724 IWineD3DTexture_GetSurfaceLevel(This->stateBlock->textures[textureNo], 0, &pSur);
725 IWineD3DSurface_SaveSnapshot(pSur, buffer);
726 IWineD3DSurface_Release(pSur);
728 FIXME("base Texture isn't of type texture %d\n", IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]));
735 TRACE("drawprim #%ld\n", primCounter);
740 /* Control goes back to the device, stateblock values may change again */
741 This->isInDraw = FALSE;
744 static void normalize_normal(float *n) {
745 float length = n[0] * n[0] + n[1] * n[1] + n[2] * n[2];
746 if (length == 0.0f) return;
747 length = sqrt(length);
748 n[0] = n[0] / length;
749 n[1] = n[1] / length;
750 n[2] = n[2] / length;
753 /* Tesselates a high order rectangular patch into single triangles using gl evaluators
755 * The problem is that OpenGL does not offer a direct way to return the tesselated primitives,
756 * and they can't be sent off for rendering directly either. Tesselating is slow, so we want
757 * to cache the patches in a vertex buffer. But more importantly, gl can't bind generated
758 * attributes to numbered shader attributes, so we have to store them and rebind them as needed
761 * To read back, the opengl feedback mode is used. This creates a problem because we want
762 * untransformed, unlit vertices, but feedback runs everything through transform and lighting.
763 * Thus disable lighting and set identity matrices to get unmodified colors and positions.
764 * To overcome clipping find the biggest x, y and z values of the vertices in the patch and scale
765 * them to [-1.0;+1.0] and set the viewport up to scale them back.
767 * Normals are more tricky: Draw white vertices with 3 directional lights, and calculate the
768 * resulting colors back to the normals.
770 * NOTE: This function activates a context for blitting, modifies matrices & viewport, but
771 * does not restore it because normally a draw follows immediately afterwards. The caller is
772 * responsible of taking care that either the gl states are restored, or the context activated
773 * for drawing to reset the lastWasBlit flag.
775 HRESULT tesselate_rectpatch(IWineD3DDeviceImpl *This,
776 struct WineD3DRectPatch *patch) {
777 unsigned int i, j, num_quads, out_vertex_size, buffer_size, d3d_out_vertex_size;
778 float max_x = 0.0f, max_y = 0.0f, max_z = 0.0f, neg_z = 0.0f;
779 struct wined3d_stream_info stream_info;
780 struct wined3d_stream_info_element *e;
781 struct wined3d_context *context;
783 const WINED3DRECTPATCH_INFO *info = &patch->RectPatchInfo;
785 GLenum feedback_type;
788 /* Simply activate the context for blitting. This disables all the things we don't want and
789 * takes care of dirtifying. Dirtifying is preferred over pushing / popping, since drawing the
790 * patch (as opposed to normal draws) will most likely need different changes anyway. */
791 context = context_acquire(This, NULL, CTXUSAGE_BLIT);
793 /* First, locate the position data. This is provided in a vertex buffer in the stateblock.
796 device_stream_info_from_declaration(This, FALSE, &stream_info, NULL);
798 e = &stream_info.elements[WINED3D_FFP_POSITION];
799 if (e->buffer_object)
801 struct wined3d_buffer *vb;
802 vb = (struct wined3d_buffer *)This->stateBlock->streamSource[e->stream_idx];
803 e->data = (BYTE *)((unsigned long)e->data + (unsigned long)buffer_get_sysmem(vb));
805 vtxStride = e->stride;
807 vtxStride * info->Stride * info->StartVertexOffsetHeight +
808 vtxStride * info->StartVertexOffsetWidth;
810 /* Not entirely sure about what happens with transformed vertices */
811 if (stream_info.position_transformed) FIXME("Transformed position in rectpatch generation\n");
813 if(vtxStride % sizeof(GLfloat)) {
814 /* glMap2f reads vertex sizes in GLfloats, the d3d stride is in bytes.
