2 * WINED3D draw functions
4 * Copyright 2002-2004 Jason Edmeades
5 * Copyright 2002-2004 Raphael Junqueira
6 * Copyright 2004 Christian Costa
7 * Copyright 2005 Oliver Stieber
8 * Copyright 2006, 2008 Henri Verbeet
9 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
10 * Copyright 2009 Henri Verbeet for CodeWeavers
12 * This library is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * This library is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with this library; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
28 #include "wined3d_private.h"
30 WINE_DEFAULT_DEBUG_CHANNEL(d3d_draw);
35 /* GL locking is done by the caller */
36 static void drawStridedFast(const struct wined3d_gl_info *gl_info, GLenum primitive_type, UINT count, UINT idx_size,
37 const void *idx_data, UINT start_idx, INT base_vertex_index)
41 GLenum idxtype = idx_size == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT;
42 if (gl_info->supported[ARB_DRAW_ELEMENTS_BASE_VERTEX])
44 GL_EXTCALL(glDrawElementsBaseVertex(primitive_type, count, idxtype,
45 (const char *)idx_data + (idx_size * start_idx), base_vertex_index));
46 checkGLcall("glDrawElementsBaseVertex");
50 glDrawElements(primitive_type, count,
51 idxtype, (const char *)idx_data + (idx_size * start_idx));
52 checkGLcall("glDrawElements");
57 glDrawArrays(primitive_type, start_idx, count);
58 checkGLcall("glDrawArrays");
63 * Actually draw using the supplied information.
64 * Slower GL version which extracts info about each vertex in turn
67 /* GL locking is done by the caller */
68 static void drawStridedSlow(const struct wined3d_device *device, const struct wined3d_context *context,
69 const struct wined3d_stream_info *si, UINT NumVertexes, GLenum glPrimType,
70 const void *idxData, UINT idxSize, UINT startIdx)
72 unsigned int textureNo = 0;
73 const WORD *pIdxBufS = NULL;
74 const DWORD *pIdxBufL = NULL;
76 const struct wined3d_state *state = &device->stateBlock->state;
77 LONG SkipnStrides = startIdx;
78 BOOL pixelShader = use_ps(state);
79 BOOL specular_fog = FALSE;
80 const BYTE *texCoords[WINED3DDP_MAXTEXCOORD];
81 const BYTE *diffuse = NULL, *specular = NULL, *normal = NULL, *position = NULL;
82 const struct wined3d_gl_info *gl_info = context->gl_info;
83 UINT texture_stages = gl_info->limits.texture_stages;
84 const struct wined3d_stream_info_element *element;
85 UINT num_untracked_materials;
88 TRACE("Using slow vertex array code\n");
90 /* Variable Initialization */
93 /* Immediate mode drawing can't make use of indices in a vbo - get the
94 * data from the index buffer. If the index buffer has no vbo (not
95 * supported or other reason), or with user pointer drawing idxData
96 * will be non-NULL. */
98 idxData = buffer_get_sysmem(state->index_buffer, gl_info);
100 if (idxSize == 2) pIdxBufS = idxData;
101 else pIdxBufL = idxData;
102 } else if (idxData) {
103 ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
107 /* Start drawing in GL */
110 if (si->use_map & (1 << WINED3D_FFP_POSITION))
112 element = &si->elements[WINED3D_FFP_POSITION];
113 position = element->data.addr;
116 if (si->use_map & (1 << WINED3D_FFP_NORMAL))
118 element = &si->elements[WINED3D_FFP_NORMAL];
119 normal = element->data.addr;
126 num_untracked_materials = context->num_untracked_materials;
127 if (si->use_map & (1 << WINED3D_FFP_DIFFUSE))
129 element = &si->elements[WINED3D_FFP_DIFFUSE];
130 diffuse = element->data.addr;
132 if (num_untracked_materials && element->format->id != WINED3DFMT_B8G8R8A8_UNORM)
133 FIXME("Implement diffuse color tracking from %s\n", debug_d3dformat(element->format->id));
137 glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
140 if (si->use_map & (1 << WINED3D_FFP_SPECULAR))
142 element = &si->elements[WINED3D_FFP_SPECULAR];
143 specular = element->data.addr;
145 /* special case where the fog density is stored in the specular alpha channel */
146 if (state->render_states[WINED3DRS_FOGENABLE]
147 && (state->render_states[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE
148 || si->elements[WINED3D_FFP_POSITION].format->id == WINED3DFMT_R32G32B32A32_FLOAT)
149 && state->render_states[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE)
151 if (gl_info->supported[EXT_FOG_COORD])
153 if (element->format->id == WINED3DFMT_B8G8R8A8_UNORM) specular_fog = TRUE;
154 else FIXME("Implement fog coordinates from %s\n", debug_d3dformat(element->format->id));
162 /* TODO: Use the fog table code from old ddraw */
163 FIXME("Implement fog for transformed vertices in software\n");
169 else if (gl_info->supported[EXT_SECONDARY_COLOR])
171 GL_EXTCALL(glSecondaryColor3fEXT)(0, 0, 0);
174 for (textureNo = 0; textureNo < texture_stages; ++textureNo)
176 int coordIdx = state->texture_states[textureNo][WINED3DTSS_TEXCOORDINDEX];
177 DWORD texture_idx = device->texUnitMap[textureNo];
179 if (!gl_info->supported[ARB_MULTITEXTURE] && textureNo > 0)
181 FIXME("Program using multiple concurrent textures which this opengl implementation doesn't support\n");
185 if (!pixelShader && !state->textures[textureNo]) continue;
187 if (texture_idx == WINED3D_UNMAPPED_STAGE) continue;
191 TRACE("tex: %d - Skip tex coords, as being system generated\n", textureNo);
194 else if (coordIdx < 0)
196 FIXME("tex: %d - Coord index %d is less than zero, expect a crash.\n", textureNo, coordIdx);
200 if (si->use_map & (1 << (WINED3D_FFP_TEXCOORD0 + coordIdx)))
202 element = &si->elements[WINED3D_FFP_TEXCOORD0 + coordIdx];
203 texCoords[coordIdx] = element->data.addr;
204 tex_mask |= (1 << textureNo);
208 TRACE("tex: %d - Skipping tex coords, as no data supplied\n", textureNo);
209 if (gl_info->supported[ARB_MULTITEXTURE])
210 GL_EXTCALL(glMultiTexCoord4fARB(GL_TEXTURE0_ARB + texture_idx, 0, 0, 0, 1));
212 glTexCoord4f(0, 0, 0, 1);
216 /* We shouldn't start this function if any VBO is involved. Should I put a safety check here?
