wined3d: Properly set the fog frag coord according to the FOGTABLEMODE.
[wine] / dlls / wined3d / arb_program_shader.c
1 /*
2  * Pixel and vertex shaders implementation using ARB_vertex_program
3  * and ARB_fragment_program GL extensions.
4  *
5  * Copyright 2002-2003 Jason Edmeades
6  * Copyright 2002-2003 Raphael Junqueira
7  * Copyright 2004 Christian Costa
8  * Copyright 2005 Oliver Stieber
9  * Copyright 2006 Ivan Gyurdiev
10  * Copyright 2006 Jason Green
11  * Copyright 2006 Henri Verbeet
12  * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13  *
14  * This library is free software; you can redistribute it and/or
15  * modify it under the terms of the GNU Lesser General Public
16  * License as published by the Free Software Foundation; either
17  * version 2.1 of the License, or (at your option) any later version.
18  *
19  * This library is distributed in the hope that it will be useful,
20  * but WITHOUT ANY WARRANTY; without even the implied warranty of
21  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
22  * Lesser General Public License for more details.
23  *
24  * You should have received a copy of the GNU Lesser General Public
25  * License along with this library; if not, write to the Free Software
26  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
27  */
28
29 #include "config.h"
30
31 #include <math.h>
32 #include <stdio.h>
33
34 #include "wined3d_private.h"
35
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d);
40
41 #define GLINFO_LOCATION      (*gl_info)
42
43 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
44  * so upload them above that
45  */
46 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
47 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
48
49 /* ARB_program_shader private data */
50 struct shader_arb_priv {
51     GLuint                  current_vprogram_id;
52     GLuint                  current_fprogram_id;
53     GLuint                  depth_blt_vprogram_id;
54     GLuint                  depth_blt_fprogram_id[tex_type_count];
55     BOOL                    use_arbfp_fixed_func;
56     struct hash_table_t     *fragment_shaders;
57 };
58
59 /********************************************************
60  * ARB_[vertex/fragment]_program helper functions follow
61  ********************************************************/
62
63 /** 
64  * Loads floating point constants into the currently set ARB_vertex/fragment_program.
65  * When constant_list == NULL, it will load all the constants.
66  *  
67  * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
68  *  or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
69  */
70 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, const WineD3D_GL_Info *gl_info,
71         GLuint target_type, unsigned int max_constants, const float *constants, char *dirty_consts)
72 {
73     DWORD shader_version = This->baseShader.reg_maps.shader_version;
74     local_constant* lconst;
75     DWORD i, j;
76     unsigned int ret;
77
78     if (TRACE_ON(d3d_shader)) {
79         for(i = 0; i < max_constants; i++) {
80             if(!dirty_consts[i]) continue;
81             TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
82                         constants[i * 4 + 0], constants[i * 4 + 1],
83                         constants[i * 4 + 2], constants[i * 4 + 3]);
84         }
85     }
86     /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
87     if (target_type == GL_FRAGMENT_PROGRAM_ARB && WINED3DSHADER_VERSION_MAJOR(shader_version) == 1)
88     {
89         float lcl_const[4];
90         for(i = 0; i < max_constants; i++) {
91             if(!dirty_consts[i]) continue;
92             dirty_consts[i] = 0;
93
94             j = 4 * i;
95             if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
96             else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
97             else lcl_const[0] = constants[j + 0];
98
99             if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
100             else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
101             else lcl_const[1] = constants[j + 1];
102
103             if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
104             else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
105             else lcl_const[2] = constants[j + 2];
106
107             if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
108             else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
109             else lcl_const[3] = constants[j + 3];
110
111             GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
112         }
113     } else {
114         if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
115             /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
116              * or just reloading *all* constants at once
117              *
118             GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
119              */
120             for(i = 0; i < max_constants; i++) {
121                 if(!dirty_consts[i]) continue;
122
123                 /* Find the next block of dirty constants */
124                 dirty_consts[i] = 0;
125                 j = i;
126                 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
127                     dirty_consts[i] = 0;
128                 }
129
130                 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
131             }
132         } else {
133             for(i = 0; i < max_constants; i++) {
134                 if(dirty_consts[i]) {
135                     dirty_consts[i] = 0;
136                     GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
137                 }
138             }
139         }
140     }
141     checkGLcall("glProgramEnvParameter4fvARB()");
142
143     /* Load immediate constants */
144     if(This->baseShader.load_local_constsF) {
145         if (TRACE_ON(d3d_shader)) {
146             LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
147                 GLfloat* values = (GLfloat*)lconst->value;
148                 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
149                         values[0], values[1], values[2], values[3]);
150             }
151         }
152         /* Immediate constants are clamped for 1.X shaders at loading times */
153         ret = 0;
154         LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
155             dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
156             ret = max(ret, lconst->idx + 1);
157             GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