815 * I don't see how the stride could not be a multiple of 4, but make sure
818 ERR("Vertex stride is not a multiple of sizeof(GLfloat)\n");
820 if(info->Basis != WINED3DBASIS_BEZIER) {
821 FIXME("Basis is %s, how to handle this?\n", debug_d3dbasis(info->Basis));
823 if(info->Degree != WINED3DDEGREE_CUBIC) {
824 FIXME("Degree is %s, how to handle this?\n", debug_d3ddegree(info->Degree));
827 /* First, get the boundary cube of the input data */
828 for(j = 0; j < info->Height; j++) {
829 for(i = 0; i < info->Width; i++) {
830 const float *v = (const float *)(data + vtxStride * i + vtxStride * info->Stride * j);
831 if(fabs(v[0]) > max_x) max_x = fabs(v[0]);
832 if(fabs(v[1]) > max_y) max_y = fabs(v[1]);
833 if(fabs(v[2]) > max_z) max_z = fabs(v[2]);
834 if(v[2] < neg_z) neg_z = v[2];
838 /* This needs some improvements in the vertex decl code */
839 FIXME("Cannot find data to generate. Only generating position and normals\n");
840 patch->has_normals = TRUE;
841 patch->has_texcoords = FALSE;
845 glMatrixMode(GL_PROJECTION);
846 checkGLcall("glMatrixMode(GL_PROJECTION)");
848 checkGLcall("glLoadIndentity()");
849 glScalef(1.0f / (max_x), 1.0f / (max_y), max_z == 0.0f ? 1.0f : 1.0f / (2.0f * max_z));
850 glTranslatef(0.0f, 0.0f, 0.5f);
851 checkGLcall("glScalef");
852 glViewport(-max_x, -max_y, 2 * (max_x), 2 * (max_y));
853 checkGLcall("glViewport");
855 /* Some states to take care of. If we're in wireframe opengl will produce lines, and confuse
856 * our feedback buffer parser
858 glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
859 checkGLcall("glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)");
860 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_FILLMODE));
861 if(patch->has_normals) {
862 static const GLfloat black[] = {0.0f, 0.0f, 0.0f, 0.0f};
863 static const GLfloat red[] = {1.0f, 0.0f, 0.0f, 0.0f};
864 static const GLfloat green[] = {0.0f, 1.0f, 0.0f, 0.0f};
865 static const GLfloat blue[] = {0.0f, 0.0f, 1.0f, 0.0f};
866 static const GLfloat white[] = {1.0f, 1.0f, 1.0f, 1.0f};
867 glEnable(GL_LIGHTING);
868 checkGLcall("glEnable(GL_LIGHTING)");
869 glLightModelfv(GL_LIGHT_MODEL_AMBIENT, black);
870 checkGLcall("glLightModel for MODEL_AMBIENT");
871 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_AMBIENT));
873 for (i = 3; i < context->gl_info->limits.lights; ++i)
875 glDisable(GL_LIGHT0 + i);
876 checkGLcall("glDisable(GL_LIGHT0 + i)");
877 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(i));
880 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(0));
881 glLightfv(GL_LIGHT0, GL_DIFFUSE, red);
882 glLightfv(GL_LIGHT0, GL_SPECULAR, black);
883 glLightfv(GL_LIGHT0, GL_AMBIENT, black);
884 glLightfv(GL_LIGHT0, GL_POSITION, red);
886 checkGLcall("Setting up light 1");
887 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(1));
888 glLightfv(GL_LIGHT1, GL_DIFFUSE, green);
889 glLightfv(GL_LIGHT1, GL_SPECULAR, black);
890 glLightfv(GL_LIGHT1, GL_AMBIENT, black);
891 glLightfv(GL_LIGHT1, GL_POSITION, green);
893 checkGLcall("Setting up light 2");
894 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(2));
895 glLightfv(GL_LIGHT2, GL_DIFFUSE, blue);
896 glLightfv(GL_LIGHT2, GL_SPECULAR, black);
897 glLightfv(GL_LIGHT2, GL_AMBIENT, black);
898 glLightfv(GL_LIGHT2, GL_POSITION, blue);
900 checkGLcall("Setting up light 3");
902 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_MATERIAL);
903 IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_COLORVERTEX));
904 glDisable(GL_COLOR_MATERIAL);
905 glMaterialfv(GL_FRONT_AND_BACK, GL_EMISSION, black);
906 glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR, black);
907 glMaterialfv(GL_FRONT_AND_BACK, GL_DIFFUSE, white);
908 checkGLcall("Setting up materials");
911 /* Enable the needed maps.