217 * Guess it's not necessary(we crash then anyway) and would only eat CPU time
220 /* For each primitive */
221 for (vx_index = 0; vx_index < NumVertexes; ++vx_index) {
222 UINT texture, tmp_tex_mask;
223 /* Blending data and Point sizes are not supported by this function. They are not supported by the fixed
224 * function pipeline at all. A Fixme for them is printed after decoding the vertex declaration
227 /* For indexed data, we need to go a few more strides in */
230 /* Indexed so work out the number of strides to skip */
232 SkipnStrides = pIdxBufS[startIdx + vx_index] + state->base_vertex_index;
234 SkipnStrides = pIdxBufL[startIdx + vx_index] + state->base_vertex_index;
237 tmp_tex_mask = tex_mask;
238 for (texture = 0; tmp_tex_mask; tmp_tex_mask >>= 1, ++texture)
244 if (!(tmp_tex_mask & 1)) continue;
246 coord_idx = state->texture_states[texture][WINED3DTSS_TEXCOORDINDEX];
247 ptr = texCoords[coord_idx] + (SkipnStrides * si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].stride);
249 texture_idx = device->texUnitMap[texture];
250 multi_texcoord_funcs[si->elements[WINED3D_FFP_TEXCOORD0 + coord_idx].format->emit_idx](
251 GL_TEXTURE0_ARB + texture_idx, ptr);
254 /* Diffuse -------------------------------- */
256 const void *ptrToCoords = diffuse + SkipnStrides * si->elements[WINED3D_FFP_DIFFUSE].stride;
258 diffuse_funcs[si->elements[WINED3D_FFP_DIFFUSE].format->emit_idx](ptrToCoords);
259 if (num_untracked_materials)
261 DWORD diffuseColor = ((const DWORD *)ptrToCoords)[0];
265 color[0] = D3DCOLOR_B_R(diffuseColor) / 255.0f;
266 color[1] = D3DCOLOR_B_G(diffuseColor) / 255.0f;
267 color[2] = D3DCOLOR_B_B(diffuseColor) / 255.0f;
268 color[3] = D3DCOLOR_B_A(diffuseColor) / 255.0f;
270 for (i = 0; i < num_untracked_materials; ++i)
272 glMaterialfv(GL_FRONT_AND_BACK, context->untracked_materials[i], color);
277 /* Specular ------------------------------- */
279 const void *ptrToCoords = specular + SkipnStrides * si->elements[WINED3D_FFP_SPECULAR].stride;
281 specular_funcs[si->elements[WINED3D_FFP_SPECULAR].format->emit_idx](ptrToCoords);
285 DWORD specularColor = *(const DWORD *)ptrToCoords;
286 GL_EXTCALL(glFogCoordfEXT((float) (specularColor >> 24)));
290 /* Normal -------------------------------- */
293 const void *ptrToCoords = normal + SkipnStrides * si->elements[WINED3D_FFP_NORMAL].stride;
294 normal_funcs[si->elements[WINED3D_FFP_NORMAL].format->emit_idx](ptrToCoords);
297 /* Position -------------------------------- */
299 const void *ptrToCoords = position + SkipnStrides * si->elements[WINED3D_FFP_POSITION].stride;
300 position_funcs[si->elements[WINED3D_FFP_POSITION].format->emit_idx](ptrToCoords);
303 /* For non indexed mode, step onto next parts */
304 if (!idxData) ++SkipnStrides;
308 checkGLcall("glEnd and previous calls");
311 /* GL locking is done by the caller */
312 static inline void send_attribute(const struct wined3d_gl_info *gl_info,
313 enum wined3d_format_id format, const UINT index, const void *ptr)
317 case WINED3DFMT_R32_FLOAT:
318 GL_EXTCALL(glVertexAttrib1fvARB(index, ptr));
320 case WINED3DFMT_R32G32_FLOAT:
321 GL_EXTCALL(glVertexAttrib2fvARB(index, ptr));
323 case WINED3DFMT_R32G32B32_FLOAT:
324 GL_EXTCALL(glVertexAttrib3fvARB(index, ptr));
326 case WINED3DFMT_R32G32B32A32_FLOAT:
327 GL_EXTCALL(glVertexAttrib4fvARB(index, ptr));
330 case WINED3DFMT_R8G8B8A8_UINT:
331 GL_EXTCALL(glVertexAttrib4ubvARB(index, ptr));
333 case WINED3DFMT_B8G8R8A8_UNORM:
334 if (gl_info->supported[ARB_VERTEX_ARRAY_BGRA])
336 const DWORD *src = ptr;
337 DWORD c = *src & 0xff00ff00;
338 c |= (*src & 0xff0000) >> 16;
339 c |= (*src & 0xff) << 16;
340 GL_EXTCALL(glVertexAttrib4NubvARB(index, (GLubyte *)&c));
343 /* else fallthrough */
344 case WINED3DFMT_R8G8B8A8_UNORM:
345 GL_EXTCALL(glVertexAttrib4NubvARB(index, ptr));
348 case WINED3DFMT_R16G16_SINT:
349 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
351 case WINED3DFMT_R16G16B16A16_SINT:
352 GL_EXTCALL(glVertexAttrib4svARB(index, ptr));
355 case WINED3DFMT_R16G16_SNORM:
357 GLshort s[4] = {((const GLshort *)ptr)[0], ((const GLshort *)ptr)[1], 0, 1};
358 GL_EXTCALL(glVertexAttrib4NsvARB(index, s));
361 case WINED3DFMT_R16G16_UNORM:
363 GLushort s[4] = {((const GLushort *)ptr)[0], ((const GLushort *)ptr)[1], 0, 1};
364 GL_EXTCALL(glVertexAttrib4NusvARB(index, s));
367 case WINED3DFMT_R16G16B16A16_SNORM:
368 GL_EXTCALL(glVertexAttrib4NsvARB(index, ptr));
370 case WINED3DFMT_R16G16B16A16_UNORM:
371 GL_EXTCALL(glVertexAttrib4NusvARB(index, ptr));
374 case WINED3DFMT_R10G10B10A2_UINT:
375 FIXME("Unsure about WINED3DDECLTYPE_UDEC3\n");
376 /*glVertexAttrib3usvARB(instancedData[j], (GLushort *) ptr); Does not exist */
378 case WINED3DFMT_R10G10B10A2_SNORM:
379 FIXME("Unsure about WINED3DDECLTYPE_DEC3N\n");
380 /*glVertexAttrib3NusvARB(instancedData[j], (GLushort *) ptr); Does not exist */
383 case WINED3DFMT_R16G16_FLOAT:
384 /* Are those 16 bit floats. C doesn't have a 16 bit float type. I could read the single bits and calculate a 4
385 * byte float according to the IEEE standard
387 if (gl_info->supported[NV_HALF_FLOAT])
389 /* Not supported by GL_ARB_half_float_vertex */
390 GL_EXTCALL(glVertexAttrib2hvNV(index, ptr));
394 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
395 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
396 GL_EXTCALL(glVertexAttrib2fARB(index, x, y));
399 case WINED3DFMT_R16G16B16A16_FLOAT:
400 if (gl_info->supported[NV_HALF_FLOAT])
402 /* Not supported by GL_ARB_half_float_vertex */
403 GL_EXTCALL(glVertexAttrib4hvNV(index, ptr));
407 float x = float_16_to_32(((const unsigned short *)ptr) + 0);
408 float y = float_16_to_32(((const unsigned short *)ptr) + 1);
409 float z = float_16_to_32(((const unsigned short *)ptr) + 2);
410 float w = float_16_to_32(((const unsigned short *)ptr) + 3);
411 GL_EXTCALL(glVertexAttrib4fARB(index, x, y, z, w));
416 ERR("Unexpected attribute format: %s\n", debug_d3dformat(format));
421 /* GL locking is done by the caller */
422 static void drawStridedSlowVs(const struct wined3d_gl_info *gl_info, const struct wined3d_state *state,
423 const struct wined3d_stream_info *si, UINT numberOfVertices, GLenum glPrimitiveType,
424 const void *idxData, UINT idxSize, UINT startIdx)
426 LONG SkipnStrides = startIdx + state->load_base_vertex_index;
427 const DWORD *pIdxBufL = NULL;
428 const WORD *pIdxBufS = NULL;
435 /* Immediate mode drawing can't make use of indices in a vbo - get the
436 * data from the index buffer. If the index buffer has no vbo (not
437 * supported or other reason), or with user pointer drawing idxData
438 * will be non-NULL. */
440 idxData = buffer_get_sysmem(state->index_buffer, gl_info);
442 if (idxSize == 2) pIdxBufS = idxData;
443 else pIdxBufL = idxData;
444 } else if (idxData) {
445 ERR("non-NULL idxData with 0 idxSize, this should never happen\n");
449 /* Start drawing in GL */
450 glBegin(glPrimitiveType);
452 for (vx_index = 0; vx_index < numberOfVertices; ++vx_index)
456 /* Indexed so work out the number of strides to skip */
458 SkipnStrides = pIdxBufS[startIdx + vx_index] + state->load_base_vertex_index;
460 SkipnStrides = pIdxBufL[startIdx + vx_index] + state->load_base_vertex_index;
463 for (i = MAX_ATTRIBS - 1; i >= 0; i--)
465 if (!(si->use_map & (1 << i))) continue;
467 ptr = si->elements[i].data.addr + si->elements[i].stride * SkipnStrides;
469 send_attribute(gl_info, si->elements[i].format->id, i, ptr);
477 /* GL locking is done by the caller */
478 static void drawStridedInstanced(const struct wined3d_gl_info *gl_info, const struct wined3d_state *state,
479 const struct wined3d_stream_info *si, UINT numberOfVertices, GLenum glPrimitiveType,
480 const void *idxData, UINT idxSize, UINT startIdx, UINT base_vertex_index)
482 UINT numInstances = 0, i;
483 int numInstancedAttribs = 0, j;
484 UINT instancedData[sizeof(si->elements) / sizeof(*si->elements) /* 16 */];
485 GLenum idxtype = idxSize == 2 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_INT;
489 /* This is a nasty thing. MSDN says no hardware supports that and apps have to use software vertex processing.
