winmm: Fix a failing mixer test on 98 and ME.
[wine] / dlls / wined3d / drawprim.c
1 /*
2  * WINED3D draw functions
3  *
4  * Copyright 2002-2004 Jason Edmeades
5  * Copyright 2002-2004 Raphael Junqueira
6  * Copyright 2004 Christian Costa
7  * Copyright 2005 Oliver Stieber
8  * Copyright 2006, 2008 Henri Verbeet
9  * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
10  *
11  * This library is free software; you can redistribute it and/or
12  * modify it under the terms of the GNU Lesser General Public
13  * License as published by the Free Software Foundation; either
14  * version 2.1 of the License, or (at your option) any later version.
15  *
16  * This library is distributed in the hope that it will be useful,
17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
19  * Lesser General Public License for more details.
20  *
21  * You should have received a copy of the GNU Lesser General Public
22  * License along with this library; if not, write to the Free Software
23  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24  */
25
26 #include "config.h"
27 #include "wined3d_private.h"
28
29 WINE_DEFAULT_DEBUG_CHANNEL(d3d_draw);
30 #define GLINFO_LOCATION This->adapter->gl_info
31
32 #include <stdio.h>
33 #include <math.h>
34
35 /* Issues the glBegin call for gl given the primitive type and count */
36 static DWORD primitiveToGl(WINED3DPRIMITIVETYPE PrimitiveType,
37                     DWORD            NumPrimitives,
38                     GLenum          *primType)
39 {
40     DWORD   NumVertexes = NumPrimitives;
41
42     switch (PrimitiveType) {
43     case WINED3DPT_POINTLIST:
44         TRACE("POINTS\n");
45         *primType   = GL_POINTS;
46         NumVertexes = NumPrimitives;
47         break;
48
49     case WINED3DPT_LINELIST:
50         TRACE("LINES\n");
51         *primType   = GL_LINES;
52         NumVertexes = NumPrimitives * 2;
53         break;
54
55     case WINED3DPT_LINESTRIP:
56         TRACE("LINE_STRIP\n");
57         *primType   = GL_LINE_STRIP;
58         NumVertexes = NumPrimitives + 1;
59         break;
60
61     case WINED3DPT_TRIANGLELIST:
62         TRACE("TRIANGLES\n");
63         *primType   = GL_TRIANGLES;
64         NumVertexes = NumPrimitives * 3;
65         break;
66
67     case WINED3DPT_TRIANGLESTRIP:
68         TRACE("TRIANGLE_STRIP\n");
69         *primType   = GL_TRIANGLE_STRIP;
70         NumVertexes = NumPrimitives + 2;
71         break;
72
73     case WINED3DPT_TRIANGLEFAN:
74         TRACE("TRIANGLE_FAN\n");
75         *primType   = GL_TRIANGLE_FAN;
76         NumVertexes = NumPrimitives + 2;
77         break;
78
79     default:
80         FIXME("Unhandled primitive\n");
81         *primType    = GL_POINTS;
82         break;
83     }
84     return NumVertexes;
85 }
86
87 static BOOL fixed_get_input(
88     BYTE usage, BYTE usage_idx,
89     unsigned int* regnum) {
90
91     *regnum = -1;
92
93     /* Those positions must have the order in the
94      * named part of the strided data */
95
96     if ((usage == WINED3DDECLUSAGE_POSITION || usage == WINED3DDECLUSAGE_POSITIONT) && usage_idx == 0)
97         *regnum = 0;
98     else if (usage == WINED3DDECLUSAGE_BLENDWEIGHT && usage_idx == 0)
99         *regnum = 1;
100     else if (usage == WINED3DDECLUSAGE_BLENDINDICES && usage_idx == 0)
101         *regnum = 2;
102     else if (usage == WINED3DDECLUSAGE_NORMAL && usage_idx == 0)
103         *regnum = 3;
104     else if (usage == WINED3DDECLUSAGE_PSIZE && usage_idx == 0)
105         *regnum = 4;
106     else if (usage == WINED3DDECLUSAGE_COLOR && usage_idx == 0)
107         *regnum = 5;
108     else if (usage == WINED3DDECLUSAGE_COLOR && usage_idx == 1)
109         *regnum = 6;
110     else if (usage == WINED3DDECLUSAGE_TEXCOORD && usage_idx < WINED3DDP_MAXTEXCOORD)
111         *regnum = 7 + usage_idx;
112     else if ((usage == WINED3DDECLUSAGE_POSITION || usage == WINED3DDECLUSAGE_POSITIONT) && usage_idx == 1)
113         *regnum = 7 + WINED3DDP_MAXTEXCOORD;
114     else if (usage == WINED3DDECLUSAGE_NORMAL && usage_idx == 1)
115         *regnum = 8 + WINED3DDP_MAXTEXCOORD;
116     else if (usage == WINED3DDECLUSAGE_TANGENT && usage_idx == 0)
117         *regnum = 9 + WINED3DDP_MAXTEXCOORD;
118     else if (usage == WINED3DDECLUSAGE_BINORMAL && usage_idx == 0)
119         *regnum = 10 + WINED3DDP_MAXTEXCOORD;
120     else if (usage == WINED3DDECLUSAGE_TESSFACTOR && usage_idx == 0)
121         *regnum = 11 + WINED3DDP_MAXTEXCOORD;
122     else if (usage == WINED3DDECLUSAGE_FOG && usage_idx == 0)
123         *regnum = 12 + WINED3DDP_MAXTEXCOORD;
124     else if (usage == WINED3DDECLUSAGE_DEPTH && usage_idx == 0)
125         *regnum = 13 + WINED3DDP_MAXTEXCOORD;
126     else if (usage == WINED3DDECLUSAGE_SAMPLE && usage_idx == 0)
127         *regnum = 14 + WINED3DDP_MAXTEXCOORD;
128
129     if (*regnum == -1) {
130         FIXME("Unsupported input stream [usage=%s, usage_idx=%u]\n",
131             debug_d3ddeclusage(usage), usage_idx);
132         return FALSE;
133     }
134     return TRUE;
135 }
136
137 void primitiveDeclarationConvertToStridedData(
138      IWineD3DDevice *iface,
139      BOOL useVertexShaderFunction,
140      WineDirect3DVertexStridedData *strided,
141      BOOL *fixup) {
142
143      /* We need to deal with frequency data!*/
144
145     BYTE  *data    = NULL;
146     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
147     IWineD3DVertexDeclarationImpl* vertexDeclaration = (IWineD3DVertexDeclarationImpl *)This->stateBlock->vertexDecl;
148     int i;
149     WINED3DVERTEXELEMENT *element;
150     DWORD stride;
151     DWORD numPreloadStreams = This->stateBlock->streamIsUP ? 0 : vertexDeclaration->num_streams;
152     DWORD *streams = vertexDeclaration->streams;
153
154     /* Check for transformed vertices, disable vertex shader if present */
155     strided->u.s.position_transformed = vertexDeclaration->position_transformed;
156     if(vertexDeclaration->position_transformed) {
157         useVertexShaderFunction = FALSE;
158     }
159
160     /* Translate the declaration into strided data */
161     for (i = 0 ; i < vertexDeclaration->declarationWNumElements - 1; ++i) {
162         GLint streamVBO = 0;
163         BOOL stride_used;
164         unsigned int idx;
165
166         element = vertexDeclaration->pDeclarationWine + i;
167         TRACE("%p Element %p (%d of %d)\n", vertexDeclaration->pDeclarationWine,
168             element,  i + 1, vertexDeclaration->declarationWNumElements - 1);
169
170         if (This->stateBlock->streamSource[element->Stream] == NULL)
171             continue;
172
173         stride  = This->stateBlock->streamStride[element->Stream];
174         if (This->stateBlock->streamIsUP) {
175             TRACE("Stream is up %d, %p\n", element->Stream, This->stateBlock->streamSource[element->Stream]);
176             streamVBO = 0;
177             data    = (BYTE *)This->stateBlock->streamSource[element->Stream];
178         } else {
179             TRACE("Stream isn't up %d, %p\n", element->Stream, This->stateBlock->streamSource[element->Stream]);
180             data    = IWineD3DVertexBufferImpl_GetMemory(This->stateBlock->streamSource[element->Stream], 0, &streamVBO);
181
182             /* Can't use vbo's if the base vertex index is negative. OpenGL doesn't accept negative offsets
183              * (or rather offsets bigger than the vbo, because the pointer is unsigned), so use system memory
184              * sources. In most sane cases the pointer - offset will still be > 0, otherwise it will wrap
185              * around to some big value. Hope that with the indices, the driver wraps it back internally. If
186              * not, drawStridedSlow is needed, including a vertex buffer path.
