iccvid: Remove dead stores (llvm/clang).
[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     const BYTE *data = NULL;
146     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
147     IWineD3DVertexDeclarationImpl* vertexDeclaration = (IWineD3DVertexDeclarationImpl *)This->stateBlock->vertexDecl;
148     unsigned int i;
149     const WINED3DVERTEXELEMENT *element;
150     DWORD stride;
151     DWORD numPreloadStreams = This->stateBlock->streamIsUP ? 0 : vertexDeclaration->num_streams;
152     const 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 (%u of %u)\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, const WineDirect3DVertexStridedData *sd, UINT NumVertexes,
290         GLenum glPrimType, const void *idxData, short idxSize, ULONG minIndex, ULONG startIdx, ULONG startVertex)
291 {
292     unsigned int               textureNo    = 0;
293     const WORD                *pIdxBufS     = NULL;
294     const DWORD               *pIdxBufL     = NULL;
295     ULONG                      vx_index;
296     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
297     const UINT *streamOffset = This->stateBlock->streamOffset;
298     long                      SkipnStrides = startVertex + This->stateBlock->loadBaseVertexIndex;
299     BOOL                      pixelShader = use_ps(This);
300
301     const BYTE *texCoords[WINED3DDP_MAXTEXCOORD];
302     const BYTE *diffuse = NULL, *specular = NULL, *normal = NULL, *position = NULL;
303
304     TRACE("Using slow vertex array code\n");
305
306     /* Variable Initialization */
307     if (idxSize != 0) {
308         /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
309          * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
310          * idxData will be != NULL
311          */
312         if(idxData == NULL) {
313             idxData = ((IWineD3DIndexBufferImpl *) This->stateBlock->pIndexData)->resource.allocatedMemory;
314         }
315
316         if (idxSize == 2) pIdxBufS = (const WORD *) idxData;
317         else pIdxBufL = (const DWORD *) idxData;
318     } else if (idxData) {
319         ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
320         return;
321     }
322
323     /* Adding the stream offset once is cheaper than doing it every iteration. Do not modify the strided data, it is a pointer
324      * to the strided Data in the device and might be needed intact on the next draw
325      */
326     for (textureNo = 0; textureNo < GL_LIMITS(texture_stages); ++textureNo) {
327         if(sd->u.s.texCoords[textureNo].lpData) {
328             texCoords[textureNo] = sd->u.s.texCoords[textureNo].lpData + streamOffset[sd->u.s.texCoords[textureNo].streamNo];
329         } else {
330             texCoords[textureNo] = NULL;
331         }
332     }
333     if(sd->u.s.diffuse.lpData) {
334         diffuse = sd->u.s.diffuse.lpData + streamOffset[sd->u.s.diffuse.streamNo];
335     }
336     if(sd->u.s.specular.lpData) {
337         specular = sd->u.s.specular.lpData + streamOffset[sd->u.s.specular.streamNo];
338     }
339     if(sd->u.s.normal.lpData) {
340         normal = sd->u.s.normal.lpData + streamOffset[sd->u.s.normal.streamNo];
341     }
342     if(sd->u.s.position.lpData) {
343         position = sd->u.s.position.lpData + streamOffset[sd->u.s.position.streamNo];
344     }
345
346     if(FIXME_ON(d3d_draw)) {
347         if(specular && This->stateBlock->renderState[WINED3DRS_FOGENABLE] &&
348            (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || sd->u.s.position_transformed )&&
349            This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
350             if(GL_SUPPORT(EXT_FOG_COORD) && sd->u.s.specular.dwType != WINED3DDECLTYPE_D3DCOLOR) {
351                 FIXME("Implement fog coordinates from %s\n", debug_d3ddecltype(sd->u.s.specular.dwType));
352             }
353         }
354         if(This->activeContext->num_untracked_materials && sd->u.s.diffuse.dwType != WINED3DDECLTYPE_D3DCOLOR) {
355             FIXME("Implement diffuse color tracking from %s\n", debug_d3ddecltype(sd->u.s.diffuse.dwType));
356         }
357     }
358
359     /* Start drawing in GL */
360     VTRACE(("glBegin(%x)\n", glPrimType));
361     glBegin(glPrimType);
362
363     /* Default settings for data that is not passed */
364     if (sd->u.s.normal.lpData == NULL) {
365         glNormal3f(0, 0, 0);
366     }
367     if(sd->u.s.diffuse.lpData == NULL) {
368         glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
369     }
370     if(sd->u.s.specular.lpData == NULL) {
371         if (GL_SUPPORT(EXT_SECONDARY_COLOR)) {
372             GL_EXTCALL(glSecondaryColor3fEXT)(0, 0, 0);
373         }
374     }
375
376     /* We shouldn't start this function if any VBO is involved. Should I put a safety check here?