490 * We don't support this for now
492 * Shouldn't be too hard to support with opengl, in theory just call glDrawArrays instead of drawElements.
493 * But the StreamSourceFreq value has a different meaning in that situation.
495 FIXME("Non-indexed instanced drawing is not supported\n");
499 /* First, figure out how many instances we have to draw */
500 for (i = 0; i < MAX_STREAMS; ++i)
502 /* Look at the streams and take the first one which matches */
503 if (state->streams[i].buffer
504 && ((state->streams[i].flags & WINED3DSTREAMSOURCE_INSTANCEDATA)
505 || (state->streams[i].flags & WINED3DSTREAMSOURCE_INDEXEDDATA)))
507 /* Use the specified number of instances from the first matched
508 * stream. A streamFreq of 0 (with INSTANCEDATA or INDEXEDDATA)
509 * is handled as 1. See d3d9/tests/visual.c-> stream_test(). */
510 numInstances = state->streams[i].frequency ? state->streams[i].frequency : 1;
515 for (i = 0; i < sizeof(si->elements) / sizeof(*si->elements); ++i)
517 if (!(si->use_map & (1 << i))) continue;
519 if (state->streams[si->elements[i].stream_idx].flags & WINED3DSTREAMSOURCE_INSTANCEDATA)
521 instancedData[numInstancedAttribs] = i;
522 numInstancedAttribs++;
526 /* now draw numInstances instances :-) */
527 for(i = 0; i < numInstances; i++) {
528 /* Specify the instanced attributes using immediate mode calls */
529 for(j = 0; j < numInstancedAttribs; j++) {
530 const BYTE *ptr = si->elements[instancedData[j]].data.addr
531 + si->elements[instancedData[j]].stride * i;
532 if (si->elements[instancedData[j]].data.buffer_object)
534 struct wined3d_buffer *vb = state->streams[si->elements[instancedData[j]].stream_idx].buffer;
535 ptr += (ULONG_PTR)buffer_get_sysmem(vb, gl_info);
538 send_attribute(gl_info, si->elements[instancedData[j]].format->id, instancedData[j], ptr);
541 if (gl_info->supported[ARB_DRAW_ELEMENTS_BASE_VERTEX])
543 GL_EXTCALL(glDrawElementsBaseVertex(glPrimitiveType, numberOfVertices, idxtype,
544 (const char *)idxData+(idxSize * startIdx), base_vertex_index));
545 checkGLcall("glDrawElementsBaseVertex");
549 glDrawElements(glPrimitiveType, numberOfVertices, idxtype,
550 (const char *)idxData+(idxSize * startIdx));
551 checkGLcall("glDrawElements");
556 static void remove_vbos(const struct wined3d_gl_info *gl_info,
557 const struct wined3d_state *state, struct wined3d_stream_info *s)
561 for (i = 0; i < (sizeof(s->elements) / sizeof(*s->elements)); ++i)
563 struct wined3d_stream_info_element *e;
565 if (!(s->use_map & (1 << i))) continue;
568 if (e->data.buffer_object)
570 struct wined3d_buffer *vb = state->streams[e->stream_idx].buffer;
571 e->data.buffer_object = 0;
572 e->data.addr = (BYTE *)((ULONG_PTR)e->data.addr + (ULONG_PTR)buffer_get_sysmem(vb, gl_info));
577 /* Routine common to the draw primitive and draw indexed primitive routines */
578 void drawPrimitive(struct wined3d_device *device, UINT index_count, UINT StartIdx, UINT idxSize, const void *idxData)
580 const struct wined3d_state *state = &device->stateBlock->state;
581 struct wined3d_context *context;
584 if (!index_count) return;
586 if (state->render_states[WINED3DRS_COLORWRITEENABLE])
588 /* Invalidate the back buffer memory so LockRect will read it the next time */
589 for (i = 0; i < device->adapter->gl_info.limits.buffers; ++i)
591 struct wined3d_surface *target = device->fb.render_targets[i];
594 surface_load_location(target, target->draw_binding, NULL);
595 surface_modify_location(target, target->draw_binding, TRUE);
600 /* Signals other modules that a drawing is in progress and the stateblock finalized */
601 device->isInDraw = TRUE;
603 context = context_acquire(device, device->fb.render_targets[0]);
606 context_release(context);
607 WARN("Invalid context, skipping draw.\n");
611 if (device->fb.depth_stencil)
613 /* Note that this depends on the context_acquire() call above to set
614 * context->render_offscreen properly. We don't currently take the
615 * Z-compare function into account, but we could skip loading the
616 * depthstencil for D3DCMP_NEVER and D3DCMP_ALWAYS as well. Also note
617 * that we never copy the stencil data.*/
618 DWORD location = context->render_offscreen ? SFLAG_DS_OFFSCREEN : SFLAG_DS_ONSCREEN;
619 if (state->render_states[WINED3DRS_ZWRITEENABLE] || state->render_states[WINED3DRS_ZENABLE])
621 struct wined3d_surface *ds = device->fb.depth_stencil;
622 RECT current_rect, draw_rect, r;
624 if (location == SFLAG_DS_ONSCREEN && ds != device->onscreen_depth_stencil)
625 device_switch_onscreen_ds(device, context, ds);
627 if (ds->flags & location)
628 SetRect(¤t_rect, 0, 0, ds->ds_current_size.cx, ds->ds_current_size.cy);
630 SetRectEmpty(¤t_rect);
632 wined3d_get_draw_rect(state, &draw_rect);
634 IntersectRect(&r, &draw_rect, ¤t_rect);
635 if (!EqualRect(&r, &draw_rect))
636 surface_load_ds_location(ds, context, location);