187              */
188             if(This->stateBlock->loadBaseVertexIndex < 0) {
189                 WARN("loadBaseVertexIndex is < 0 (%d), not using vbos\n", This->stateBlock->loadBaseVertexIndex);
190                 streamVBO = 0;
191                 data = ((IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[element->Stream])->resource.allocatedMemory;
192                 if((UINT_PTR)data < -This->stateBlock->loadBaseVertexIndex * stride) {
193                     FIXME("System memory vertex data load offset is negative!\n");
194                 }
195             }
196
197             if(fixup) {
198                 if( streamVBO != 0) *fixup = TRUE;
199                 else if(*fixup && !useVertexShaderFunction &&
200                        (element->Usage == WINED3DDECLUSAGE_COLOR ||
201                         element->Usage == WINED3DDECLUSAGE_POSITIONT)) {
202                     static BOOL warned = FALSE;
203                     if(!warned) {
204                         /* This may be bad with the fixed function pipeline */
205                         FIXME("Missing vbo streams with unfixed colors or transformed position, expect problems\n");
206                         warned = TRUE;
207                     }
208                 }
209             }
210         }
211         data += element->Offset;
212
213         TRACE("Offset %d Stream %d UsageIndex %d\n", element->Offset, element->Stream, element->UsageIndex);
214
215         if (useVertexShaderFunction)
216             stride_used = vshader_get_input(This->stateBlock->vertexShader,
217                 element->Usage, element->UsageIndex, &idx);
218         else
219             stride_used = fixed_get_input(element->Usage, element->UsageIndex, &idx);
220
221         if (stride_used) {
222             TRACE("Load %s array %u [usage=%s, usage_idx=%u, "
223                     "stream=%u, offset=%u, stride=%u, type=%s, VBO=%u]\n",
224                     useVertexShaderFunction? "shader": "fixed function", idx,
225                     debug_d3ddeclusage(element->Usage), element->UsageIndex,
226                     element->Stream, element->Offset, stride, debug_d3ddecltype(element->Type), streamVBO);
227
228             if (!useVertexShaderFunction && !vertexDeclaration->ffp_valid[i]) {
229                 WARN("Skipping unsupported fixed function element of type %s and usage %s\n",
230                     debug_d3ddecltype(element->Type), debug_d3ddeclusage(element->Usage));
231             } else {
232                 strided->u.input[idx].lpData = data;
233                 strided->u.input[idx].dwType = element->Type;
234                 strided->u.input[idx].dwStride = stride;
235                 strided->u.input[idx].VBO = streamVBO;
236                 strided->u.input[idx].streamNo = element->Stream;
237             }
238         }
239     }
240     /* Now call PreLoad on all the vertex buffers. In the very rare case
241      * that the buffers stopps converting PreLoad will dirtify the VDECL again.
242      * The vertex buffer can now use the strided structure in the device instead of finding its
243      * own again.
244      *
245      * NULL streams won't be recorded in the array, UP streams won't be either. A stream is only
246      * once in there.
247      */
248     for(i=0; i < numPreloadStreams; i++) {
249         IWineD3DVertexBuffer *vb = This->stateBlock->streamSource[streams[i]];
250         if(vb) {
251             IWineD3DVertexBuffer_PreLoad(vb);
252         }
253     }
254 }
255
256 static void drawStridedFast(IWineD3DDevice *iface,UINT numberOfVertices, GLenum glPrimitiveType,
257                      const void *idxData, short idxSize, ULONG minIndex, ULONG startIdx, ULONG startVertex) {
258     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
259
260     if (idxSize != 0 /* This crashes sometimes!*/) {
261         TRACE("(%p) : glElements(%x, %d, %d, ...)\n", This, glPrimitiveType, numberOfVertices, minIndex);
262         idxData = idxData == (void *)-1 ? NULL : idxData;
263 #if 1
264         glDrawElements(glPrimitiveType, numberOfVertices, idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
265                      (const char *)idxData+(idxSize * startIdx));
266         checkGLcall("glDrawElements");
267 #else /* using drawRangeElements may be faster */
268
269         glDrawRangeElements(glPrimitiveType, minIndex, minIndex + numberOfVertices - 1, numberOfVertices,
270                       idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
271                       (const char *)idxData+(idxSize * startIdx));
272         checkGLcall("glDrawRangeElements");
273 #endif
274
275     } else {
276         TRACE("(%p) : glDrawArrays(%#x, %d, %d)\n", This, glPrimitiveType, startVertex, numberOfVertices);
277         glDrawArrays(glPrimitiveType, startVertex, numberOfVertices);
278         checkGLcall("glDrawArrays");
279     }
280
281     return;
282 }
283
284 /*
285  * Actually draw using the supplied information.
286  * Slower GL version which extracts info about each vertex in turn
287  */
288
289 static void drawStridedSlow(IWineD3DDevice *iface, WineDirect3DVertexStridedData *sd,
290                      UINT NumVertexes, GLenum glPrimType,
291                      const void *idxData, short idxSize, ULONG minIndex, ULONG startIdx, ULONG startVertex) {
292
293     unsigned int               textureNo    = 0;
294     const WORD                *pIdxBufS     = NULL;
295     const DWORD               *pIdxBufL     = NULL;
296     LONG                       vx_index;
297     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
298     UINT *streamOffset = This->stateBlock->streamOffset;
299     long                      SkipnStrides = startVertex + This->stateBlock->loadBaseVertexIndex;
300     BOOL                      pixelShader = use_ps(This);
301
302     BYTE *texCoords[WINED3DDP_MAXTEXCOORD];
303     BYTE *diffuse = NULL, *specular = NULL, *normal = NULL, *position = NULL;
304
305     TRACE("Using slow vertex array code\n");
306
307     /* Variable Initialization */
308     if (idxSize != 0) {
309         /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
310          * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
311          * idxData will be != NULL
312          */
313         if(idxData == NULL) {
314             idxData = ((IWineD3DIndexBufferImpl *) This->stateBlock->pIndexData)->resource.allocatedMemory;
315         }
316
317         if (idxSize == 2) pIdxBufS = (const WORD *) idxData;
318         else pIdxBufL = (const DWORD *) idxData;
319     } else if (idxData) {
320         ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
321         return;
322     }
323
324     /* Adding the stream offset once is cheaper than doing it every iteration. Do not modify the strided data, it is a pointer
325      * to the strided Data in the device and might be needed intact on the next draw
326      */
327     for (textureNo = 0; textureNo < GL_LIMITS(texture_stages); ++textureNo) {
328         if(sd->u.s.texCoords[textureNo].lpData) {
329             texCoords[textureNo] = sd->u.s.texCoords[textureNo].lpData + streamOffset[sd->u.s.texCoords[textureNo].streamNo];
330         } else {
331             texCoords[textureNo] = NULL;
332         }
333     }
334     if(sd->u.s.diffuse.lpData) {
335         diffuse = sd->u.s.diffuse.lpData + streamOffset[sd->u.s.diffuse.streamNo];
336     }
337     if(sd->u.s.specular.lpData) {
338         specular = sd->u.s.specular.lpData + streamOffset[sd->u.s.specular.streamNo];
339     }
340     if(sd->u.s.normal.lpData) {
341         normal = sd->u.s.normal.lpData + streamOffset[sd->u.s.normal.streamNo];
342     }
343     if(sd->u.s.position.lpData) {
344         position = sd->u.s.position.lpData + streamOffset[sd->u.s.position.streamNo];
345     }
346
347     if(FIXME_ON(d3d_draw)) {
348         if(specular && This->stateBlock->renderState[WINED3DRS_FOGENABLE] &&
349            (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || sd->u.s.position_transformed )&&
350            This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
351             if(GL_SUPPORT(EXT_FOG_COORD) && sd->u.s.specular.dwType != WINED3DDECLTYPE_D3DCOLOR) {
352                 FIXME("Implement fog coordinates from %s\n", debug_d3ddecltype(sd->u.s.specular.dwType));
353             }
354         }
355         if(This->activeContext->num_untracked_materials && sd->u.s.diffuse.dwType != WINED3DDECLTYPE_D3DCOLOR) {
356             FIXME("Implement diffuse color tracking from %s\n", debug_d3ddecltype(sd->u.s.diffuse.dwType));
357         }
358     }
359
360     /* Start drawing in GL */
361     VTRACE(("glBegin(%x)\n", glPrimType));
362     glBegin(glPrimType);
363
364     /* Default settings for data that is not passed */
365     if (sd->u.s.normal.lpData == NULL) {
366         glNormal3f(0, 0, 0);
367     }
368     if(sd->u.s.diffuse.lpData == NULL) {
369         glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
370     }
371     if(sd->u.s.specular.lpData == NULL) {
372         if (GL_SUPPORT(EXT_SECONDARY_COLOR)) {
373             GL_EXTCALL(glSecondaryColor3fEXT)(0, 0, 0);
374         }
375     }
376
377     /* We shouldn't start this function if any VBO is involved. Should I put a safety check here?