377      * Guess it's not necessary(we crash then anyway) and would only eat CPU time
378      */
379
380     /* For each primitive */
381     for (vx_index = 0; vx_index < NumVertexes; ++vx_index) {
382         /* Blending data and Point sizes are not supported by this function. They are not supported by the fixed
383          * function pipeline at all. A Fixme for them is printed after decoding the vertex declaration
384          */
385
386         /* For indexed data, we need to go a few more strides in */
387         if (idxData != NULL) {
388
389             /* Indexed so work out the number of strides to skip */
390             if (idxSize == 2) {
391                 VTRACE(("Idx for vertex %u = %u\n", vx_index, pIdxBufS[startIdx+vx_index]));
392                 SkipnStrides = pIdxBufS[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
393             } else {
394                 VTRACE(("Idx for vertex %u = %u\n", vx_index, pIdxBufL[startIdx+vx_index]));
395                 SkipnStrides = pIdxBufL[startIdx + vx_index] + This->stateBlock->loadBaseVertexIndex;
396             }
397         }
398
399         /* Texture coords --------------------------- */
400         for (textureNo = 0; textureNo < GL_LIMITS(texture_stages); ++textureNo) {
401
402             if (!GL_SUPPORT(ARB_MULTITEXTURE) && textureNo > 0) {
403                 FIXME("Program using multiple concurrent textures which this opengl implementation doesn't support\n");
404                 continue ;
405             }
406
407             /* Query tex coords */
408             if (This->stateBlock->textures[textureNo] != NULL || pixelShader) {
409                 int    coordIdx = This->stateBlock->textureState[textureNo][WINED3DTSS_TEXCOORDINDEX];
410                 int texture_idx = This->texUnitMap[textureNo];
411                 const void *ptrToCoords;
412
413                 if (coordIdx > 7) {
414                     VTRACE(("tex: %d - Skip tex coords, as being system generated\n", textureNo));
415                     continue;
416                 } else if (coordIdx < 0) {
417                     FIXME("tex: %d - Coord index %d is less than zero, expect a crash.\n", textureNo, coordIdx);
418                     continue;
419                 }
420
421                 if (texture_idx == -1) continue;
422
423                 if (texCoords[coordIdx] == NULL) {
424                     TRACE("tex: %d - Skipping tex coords, as no data supplied\n", textureNo);
425                     if (GL_SUPPORT(ARB_MULTITEXTURE)) {
426                         GL_EXTCALL(glMultiTexCoord4fARB(GL_TEXTURE0_ARB + texture_idx, 0, 0, 0, 1));
427                     } else {
428                         glTexCoord4f(0, 0, 0, 1);
429                     }
430                     continue;
431                 }
432
433                 ptrToCoords = texCoords[coordIdx] + (SkipnStrides * sd->u.s.texCoords[coordIdx].dwStride);
434                 if (GL_SUPPORT(ARB_MULTITEXTURE))
435                     multi_texcoord_funcs[sd->u.s.texCoords[coordIdx].dwType](GL_TEXTURE0_ARB + texture_idx, ptrToCoords);
436                 else
437                     texcoord_funcs[sd->u.s.texCoords[coordIdx].dwType](ptrToCoords);
438             }
439         } /* End of textures */
440
441         /* Diffuse -------------------------------- */
442         if (diffuse) {
443             const void *ptrToCoords = diffuse + SkipnStrides * sd->u.s.diffuse.dwStride;
444
445             diffuse_funcs[sd->u.s.diffuse.dwType](ptrToCoords);
446             if(This->activeContext->num_untracked_materials) {
447                 DWORD diffuseColor = ((const DWORD *)ptrToCoords)[0];
448                 unsigned char i;
449                 float color[4];
450
451                 color[0] = D3DCOLOR_B_R(diffuseColor) / 255.0;
452                 color[1] = D3DCOLOR_B_G(diffuseColor) / 255.0;
453                 color[2] = D3DCOLOR_B_B(diffuseColor) / 255.0;
454                 color[3] = D3DCOLOR_B_A(diffuseColor) / 255.0;
455
456                 for(i = 0; i < This->activeContext->num_untracked_materials; i++) {
457                     glMaterialfv(GL_FRONT_AND_BACK, This->activeContext->untracked_materials[i], color);
458                 }
459             }
460         }
461
462         /* Specular ------------------------------- */
463         if (specular) {
464             const void *ptrToCoords = specular + SkipnStrides * sd->u.s.specular.dwStride;
465
466             /* special case where the fog density is stored in the specular alpha channel */
467             if(This->stateBlock->renderState[WINED3DRS_FOGENABLE] &&
468               (This->stateBlock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || sd->u.s.position.dwType == WINED3DDECLTYPE_FLOAT4 )&&
469               This->stateBlock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
470                 if(GL_SUPPORT(EXT_FOG_COORD)) {
471                     DWORD specularColor = ((const DWORD *)ptrToCoords)[0];
472                     GL_EXTCALL(glFogCoordfEXT(specularColor >> 24));
473                 } else {
474                     static BOOL warned = FALSE;
475                     if(!warned) {
476                         /* TODO: Use the fog table code from old ddraw */
477                         FIXME("Implement fog for transformed vertices in software\n");
478                         warned = TRUE;
479                     }
480                 }
481             }
482
483             specular_funcs[sd->u.s.specular.dwType](ptrToCoords);
484         }
485
486         /* Normal -------------------------------- */
487         if (normal != NULL) {
488             const void *ptrToCoords = normal + SkipnStrides * sd->u.s.normal.dwStride;
489             normal_funcs[sd->u.s.normal.dwType](ptrToCoords);
490         }
491
492         /* Position -------------------------------- */
493         if (position) {
494             const void *ptrToCoords = position + SkipnStrides * sd->u.s.position.dwStride;
495             position_funcs[sd->u.s.position.dwType](ptrToCoords);
496         }
497
498         /* For non indexed mode, step onto next parts */
499         if (idxData == NULL) {
500             ++SkipnStrides;
501         }
502     }
503
504     glEnd();
505     checkGLcall("glEnd and previous calls");
506 }
507
508 static inline void send_attribute(IWineD3DDeviceImpl *This, const DWORD type, const UINT index, const void *ptr) {
509     switch(type) {
510         case WINED3DDECLTYPE_FLOAT1:
511             GL_EXTCALL(glVertexAttrib1fvARB(index, (const float *)ptr));
512             break;
513         case WINED3DDECLTYPE_FLOAT2:
514             GL_EXTCALL(glVertexAttrib2fvARB(index, (const float *)ptr));
515             break;
516         case WINED3DDECLTYPE_FLOAT3:
517             GL_EXTCALL(glVertexAttrib3fvARB(index, (const float *)ptr));
518             break;
519         case WINED3DDECLTYPE_FLOAT4:
520             GL_EXTCALL(glVertexAttrib4fvARB(index, (const float *)ptr));