158         }
159         checkGLcall("glProgramEnvParameter4fvARB()");
160         return ret; /* The loaded immediate constants need reloading for the next shader */
161     } else {
162         return 0; /* No constants are dirty now */
163     }
164 }
165
166 /**
167  * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
168  * 
169  * We only support float constants in ARB at the moment, so don't 
170  * worry about the Integers or Booleans
171  */
172 static void shader_arb_load_constants(
173     IWineD3DDevice* device,
174     char usePixelShader,
175     char useVertexShader) {
176    
177     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device; 
178     IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
179     const WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
180     unsigned char i;
181
182     if (useVertexShader) {
183         IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
184
185         /* Load DirectX 9 float constants for vertex shader */
186         deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
187                 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
188                 deviceImpl->highest_dirty_vs_const,
189                 stateBlock->vertexShaderConstantF,
190                 deviceImpl->activeContext->vshader_const_dirty);
191
192         /* Upload the position fixup */
193         GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
194     }
195
196     if (usePixelShader) {
197
198         IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
199         IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
200
201         /* Load DirectX 9 float constants for pixel shader */
202         deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
203                 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
204                 deviceImpl->highest_dirty_ps_const,
205                 stateBlock->pixelShaderConstantF,
206                 deviceImpl->activeContext->pshader_const_dirty);
207
208         for(i = 0; i < psi->numbumpenvmatconsts; i++) {
209             /* The state manager takes care that this function is always called if the bump env matrix changes
210              */
211             const float *data = (const float *)&stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
212             GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
213             deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
214
215             if(psi->luminanceconst[i].const_num != -1) {
216                 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
217                  * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
218                  * don't care about them. The pointers are valid for sure because the stateblock is bigger.
219                  * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
220                  */
221                 const float *scale = (const float *)&stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
222                 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
223                 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
224             }
225         }
226     }
227 }
228
229 static void shader_arb_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count)
230 {
231     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
232
233     /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
234      * context. On a context switch the old context will be fully dirtified */
235     memset(This->activeContext->vshader_const_dirty + start, 1,
236             sizeof(*This->activeContext->vshader_const_dirty) * count);
237     This->highest_dirty_vs_const = max(This->highest_dirty_vs_const, start + count + 1);
238 }
239
240 static void shader_arb_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count)
241 {
242     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
243
244     /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
245      * context. On a context switch the old context will be fully dirtified */
246     memset(This->activeContext->pshader_const_dirty + start, 1,
247             sizeof(*This->activeContext->pshader_const_dirty) * count);
248     This->highest_dirty_ps_const = max(This->highest_dirty_ps_const, start + count + 1);
249 }
250
251 /* Generate the variable & register declarations for the ARB_vertex_program output target */
252 static void shader_generate_arb_declarations(IWineD3DBaseShader *iface, const shader_reg_maps *reg_maps,
253         SHADER_BUFFER *buffer, const WineD3D_GL_Info *gl_info)
254 {
255     IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
256     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
257     DWORD i, cur;
258     char pshader = shader_is_pshader_version(reg_maps->shader_version);
259     unsigned max_constantsF = min(This->baseShader.limits.constant_float, 
260             (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
261     UINT extra_constants_needed = 0;
262     const local_constant *lconst;
263
264     /* Temporary Output register */
265     shader_addline(buffer, "TEMP TMP_OUT;\n");
266
267     for(i = 0; i < This->baseShader.limits.temporary; i++) {
268         if (reg_maps->temporary[i])
269             shader_addline(buffer, "TEMP R%u;\n", i);
270     }
271
272     for (i = 0; i < This->baseShader.limits.address; i++) {
273         if (reg_maps->address[i])
274             shader_addline(buffer, "ADDRESS A%d;\n", i);
275     }
276
277     for(i = 0; i < This->baseShader.limits.texcoord; i++) {
278         if (reg_maps->texcoord[i])
279             shader_addline(buffer,"TEMP T%u;\n", i);
280     }
281
282     /* Texture coordinate registers must be pre-loaded */
283     for (i = 0; i < This->baseShader.limits.texcoord; i++) {
284         if (reg_maps->texcoord[i])
285             shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
286     }
287
288     for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
289         IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
290         if(!reg_maps->bumpmat[i]) continue;
291
292         cur = ps->numbumpenvmatconsts;
293         ps->bumpenvmatconst[cur].const_num = -1;
294         ps->bumpenvmatconst[cur].texunit = i;
295         ps->luminanceconst[cur].const_num = -1;
296         ps->luminanceconst[cur].texunit = i;
297
298         /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
299          * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
300          * bump mapping.