378      * Guess it's not necessary(we crash then anyway) and would only eat CPU time
379      */
380
381     /* For each primitive */
382     for (vx_index = 0; vx_index < NumVertexes; ++vx_index) {
383         /* Blending data and Point sizes are not supported by this function. They are not supported by the fixed
384          * function pipeline at all. A Fixme for them is printed after decoding the vertex declaration
385          */
386
387         /* For indexed data, we need to go a few more strides in */
388         if (idxData != NULL) {
389
390             /* Indexed so work out the number of strides to skip */
391             if (idxSize == 2) {
392                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufS[startIdx+vx_index]));
393                 SkipnStrides = pIdxBufS[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
394             } else {
395                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufL[startIdx+vx_index]));
396                 SkipnStrides = pIdxBufL[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
397             }
398         }
399
400         /* Texture coords --------------------------- */
401         for (textureNo = 0; textureNo < GL_LIMITS(texture_stages); ++textureNo) {
402
403             if (!GL_SUPPORT(ARB_MULTITEXTURE) && textureNo > 0) {
404                 FIXME("Program using multiple concurrent textures which this opengl implementation doesn't support\n");
405                 continue ;
406             }
407
408             /* Query tex coords */
409             if (This->stateBlock->textures[textureNo] != NULL || pixelShader) {
410                 int    coordIdx = This->stateBlock->textureState[textureNo][WINED3DTSS_TEXCOORDINDEX];
411                 int texture_idx = This->texUnitMap[textureNo];
412                 void *ptrToCoords;
413
414                 if (coordIdx > 7) {
415                     VTRACE(("tex: %d - Skip tex coords, as being system generated\n", textureNo));
416                     continue;
417                 } else if (coordIdx < 0) {
418                     FIXME("tex: %d - Coord index %d is less than zero, expect a crash.\n", textureNo, coordIdx);
419                     continue;
420                 }
421
422                 if (texture_idx == -1) continue;
423
424                 if (texCoords[coordIdx] == NULL) {
425                     TRACE("tex: %d - Skipping tex coords, as no data supplied\n", textureNo);
426                     if (GL_SUPPORT(ARB_MULTITEXTURE)) {
427                         GL_EXTCALL(glMultiTexCoord4fARB(GL_TEXTURE0_ARB + texture_idx, 0, 0, 0, 1));
428                     } else {
429                         glTexCoord4f(0, 0, 0, 1);
430                     }
431                     continue;
432                 }
433
434                 ptrToCoords = texCoords[coordIdx] + (SkipnStrides * sd->u.s.texCoords[coordIdx].dwStride);
435                 if (GL_SUPPORT(ARB_MULTITEXTURE))
436                     multi_texcoord_funcs[sd->u.s.texCoords[coordIdx].dwType](GL_TEXTURE0_ARB + texture_idx, ptrToCoords);
437                 else
438                     texcoord_funcs[sd->u.s.texCoords[coordIdx].dwType](ptrToCoords);
439             }
440         } /* End of textures */
441
442         /* Diffuse -------------------------------- */
443         if (diffuse) {
444             DWORD *ptrToCoords = (DWORD *)(diffuse + (SkipnStrides * sd->u.s.diffuse.dwStride));
445
446             diffuse_funcs[sd->u.s.diffuse.dwType]((void *) ptrToCoords);
447             if(This->activeContext->num_untracked_materials) {
448                 DWORD diffuseColor = ptrToCoords[0];
449                 unsigned char i;
450                 float color[4];
451
452                 color[0] = D3DCOLOR_B_R(diffuseColor) / 255.0;
453                 color[1] = D3DCOLOR_B_G(diffuseColor) / 255.0;
454                 color[2] = D3DCOLOR_B_B(diffuseColor) / 255.0;
455                 color[3] = D3DCOLOR_B_A(diffuseColor) / 255.0;
456
457                 for(i = 0; i < This->activeContext->num_untracked_materials; i++) {
458                     glMaterialfv(GL_FRONT_AND_BACK, This->activeContext->untracked_materials[i], color);
459                 }
460             }
461         }
462
463         /* Specular ------------------------------- */
464         if (specular) {
465             DWORD *ptrToCoords = (DWORD *)(specular + (SkipnStrides * sd->u.s.specular.dwStride));
466
467             /* special case where the fog density is stored in the specular alpha channel */
468             if(This->stateBlock->renderState[WINED3DRS_FOGENABLE] &&
469               (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || sd->u.s.position.dwType == WINED3DDECLTYPE_FLOAT4 )&&
470               This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
471                 if(GL_SUPPORT(EXT_FOG_COORD)) {
472                     DWORD specularColor = ptrToCoords[0];
473                     GL_EXTCALL(glFogCoordfEXT(specularColor >> 24));
474                 } else {
475                     static BOOL warned = FALSE;
476                     if(!warned) {
477                         /* TODO: Use the fog table code from old ddraw */
478                         FIXME("Implement fog for transformed vertices in software\n");
479                         warned = TRUE;
480                     }
481                 }
482             }
483
484             specular_funcs[sd->u.s.specular.dwType]((void *) ptrToCoords);
485         }
486
487         /* Normal -------------------------------- */
488         if (normal != NULL) {
489             float *ptrToCoords = (float *)(normal + (SkipnStrides * sd->u.s.normal.dwStride));
490             normal_funcs[sd->u.s.normal.dwType](ptrToCoords);
491         }
492
493         /* Position -------------------------------- */
494         if (position) {
495             float *ptrToCoords = (float *)(position + (SkipnStrides * sd->u.s.position.dwStride));
496             position_funcs[sd->u.s.position.dwType](ptrToCoords);
497         }
498
499         /* For non indexed mode, step onto next parts */
500         if (idxData == NULL) {
501             ++SkipnStrides;
502         }
503     }
504
505     glEnd();
506     checkGLcall("glEnd and previous calls");
507 }
508
509 static inline void send_attribute(IWineD3DDeviceImpl *This, const DWORD type, const UINT index, const void *ptr) {
510     switch(type) {
511         case WINED3DDECLTYPE_FLOAT1:
512             GL_EXTCALL(glVertexAttrib1fvARB(index, (float *) ptr));
513             break;
514         case WINED3DDECLTYPE_FLOAT2:
515             GL_EXTCALL(glVertexAttrib2fvARB(index, (float *) ptr));
516             break;
517         case WINED3DDECLTYPE_FLOAT3:
518             GL_EXTCALL(glVertexAttrib3fvARB(index, (float *) ptr));
519             break;
520         case WINED3DDECLTYPE_FLOAT4:
521             GL_EXTCALL(glVertexAttrib4fvARB(index, (float *) ptr));
522             break;
523
524         case WINED3DDECLTYPE_UBYTE4:
525             GL_EXTCALL(glVertexAttrib4ubvARB(index, ptr));
526             break;
527         case WINED3DDECLTYPE_UBYTE4N:
528         case WINED3DDECLTYPE_D3DCOLOR:
529             GL_EXTCALL(glVertexAttrib4NubvARB(index, ptr));
530             break;
531
532         case WINED3DDECLTYPE_SHORT2:
533             GL_EXTCALL(glVertexAttrib4svARB(index, (GLshort *) ptr));
534             break;
535         case WINED3DDECLTYPE_SHORT4:
536             GL_EXTCALL(glVertexAttrib4svARB(index, (GLshort *) ptr));
537             break;
538
539         case WINED3DDECLTYPE_SHORT2N:
540         {
541             GLshort s[4] = {((short *) ptr)[0], ((short *) ptr)[1], 0, 1};
542             GL_EXTCALL(glVertexAttrib4NsvARB(index, s));
543             break;
544         }
545         case WINED3DDECLTYPE_USHORT2N:
546         {
547             GLushort s[4] = {((unsigned short *) ptr)[0], ((unsigned short *) ptr)[1], 0, 1};
548             GL_EXTCALL(glVertexAttrib4NusvARB(index, s));
549             break;
550         }
551         case WINED3DDECLTYPE_SHORT4N:
552             GL_EXTCALL(glVertexAttrib4NsvARB(index, (GLshort *) ptr));
553             break;
554         case WINED3DDECLTYPE_USHORT4N:
555             GL_EXTCALL(glVertexAttrib4NusvARB(index, (GLushort *) ptr));
556             break;
557
558         case WINED3DDECLTYPE_UDEC3:
559             FIXME("Unsure about WINED3DDECLTYPE_UDEC3\n");
560             /*glVertexAttrib3usvARB(instancedData[j], (GLushort *) ptr); Does not exist */
561             break;
562         case WINED3DDECLTYPE_DEC3N:
563             FIXME("Unsure about WINED3DDECLTYPE_DEC3N\n");
564             /*glVertexAttrib3NusvARB(instancedData[j], (GLushort *) ptr); Does not exist */
565             break;
566
567         case WINED3DDECLTYPE_FLOAT16_2:
568             /* Are those 16 bit floats. C doesn't have a 16 bit float type. I could read the single bits and calculate a 4
569              * byte float according to the IEEE standard
570              */
571             if (GL_SUPPORT(NV_HALF_FLOAT)) {
572                 GL_EXTCALL(glVertexAttrib2hvNV(index, (GLhalfNV *)ptr));
573             } else {
574                 float x = float_16_to_32(((unsigned short *) ptr) + 0);
575                 float y = float_16_to_32(((unsigned short *) ptr) + 1);
576                 GL_EXTCALL(glVertexAttrib2fARB(index, x, y));
577             }
578             break;
579         case WINED3DDECLTYPE_FLOAT16_4:
580             if (GL_SUPPORT(NV_HALF_FLOAT)) {
581                 GL_EXTCALL(glVertexAttrib4hvNV(index, (GLhalfNV *)ptr));
582             } else {
583                 float x = float_16_to_32(((unsigned short *) ptr) + 0);
584                 float y = float_16_to_32(((unsigned short *) ptr) + 1);
585                 float z = float_16_to_32(((unsigned short *) ptr) + 2);
586                 float w = float_16_to_32(((unsigned short *) ptr) + 3);
587                 GL_EXTCALL(glVertexAttrib4fARB(index, x, y, z, w));
588             }
589             break;
590
591         case WINED3DDECLTYPE_UNUSED:
592         default:
593             ERR("Unexpected attribute declaration: %d\n", type);
594             break;
595     }
596 }
597
598 static void drawStridedSlowVs(IWineD3DDevice *iface, WineDirect3DVertexStridedData *sd, UINT numberOfVertices,
599                               GLenum glPrimitiveType, const void *idxData, short idxSize, ULONG minIndex, ULONG startIdx,
600                               ULONG startVertex) {
601
602     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
603     long                      SkipnStrides = startVertex + This->stateBlock->loadBaseVertexIndex;
604     const WORD                *pIdxBufS     = NULL;
605     const DWORD               *pIdxBufL     = NULL;
606     LONG                       vx_index;
607     int i;
608     IWineD3DStateBlockImpl *stateblock = This->stateBlock;
609     BYTE *ptr;
610
611     if (idxSize != 0) {
612         /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
613          * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
614          * idxData will be != NULL
615          */
616         if(idxData == NULL) {
617             idxData = ((IWineD3DIndexBufferImpl *) stateblock->pIndexData)->resource.allocatedMemory;
618         }
619
620         if (idxSize == 2) pIdxBufS = (const WORD *) idxData;
621         else pIdxBufL = (const DWORD *) idxData;
622     }
623
624     /* Start drawing in GL */
625     VTRACE(("glBegin(%x)\n", glPrimitiveType));
626     glBegin(glPrimitiveType);
627
628     for (vx_index = 0; vx_index < numberOfVertices; ++vx_index) {
629         if (idxData != NULL) {
630
631             /* Indexed so work out the number of strides to skip */
632             if (idxSize == 2) {
633                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufS[startIdx+vx_index]));
634                 SkipnStrides = pIdxBufS[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
635             } else {
636                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufL[startIdx+vx_index]));
637                 SkipnStrides = pIdxBufL[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
638             }
639         }
640
641         for(i = MAX_ATTRIBS - 1; i >= 0; i--) {
642             if(!sd->u.input[i].lpData) continue;
643
644             ptr = sd->u.input[i].lpData +
645                   sd->u.input[i].dwStride * SkipnStrides +
646                   stateblock->streamOffset[sd->u.input[i].streamNo];
647
648             send_attribute(This, sd->u.input[i].dwType, i, ptr);
649         }
650         SkipnStrides++;
651     }
652
653     glEnd();
654 }
655
656 void depth_blt(IWineD3DDevice *iface, GLuint texture, GLsizei w, GLsizei h) {
657     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
658     GLint old_binding = 0;
659
660     glPushAttrib(GL_ENABLE_BIT | GL_DEPTH_BUFFER_BIT | GL_COLOR_BUFFER_BIT | GL_VIEWPORT_BIT);
661
662     glDisable(GL_CULL_FACE);
663     glEnable(GL_BLEND);
664     glDisable(GL_ALPHA_TEST);
665     glDisable(GL_SCISSOR_TEST);
666     glDisable(GL_STENCIL_TEST);
667     glEnable(GL_DEPTH_TEST);
668     glDepthFunc(GL_ALWAYS);
669     glDepthMask(GL_TRUE);
670     glBlendFunc(GL_ZERO, GL_ONE);
671     glViewport(0, 0, w, h);
672
673     GL_EXTCALL(glActiveTextureARB(GL_TEXTURE0_ARB));
674     glGetIntegerv(GL_TEXTURE_BINDING_2D, &old_binding);
675     glBindTexture(GL_TEXTURE_2D, texture);
676
677     This->shader_backend->shader_select_depth_blt(iface);
678
679     glBegin(GL_TRIANGLE_STRIP);
680     glVertex2f(-1.0f, -1.0f);
681     glVertex2f(1.0f, -1.0f);
682     glVertex2f(-1.0f, 1.0f);
683     glVertex2f(1.0f, 1.0f);
684     glEnd();
685
686     glBindTexture(GL_TEXTURE_2D, old_binding);
687
688     glPopAttrib();
689
690     This->shader_backend->shader_deselect_depth_blt(iface);
691 }
692
693 static inline void drawStridedInstanced(IWineD3DDevice *iface, WineDirect3DVertexStridedData *sd, UINT numberOfVertices,
694                                  GLenum glPrimitiveType, const void *idxData, short idxSize, ULONG minIndex,
695                                  ULONG startIdx, ULONG startVertex) {
696     UINT numInstances = 0;
697     int numInstancedAttribs = 0, i, j;
698     UINT instancedData[sizeof(sd->u.input) / sizeof(sd->u.input[0]) /* 16 */];
699     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
700     IWineD3DStateBlockImpl *stateblock = This->stateBlock;
701
702     if (idxSize == 0) {
703         /* This is a nasty thing. MSDN says no hardware supports that and apps have to use software vertex processing.
704          * We don't support this for now
705          *
706          * Shouldn't be too hard to support with opengl, in theory just call glDrawArrays instead of drawElements.
707          * But the StreamSourceFreq value has a different meaning in that situation.