521             break;
522
523         case WINED3DDECLTYPE_UBYTE4:
524             GL_EXTCALL(glVertexAttrib4ubvARB(index, ptr));
525             break;
526         case WINED3DDECLTYPE_UBYTE4N:
527         case WINED3DDECLTYPE_D3DCOLOR:
528             GL_EXTCALL(glVertexAttrib4NubvARB(index, ptr));
529             break;
530
531         case WINED3DDECLTYPE_SHORT2:
532             GL_EXTCALL(glVertexAttrib4svARB(index, (const GLshort *)ptr));
533             break;
534         case WINED3DDECLTYPE_SHORT4:
535             GL_EXTCALL(glVertexAttrib4svARB(index, (const GLshort *)ptr));
536             break;
537
538         case WINED3DDECLTYPE_SHORT2N:
539         {
540             GLshort s[4] = {((const GLshort *)ptr)[0], ((const GLshort *)ptr)[1], 0, 1};
541             GL_EXTCALL(glVertexAttrib4NsvARB(index, s));
542             break;
543         }
544         case WINED3DDECLTYPE_USHORT2N:
545         {
546             GLushort s[4] = {((const GLushort *)ptr)[0], ((const GLushort *)ptr)[1], 0, 1};
547             GL_EXTCALL(glVertexAttrib4NusvARB(index, s));
548             break;
549         }
550         case WINED3DDECLTYPE_SHORT4N:
551             GL_EXTCALL(glVertexAttrib4NsvARB(index, (const GLshort *)ptr));
552             break;
553         case WINED3DDECLTYPE_USHORT4N:
554             GL_EXTCALL(glVertexAttrib4NusvARB(index, (const GLushort *)ptr));
555             break;
556
557         case WINED3DDECLTYPE_UDEC3:
558             FIXME("Unsure about WINED3DDECLTYPE_UDEC3\n");
559             /*glVertexAttrib3usvARB(instancedData[j], (GLushort *) ptr); Does not exist */
560             break;
561         case WINED3DDECLTYPE_DEC3N:
562             FIXME("Unsure about WINED3DDECLTYPE_DEC3N\n");
563             /*glVertexAttrib3NusvARB(instancedData[j], (GLushort *) ptr); Does not exist */
564             break;
565
566         case WINED3DDECLTYPE_FLOAT16_2:
567             /* Are those 16 bit floats. C doesn't have a 16 bit float type. I could read the single bits and calculate a 4
568              * byte float according to the IEEE standard
569              */
570             if (GL_SUPPORT(NV_HALF_FLOAT)) {
571                 GL_EXTCALL(glVertexAttrib2hvNV(index, (const GLhalfNV *)ptr));
572             } else {
573                 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
574                 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
575                 GL_EXTCALL(glVertexAttrib2fARB(index, x, y));
576             }
577             break;
578         case WINED3DDECLTYPE_FLOAT16_4:
579             if (GL_SUPPORT(NV_HALF_FLOAT)) {
580                 GL_EXTCALL(glVertexAttrib4hvNV(index, (const GLhalfNV *)ptr));
581             } else {
582                 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
583                 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
584                 float z = float_16_to_32(((const unsigned short *)ptr) + 2);
585                 float w = float_16_to_32(((const unsigned short *)ptr) + 3);
586                 GL_EXTCALL(glVertexAttrib4fARB(index, x, y, z, w));
587             }
588             break;
589
590         case WINED3DDECLTYPE_UNUSED:
591         default:
592             ERR("Unexpected attribute declaration: %d\n", type);
593             break;
594     }
595 }
596
597 static void drawStridedSlowVs(IWineD3DDevice *iface, const WineDirect3DVertexStridedData *sd, UINT numberOfVertices,
598         GLenum glPrimitiveType, const void *idxData, short idxSize, ULONG minIndex, ULONG startIdx, ULONG startVertex)
599 {
600     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
601     long                      SkipnStrides = startVertex + This->stateBlock->loadBaseVertexIndex;
602     const WORD                *pIdxBufS     = NULL;
603     const DWORD               *pIdxBufL     = NULL;
604     ULONG                      vx_index;
605     int i;
606     IWineD3DStateBlockImpl *stateblock = This->stateBlock;
607     const BYTE *ptr;
608
609     if (idxSize != 0) {
610         /* Immediate mode drawing can't make use of indices in a vbo - get the data from the index buffer.
611          * If the index buffer has no vbo(not supported or other reason), or with user pointer drawing
612          * idxData will be != NULL
613          */
614         if(idxData == NULL) {
615             idxData = ((IWineD3DIndexBufferImpl *) stateblock->pIndexData)->resource.allocatedMemory;
616         }
617
618         if (idxSize == 2) pIdxBufS = (const WORD *) idxData;
619         else pIdxBufL = (const DWORD *) idxData;
620     } else if (idxData) {
621         ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
622         return;
623     }
624
625     /* Start drawing in GL */
626     VTRACE(("glBegin(%x)\n", glPrimitiveType));
627     glBegin(glPrimitiveType);
628
629     for (vx_index = 0; vx_index < numberOfVertices; ++vx_index) {
630         if (idxData != NULL) {
631
632             /* Indexed so work out the number of strides to skip */
633             if (idxSize == 2) {
634                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufS[startIdx+vx_index]));
635                 SkipnStrides = pIdxBufS[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
636             } else {
637                 VTRACE(("Idx for vertex %d = %d\n", vx_index, pIdxBufL[startIdx+vx_index]));
638                 SkipnStrides = pIdxBufL[startIdx + vx_index] + stateblock->loadBaseVertexIndex;
639             }
640         }
641
642         for(i = MAX_ATTRIBS - 1; i >= 0; i--) {
643             if(!sd->u.input[i].lpData) continue;
644
645             ptr = sd->u.input[i].lpData +
646                   sd->u.input[i].dwStride * SkipnStrides +
647                   stateblock->streamOffset[sd->u.input[i].streamNo];
648
649             send_attribute(This, sd->u.input[i].dwType, i, ptr);
650         }
651         SkipnStrides++;
652     }
653
654     glEnd();
655 }
656
657 static inline void drawStridedInstanced(IWineD3DDevice *iface, const WineDirect3DVertexStridedData *sd,
658         UINT numberOfVertices, GLenum glPrimitiveType, const void *idxData, short idxSize, ULONG minIndex,
659         ULONG startIdx, ULONG startVertex)
660 {
661     UINT numInstances = 0, i;
662     int numInstancedAttribs = 0, j;
663     UINT instancedData[sizeof(sd->u.input) / sizeof(sd->u.input[0]) /* 16 */];
664     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
665     IWineD3DStateBlockImpl *stateblock = This->stateBlock;
666
667     if (idxSize == 0) {
668         /* This is a nasty thing. MSDN says no hardware supports that and apps have to use software vertex processing.
669          * We don't support this for now
670          *
671          * Shouldn't be too hard to support with opengl, in theory just call glDrawArrays instead of drawElements.
672          * But the StreamSourceFreq value has a different meaning in that situation.