912 * GL_MAP2_VERTEX_3 is needed for positional data.
913 * GL_AUTO_NORMAL to generate normals from the position. Do not use GL_MAP2_NORMAL.
914 * GL_MAP2_TEXTURE_COORD_4 for texture coords
916 num_quads = ceilf(patch->numSegs[0]) * ceilf(patch->numSegs[1]);
917 out_vertex_size = 3 /* position */;
918 d3d_out_vertex_size = 3;
919 glEnable(GL_MAP2_VERTEX_3);
920 if(patch->has_normals && patch->has_texcoords) {
921 FIXME("Texcoords not handled yet\n");
922 feedback_type = GL_3D_COLOR_TEXTURE;
923 out_vertex_size += 8;
924 d3d_out_vertex_size += 7;
925 glEnable(GL_AUTO_NORMAL);
926 glEnable(GL_MAP2_TEXTURE_COORD_4);
927 } else if(patch->has_texcoords) {
928 FIXME("Texcoords not handled yet\n");
929 feedback_type = GL_3D_COLOR_TEXTURE;
930 out_vertex_size += 7;
931 d3d_out_vertex_size += 4;
932 glEnable(GL_MAP2_TEXTURE_COORD_4);
933 } else if(patch->has_normals) {
934 feedback_type = GL_3D_COLOR;
935 out_vertex_size += 4;
936 d3d_out_vertex_size += 3;
937 glEnable(GL_AUTO_NORMAL);
939 feedback_type = GL_3D;
941 checkGLcall("glEnable vertex attrib generation");
943 buffer_size = num_quads * out_vertex_size * 2 /* triangle list */ * 3 /* verts per tri */
944 + 4 * num_quads /* 2 triangle markers per quad + num verts in tri */;
945 feedbuffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, buffer_size * sizeof(float) * 8);
947 glMap2f(GL_MAP2_VERTEX_3,
948 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
949 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
950 (const GLfloat *)data);
951 checkGLcall("glMap2f");
952 if(patch->has_texcoords) {
953 glMap2f(GL_MAP2_TEXTURE_COORD_4,
954 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
955 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
956 (const GLfloat *)data);
957 checkGLcall("glMap2f");
959 glMapGrid2f(ceilf(patch->numSegs[0]), 0.0f, 1.0f, ceilf(patch->numSegs[1]), 0.0f, 1.0f);
960 checkGLcall("glMapGrid2f");
962 glFeedbackBuffer(buffer_size * 2, feedback_type, feedbuffer);
963 checkGLcall("glFeedbackBuffer");
964 glRenderMode(GL_FEEDBACK);
966 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
967 checkGLcall("glEvalMesh2");
969 i = glRenderMode(GL_RENDER);
972 ERR("Feedback failed. Expected %d elements back\n", buffer_size);
973 HeapFree(GetProcessHeap(), 0, feedbuffer);
974 context_release(context);
975 return WINED3DERR_DRIVERINTERNALERROR;
976 } else if(i != buffer_size) {
978 ERR("Unexpected amount of elements returned. Expected %d, got %d\n", buffer_size, i);
979 HeapFree(GetProcessHeap(), 0, feedbuffer);
980 context_release(context);
981 return WINED3DERR_DRIVERINTERNALERROR;
983 TRACE("Got %d elements as expected\n", i);
986 HeapFree(GetProcessHeap(), 0, patch->mem);
987 patch->mem = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, num_quads * 6 * d3d_out_vertex_size * sizeof(float) * 8);
989 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
990 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
991 ERR("Unexpected token: %f\n", feedbuffer[j]);
994 if(feedbuffer[j + 1] != 3) {
995 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
998 /* Somehow there are different ideas about back / front facing, so fix up the
1001 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 2 + 2]; /* x, triangle 2 */
1002 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 2 + 3]; /* y, triangle 2 */
1003 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 2 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 3 */
1004 if(patch->has_normals) {
1005 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 2 + 5];
1006 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 2 + 6];
1007 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 2 + 7];
1009 i += d3d_out_vertex_size;
1011 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 1 + 2]; /* x, triangle 2 */
1012 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 1 + 3]; /* y, triangle 2 */
1013 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 1 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 2 */
1014 if(patch->has_normals) {
1015 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 1 + 5];