708          */
709         FIXME("Non-indexed instanced drawing is not supported\n");
710         return;
711     }
712
713     TRACE("(%p) : glElements(%x, %d, %d, ...)\n", This, glPrimitiveType, numberOfVertices, minIndex);
714     idxData = idxData == (void *)-1 ? NULL : idxData;
715
716     /* First, figure out how many instances we have to draw */
717     for(i = 0; i < MAX_STREAMS; i++) {
718         /* Look at the streams and take the first one which matches */
719         if(((stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INSTANCEDATA) || (stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INDEXEDDATA)) && stateblock->streamSource[i]) {
720             /* D3D9 could set streamFreq 0 with (INSTANCEDATA or INDEXEDDATA) and then it is handled as 1. See d3d9/tests/visual.c-> stream_test() */
721             if(stateblock->streamFreq[i] == 0){
722                 numInstances = 1;
723             } else {
724                 numInstances = stateblock->streamFreq[i]; /* use the specified number of instances from the first matched stream. See d3d9/tests/visual.c-> stream_test() */
725             }
726             break; /* break, because only the first suitable value is interesting */
727         }
728     }
729
730     for(i = 0; i < sizeof(sd->u.input) / sizeof(sd->u.input[0]); i++) {
731         if(stateblock->streamFlags[sd->u.input[i].streamNo] & WINED3DSTREAMSOURCE_INSTANCEDATA) {
732             instancedData[numInstancedAttribs] = i;
733             numInstancedAttribs++;
734         }
735     }
736
737     /* now draw numInstances instances :-) */
738     for(i = 0; i < numInstances; i++) {
739         /* Specify the instanced attributes using immediate mode calls */
740         for(j = 0; j < numInstancedAttribs; j++) {
741             BYTE *ptr = sd->u.input[instancedData[j]].lpData +
742                         sd->u.input[instancedData[j]].dwStride * i +
743                         stateblock->streamOffset[sd->u.input[instancedData[j]].streamNo];
744             if(sd->u.input[instancedData[j]].VBO) {
745                 IWineD3DVertexBufferImpl *vb = (IWineD3DVertexBufferImpl *) stateblock->streamSource[sd->u.input[instancedData[j]].streamNo];
746                 ptr += (long) vb->resource.allocatedMemory;
747             }
748
749             send_attribute(This, sd->u.input[instancedData[j]].dwType, instancedData[j], ptr);
750         }
751
752         glDrawElements(glPrimitiveType, numberOfVertices, idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
753                     (const char *)idxData+(idxSize * startIdx));
754         checkGLcall("glDrawElements");
755     }
756 }
757
758 static inline void remove_vbos(IWineD3DDeviceImpl *This, WineDirect3DVertexStridedData *s) {
759     unsigned char i;
760     IWineD3DVertexBufferImpl *vb;
761
762     if(s->u.s.position.VBO) {
763         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.position.streamNo];
764         s->u.s.position.VBO = 0;
765         s->u.s.position.lpData = (BYTE *) ((unsigned long) s->u.s.position.lpData + (unsigned long) vb->resource.allocatedMemory);
766     }
767     if(s->u.s.blendWeights.VBO) {
768         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.blendWeights.streamNo];
769         s->u.s.blendWeights.VBO = 0;
770         s->u.s.blendWeights.lpData = (BYTE *) ((unsigned long) s->u.s.blendWeights.lpData + (unsigned long) vb->resource.allocatedMemory);
771     }
772     if(s->u.s.blendMatrixIndices.VBO) {
773         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.blendMatrixIndices.streamNo];
774         s->u.s.blendMatrixIndices.VBO = 0;
775         s->u.s.blendMatrixIndices.lpData = (BYTE *) ((unsigned long) s->u.s.blendMatrixIndices.lpData + (unsigned long) vb->resource.allocatedMemory);
776     }
777     if(s->u.s.normal.VBO) {
778         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.normal.streamNo];
779         s->u.s.normal.VBO = 0;
780         s->u.s.normal.lpData = (BYTE *) ((unsigned long) s->u.s.normal.lpData + (unsigned long) vb->resource.allocatedMemory);
781     }
782     if(s->u.s.pSize.VBO) {
783         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.pSize.streamNo];
784         s->u.s.pSize.VBO = 0;
785         s->u.s.pSize.lpData = (BYTE *) ((unsigned long) s->u.s.pSize.lpData + (unsigned long) vb->resource.allocatedMemory);
786     }
787     if(s->u.s.diffuse.VBO) {
788         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.diffuse.streamNo];
789         s->u.s.diffuse.VBO = 0;
790         s->u.s.diffuse.lpData = (BYTE *) ((unsigned long) s->u.s.diffuse.lpData + (unsigned long) vb->resource.allocatedMemory);
791     }
792     if(s->u.s.specular.VBO) {
793         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.specular.streamNo];
794         s->u.s.specular.VBO = 0;
795         s->u.s.specular.lpData = (BYTE *) ((unsigned long) s->u.s.specular.lpData + (unsigned long) vb->resource.allocatedMemory);
796     }
797     for(i = 0; i < WINED3DDP_MAXTEXCOORD; i++) {
798         if(s->u.s.texCoords[i].VBO) {
799             vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.texCoords[i].streamNo];
800             s->u.s.texCoords[i].VBO = 0;
801             s->u.s.texCoords[i].lpData = (BYTE *) ((unsigned long) s->u.s.texCoords[i].lpData + (unsigned long) vb->resource.allocatedMemory);
802         }
803     }
804     if(s->u.s.position2.VBO) {
805         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.position2.streamNo];
806         s->u.s.position2.VBO = 0;
807         s->u.s.position2.lpData = (BYTE *) ((unsigned long) s->u.s.position2.lpData + (unsigned long) vb->resource.allocatedMemory);
808     }
809     if(s->u.s.normal2.VBO) {
810         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.normal2.streamNo];
811         s->u.s.normal2.VBO = 0;
812         s->u.s.normal2.lpData = (BYTE *) ((unsigned long) s->u.s.normal2.lpData + (unsigned long) vb->resource.allocatedMemory);
813     }
814     if(s->u.s.tangent.VBO) {
815         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.tangent.streamNo];
816         s->u.s.tangent.VBO = 0;
817         s->u.s.tangent.lpData = (BYTE *) ((unsigned long) s->u.s.tangent.lpData + (unsigned long) vb->resource.allocatedMemory);
818     }
819     if(s->u.s.binormal.VBO) {
820         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.binormal.streamNo];
821         s->u.s.binormal.VBO = 0;
822         s->u.s.binormal.lpData = (BYTE *) ((unsigned long) s->u.s.binormal.lpData + (unsigned long) vb->resource.allocatedMemory);
823     }
824     if(s->u.s.tessFactor.VBO) {
825         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.tessFactor.streamNo];
826         s->u.s.tessFactor.VBO = 0;
827         s->u.s.tessFactor.lpData = (BYTE *) ((unsigned long) s->u.s.tessFactor.lpData + (unsigned long) vb->resource.allocatedMemory);
828     }
829     if(s->u.s.fog.VBO) {
830         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.fog.streamNo];
831         s->u.s.fog.VBO = 0;
832         s->u.s.fog.lpData = (BYTE *) ((unsigned long) s->u.s.fog.lpData + (unsigned long) vb->resource.allocatedMemory);
833     }
834     if(s->u.s.depth.VBO) {
835         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.depth.streamNo];
836         s->u.s.depth.VBO = 0;
837         s->u.s.depth.lpData = (BYTE *) ((unsigned long) s->u.s.depth.lpData + (unsigned long) vb->resource.allocatedMemory);
838     }
839     if(s->u.s.sample.VBO) {
840         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.sample.streamNo];
841         s->u.s.sample.VBO = 0;
842         s->u.s.sample.lpData = (BYTE *) ((unsigned long) s->u.s.sample.lpData + (unsigned long) vb->resource.allocatedMemory);
843     }
844 }
845
846 /* Routine common to the draw primitive and draw indexed primitive routines */
847 void drawPrimitive(IWineD3DDevice *iface,
848                    int PrimitiveType,
849                    long NumPrimitives,
850                    /* for Indexed: */
851                    long  StartVertexIndex,
852                    UINT  numberOfVertices,
853                    long  StartIdx,
854                    short idxSize,
855                    const void *idxData,
856                    int   minIndex) {
857
858     IWineD3DDeviceImpl           *This = (IWineD3DDeviceImpl *)iface;
859     IWineD3DSurfaceImpl          *target;
860     int i;
861
862     if (NumPrimitives == 0) return;
863
864     /* Invalidate the back buffer memory so LockRect will read it the next time */
865     for(i = 0; i < GL_LIMITS(buffers); i++) {
866         target = (IWineD3DSurfaceImpl *) This->render_targets[i];
867         if (target) {
868             IWineD3DSurface_LoadLocation((IWineD3DSurface *) target, SFLAG_INDRAWABLE, NULL);
869             IWineD3DSurface_ModifyLocation((IWineD3DSurface *) target, SFLAG_INDRAWABLE, TRUE);
870         }
871     }
872
873     /* Signals other modules that a drawing is in progress and the stateblock finalized */