673          */
674         FIXME("Non-indexed instanced drawing is not supported\n");
675         return;
676     }
677
678     TRACE("(%p) : glElements(%x, %d, %d, ...)\n", This, glPrimitiveType, numberOfVertices, minIndex);
679     idxData = idxData == (void *)-1 ? NULL : idxData;
680
681     /* First, figure out how many instances we have to draw */
682     for(i = 0; i < MAX_STREAMS; i++) {
683         /* Look at the streams and take the first one which matches */
684         if(((stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INSTANCEDATA) || (stateblock->streamFlags[i] & WINED3DSTREAMSOURCE_INDEXEDDATA)) && stateblock->streamSource[i]) {
685             /* D3D9 could set streamFreq 0 with (INSTANCEDATA or INDEXEDDATA) and then it is handled as 1. See d3d9/tests/visual.c-> stream_test() */
686             if(stateblock->streamFreq[i] == 0){
687                 numInstances = 1;
688             } else {
689                 numInstances = stateblock->streamFreq[i]; /* use the specified number of instances from the first matched stream. See d3d9/tests/visual.c-> stream_test() */
690             }
691             break; /* break, because only the first suitable value is interesting */
692         }
693     }
694
695     for(i = 0; i < sizeof(sd->u.input) / sizeof(sd->u.input[0]); i++) {
696         if(stateblock->streamFlags[sd->u.input[i].streamNo] & WINED3DSTREAMSOURCE_INSTANCEDATA) {
697             instancedData[numInstancedAttribs] = i;
698             numInstancedAttribs++;
699         }
700     }
701
702     /* now draw numInstances instances :-) */
703     for(i = 0; i < numInstances; i++) {
704         /* Specify the instanced attributes using immediate mode calls */
705         for(j = 0; j < numInstancedAttribs; j++) {
706             const BYTE *ptr = sd->u.input[instancedData[j]].lpData +
707                         sd->u.input[instancedData[j]].dwStride * i +
708                         stateblock->streamOffset[sd->u.input[instancedData[j]].streamNo];
709             if(sd->u.input[instancedData[j]].VBO) {
710                 IWineD3DVertexBufferImpl *vb = (IWineD3DVertexBufferImpl *) stateblock->streamSource[sd->u.input[instancedData[j]].streamNo];
711                 ptr += (long) vb->resource.allocatedMemory;
712             }
713
714             send_attribute(This, sd->u.input[instancedData[j]].dwType, instancedData[j], ptr);
715         }
716
717         glDrawElements(glPrimitiveType, numberOfVertices, idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT,
718                     (const char *)idxData+(idxSize * startIdx));
719         checkGLcall("glDrawElements");
720     }
721 }
722
723 static inline void remove_vbos(IWineD3DDeviceImpl *This, WineDirect3DVertexStridedData *s) {
724     unsigned char i;
725     IWineD3DVertexBufferImpl *vb;
726
727     if(s->u.s.position.VBO) {
728         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.position.streamNo];
729         s->u.s.position.VBO = 0;
730         s->u.s.position.lpData = (BYTE *) ((unsigned long) s->u.s.position.lpData + (unsigned long) vb->resource.allocatedMemory);
731     }
732     if(s->u.s.blendWeights.VBO) {
733         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.blendWeights.streamNo];
734         s->u.s.blendWeights.VBO = 0;
735         s->u.s.blendWeights.lpData = (BYTE *) ((unsigned long) s->u.s.blendWeights.lpData + (unsigned long) vb->resource.allocatedMemory);
736     }
737     if(s->u.s.blendMatrixIndices.VBO) {
738         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.blendMatrixIndices.streamNo];
739         s->u.s.blendMatrixIndices.VBO = 0;
740         s->u.s.blendMatrixIndices.lpData = (BYTE *) ((unsigned long) s->u.s.blendMatrixIndices.lpData + (unsigned long) vb->resource.allocatedMemory);
741     }
742     if(s->u.s.normal.VBO) {
743         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.normal.streamNo];
744         s->u.s.normal.VBO = 0;
745         s->u.s.normal.lpData = (BYTE *) ((unsigned long) s->u.s.normal.lpData + (unsigned long) vb->resource.allocatedMemory);
746     }
747     if(s->u.s.pSize.VBO) {
748         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.pSize.streamNo];
749         s->u.s.pSize.VBO = 0;
750         s->u.s.pSize.lpData = (BYTE *) ((unsigned long) s->u.s.pSize.lpData + (unsigned long) vb->resource.allocatedMemory);
751     }
752     if(s->u.s.diffuse.VBO) {
753         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.diffuse.streamNo];
754         s->u.s.diffuse.VBO = 0;
755         s->u.s.diffuse.lpData = (BYTE *) ((unsigned long) s->u.s.diffuse.lpData + (unsigned long) vb->resource.allocatedMemory);
756     }
757     if(s->u.s.specular.VBO) {
758         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.specular.streamNo];
759         s->u.s.specular.VBO = 0;
760         s->u.s.specular.lpData = (BYTE *) ((unsigned long) s->u.s.specular.lpData + (unsigned long) vb->resource.allocatedMemory);
761     }
762     for(i = 0; i < WINED3DDP_MAXTEXCOORD; i++) {
763         if(s->u.s.texCoords[i].VBO) {
764             vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.texCoords[i].streamNo];
765             s->u.s.texCoords[i].VBO = 0;
766             s->u.s.texCoords[i].lpData = (BYTE *) ((unsigned long) s->u.s.texCoords[i].lpData + (unsigned long) vb->resource.allocatedMemory);
767         }
768     }
769     if(s->u.s.position2.VBO) {
770         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.position2.streamNo];
771         s->u.s.position2.VBO = 0;
772         s->u.s.position2.lpData = (BYTE *) ((unsigned long) s->u.s.position2.lpData + (unsigned long) vb->resource.allocatedMemory);
773     }
774     if(s->u.s.normal2.VBO) {
775         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.normal2.streamNo];
776         s->u.s.normal2.VBO = 0;
777         s->u.s.normal2.lpData = (BYTE *) ((unsigned long) s->u.s.normal2.lpData + (unsigned long) vb->resource.allocatedMemory);
778     }
779     if(s->u.s.tangent.VBO) {
780         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.tangent.streamNo];
781         s->u.s.tangent.VBO = 0;
782         s->u.s.tangent.lpData = (BYTE *) ((unsigned long) s->u.s.tangent.lpData + (unsigned long) vb->resource.allocatedMemory);
783     }
784     if(s->u.s.binormal.VBO) {
785         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.binormal.streamNo];
786         s->u.s.binormal.VBO = 0;
787         s->u.s.binormal.lpData = (BYTE *) ((unsigned long) s->u.s.binormal.lpData + (unsigned long) vb->resource.allocatedMemory);
788     }
789     if(s->u.s.tessFactor.VBO) {
790         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.tessFactor.streamNo];
791         s->u.s.tessFactor.VBO = 0;
792         s->u.s.tessFactor.lpData = (BYTE *) ((unsigned long) s->u.s.tessFactor.lpData + (unsigned long) vb->resource.allocatedMemory);
793     }
794     if(s->u.s.fog.VBO) {
795         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.fog.streamNo];
796         s->u.s.fog.VBO = 0;
797         s->u.s.fog.lpData = (BYTE *) ((unsigned long) s->u.s.fog.lpData + (unsigned long) vb->resource.allocatedMemory);
798     }
799     if(s->u.s.depth.VBO) {
800         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.depth.streamNo];
801         s->u.s.depth.VBO = 0;
802         s->u.s.depth.lpData = (BYTE *) ((unsigned long) s->u.s.depth.lpData + (unsigned long) vb->resource.allocatedMemory);
803     }
804     if(s->u.s.sample.VBO) {
805         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[s->u.s.sample.streamNo];
806         s->u.s.sample.VBO = 0;
807         s->u.s.sample.lpData = (BYTE *) ((unsigned long) s->u.s.sample.lpData + (unsigned long) vb->resource.allocatedMemory);
808     }
809 }
810
811 /* Routine common to the draw primitive and draw indexed primitive routines */
812 void drawPrimitive(IWineD3DDevice *iface,
813                    int PrimitiveType,
814                    long NumPrimitives,
815                    /* for Indexed: */
816                    long  StartVertexIndex,
817                    UINT  numberOfVertices,
818                    long  StartIdx,
819                    short idxSize,