638 if (state->render_states[WINED3DRS_ZWRITEENABLE])
640 surface_modify_ds_location(ds, location, ds->ds_current_size.cx, ds->ds_current_size.cy);
641 surface_modify_location(ds, ds->draw_binding, TRUE);
646 if (!context_apply_draw_state(context, device))
648 context_release(context);
649 WARN("Unable to apply draw state, skipping draw.\n");
653 if ((!context->gl_info->supported[WINED3D_GL_VERSION_2_0]
654 || (!glPointParameteri && !context->gl_info->supported[NV_POINT_SPRITE]))
655 && context->render_offscreen
656 && state->render_states[WINED3DRS_POINTSPRITEENABLE]
657 && state->gl_primitive_type == GL_POINTS)
659 FIXME("Point sprite coordinate origin switching not supported.\n");
662 /* Ok, we will be updating the screen from here onwards so grab the lock */
665 GLenum glPrimType = state->gl_primitive_type;
666 INT base_vertex_index = state->base_vertex_index;
667 BOOL emulation = FALSE;
668 const struct wined3d_stream_info *stream_info = &device->strided_streams;
669 struct wined3d_stream_info stridedlcl;
673 if (!stream_info->position_transformed && context->num_untracked_materials
674 && state->render_states[WINED3DRS_LIGHTING])
678 FIXME("Using software emulation because not all material properties could be tracked\n");
681 TRACE("Using software emulation because not all material properties could be tracked\n");
685 else if (context->fog_coord && state->render_states[WINED3DRS_FOGENABLE])
687 /* Either write a pipeline replacement shader or convert the specular alpha from unsigned byte
688 * to a float in the vertex buffer
692 FIXME("Using software emulation because manual fog coordinates are provided\n");
695 TRACE("Using software emulation because manual fog coordinates are provided\n");
701 stream_info = &stridedlcl;
702 memcpy(&stridedlcl, &device->strided_streams, sizeof(stridedlcl));
703 remove_vbos(context->gl_info, state, &stridedlcl);
707 if (device->useDrawStridedSlow || emulation)
709 /* Immediate mode drawing */
714 FIXME("Using immediate mode with vertex shaders for half float emulation\n");
717 TRACE("Using immediate mode with vertex shaders for half float emulation\n");
719 drawStridedSlowVs(context->gl_info, state, stream_info,
720 index_count, glPrimType, idxData, idxSize, StartIdx);
724 drawStridedSlow(device, context, stream_info, index_count,
725 glPrimType, idxData, idxSize, StartIdx);
728 else if (device->instancedDraw)
730 /* Instancing emulation with mixing immediate mode and arrays */
731 drawStridedInstanced(context->gl_info, state, stream_info,
732 index_count, glPrimType, idxData, idxSize, StartIdx, base_vertex_index);
736 drawStridedFast(context->gl_info, glPrimType, index_count, idxSize, idxData, StartIdx, base_vertex_index);
740 /* Finished updating the screen, restore lock */
743 for(i = 0; i < device->num_buffer_queries; ++i)
745 wined3d_event_query_issue(device->buffer_queries[i], device);
748 if (wined3d_settings.strict_draw_ordering) wglFlush(); /* Flush to ensure ordering across contexts. */
750 context_release(context);
752 TRACE("Done all gl drawing\n");
754 /* Control goes back to the device, stateblock values may change again */
755 device->isInDraw = FALSE;
758 static void normalize_normal(float *n) {
759 float length = n[0] * n[0] + n[1] * n[1] + n[2] * n[2];
760 if (length == 0.0f) return;
761 length = sqrtf(length);
762 n[0] = n[0] / length;
763 n[1] = n[1] / length;
764 n[2] = n[2] / length;
767 /* Tesselates a high order rectangular patch into single triangles using gl evaluators
769 * The problem is that OpenGL does not offer a direct way to return the tesselated primitives,
770 * and they can't be sent off for rendering directly either. Tesselating is slow, so we want
771 * to cache the patches in a vertex buffer. But more importantly, gl can't bind generated
772 * attributes to numbered shader attributes, so we have to store them and rebind them as needed
775 * To read back, the opengl feedback mode is used. This creates a problem because we want
776 * untransformed, unlit vertices, but feedback runs everything through transform and lighting.
777 * Thus disable lighting and set identity matrices to get unmodified colors and positions.
778 * To overcome clipping find the biggest x, y and z values of the vertices in the patch and scale
779 * them to [-1.0;+1.0] and set the viewport up to scale them back.
781 * Normals are more tricky: Draw white vertices with 3 directional lights, and calculate the
782 * resulting colors back to the normals.
784 * NOTE: This function activates a context for blitting, modifies matrices & viewport, but
785 * does not restore it because normally a draw follows immediately afterwards. The caller is
786 * responsible of taking care that either the gl states are restored, or the context activated
787 * for drawing to reset the lastWasBlit flag.