874     This->isInDraw = TRUE;
875
876     ActivateContext(This, This->render_targets[0], CTXUSAGE_DRAWPRIM);
877
878     if (This->stencilBufferTarget) {
879         /* Note that this depends on the ActivateContext call above to set
880          * This->render_offscreen properly */
881         DWORD location = This->render_offscreen ? SFLAG_DS_OFFSCREEN : SFLAG_DS_ONSCREEN;
882         surface_load_ds_location(This->stencilBufferTarget, location);
883         surface_modify_ds_location(This->stencilBufferTarget, location);
884     }
885
886     /* Ok, we will be updating the screen from here onwards so grab the lock */
887     ENTER_GL();
888     {
889         GLenum glPrimType;
890         BOOL emulation = FALSE;
891         WineDirect3DVertexStridedData *strided = &This->strided_streams;
892         WineDirect3DVertexStridedData stridedlcl;
893         /* Ok, Work out which primitive is requested and how many vertexes that
894            will be                                                              */
895         UINT calculatedNumberOfindices = primitiveToGl(PrimitiveType, NumPrimitives, &glPrimType);
896         if (numberOfVertices == 0 )
897             numberOfVertices = calculatedNumberOfindices;
898
899         if(!use_vs(This)) {
900             if(!This->strided_streams.u.s.position_transformed && This->activeContext->num_untracked_materials &&
901                 This->stateBlock->renderState[WINED3DRS_LIGHTING]) {
902                 static BOOL first = TRUE;
903                 if(first) {
904                     FIXME("Using software emulation because not all material properties could be tracked\n");
905                     first = FALSE;
906                 } else {
907                     TRACE("Using software emulation because not all material properties could be tracked\n");
908                 }
909                 emulation = TRUE;
910             }
911             else if(This->activeContext->fog_coord && This->stateBlock->renderState[WINED3DRS_FOGENABLE]) {
912                 /* Either write a pipeline replacement shader or convert the specular alpha from unsigned byte
913                  * to a float in the vertex buffer
914                  */
915                 static BOOL first = TRUE;
916                 if(first) {
917                     FIXME("Using software emulation because manual fog coordinates are provided\n");
918                     first = FALSE;
919                 } else {
920                     TRACE("Using software emulation because manual fog coordinates are provided\n");
921                 }
922                 emulation = TRUE;
923             }
924
925             if(emulation) {
926                 strided = &stridedlcl;
927                 memcpy(&stridedlcl, &This->strided_streams, sizeof(stridedlcl));
928                 remove_vbos(This, &stridedlcl);
929             }
930         }
931
932         if (This->useDrawStridedSlow || emulation) {
933             /* Immediate mode drawing */
934             if(use_vs(This)) {
935                 static BOOL first = TRUE;
936                 if(first) {
937                     FIXME("Using immediate mode with vertex shaders for half float emulation\n");
938                     first = FALSE;
939                 } else {
940                     TRACE("Using immediate mode with vertex shaders for half float emulation\n");
941                 }
942                 drawStridedSlowVs(iface, strided, calculatedNumberOfindices, glPrimType,
943                                   idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
944             } else {
945                 drawStridedSlow(iface, strided, calculatedNumberOfindices,
946                                 glPrimType, idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
947             }
948         } else if(This->instancedDraw) {
949             /* Instancing emulation with mixing immediate mode and arrays */
950             drawStridedInstanced(iface, &This->strided_streams, calculatedNumberOfindices, glPrimType,
951                             idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
952         } else {
953             drawStridedFast(iface, calculatedNumberOfindices, glPrimType,
954                             idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
955         }
956     }
957
958     /* Finished updating the screen, restore lock */
959     LEAVE_GL();
960     TRACE("Done all gl drawing\n");
961
962     /* Diagnostics */
963 #ifdef SHOW_FRAME_MAKEUP
964     {
965         static long int primCounter = 0;
966         /* NOTE: set primCounter to the value reported by drawprim 
967            before you want to to write frame makeup to /tmp */
968         if (primCounter >= 0) {
969             WINED3DLOCKED_RECT r;
970             char buffer[80];
971             IWineD3DSurface_LockRect(This->render_targets[0], &r, NULL, WINED3DLOCK_READONLY);
972             sprintf(buffer, "/tmp/backbuffer_%ld.tga", primCounter);
973             TRACE("Saving screenshot %s\n", buffer);
974             IWineD3DSurface_SaveSnapshot(This->render_targets[0], buffer);
975             IWineD3DSurface_UnlockRect(This->render_targets[0]);
976
977 #ifdef SHOW_TEXTURE_MAKEUP
978            {
979             IWineD3DSurface *pSur;
980             int textureNo;
981             for (textureNo = 0; textureNo < MAX_COMBINED_SAMPLERS; ++textureNo) {
982                 if (This->stateBlock->textures[textureNo] != NULL) {
983                     sprintf(buffer, "/tmp/texture_%p_%ld_%d.tga", This->stateBlock->textures[textureNo], primCounter, textureNo);
984                     TRACE("Saving texture %s\n", buffer);
985                     if (IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]) == WINED3DRTYPE_TEXTURE) {
986                             IWineD3DTexture_GetSurfaceLevel((IWineD3DTexture *)This->stateBlock->textures[textureNo], 0, &pSur);
987                             IWineD3DSurface_SaveSnapshot(pSur, buffer);
988                             IWineD3DSurface_Release(pSur);
989                     } else  {
990                         FIXME("base Texture isn't of type texture %d\n", IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]));
991                     }
992                 }
993             }
994            }
995 #endif
996         }
997         TRACE("drawprim #%ld\n", primCounter);
998         ++primCounter;
999     }
1000 #endif
1001
1002     /* Control goes back to the device, stateblock values may change again */
1003     This->isInDraw = FALSE;
1004 }
1005
1006 static void normalize_normal(float *n) {
1007     float length = n[0] * n[0] + n[1] * n[1] + n[2] * n[2];
1008     if(length == 0.0) return;
1009     length = sqrt(length);
1010     n[0] = n[0] / length;
1011     n[1] = n[1] / length;
1012     n[2] = n[2] / length;
1013 }
1014
1015 /* Tesselates a high order rectangular patch into single triangles using gl evaluators
1016  *
1017  * The problem is that OpenGL does not offer a direct way to return the tesselated primitives,
1018  * and they can't be sent off for rendering directly either. Tesselating is slow, so we want
1019  * to cache the patches in a vertex buffer. But more importantly, gl can't bind generated
1020  * attributes to numbered shader attributes, so we have to store them and rebind them as needed
1021  * in drawprim.
1022  *
1023  * To read back, the opengl feedback mode is used. This creates a problem because we want
1024  * untransformed, unlit vertices, but feedback runs everything through transform and lighting.
1025  * Thus disable lighting and set identity matrices to get unmodified colors and positions.
1026  * To overcome clipping find the biggest x, y and z values of the vertices in the patch and scale
1027  * them to [-1.0;+1.0] and set the viewport up to scale them back.
1028  *
1029  * Normals are more tricky: Draw white vertices with 3 directional lights, and calculate the
1030  * resulting colors back to the normals.
1031  *
1032  * NOTE: This function activates a context for blitting, modifies matrices & viewport, but
1033  * does not restore it because normally a draw follows immediately afterwards. The caller is
1034  * responsible of taking care that either the gl states are restored, or the context activated
1035  * for drawing to reset the lastWasBlit flag.
1036  */
1037 HRESULT tesselate_rectpatch(IWineD3DDeviceImpl *This,
1038                             struct WineD3DRectPatch *patch) {
1039     unsigned int i, j, num_quads, out_vertex_size, buffer_size, d3d_out_vertex_size;
1040     float max_x = 0.0, max_y = 0.0, max_z = 0.0, neg_z = 0.0;
1041     WineDirect3DVertexStridedData strided;
1042     BYTE *data;
1043     WINED3DRECTPATCH_INFO *info = &patch->RectPatchInfo;
1044     DWORD vtxStride;
1045     GLenum feedback_type;
1046     GLfloat *feedbuffer;
1047
1048     /* First, locate the position data. This is provided in a vertex buffer in the stateblock.