820                    const void *idxData,
821                    int   minIndex) {
822
823     IWineD3DDeviceImpl           *This = (IWineD3DDeviceImpl *)iface;
824     IWineD3DSurfaceImpl          *target;
825     unsigned int i;
826
827     if (NumPrimitives == 0) return;
828
829     /* Invalidate the back buffer memory so LockRect will read it the next time */
830     for(i = 0; i < GL_LIMITS(buffers); i++) {
831         target = (IWineD3DSurfaceImpl *) This->render_targets[i];
832         if (target) {
833             IWineD3DSurface_LoadLocation((IWineD3DSurface *) target, SFLAG_INDRAWABLE, NULL);
834             IWineD3DSurface_ModifyLocation((IWineD3DSurface *) target, SFLAG_INDRAWABLE, TRUE);
835         }
836     }
837
838     /* Signals other modules that a drawing is in progress and the stateblock finalized */
839     This->isInDraw = TRUE;
840
841     ActivateContext(This, This->render_targets[0], CTXUSAGE_DRAWPRIM);
842
843     if (This->stencilBufferTarget) {
844         /* Note that this depends on the ActivateContext call above to set
845          * This->render_offscreen properly */
846         DWORD location = This->render_offscreen ? SFLAG_DS_OFFSCREEN : SFLAG_DS_ONSCREEN;
847         surface_load_ds_location(This->stencilBufferTarget, location);
848         surface_modify_ds_location(This->stencilBufferTarget, location);
849     }
850
851     /* Ok, we will be updating the screen from here onwards so grab the lock */
852     ENTER_GL();
853     {
854         GLenum glPrimType;
855         BOOL emulation = FALSE;
856         const WineDirect3DVertexStridedData *strided = &This->strided_streams;
857         WineDirect3DVertexStridedData stridedlcl;
858         /* Ok, Work out which primitive is requested and how many vertexes that
859            will be                                                              */
860         UINT calculatedNumberOfindices = primitiveToGl(PrimitiveType, NumPrimitives, &glPrimType);
861         if (numberOfVertices == 0 )
862             numberOfVertices = calculatedNumberOfindices;
863
864         if(!use_vs(This)) {
865             if(!This->strided_streams.u.s.position_transformed && This->activeContext->num_untracked_materials &&
866                 This->stateBlock->renderState[WINED3DRS_LIGHTING]) {
867                 static BOOL warned;
868                 if (!warned) {
869                     FIXME("Using software emulation because not all material properties could be tracked\n");
870                     warned = TRUE;
871                 } else {
872                     TRACE("Using software emulation because not all material properties could be tracked\n");
873                 }
874                 emulation = TRUE;
875             }
876             else if(This->activeContext->fog_coord && This->stateBlock->renderState[WINED3DRS_FOGENABLE]) {
877                 /* Either write a pipeline replacement shader or convert the specular alpha from unsigned byte
878                  * to a float in the vertex buffer
879                  */
880                 static BOOL warned;
881                 if (!warned) {
882                     FIXME("Using software emulation because manual fog coordinates are provided\n");
883                     warned = TRUE;
884                 } else {
885                     TRACE("Using software emulation because manual fog coordinates are provided\n");
886                 }
887                 emulation = TRUE;
888             }
889
890             if(emulation) {
891                 strided = &stridedlcl;
892                 memcpy(&stridedlcl, &This->strided_streams, sizeof(stridedlcl));
893                 remove_vbos(This, &stridedlcl);
894             }
895         }
896
897         if (This->useDrawStridedSlow || emulation) {
898             /* Immediate mode drawing */
899             if(use_vs(This)) {
900                 static BOOL warned;
901                 if (!warned) {
902                     FIXME("Using immediate mode with vertex shaders for half float emulation\n");
903                     warned = TRUE;
904                 } else {
905                     TRACE("Using immediate mode with vertex shaders for half float emulation\n");
906                 }
907                 drawStridedSlowVs(iface, strided, calculatedNumberOfindices, glPrimType,
908                                   idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
909             } else {
910                 drawStridedSlow(iface, strided, calculatedNumberOfindices,
911                                 glPrimType, idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
912             }
913         } else if(This->instancedDraw) {
914             /* Instancing emulation with mixing immediate mode and arrays */
915             drawStridedInstanced(iface, &This->strided_streams, calculatedNumberOfindices, glPrimType,
916                             idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
917         } else {
918             drawStridedFast(iface, calculatedNumberOfindices, glPrimType,
919                             idxData, idxSize, minIndex, StartIdx, StartVertexIndex);
920         }
921     }
922
923     /* Finished updating the screen, restore lock */
924     LEAVE_GL();
925     TRACE("Done all gl drawing\n");
926
927     /* Diagnostics */
928 #ifdef SHOW_FRAME_MAKEUP
929     {
930         static long int primCounter = 0;
931         /* NOTE: set primCounter to the value reported by drawprim 
932            before you want to to write frame makeup to /tmp */
933         if (primCounter >= 0) {
934             WINED3DLOCKED_RECT r;
935             char buffer[80];
936             IWineD3DSurface_LockRect(This->render_targets[0], &r, NULL, WINED3DLOCK_READONLY);
937             sprintf(buffer, "/tmp/backbuffer_%ld.tga", primCounter);
938             TRACE("Saving screenshot %s\n", buffer);
939             IWineD3DSurface_SaveSnapshot(This->render_targets[0], buffer);
940             IWineD3DSurface_UnlockRect(This->render_targets[0]);
941
942 #ifdef SHOW_TEXTURE_MAKEUP
943            {
944             IWineD3DSurface *pSur;
945             int textureNo;
946             for (textureNo = 0; textureNo < MAX_COMBINED_SAMPLERS; ++textureNo) {
947                 if (This->stateBlock->textures[textureNo] != NULL) {
948                     sprintf(buffer, "/tmp/texture_%p_%ld_%d.tga", This->stateBlock->textures[textureNo], primCounter, textureNo);
949                     TRACE("Saving texture %s\n", buffer);
950                     if (IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]) == WINED3DRTYPE_TEXTURE) {
951                             IWineD3DTexture_GetSurfaceLevel((IWineD3DTexture *)This->stateBlock->textures[textureNo], 0, &pSur);
952                             IWineD3DSurface_SaveSnapshot(pSur, buffer);
953                             IWineD3DSurface_Release(pSur);
954                     } else  {
955                         FIXME("base Texture isn't of type texture %d\n", IWineD3DBaseTexture_GetType(This->stateBlock->textures[textureNo]));
956                     }
957                 }
958             }
959            }
960 #endif
961         }
962         TRACE("drawprim #%ld\n", primCounter);
963         ++primCounter;
964     }
965 #endif
966
967     /* Control goes back to the device, stateblock values may change again */
968     This->isInDraw = FALSE;
969 }
970
971 static void normalize_normal(float *n) {
972     float length = n[0] * n[0] + n[1] * n[1] + n[2] * n[2];
973     if(length == 0.0) return;
974     length = sqrt(length);
975     n[0] = n[0] / length;
976     n[1] = n[1] / length;
977     n[2] = n[2] / length;
978 }
979
980 /* Tesselates a high order rectangular patch into single triangles using gl evaluators
981  *
982  * The problem is that OpenGL does not offer a direct way to return the tesselated primitives,
983  * and they can't be sent off for rendering directly either. Tesselating is slow, so we want
984  * to cache the patches in a vertex buffer. But more importantly, gl can't bind generated
985  * attributes to numbered shader attributes, so we have to store them and rebind them as needed
986  * in drawprim.