789 HRESULT tesselate_rectpatch(struct wined3d_device *This, struct WineD3DRectPatch *patch)
791 unsigned int i, j, num_quads, out_vertex_size, buffer_size, d3d_out_vertex_size;
792 float max_x = 0.0f, max_y = 0.0f, max_z = 0.0f, neg_z = 0.0f;
793 struct wined3d_state *state = &This->stateBlock->state;
794 struct wined3d_stream_info stream_info;
795 struct wined3d_stream_info_element *e;
796 struct wined3d_context *context;
797 struct wined3d_shader *vs;
799 const WINED3DRECTPATCH_INFO *info = &patch->RectPatchInfo;
801 GLenum feedback_type;
804 /* Simply activate the context for blitting. This disables all the things we don't want and
805 * takes care of dirtifying. Dirtifying is preferred over pushing / popping, since drawing the
806 * patch (as opposed to normal draws) will most likely need different changes anyway. */
807 context = context_acquire(This, NULL);
808 context_apply_blit_state(context, This);
810 /* First, locate the position data. This is provided in a vertex buffer in
811 * the stateblock. Beware of VBOs. */
812 vs = state->vertex_shader;
813 state->vertex_shader = NULL;
814 device_stream_info_from_declaration(This, &stream_info, NULL);
815 state->vertex_shader = vs;
817 e = &stream_info.elements[WINED3D_FFP_POSITION];
818 if (e->data.buffer_object)
820 struct wined3d_buffer *vb = state->streams[e->stream_idx].buffer;
821 e->data.addr = (BYTE *)((ULONG_PTR)e->data.addr + (ULONG_PTR)buffer_get_sysmem(vb, context->gl_info));
823 vtxStride = e->stride;
824 data = e->data.addr +
825 vtxStride * info->Stride * info->StartVertexOffsetHeight +
826 vtxStride * info->StartVertexOffsetWidth;
828 /* Not entirely sure about what happens with transformed vertices */
829 if (stream_info.position_transformed) FIXME("Transformed position in rectpatch generation\n");
831 if(vtxStride % sizeof(GLfloat)) {
832 /* glMap2f reads vertex sizes in GLfloats, the d3d stride is in bytes.
833 * I don't see how the stride could not be a multiple of 4, but make sure
836 ERR("Vertex stride is not a multiple of sizeof(GLfloat)\n");
838 if(info->Basis != WINED3DBASIS_BEZIER) {
839 FIXME("Basis is %s, how to handle this?\n", debug_d3dbasis(info->Basis));
841 if(info->Degree != WINED3DDEGREE_CUBIC) {
842 FIXME("Degree is %s, how to handle this?\n", debug_d3ddegree(info->Degree));
845 /* First, get the boundary cube of the input data */
846 for(j = 0; j < info->Height; j++) {
847 for(i = 0; i < info->Width; i++) {
848 const float *v = (const float *)(data + vtxStride * i + vtxStride * info->Stride * j);
849 if(fabs(v[0]) > max_x) max_x = fabsf(v[0]);
850 if(fabs(v[1]) > max_y) max_y = fabsf(v[1]);
851 if(fabs(v[2]) > max_z) max_z = fabsf(v[2]);
852 if(v[2] < neg_z) neg_z = v[2];
856 /* This needs some improvements in the vertex decl code */
857 FIXME("Cannot find data to generate. Only generating position and normals\n");
858 patch->has_normals = TRUE;
859 patch->has_texcoords = FALSE;
863 glMatrixMode(GL_PROJECTION);
864 checkGLcall("glMatrixMode(GL_PROJECTION)");
866 checkGLcall("glLoadIdentity()");
867 glScalef(1.0f / (max_x), 1.0f / (max_y), max_z == 0.0f ? 1.0f : 1.0f / (2.0f * max_z));
868 glTranslatef(0.0f, 0.0f, 0.5f);
869 checkGLcall("glScalef");
870 glViewport(-max_x, -max_y, 2 * (max_x), 2 * (max_y));
871 checkGLcall("glViewport");
873 /* Some states to take care of. If we're in wireframe opengl will produce lines, and confuse
874 * our feedback buffer parser
876 glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
877 checkGLcall("glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)");
878 context_invalidate_state(context, STATE_RENDER(WINED3DRS_FILLMODE));
879 if (patch->has_normals)
881 static const GLfloat black[] = {0.0f, 0.0f, 0.0f, 0.0f};
882 static const GLfloat red[] = {1.0f, 0.0f, 0.0f, 0.0f};
883 static const GLfloat green[] = {0.0f, 1.0f, 0.0f, 0.0f};
884 static const GLfloat blue[] = {0.0f, 0.0f, 1.0f, 0.0f};
885 static const GLfloat white[] = {1.0f, 1.0f, 1.0f, 1.0f};
886 glEnable(GL_LIGHTING);
887 checkGLcall("glEnable(GL_LIGHTING)");
888 glLightModelfv(GL_LIGHT_MODEL_AMBIENT, black);
889 checkGLcall("glLightModel for MODEL_AMBIENT");
890 context_invalidate_state(context, STATE_RENDER(WINED3DRS_AMBIENT));
892 for (i = 3; i < context->gl_info->limits.lights; ++i)
894 glDisable(GL_LIGHT0 + i);
895 checkGLcall("glDisable(GL_LIGHT0 + i)");
896 context_invalidate_state(context, STATE_ACTIVELIGHT(i));
899 context_invalidate_state(context, STATE_ACTIVELIGHT(0));
900 glLightfv(GL_LIGHT0, GL_DIFFUSE, red);
901 glLightfv(GL_LIGHT0, GL_SPECULAR, black);
902 glLightfv(GL_LIGHT0, GL_AMBIENT, black);
903 glLightfv(GL_LIGHT0, GL_POSITION, red);
905 checkGLcall("Setting up light 1");
906 context_invalidate_state(context, STATE_ACTIVELIGHT(1));
907 glLightfv(GL_LIGHT1, GL_DIFFUSE, green);
908 glLightfv(GL_LIGHT1, GL_SPECULAR, black);
909 glLightfv(GL_LIGHT1, GL_AMBIENT, black);
910 glLightfv(GL_LIGHT1, GL_POSITION, green);
912 checkGLcall("Setting up light 2");
913 context_invalidate_state(context, STATE_ACTIVELIGHT(2));
914 glLightfv(GL_LIGHT2, GL_DIFFUSE, blue);
915 glLightfv(GL_LIGHT2, GL_SPECULAR, black);
916 glLightfv(GL_LIGHT2, GL_AMBIENT, black);
917 glLightfv(GL_LIGHT2, GL_POSITION, blue);
919 checkGLcall("Setting up light 3");
921 context_invalidate_state(context, STATE_MATERIAL);
922 context_invalidate_state(context, STATE_RENDER(WINED3DRS_COLORVERTEX));
923 glDisable(GL_COLOR_MATERIAL);
924 glMaterialfv(GL_FRONT_AND_BACK, GL_EMISSION, black);
925 glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR, black);
926 glMaterialfv(GL_FRONT_AND_BACK, GL_DIFFUSE, white);
927 checkGLcall("Setting up materials");
930 /* Enable the needed maps.