1049      * Beware of vbos
1050      */
1051     memset(&strided, 0, sizeof(strided));
1052     primitiveDeclarationConvertToStridedData((IWineD3DDevice *) This, FALSE, &strided, NULL);
1053     if(strided.u.s.position.VBO) {
1054         IWineD3DVertexBufferImpl *vb;
1055         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[strided.u.s.position.streamNo];
1056         strided.u.s.position.lpData = (BYTE *) ((unsigned long) strided.u.s.position.lpData +
1057                                                 (unsigned long) vb->resource.allocatedMemory);
1058     }
1059     vtxStride = strided.u.s.position.dwStride;
1060     data = strided.u.s.position.lpData +
1061            vtxStride * info->Stride * info->StartVertexOffsetHeight +
1062            vtxStride * info->StartVertexOffsetWidth;
1063
1064     /* Not entirely sure about what happens with transformed vertices */
1065     if(strided.u.s.position_transformed) {
1066         FIXME("Transformed position in rectpatch generation\n");
1067     }
1068     if(vtxStride % sizeof(GLfloat)) {
1069         /* glMap2f reads vertex sizes in GLfloats, the d3d stride is in bytes.
1070          * I don't see how the stride could not be a multiple of 4, but make sure
1071          * to check it
1072          */
1073         ERR("Vertex stride is not a multiple of sizeof(GLfloat)\n");
1074     }
1075     if(info->Basis != WINED3DBASIS_BEZIER) {
1076         FIXME("Basis is %s, how to handle this?\n", debug_d3dbasis(info->Basis));
1077     }
1078     if(info->Degree != WINED3DDEGREE_CUBIC) {
1079         FIXME("Degree is %s, how to handle this?\n", debug_d3ddegree(info->Degree));
1080     }
1081
1082     /* First, get the boundary cube of the input data */
1083     for(j = 0; j < info->Height; j++) {
1084         for(i = 0; i < info->Width; i++) {
1085             float *v = (float *) (data + vtxStride * i + vtxStride * info->Stride * j);
1086             if(fabs(v[0]) > max_x) max_x = fabs(v[0]);
1087             if(fabs(v[1]) > max_y) max_y = fabs(v[1]);
1088             if(fabs(v[2]) > max_z) max_z = fabs(v[2]);
1089             if(v[2] < neg_z) neg_z = v[2];
1090         }
1091     }
1092
1093     /* This needs some improvements in the vertex decl code */
1094     FIXME("Cannot find data to generate. Only generating position and normals\n");
1095     patch->has_normals = TRUE;
1096     patch->has_texcoords = FALSE;
1097
1098     /* Simply activate the context for blitting. This disables all the things we don't want and
1099      * takes care of dirtifying. Dirtifying is preferred over pushing / popping, since drawing the
1100      * patch (as opposed to normal draws) will most likely need different changes anyway
1101      */
1102     ActivateContext(This, This->lastActiveRenderTarget, CTXUSAGE_BLIT);
1103     ENTER_GL();
1104
1105     glMatrixMode(GL_PROJECTION);
1106     checkGLcall("glMatrixMode(GL_PROJECTION)");
1107     glLoadIdentity();
1108     checkGLcall("glLoadIndentity()");
1109     glScalef(1 / (max_x) , 1 / (max_y), max_z == 0 ? 1 : 1 / ( 2 * max_z));
1110     glTranslatef(0, 0, 0.5);
1111     checkGLcall("glScalef");
1112     glViewport(-max_x, -max_y, 2 * (max_x), 2 * (max_y));
1113     checkGLcall("glViewport");
1114
1115     /* Some states to take care of. If we're in wireframe opengl will produce lines, and confuse
1116      * our feedback buffer parser
1117      */
1118     glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
1119     checkGLcall("glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)");
1120     IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_FILLMODE));
1121     if(patch->has_normals) {
1122         float black[4] = {0, 0, 0, 0};
1123         float red[4]   = {1, 0, 0, 0};
1124         float green[4] = {0, 1, 0, 0};
1125         float blue[4]  = {0, 0, 1, 0};
1126         float white[4] = {1, 1, 1, 1};
1127         glEnable(GL_LIGHTING);
1128         checkGLcall("glEnable(GL_LIGHTING)");
1129         glLightModelfv(GL_LIGHT_MODEL_AMBIENT, black);
1130         checkGLcall("glLightModel for MODEL_AMBIENT");
1131         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_AMBIENT));
1132
1133         for(i = 3; i < GL_LIMITS(lights); i++) {
1134             glDisable(GL_LIGHT0 + i);
1135             checkGLcall("glDisable(GL_LIGHT0 + i)");
1136             IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(i));
1137         }
1138
1139         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(0));
1140         glLightfv(GL_LIGHT0, GL_DIFFUSE, red);
1141         glLightfv(GL_LIGHT0, GL_SPECULAR, black);
1142         glLightfv(GL_LIGHT0, GL_AMBIENT, black);
1143         glLightfv(GL_LIGHT0, GL_POSITION, red);
1144         glEnable(GL_LIGHT0);
1145         checkGLcall("Setting up light 1\n");
1146         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(1));
1147         glLightfv(GL_LIGHT1, GL_DIFFUSE, green);
1148         glLightfv(GL_LIGHT1, GL_SPECULAR, black);
1149         glLightfv(GL_LIGHT1, GL_AMBIENT, black);
1150         glLightfv(GL_LIGHT1, GL_POSITION, green);
1151         glEnable(GL_LIGHT1);
1152         checkGLcall("Setting up light 2\n");
1153         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(2));
1154         glLightfv(GL_LIGHT2, GL_DIFFUSE, blue);
1155         glLightfv(GL_LIGHT2, GL_SPECULAR, black);
1156         glLightfv(GL_LIGHT2, GL_AMBIENT, black);
1157         glLightfv(GL_LIGHT2, GL_POSITION, blue);
1158         glEnable(GL_LIGHT2);
1159         checkGLcall("Setting up light 3\n");
1160
1161         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_MATERIAL);
1162         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_COLORVERTEX));
1163         glDisable(GL_COLOR_MATERIAL);
1164         glMaterialfv(GL_FRONT_AND_BACK, GL_EMISSION, black);
1165         glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR, black);
1166         glMaterialfv(GL_FRONT_AND_BACK, GL_DIFFUSE, white);
1167         checkGLcall("Setting up materials\n");
1168     }
1169
1170     /* Enable the needed maps.
1171      * GL_MAP2_VERTEX_3 is needed for positional data.
1172      * GL_AUTO_NORMAL to generate normals from the position. Do not use GL_MAP2_NORMAL.