301          */
302         if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
303             ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
304             shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
305                            i, ps->bumpenvmatconst[cur].const_num);
306             extra_constants_needed++;
307
308             if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
309                 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
310                 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
311                                i, ps->luminanceconst[cur].const_num);
312                 extra_constants_needed++;
313             } else if(reg_maps->luminanceparams) {
314                 FIXME("No free constant to load the luminance parameters\n");
315             }
316         } else {
317             FIXME("No free constant found to load environment bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
318         }
319
320         ps->numbumpenvmatconsts = cur + 1;
321     }
322
323     if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
324         shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
325                         srgb_mul_low, srgb_mul_low, srgb_mul_low);
326         shader_addline(buffer, "PARAM srgb_comparison =  {%f, %f, %f, %f};\n",
327                         srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
328         shader_addline(buffer, "PARAM srgb_pow =  {%f, %f, %f, 1.0};\n",
329                        srgb_pow, srgb_pow, srgb_pow);
330         shader_addline(buffer, "PARAM srgb_mul_hi =  {%f, %f, %f, 1.0};\n",
331                        srgb_mul_high, srgb_mul_high, srgb_mul_high);
332         shader_addline(buffer, "PARAM srgb_sub_hi =  {%f, %f, %f, 0.0};\n",
333                        srgb_sub_high, srgb_sub_high, srgb_sub_high);
334     }
335
336     /* Load local constants using the program-local space,
337      * this avoids reloading them each time the shader is used
338      */
339     if(!This->baseShader.load_local_constsF) {
340         LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
341             shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
342                            lconst->idx);
343         }
344     }
345
346     /* we use the array-based constants array if the local constants are marked for loading,
347      * because then we use indirect addressing, or when the local constant list is empty,
348      * because then we don't know if we're using indirect addressing or not. If we're hardcoding
349      * local constants do not declare the loaded constants as an array because ARB compilers usually
350      * do not optimize unused constants away
351      */
352     if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
353         /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
354         shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
355                     max_constantsF, max_constantsF - 1);
356     } else {
357         for(i = 0; i < max_constantsF; i++) {
358             if(!shader_constant_is_local(This, i)) {
359                 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
360             }
361         }
362     }
363 }
364
365 static const char * const shift_tab[] = {
366     "dummy",     /*  0 (none) */
367     "coefmul.x", /*  1 (x2)   */
368     "coefmul.y", /*  2 (x4)   */
369     "coefmul.z", /*  3 (x8)   */
370     "coefmul.w", /*  4 (x16)  */
371     "dummy",     /*  5 (x32)  */
372     "dummy",     /*  6 (x64)  */
373     "dummy",     /*  7 (x128) */
374     "dummy",     /*  8 (d256) */
375     "dummy",     /*  9 (d128) */
376     "dummy",     /* 10 (d64)  */
377     "dummy",     /* 11 (d32)  */
378     "coefdiv.w", /* 12 (d16)  */
379     "coefdiv.z", /* 13 (d8)   */
380     "coefdiv.y", /* 14 (d4)   */
381     "coefdiv.x"  /* 15 (d2)   */
382 };
383
384 static void shader_arb_get_write_mask(const SHADER_OPCODE_ARG *arg, const DWORD param, char *write_mask)
385 {
386     char *ptr = write_mask;
387     char vshader = shader_is_vshader_version(arg->reg_maps->shader_version);
388
389     if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
390         *ptr++ = '.';
391         *ptr++ = 'x';
392     } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
393         *ptr++ = '.';
394         if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
395         if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
396         if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