931 * GL_MAP2_VERTEX_3 is needed for positional data.
932 * GL_AUTO_NORMAL to generate normals from the position. Do not use GL_MAP2_NORMAL.
933 * GL_MAP2_TEXTURE_COORD_4 for texture coords
935 num_quads = ceilf(patch->numSegs[0]) * ceilf(patch->numSegs[1]);
936 out_vertex_size = 3 /* position */;
937 d3d_out_vertex_size = 3;
938 glEnable(GL_MAP2_VERTEX_3);
939 if(patch->has_normals && patch->has_texcoords) {
940 FIXME("Texcoords not handled yet\n");
941 feedback_type = GL_3D_COLOR_TEXTURE;
942 out_vertex_size += 8;
943 d3d_out_vertex_size += 7;
944 glEnable(GL_AUTO_NORMAL);
945 glEnable(GL_MAP2_TEXTURE_COORD_4);
946 } else if(patch->has_texcoords) {
947 FIXME("Texcoords not handled yet\n");
948 feedback_type = GL_3D_COLOR_TEXTURE;
949 out_vertex_size += 7;
950 d3d_out_vertex_size += 4;
951 glEnable(GL_MAP2_TEXTURE_COORD_4);
952 } else if(patch->has_normals) {
953 feedback_type = GL_3D_COLOR;
954 out_vertex_size += 4;
955 d3d_out_vertex_size += 3;
956 glEnable(GL_AUTO_NORMAL);
958 feedback_type = GL_3D;
960 checkGLcall("glEnable vertex attrib generation");
962 buffer_size = num_quads * out_vertex_size * 2 /* triangle list */ * 3 /* verts per tri */
963 + 4 * num_quads /* 2 triangle markers per quad + num verts in tri */;
964 feedbuffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, buffer_size * sizeof(float) * 8);
966 glMap2f(GL_MAP2_VERTEX_3,
967 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
968 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
969 (const GLfloat *)data);
970 checkGLcall("glMap2f");
971 if(patch->has_texcoords) {
972 glMap2f(GL_MAP2_TEXTURE_COORD_4,
973 0.0f, 1.0f, vtxStride / sizeof(float), info->Width,
974 0.0f, 1.0f, info->Stride * vtxStride / sizeof(float), info->Height,
975 (const GLfloat *)data);
976 checkGLcall("glMap2f");
978 glMapGrid2f(ceilf(patch->numSegs[0]), 0.0f, 1.0f, ceilf(patch->numSegs[1]), 0.0f, 1.0f);
979 checkGLcall("glMapGrid2f");
981 glFeedbackBuffer(buffer_size * 2, feedback_type, feedbuffer);
982 checkGLcall("glFeedbackBuffer");
983 glRenderMode(GL_FEEDBACK);
985 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
986 checkGLcall("glEvalMesh2");
988 i = glRenderMode(GL_RENDER);
991 ERR("Feedback failed. Expected %d elements back\n", buffer_size);
992 HeapFree(GetProcessHeap(), 0, feedbuffer);
993 context_release(context);
994 return WINED3DERR_DRIVERINTERNALERROR;
995 } else if(i != buffer_size) {
997 ERR("Unexpected amount of elements returned. Expected %d, got %d\n", buffer_size, i);
998 HeapFree(GetProcessHeap(), 0, feedbuffer);
999 context_release(context);
1000 return WINED3DERR_DRIVERINTERNALERROR;
1002 TRACE("Got %d elements as expected\n", i);
1005 HeapFree(GetProcessHeap(), 0, patch->mem);
1006 patch->mem = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, num_quads * 6 * d3d_out_vertex_size * sizeof(float) * 8);
1008 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1009 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1010 ERR("Unexpected token: %f\n", feedbuffer[j]);
1013 if(feedbuffer[j + 1] != 3) {
1014 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1017 /* Somehow there are different ideas about back / front facing, so fix up the
1020 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 2 + 2]; /* x, triangle 2 */
1021 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 2 + 3]; /* y, triangle 2 */
1022 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 2 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 3 */
1023 if(patch->has_normals) {
1024 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 2 + 5];
1025 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 2 + 6];
1026 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 2 + 7];
1028 i += d3d_out_vertex_size;
1030 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 1 + 2]; /* x, triangle 2 */
1031 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 1 + 3]; /* y, triangle 2 */
1032 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 1 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 2 */
1033 if(patch->has_normals) {
1034 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 1 + 5];