987  *
988  * To read back, the opengl feedback mode is used. This creates a problem because we want
989  * untransformed, unlit vertices, but feedback runs everything through transform and lighting.
990  * Thus disable lighting and set identity matrices to get unmodified colors and positions.
991  * To overcome clipping find the biggest x, y and z values of the vertices in the patch and scale
992  * them to [-1.0;+1.0] and set the viewport up to scale them back.
993  *
994  * Normals are more tricky: Draw white vertices with 3 directional lights, and calculate the
995  * resulting colors back to the normals.
996  *
997  * NOTE: This function activates a context for blitting, modifies matrices & viewport, but
998  * does not restore it because normally a draw follows immediately afterwards. The caller is
999  * responsible of taking care that either the gl states are restored, or the context activated
1000  * for drawing to reset the lastWasBlit flag.
1001  */
1002 HRESULT tesselate_rectpatch(IWineD3DDeviceImpl *This,
1003                             struct WineD3DRectPatch *patch) {
1004     unsigned int i, j, num_quads, out_vertex_size, buffer_size, d3d_out_vertex_size;
1005     float max_x = 0.0, max_y = 0.0, max_z = 0.0, neg_z = 0.0;
1006     WineDirect3DVertexStridedData strided;
1007     const BYTE *data;
1008     const WINED3DRECTPATCH_INFO *info = &patch->RectPatchInfo;
1009     DWORD vtxStride;
1010     GLenum feedback_type;
1011     GLfloat *feedbuffer;
1012
1013     /* First, locate the position data. This is provided in a vertex buffer in the stateblock.
1014      * Beware of vbos
1015      */
1016     memset(&strided, 0, sizeof(strided));
1017     primitiveDeclarationConvertToStridedData((IWineD3DDevice *) This, FALSE, &strided, NULL);
1018     if(strided.u.s.position.VBO) {
1019         IWineD3DVertexBufferImpl *vb;
1020         vb = (IWineD3DVertexBufferImpl *) This->stateBlock->streamSource[strided.u.s.position.streamNo];
1021         strided.u.s.position.lpData = (BYTE *) ((unsigned long) strided.u.s.position.lpData +
1022                                                 (unsigned long) vb->resource.allocatedMemory);
1023     }
1024     vtxStride = strided.u.s.position.dwStride;
1025     data = strided.u.s.position.lpData +
1026            vtxStride * info->Stride * info->StartVertexOffsetHeight +
1027            vtxStride * info->StartVertexOffsetWidth;
1028
1029     /* Not entirely sure about what happens with transformed vertices */
1030     if(strided.u.s.position_transformed) {
1031         FIXME("Transformed position in rectpatch generation\n");
1032     }
1033     if(vtxStride % sizeof(GLfloat)) {
1034         /* glMap2f reads vertex sizes in GLfloats, the d3d stride is in bytes.
1035          * I don't see how the stride could not be a multiple of 4, but make sure
1036          * to check it
1037          */
1038         ERR("Vertex stride is not a multiple of sizeof(GLfloat)\n");
1039     }
1040     if(info->Basis != WINED3DBASIS_BEZIER) {
1041         FIXME("Basis is %s, how to handle this?\n", debug_d3dbasis(info->Basis));
1042     }
1043     if(info->Degree != WINED3DDEGREE_CUBIC) {
1044         FIXME("Degree is %s, how to handle this?\n", debug_d3ddegree(info->Degree));
1045     }
1046
1047     /* First, get the boundary cube of the input data */
1048     for(j = 0; j < info->Height; j++) {
1049         for(i = 0; i < info->Width; i++) {
1050             const float *v = (const float *)(data + vtxStride * i + vtxStride * info->Stride * j);
1051             if(fabs(v[0]) > max_x) max_x = fabs(v[0]);
1052             if(fabs(v[1]) > max_y) max_y = fabs(v[1]);
1053             if(fabs(v[2]) > max_z) max_z = fabs(v[2]);
1054             if(v[2] < neg_z) neg_z = v[2];
1055         }
1056     }
1057
1058     /* This needs some improvements in the vertex decl code */
1059     FIXME("Cannot find data to generate. Only generating position and normals\n");
1060     patch->has_normals = TRUE;
1061     patch->has_texcoords = FALSE;
1062
1063     /* Simply activate the context for blitting. This disables all the things we don't want and
1064      * takes care of dirtifying. Dirtifying is preferred over pushing / popping, since drawing the
1065      * patch (as opposed to normal draws) will most likely need different changes anyway
1066      */
1067     ActivateContext(This, This->lastActiveRenderTarget, CTXUSAGE_BLIT);
1068     ENTER_GL();
1069
1070     glMatrixMode(GL_PROJECTION);
1071     checkGLcall("glMatrixMode(GL_PROJECTION)");
1072     glLoadIdentity();
1073     checkGLcall("glLoadIndentity()");
1074     glScalef(1 / (max_x) , 1 / (max_y), max_z == 0 ? 1 : 1 / ( 2 * max_z));
1075     glTranslatef(0, 0, 0.5);
1076     checkGLcall("glScalef");
1077     glViewport(-max_x, -max_y, 2 * (max_x), 2 * (max_y));
1078     checkGLcall("glViewport");
1079
1080     /* Some states to take care of. If we're in wireframe opengl will produce lines, and confuse
1081      * our feedback buffer parser
1082      */
1083     glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
1084     checkGLcall("glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)");
1085     IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_FILLMODE));
1086     if(patch->has_normals) {
1087         const GLfloat black[4] = {0, 0, 0, 0};
1088         const GLfloat red[4]   = {1, 0, 0, 0};
1089         const GLfloat green[4] = {0, 1, 0, 0};
1090         const GLfloat blue[4]  = {0, 0, 1, 0};
1091         const GLfloat white[4] = {1, 1, 1, 1};
1092         glEnable(GL_LIGHTING);
1093         checkGLcall("glEnable(GL_LIGHTING)");
1094         glLightModelfv(GL_LIGHT_MODEL_AMBIENT, black);
1095         checkGLcall("glLightModel for MODEL_AMBIENT");
1096         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_AMBIENT));
1097
1098         for(i = 3; i < GL_LIMITS(lights); i++) {
1099             glDisable(GL_LIGHT0 + i);
1100             checkGLcall("glDisable(GL_LIGHT0 + i)");
1101             IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(i));
1102         }
1103
1104         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(0));