1016 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 1 + 6];
1017 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 1 + 7];
1019 i += d3d_out_vertex_size;
1021 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 0 + 2]; /* x, triangle 1 */
1022 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 0 + 3]; /* y, triangle 1 */
1023 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 0 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 1 */
1024 if(patch->has_normals) {
1025 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 0 + 5];
1026 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 0 + 6];
1027 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 0 + 7];
1029 i += d3d_out_vertex_size;
1032 if(patch->has_normals) {
1033 /* Now do the same with reverse light directions */
1034 static const GLfloat x[] = {-1.0f, 0.0f, 0.0f, 0.0f};
1035 static const GLfloat y[] = { 0.0f, -1.0f, 0.0f, 0.0f};
1036 static const GLfloat z[] = { 0.0f, 0.0f, -1.0f, 0.0f};
1037 glLightfv(GL_LIGHT0, GL_POSITION, x);
1038 glLightfv(GL_LIGHT1, GL_POSITION, y);
1039 glLightfv(GL_LIGHT2, GL_POSITION, z);
1040 checkGLcall("Setting up reverse light directions");
1042 glRenderMode(GL_FEEDBACK);
1043 checkGLcall("glRenderMode(GL_FEEDBACK)");
1044 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1045 checkGLcall("glEvalMesh2");
1046 i = glRenderMode(GL_RENDER);
1047 checkGLcall("glRenderMode(GL_RENDER)");
1050 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1051 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1052 ERR("Unexpected token: %f\n", feedbuffer[j]);
1055 if(feedbuffer[j + 1] != 3) {
1056 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1059 if(patch->mem[i + 3] == 0.0f)
1060 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 2 + 5];
1061 if(patch->mem[i + 4] == 0.0f)
1062 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 2 + 6];
1063 if(patch->mem[i + 5] == 0.0f)
1064 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 2 + 7];
1065 normalize_normal(patch->mem + i + 3);
1066 i += d3d_out_vertex_size;
1068 if(patch->mem[i + 3] == 0.0f)
1069 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 1 + 5];
1070 if(patch->mem[i + 4] == 0.0f)
1071 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 1 + 6];
1072 if(patch->mem[i + 5] == 0.0f)
1073 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 1 + 7];
1074 normalize_normal(patch->mem + i + 3);
1075 i += d3d_out_vertex_size;
1077 if(patch->mem[i + 3] == 0.0f)
1078 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 0 + 5];
1079 if(patch->mem[i + 4] == 0.0f)
1080 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 0 + 6];
1081 if(patch->mem[i + 5] == 0.0f)
1082 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 0 + 7];
1083 normalize_normal(patch->mem + i + 3);
1084 i += d3d_out_vertex_size;
1088 glDisable(GL_MAP2_VERTEX_3);
1089 glDisable(GL_AUTO_NORMAL);
1090 glDisable(GL_MAP2_NORMAL);
1091 glDisable(GL_MAP2_TEXTURE_COORD_4);
1092 checkGLcall("glDisable vertex attrib generation");
1095 context_release(context);
1097 HeapFree(GetProcessHeap(), 0, feedbuffer);
1099 vtxStride = 3 * sizeof(float);
1100 if(patch->has_normals) {
1101 vtxStride += 3 * sizeof(float);
1103 if(patch->has_texcoords) {
1104 vtxStride += 4 * sizeof(float);
1106 memset(&patch->strided, 0, sizeof(patch->strided));
1107 patch->strided.position.format = WINED3DFMT_R32G32B32_FLOAT;
1108 patch->strided.position.lpData = (BYTE *) patch->mem;
1109 patch->strided.position.dwStride = vtxStride;
1111 if(patch->has_normals) {
1112 patch->strided.normal.format = WINED3DFMT_R32G32B32_FLOAT;
1113 patch->strided.normal.lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1114 patch->strided.normal.dwStride = vtxStride;
1116 if(patch->has_texcoords) {
1117 patch->strided.texCoords[0].format = WINED3DFMT_R32G32B32A32_FLOAT;
1118 patch->strided.texCoords[0].lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1119 if(patch->has_normals) {
1120 patch->strided.texCoords[0].lpData += 3 * sizeof(float);
1122 patch->strided.texCoords[0].dwStride = vtxStride;