1173      * GL_MAP2_TEXTURE_COORD_4 for texture coords
1174      */
1175     num_quads = ceilf(patch->numSegs[0]) * ceilf(patch->numSegs[1]);
1176     out_vertex_size = 3 /* position */;
1177     d3d_out_vertex_size = 3;
1178     glEnable(GL_MAP2_VERTEX_3);
1179     if(patch->has_normals && patch->has_texcoords) {
1180         FIXME("Texcoords not handled yet\n");
1181         feedback_type = GL_3D_COLOR_TEXTURE;
1182         out_vertex_size += 8;
1183         d3d_out_vertex_size += 7;
1184         glEnable(GL_AUTO_NORMAL);
1185         glEnable(GL_MAP2_TEXTURE_COORD_4);
1186     } else if(patch->has_texcoords) {
1187         FIXME("Texcoords not handled yet\n");
1188         feedback_type = GL_3D_COLOR_TEXTURE;
1189         out_vertex_size += 7;
1190         d3d_out_vertex_size += 4;
1191         glEnable(GL_MAP2_TEXTURE_COORD_4);
1192     } else if(patch->has_normals) {
1193         feedback_type = GL_3D_COLOR;
1194         out_vertex_size += 4;
1195         d3d_out_vertex_size += 3;
1196         glEnable(GL_AUTO_NORMAL);
1197     } else {
1198         feedback_type = GL_3D;
1199     }
1200     checkGLcall("glEnable vertex attrib generation");
1201
1202     buffer_size = num_quads * out_vertex_size * 2 /* triangle list */ * 3 /* verts per tri */
1203                    + 4 * num_quads /* 2 triangle markers per quad + num verts in tri */;
1204     feedbuffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, buffer_size * sizeof(float) * 8);
1205
1206     glMap2f(GL_MAP2_VERTEX_3,
1207             0, 1, vtxStride / sizeof(float), info->Width,
1208             0, 1, info->Stride * vtxStride / sizeof(float), info->Height,
1209             (float *) data);
1210     checkGLcall("glMap2f");
1211     if(patch->has_texcoords) {
1212         glMap2f(GL_MAP2_TEXTURE_COORD_4,
1213                 0, 1, vtxStride / sizeof(float), info->Width,
1214                 0, 1, info->Stride * vtxStride / sizeof(float), info->Height,
1215                 (float *) data);
1216         checkGLcall("glMap2f");
1217     }
1218     glMapGrid2f(ceilf(patch->numSegs[0]), 0.0, 1.0, ceilf(patch->numSegs[1]), 0.0, 1.0);
1219     checkGLcall("glMapGrid2f");
1220
1221     glFeedbackBuffer(buffer_size * 2, feedback_type, feedbuffer);
1222     checkGLcall("glFeedbackBuffer");
1223     glRenderMode(GL_FEEDBACK);
1224
1225     glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1226     checkGLcall("glEvalMesh2\n");
1227
1228     i = glRenderMode(GL_RENDER);
1229     if(i == -1) {
1230         LEAVE_GL();
1231         ERR("Feedback failed. Expected %d elements back\n", buffer_size);
1232         Sleep(10000);
1233         HeapFree(GetProcessHeap(), 0, feedbuffer);
1234         return WINED3DERR_DRIVERINTERNALERROR;
1235     } else if(i != buffer_size) {
1236         LEAVE_GL();
1237         ERR("Unexpected amount of elements returned. Expected %d, got %d\n", buffer_size, i);
1238         Sleep(10000);
1239         HeapFree(GetProcessHeap(), 0, feedbuffer);
1240         return WINED3DERR_DRIVERINTERNALERROR;
1241     } else {
1242         TRACE("Got %d elements as expected\n", i);
1243     }
1244
1245     HeapFree(GetProcessHeap(), 0, patch->mem);
1246     patch->mem = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, num_quads * 6 * d3d_out_vertex_size * sizeof(float) * 8);
1247     i = 0;
1248     for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1249         if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1250             ERR("Unexpected token: %f\n", feedbuffer[j]);
1251             continue;
1252         }
1253         if(feedbuffer[j + 1] != 3) {
1254             ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1255             continue;
1256         }
1257         /* Somehow there are different ideas about back / front facing, so fix up the
1258          * vertex order
1259          */
1260         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 2 + 2]; /* x, triangle 2 */
1261         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 2 + 3]; /* y, triangle 2 */
1262         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 2 + 4] - 0.5) * 4 * max_z; /* z, triangle 3 */
1263         if(patch->has_normals) {
1264             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 2 + 5];
1265             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 2 + 6];
1266             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 2 + 7];
1267         }
1268         i += d3d_out_vertex_size;
1269
1270         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 1 + 2]; /* x, triangle 2 */
1271         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 1 + 3]; /* y, triangle 2 */
1272         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 1 + 4] - 0.5) * 4 * max_z; /* z, triangle 2 */
1273         if(patch->has_normals) {
1274             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 1 + 5];
1275             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 1 + 6];
1276             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 1 + 7];
1277         }
1278         i += d3d_out_vertex_size;
1279
1280         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 0 + 2]; /* x, triangle 1 */
1281         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 0 + 3]; /* y, triangle 1 */
1282         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 0 + 4] - 0.5) * 4 * max_z; /* z, triangle 1 */
1283         if(patch->has_normals) {
1284             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 0 + 5];
1285             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 0 + 6];
1286             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 0 + 7];
1287         }
1288         i += d3d_out_vertex_size;
1289     }
1290
1291     if(patch->has_normals) {
1292         /* Now do the same with reverse light directions */
1293         float x[4] = {-1,  0,  0, 0};
1294         float y[4] = { 0, -1,  0, 0};
1295         float z[4] = { 0,  0, -1, 0};
1296         glLightfv(GL_LIGHT0, GL_POSITION, x);
1297         glLightfv(GL_LIGHT1, GL_POSITION, y);
1298         glLightfv(GL_LIGHT2, GL_POSITION, z);
1299         checkGLcall("Setting up reverse light directions\n");
1300
1301         glRenderMode(GL_FEEDBACK);
1302         checkGLcall("glRenderMode(GL_FEEDBACK)");
1303         glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1304         checkGLcall("glEvalMesh2\n");
1305         i = glRenderMode(GL_RENDER);
1306         checkGLcall("glRenderMode(GL_RENDER)");
1307
1308         i = 0;
1309         for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1310             if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1311                 ERR("Unexpected token: %f\n", feedbuffer[j]);
1312                 continue;
1313             }
1314             if(feedbuffer[j + 1] != 3) {
1315                 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1316                 continue;
1317             }
1318             if(patch->mem[i + 3] == 0.0)
1319                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 2 + 5];
1320             if(patch->mem[i + 4] == 0.0)
1321                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 2 + 6];
1322             if(patch->mem[i + 5] == 0.0)
1323                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 2 + 7];
1324             normalize_normal(patch->mem + i + 3);
1325             i += d3d_out_vertex_size;
1326
1327             if(patch->mem[i + 3] == 0.0)
1328                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 1 + 5];
1329             if(patch->mem[i + 4] == 0.0)
1330                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 1 + 6];
1331             if(patch->mem[i + 5] == 0.0)
1332                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 1 + 7];
1333             normalize_normal(patch->mem + i + 3);
1334             i += d3d_out_vertex_size;
1335
1336             if(patch->mem[i + 3] == 0.0)
1337                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 0 + 5];
1338             if(patch->mem[i + 4] == 0.0)
1339                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 0 + 6];
1340             if(patch->mem[i + 5] == 0.0)
1341                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 0 + 7];
1342             normalize_normal(patch->mem + i + 3);
1343             i += d3d_out_vertex_size;
1344         }
1345     }
1346
1347     glDisable(GL_MAP2_VERTEX_3);
1348     glDisable(GL_AUTO_NORMAL);
1349     glDisable(GL_MAP2_NORMAL);
1350     glDisable(GL_MAP2_TEXTURE_COORD_4);
1351     checkGLcall("glDisable vertex attrib generation");
1352     LEAVE_GL();
1353
1354     HeapFree(GetProcessHeap(), 0, feedbuffer);
1355
1356     vtxStride = 3 * sizeof(float);
1357     if(patch->has_normals) {
1358         vtxStride += 3 * sizeof(float);
1359     }
1360     if(patch->has_texcoords) {
1361         vtxStride += 4 * sizeof(float);
1362     }
1363     memset(&patch->strided, 0, sizeof(&patch->strided));
1364     patch->strided.u.s.position.lpData = (BYTE *) patch->mem;
1365     patch->strided.u.s.position.dwStride = vtxStride;
1366     patch->strided.u.s.position.dwType = WINED3DDECLTYPE_FLOAT3;
1367     patch->strided.u.s.position.streamNo = 255;
1368
1369     if(patch->has_normals) {
1370         patch->strided.u.s.normal.lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1371         patch->strided.u.s.normal.dwStride = vtxStride;
1372         patch->strided.u.s.normal.dwType = WINED3DDECLTYPE_FLOAT3;
1373         patch->strided.u.s.normal.streamNo = 255;
1374     }
1375     if(patch->has_texcoords) {
1376         patch->strided.u.s.texCoords[0].lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1377         if(patch->has_normals) {
1378             patch->strided.u.s.texCoords[0].lpData += 3 * sizeof(float);
1379         }
1380         patch->strided.u.s.texCoords[0].dwStride = vtxStride;
1381         patch->strided.u.s.texCoords[0].dwType = WINED3DDECLTYPE_FLOAT4;
1382         /* MAX_STREAMS index points to an unused element in stateblock->streamOffsets which
1383          * always remains set to 0. Windows uses stream 255 here, but this is not visible to the
1384          * application.
1385          */
1386         patch->strided.u.s.texCoords[0].streamNo = MAX_STREAMS;
1387     }
1388
1389     return WINED3D_OK;
1390 }