1105         glLightfv(GL_LIGHT0, GL_DIFFUSE, red);
1106         glLightfv(GL_LIGHT0, GL_SPECULAR, black);
1107         glLightfv(GL_LIGHT0, GL_AMBIENT, black);
1108         glLightfv(GL_LIGHT0, GL_POSITION, red);
1109         glEnable(GL_LIGHT0);
1110         checkGLcall("Setting up light 1\n");
1111         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(1));
1112         glLightfv(GL_LIGHT1, GL_DIFFUSE, green);
1113         glLightfv(GL_LIGHT1, GL_SPECULAR, black);
1114         glLightfv(GL_LIGHT1, GL_AMBIENT, black);
1115         glLightfv(GL_LIGHT1, GL_POSITION, green);
1116         glEnable(GL_LIGHT1);
1117         checkGLcall("Setting up light 2\n");
1118         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_ACTIVELIGHT(2));
1119         glLightfv(GL_LIGHT2, GL_DIFFUSE, blue);
1120         glLightfv(GL_LIGHT2, GL_SPECULAR, black);
1121         glLightfv(GL_LIGHT2, GL_AMBIENT, black);
1122         glLightfv(GL_LIGHT2, GL_POSITION, blue);
1123         glEnable(GL_LIGHT2);
1124         checkGLcall("Setting up light 3\n");
1125
1126         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_MATERIAL);
1127         IWineD3DDeviceImpl_MarkStateDirty(This, STATE_RENDER(WINED3DRS_COLORVERTEX));
1128         glDisable(GL_COLOR_MATERIAL);
1129         glMaterialfv(GL_FRONT_AND_BACK, GL_EMISSION, black);
1130         glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR, black);
1131         glMaterialfv(GL_FRONT_AND_BACK, GL_DIFFUSE, white);
1132         checkGLcall("Setting up materials\n");
1133     }
1134
1135     /* Enable the needed maps.
1136      * GL_MAP2_VERTEX_3 is needed for positional data.
1137      * GL_AUTO_NORMAL to generate normals from the position. Do not use GL_MAP2_NORMAL.
1138      * GL_MAP2_TEXTURE_COORD_4 for texture coords
1139      */
1140     num_quads = ceilf(patch->numSegs[0]) * ceilf(patch->numSegs[1]);
1141     out_vertex_size = 3 /* position */;
1142     d3d_out_vertex_size = 3;
1143     glEnable(GL_MAP2_VERTEX_3);
1144     if(patch->has_normals && patch->has_texcoords) {
1145         FIXME("Texcoords not handled yet\n");
1146         feedback_type = GL_3D_COLOR_TEXTURE;
1147         out_vertex_size += 8;
1148         d3d_out_vertex_size += 7;
1149         glEnable(GL_AUTO_NORMAL);
1150         glEnable(GL_MAP2_TEXTURE_COORD_4);
1151     } else if(patch->has_texcoords) {
1152         FIXME("Texcoords not handled yet\n");
1153         feedback_type = GL_3D_COLOR_TEXTURE;
1154         out_vertex_size += 7;
1155         d3d_out_vertex_size += 4;
1156         glEnable(GL_MAP2_TEXTURE_COORD_4);
1157     } else if(patch->has_normals) {
1158         feedback_type = GL_3D_COLOR;
1159         out_vertex_size += 4;
1160         d3d_out_vertex_size += 3;
1161         glEnable(GL_AUTO_NORMAL);
1162     } else {
1163         feedback_type = GL_3D;
1164     }
1165     checkGLcall("glEnable vertex attrib generation");
1166
1167     buffer_size = num_quads * out_vertex_size * 2 /* triangle list */ * 3 /* verts per tri */
1168                    + 4 * num_quads /* 2 triangle markers per quad + num verts in tri */;
1169     feedbuffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, buffer_size * sizeof(float) * 8);
1170
1171     glMap2f(GL_MAP2_VERTEX_3,
1172             0, 1, vtxStride / sizeof(float), info->Width,
1173             0, 1, info->Stride * vtxStride / sizeof(float), info->Height,
1174             (const GLfloat *)data);
1175     checkGLcall("glMap2f");
1176     if(patch->has_texcoords) {
1177         glMap2f(GL_MAP2_TEXTURE_COORD_4,
1178                 0, 1, vtxStride / sizeof(float), info->Width,
1179                 0, 1, info->Stride * vtxStride / sizeof(float), info->Height,
1180                 (const GLfloat *)data);
1181         checkGLcall("glMap2f");
1182     }
1183     glMapGrid2f(ceilf(patch->numSegs[0]), 0.0, 1.0, ceilf(patch->numSegs[1]), 0.0, 1.0);
1184     checkGLcall("glMapGrid2f");
1185
1186     glFeedbackBuffer(buffer_size * 2, feedback_type, feedbuffer);
1187     checkGLcall("glFeedbackBuffer");
1188     glRenderMode(GL_FEEDBACK);
1189
1190     glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1191     checkGLcall("glEvalMesh2\n");
1192
1193     i = glRenderMode(GL_RENDER);
1194     if(i == -1) {
1195         LEAVE_GL();
1196         ERR("Feedback failed. Expected %d elements back\n", buffer_size);
1197         Sleep(10000);
1198         HeapFree(GetProcessHeap(), 0, feedbuffer);
1199         return WINED3DERR_DRIVERINTERNALERROR;
1200     } else if(i != buffer_size) {
1201         LEAVE_GL();
1202         ERR("Unexpected amount of elements returned. Expected %d, got %d\n", buffer_size, i);
1203         Sleep(10000);
1204         HeapFree(GetProcessHeap(), 0, feedbuffer);
1205         return WINED3DERR_DRIVERINTERNALERROR;
1206     } else {
1207         TRACE("Got %d elements as expected\n", i);
1208     }
1209
1210     HeapFree(GetProcessHeap(), 0, patch->mem);
1211     patch->mem = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, num_quads * 6 * d3d_out_vertex_size * sizeof(float) * 8);
1212     i = 0;
1213     for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1214         if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1215             ERR("Unexpected token: %f\n", feedbuffer[j]);
1216             continue;
1217         }
1218         if(feedbuffer[j + 1] != 3) {
1219             ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1220             continue;
1221         }
1222         /* Somehow there are different ideas about back / front facing, so fix up the
1223          * vertex order
1224          */
1225         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 2 + 2]; /* x, triangle 2 */
1226         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 2 + 3]; /* y, triangle 2 */
1227         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 2 + 4] - 0.5) * 4 * max_z; /* z, triangle 3 */
1228         if(patch->has_normals) {
1229             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 2 + 5];
1230             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 2 + 6];