397         if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
398     }
399
400     *ptr = '\0';
401 }
402
403 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
404     /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
405      * but addressed as "rgba". To fix this we need to swap the register's x
406      * and z components. */
407     const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
408     char *ptr = swizzle_str;
409
410     /* swizzle bits fields: wwzzyyxx */
411     DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
412     DWORD swizzle_x = swizzle & 0x03;
413     DWORD swizzle_y = (swizzle >> 2) & 0x03;
414     DWORD swizzle_z = (swizzle >> 4) & 0x03;
415     DWORD swizzle_w = (swizzle >> 6) & 0x03;
416
417     /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
418      * generate a swizzle string. Unless we need to our own swizzling. */
419     if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
420         *ptr++ = '.';
421         if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
422             *ptr++ = swizzle_chars[swizzle_x];
423         } else {
424             *ptr++ = swizzle_chars[swizzle_x];
425             *ptr++ = swizzle_chars[swizzle_y];
426             *ptr++ = swizzle_chars[swizzle_z];
427             *ptr++ = swizzle_chars[swizzle_w];
428         }
429     }
430
431     *ptr = '\0';
432 }
433
434 static void pshader_get_register_name(IWineD3DBaseShader* iface,
435     const DWORD param, char* regstr) {
436
437     DWORD reg = param & WINED3DSP_REGNUM_MASK;
438     DWORD regtype = shader_get_regtype(param);
439     IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
440
441     switch (regtype) {
442     case WINED3DSPR_TEMP:
443         sprintf(regstr, "R%u", reg);
444     break;
445     case WINED3DSPR_INPUT:
446         if (reg==0) {
447             strcpy(regstr, "fragment.color.primary");
448         } else {
449             strcpy(regstr, "fragment.color.secondary");
450         }
451     break;
452     case WINED3DSPR_CONST:
453         if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
454             sprintf(regstr, "C[%u]", reg);
455         } else {
456             sprintf(regstr, "C%u", reg);
457         }
458     break;
459     case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
460         sprintf(regstr,"T%u", reg);
461     break;
462     case WINED3DSPR_COLOROUT:
463         if (reg == 0)
464             sprintf(regstr, "TMP_COLOR");
465         else {
466             /* TODO: See GL_ARB_draw_buffers */
467             FIXME("Unsupported write to render target %u\n", reg);
468             sprintf(regstr, "unsupported_register");
469         }
470     break;
471     case WINED3DSPR_DEPTHOUT:
472         sprintf(regstr, "result.depth");
473     break;
474     case WINED3DSPR_ATTROUT:
475         sprintf(regstr, "oD[%u]", reg);
476     break;
477     case WINED3DSPR_TEXCRDOUT:
478         sprintf(regstr, "oT[%u]", reg);
479     break;
480     default:
481         FIXME("Unhandled register name Type(%d)\n", regtype);
482         sprintf(regstr, "unrecognized_register");
483     break;
484     }
485 }
486
487 /* TODO: merge with pixel shader */
488 static void vshader_program_add_param(const SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine)
489 {
490   IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
491
492   /* oPos, oFog and oPts in D3D */
493   static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
494
495   DWORD reg = param & WINED3DSP_REGNUM_MASK;
496   DWORD regtype = shader_get_regtype(param);
497   char  tmpReg[255];
498   BOOL is_color = FALSE;
499
500   if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
501       strcat(hwLine, " -");
502   } else {
503       strcat(hwLine, " ");
504   }
505
506   switch (regtype) {
507   case WINED3DSPR_TEMP:
508     sprintf(tmpReg, "R%u", reg);
509     strcat(hwLine, tmpReg);
510     break;
511   case WINED3DSPR_INPUT:
512
513     if (This->cur_args->swizzle_map & (1 << reg)) is_color = TRUE;
514
515     sprintf(tmpReg, "vertex.attrib[%u]", reg);
516     strcat(hwLine, tmpReg);
517     break;
518   case WINED3DSPR_CONST:
519       if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
520           if(reg >= This->rel_offset) {
521               sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
522           } else {
523               sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
524           }
525       } else {