1035 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 1 + 6];
1036 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 1 + 7];
1038 i += d3d_out_vertex_size;
1040 patch->mem[i + 0] = feedbuffer[j + out_vertex_size * 0 + 2]; /* x, triangle 1 */
1041 patch->mem[i + 1] = feedbuffer[j + out_vertex_size * 0 + 3]; /* y, triangle 1 */
1042 patch->mem[i + 2] = (feedbuffer[j + out_vertex_size * 0 + 4] - 0.5f) * 4.0f * max_z; /* z, triangle 1 */
1043 if(patch->has_normals) {
1044 patch->mem[i + 3] = feedbuffer[j + out_vertex_size * 0 + 5];
1045 patch->mem[i + 4] = feedbuffer[j + out_vertex_size * 0 + 6];
1046 patch->mem[i + 5] = feedbuffer[j + out_vertex_size * 0 + 7];
1048 i += d3d_out_vertex_size;
1051 if(patch->has_normals) {
1052 /* Now do the same with reverse light directions */
1053 static const GLfloat x[] = {-1.0f, 0.0f, 0.0f, 0.0f};
1054 static const GLfloat y[] = { 0.0f, -1.0f, 0.0f, 0.0f};
1055 static const GLfloat z[] = { 0.0f, 0.0f, -1.0f, 0.0f};
1056 glLightfv(GL_LIGHT0, GL_POSITION, x);
1057 glLightfv(GL_LIGHT1, GL_POSITION, y);
1058 glLightfv(GL_LIGHT2, GL_POSITION, z);
1059 checkGLcall("Setting up reverse light directions");
1061 glRenderMode(GL_FEEDBACK);
1062 checkGLcall("glRenderMode(GL_FEEDBACK)");
1063 glEvalMesh2(GL_FILL, 0, ceilf(patch->numSegs[0]), 0, ceilf(patch->numSegs[1]));
1064 checkGLcall("glEvalMesh2");
1065 i = glRenderMode(GL_RENDER);
1066 checkGLcall("glRenderMode(GL_RENDER)");
1069 for(j = 0; j < buffer_size; j += (3 /* num verts */ * out_vertex_size + 2 /* tri marker */)) {
1070 if(feedbuffer[j] != GL_POLYGON_TOKEN) {
1071 ERR("Unexpected token: %f\n", feedbuffer[j]);
1074 if(feedbuffer[j + 1] != 3) {
1075 ERR("Unexpected polygon: %f corners\n", feedbuffer[j + 1]);
1078 if(patch->mem[i + 3] == 0.0f)
1079 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 2 + 5];
1080 if(patch->mem[i + 4] == 0.0f)
1081 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 2 + 6];
1082 if(patch->mem[i + 5] == 0.0f)
1083 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 2 + 7];
1084 normalize_normal(patch->mem + i + 3);
1085 i += d3d_out_vertex_size;
1087 if(patch->mem[i + 3] == 0.0f)
1088 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 1 + 5];
1089 if(patch->mem[i + 4] == 0.0f)
1090 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 1 + 6];
1091 if(patch->mem[i + 5] == 0.0f)
1092 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 1 + 7];
1093 normalize_normal(patch->mem + i + 3);
1094 i += d3d_out_vertex_size;
1096 if(patch->mem[i + 3] == 0.0f)
1097 patch->mem[i + 3] = -feedbuffer[j + out_vertex_size * 0 + 5];
1098 if(patch->mem[i + 4] == 0.0f)
1099 patch->mem[i + 4] = -feedbuffer[j + out_vertex_size * 0 + 6];
1100 if(patch->mem[i + 5] == 0.0f)
1101 patch->mem[i + 5] = -feedbuffer[j + out_vertex_size * 0 + 7];
1102 normalize_normal(patch->mem + i + 3);
1103 i += d3d_out_vertex_size;
1107 glDisable(GL_MAP2_VERTEX_3);
1108 glDisable(GL_AUTO_NORMAL);
1109 glDisable(GL_MAP2_NORMAL);
1110 glDisable(GL_MAP2_TEXTURE_COORD_4);
1111 checkGLcall("glDisable vertex attrib generation");
1114 context_release(context);
1116 HeapFree(GetProcessHeap(), 0, feedbuffer);
1118 vtxStride = 3 * sizeof(float);
1119 if(patch->has_normals) {
1120 vtxStride += 3 * sizeof(float);
1122 if(patch->has_texcoords) {
1123 vtxStride += 4 * sizeof(float);
1125 memset(&patch->strided, 0, sizeof(patch->strided));
1126 patch->strided.position.format = WINED3DFMT_R32G32B32_FLOAT;
1127 patch->strided.position.lpData = (BYTE *) patch->mem;
1128 patch->strided.position.dwStride = vtxStride;
1130 if(patch->has_normals) {
1131 patch->strided.normal.format = WINED3DFMT_R32G32B32_FLOAT;
1132 patch->strided.normal.lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1133 patch->strided.normal.dwStride = vtxStride;
1135 if(patch->has_texcoords) {
1136 patch->strided.texCoords[0].format = WINED3DFMT_R32G32B32A32_FLOAT;
1137 patch->strided.texCoords[0].lpData = (BYTE *) patch->mem + 3 * sizeof(float) /* pos */;
1138 if(patch->has_normals) {
1139 patch->strided.texCoords[0].lpData += 3 * sizeof(float);
1141 patch->strided.texCoords[0].dwStride = vtxStride;