1231             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 2 + 7];
1232         }
1233         i += d3d_out_vertex_size;
1234
1235         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 1 + 2]; /* x, triangle 2 */
1236         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 1 + 3]; /* y, triangle 2 */
1237         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 1 + 4] - 0.5) * 4 * max_z; /* z, triangle 2 */
1238         if(patch->has_normals) {
1239             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 1 + 5];
1240             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 1 + 6];
1241             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 1 + 7];
1242         }
1243         i += d3d_out_vertex_size;
1244
1245         patch->mem[i + 0] =  feedbuffer[j + out_vertex_size * 0 + 2]; /* x, triangle 1 */
1246         patch->mem[i + 1] =  feedbuffer[j + out_vertex_size * 0 + 3]; /* y, triangle 1 */
1247         patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 0 + 4] - 0.5) * 4 * max_z; /* z, triangle 1 */
1248         if(patch->has_normals) {
1249             patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 0 + 5];
1250             patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 0 + 6];
1251             patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 0 + 7];
1252         }
1253         i += d3d_out_vertex_size;
1254     }
1255
1256     if(patch->has_normals) {
1257         /* Now do the same with reverse light directions */
1258         const GLfloat x[4] = {-1,  0,  0, 0};
1259         const GLfloat y[4] = { 0, -1,  0, 0};
1260         const GLfloat z[4] = { 0,  0, -1, 0};
1261         glLightfv(GL_LIGHT0, GL_POSITION, x);
1262         glLightfv(GL_LIGHT1, GL_POSITION, y);
1263         glLightfv(GL_LIGHT2, GL_POSITION, z);
1264         checkGLcall("Setting up reverse light directions\n");
1265
1266         glRenderMode(GL_FEEDBACK);
1267         checkGLcall("glRenderMode(GL_FEEDBACK)");
1268         glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1269         checkGLcall("glEvalMesh2\n");
1270         i = glRenderMode(GL_RENDER);
1271         checkGLcall("glRenderMode(GL_RENDER)");
1272
1273         i = 0;
1274         for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1275             if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1276                 ERR("Unexpected token: %f\n", feedbuffer[j]);
1277                 continue;
1278             }
1279             if(feedbuffer[j + 1] != 3) {
1280                 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1281                 continue;
1282             }
1283             if(patch->mem[i + 3] == 0.0)
1284                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 2 + 5];
1285             if(patch->mem[i + 4] == 0.0)
1286                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 2 + 6];
1287             if(patch->mem[i + 5] == 0.0)
1288                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 2 + 7];
1289             normalize_normal(patch->mem + i + 3);
1290             i += d3d_out_vertex_size;
1291
1292             if(patch->mem[i + 3] == 0.0)
1293                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 1 + 5];
1294             if(patch->mem[i + 4] == 0.0)
1295                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 1 + 6];
1296             if(patch->mem[i + 5] == 0.0)
1297                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 1 + 7];
1298             normalize_normal(patch->mem + i + 3);
1299             i += d3d_out_vertex_size;
1300
1301             if(patch->mem[i + 3] == 0.0)
1302                 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 0 + 5];
1303             if(patch->mem[i + 4] == 0.0)
1304                 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 0 + 6];
1305             if(patch->mem[i + 5] == 0.0)
1306                 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 0 + 7];
1307             normalize_normal(patch->mem + i + 3);
1308             i += d3d_out_vertex_size;
1309         }
1310     }
1311
1312     glDisable(GL_MAP2_VERTEX_3);
1313     glDisable(GL_AUTO_NORMAL);
1314     glDisable(GL_MAP2_NORMAL);
1315     glDisable(GL_MAP2_TEXTURE_COORD_4);
1316     checkGLcall("glDisable vertex attrib generation");
1317     LEAVE_GL();
1318
1319     HeapFree(GetProcessHeap(), 0, feedbuffer);
1320
1321     vtxStride = 3 * sizeof(float);
1322     if(patch->has_normals) {
1323         vtxStride += 3 * sizeof(float);
1324     }
1325     if(patch->has_texcoords) {
1326         vtxStride += 4 * sizeof(float);
1327     }
1328     memset(&patch->strided, 0, sizeof(&patch->strided));
1329     patch->strided.u.s.position.lpData = (BYTE *) patch->mem;
1330     patch->strided.u.s.position.dwStride = vtxStride;
1331     patch->strided.u.s.position.dwType = WINED3DDECLTYPE_FLOAT3;
1332     patch->strided.u.s.position.streamNo = 255;
1333
1334     if(patch->has_normals) {
1335         patch->strided.u.s.normal.lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1336         patch->strided.u.s.normal.dwStride = vtxStride;
1337         patch->strided.u.s.normal.dwType = WINED3DDECLTYPE_FLOAT3;
1338         patch->strided.u.s.normal.streamNo = 255;
1339     }
1340     if(patch->has_texcoords) {
1341         patch->strided.u.s.texCoords[0].lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1342         if(patch->has_normals) {
1343             patch->strided.u.s.texCoords[0].lpData += 3 * sizeof(float);
1344         }
1345         patch->strided.u.s.texCoords[0].dwStride = vtxStride;
1346         patch->strided.u.s.texCoords[0].dwType = WINED3DDECLTYPE_FLOAT4;
1347         /* MAX_STREAMS index points to an unused element in stateblock->streamOffsets which
1348          * always remains set to 0. Windows uses stream 255 here, but this is not visible to the
1349          * application.
1350          */
1351         patch->strided.u.s.texCoords[0].streamNo = MAX_STREAMS;
1352     }
1353
1354     return WINED3D_OK;
1355 }