526           if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
527               sprintf(tmpReg, "C[%u]", reg);
528           } else {
529               sprintf(tmpReg, "C%u", reg);
530           }
531       }
532     strcat(hwLine, tmpReg);
533     break;
534   case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
535     sprintf(tmpReg, "A%u", reg);
536     strcat(hwLine, tmpReg);
537     break;
538   case WINED3DSPR_RASTOUT:
539     sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
540     strcat(hwLine, tmpReg);
541     break;
542   case WINED3DSPR_ATTROUT:
543     if (reg==0) {
544        strcat(hwLine, "result.color.primary");
545     } else {
546        strcat(hwLine, "result.color.secondary");
547     }
548     break;
549   case WINED3DSPR_TEXCRDOUT:
550     sprintf(tmpReg, "result.texcoord[%u]", reg);
551     strcat(hwLine, tmpReg);
552     break;
553   default:
554     FIXME("Unknown reg type %d %d\n", regtype, reg);
555     strcat(hwLine, "unrecognized_register");
556     break;
557   }
558
559   if (!is_input) {
560     char write_mask[6];
561     shader_arb_get_write_mask(arg, param, write_mask);
562     strcat(hwLine, write_mask);
563   } else {
564     char swizzle[6];
565     shader_arb_get_swizzle(param, is_color, swizzle);
566     strcat(hwLine, swizzle);
567   }
568 }
569
570 static void shader_hw_sample(const SHADER_OPCODE_ARG *arg, DWORD sampler_idx, const char *dst_str,
571         const char *coord_reg, BOOL projected, BOOL bias)
572 {
573     SHADER_BUFFER* buffer = arg->buffer;
574     DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
575     const char *tex_type;
576
577     switch(sampler_type) {
578         case WINED3DSTT_1D:
579             tex_type = "1D";
580             break;
581
582         case WINED3DSTT_2D:
583         {
584             IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
585             IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
586             if(device->stateBlock->textures[sampler_idx] &&
587                IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
588                 tex_type = "RECT";
589             } else {
590                 tex_type = "2D";
591             }
592             break;
593         }
594
595         case WINED3DSTT_VOLUME:
596             tex_type = "3D";
597             break;
598
599         case WINED3DSTT_CUBE:
600             tex_type = "CUBE";
601             break;
602
603         default:
604             ERR("Unexpected texture type %d\n", sampler_type);
605             tex_type = "";
606     }
607
608     if (bias) {
609         /* Shouldn't be possible, but let's check for it */
610         if(projected) FIXME("Biased and Projected texture sampling\n");
611         /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
612         shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
613     } else if (projected) {
614         shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
615     } else {
616         shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
617     }
618 }
619
620 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source)
621 {
622     switch(channel_source)
623     {
624         case CHANNEL_SOURCE_ZERO: return "0";
625         case CHANNEL_SOURCE_ONE: return "1";
626         case CHANNEL_SOURCE_X: return "x";
627         case CHANNEL_SOURCE_Y: return "y";
628         case CHANNEL_SOURCE_Z: return "z";
629         case CHANNEL_SOURCE_W: return "w";
630         default:
631             FIXME("Unhandled channel source %#x\n", channel_source);
632             return "undefined";
633     }
634 }
635
636 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, DWORD dst_mask,
637         const char *one, const char *two, struct color_fixup_desc fixup)
638 {
639     DWORD mask;
640
641     if (is_yuv_fixup(fixup))
642     {
643         enum yuv_fixup yuv_fixup = get_yuv_fixup(fixup);
644         FIXME("YUV fixup (%#x) not supported\n", yuv_fixup);
645         return;
646     }
647
648     mask = 0;
649     if (fixup.x_source != CHANNEL_SOURCE_X) mask |= WINED3DSP_WRITEMASK_0;
650     if (fixup.y_source != CHANNEL_SOURCE_Y) mask |= WINED3DSP_WRITEMASK_1;
651     if (fixup.z_source != CHANNEL_SOURCE_Z) mask |= WINED3DSP_WRITEMASK_2;
652     if (fixup.w_source != CHANNEL_SOURCE_W) mask |= WINED3DSP_WRITEMASK_3;
653     mask &= dst_mask;
654
655     if (mask)
656     {
657         shader_addline(buffer, "SWZ %s, %s, %s, %s, %s, %s;\n", reg, reg,
658                 shader_arb_get_fixup_swizzle(fixup.x_source), shader_arb_get_fixup_swizzle(fixup.y_source),
659                 shader_arb_get_fixup_swizzle(fixup.z_source), shader_arb_get_fixup_swizzle(fixup.w_source));
660     }
661
662     mask = 0;
663     if (fixup.x_sign_fixup) mask |= WINED3DSP_WRITEMASK_0;
664     if (fixup.y_sign_fixup) mask |= WINED3DSP_WRITEMASK_1;
665     if (fixup.z_sign_fixup) mask |= WINED3DSP_WRITEMASK_2;
666     if (fixup.w_sign_fixup) mask |= WINED3DSP_WRITEMASK_3;
667     mask &= dst_mask;
668
669     if (mask)
670     {
671         char reg_mask[6];
672         char *ptr = reg_mask;
673
674         if (mask != WINED3DSP_WRITEMASK_ALL)
675         {
676             *ptr++ = '.';
677             if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
678             if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
679             if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
680             if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
681         }
682         *ptr = '\0';
683
684         shader_addline(buffer, "MAD %s%s, %s, %s, -%s;\n", reg, reg_mask, reg, two, one);
685     }
686 }
687
688 static void shader_arb_color_correction(const struct SHADER_OPCODE_ARG* arg, struct color_fixup_desc fixup)
689 {
690     char reg[256];
691     pshader_get_register_name(arg->shader, arg->dst, reg);
692     gen_color_correction(arg->buffer, reg, arg->dst & WINED3DSP_WRITEMASK_ALL, "one", "coefmul.x", fixup);
693 }
694
695 static void pshader_gen_input_modifier_line (
696     IWineD3DBaseShader *iface,
697     SHADER_BUFFER* buffer,
698     const DWORD instr,
699     int tmpreg,
700     char *outregstr) {
701
702     /* Generate a line that does the input modifier computation and return the input register to use */
703     char regstr[256];
704     char swzstr[20];
705     int insert_line;
706
707     /* Assume a new line will be added */
708     insert_line = 1;
709
710     /* Get register name */
711     pshader_get_register_name(iface, instr, regstr);
712     shader_arb_get_swizzle(instr, FALSE, swzstr);
713
714     switch (instr & WINED3DSP_SRCMOD_MASK) {
715     case WINED3DSPSM_NONE:
716         sprintf(outregstr, "%s%s", regstr, swzstr);
717         insert_line = 0;
718         break;
719     case WINED3DSPSM_NEG:
720         sprintf(outregstr, "-%s%s", regstr, swzstr);
721         insert_line = 0;
722         break;
723     case WINED3DSPSM_BIAS:
724         shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
725         break;
726     case WINED3DSPSM_BIASNEG:
727         shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
728         break;
729     case WINED3DSPSM_SIGN:
730         shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
731         break;
732     case WINED3DSPSM_SIGNNEG:
733         shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
734         break;
735     case WINED3DSPSM_COMP:
736         shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
737         break;
738     case WINED3DSPSM_X2:
739         shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
740         break;
741     case WINED3DSPSM_X2NEG:
742         shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
743         break;
744     case WINED3DSPSM_DZ:
745         shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
746         shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
747         break;
748     case WINED3DSPSM_DW:
749         shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
750         shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
751         break;
752     default:
753         sprintf(outregstr, "%s%s", regstr, swzstr);
754         insert_line = 0;
755     }
756
757     /* Return modified or original register, with swizzle */
758     if (insert_line)
759         sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
760 }
761
762 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER *buffer, int saturate, const char *write_mask,
763         int shift, const char *regstr)
764 {
765     /* Generate a line that does the output modifier computation */
766     shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
767         regstr, write_mask, regstr, shift_tab[shift]);
768 }
769
770 static void pshader_hw_bem(const SHADER_OPCODE_ARG *arg)
771 {
772     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
773
774     SHADER_BUFFER* buffer = arg->buffer;
775     char dst_name[50];
776     char src_name[2][50];
777     char dst_wmask[20];
778     DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
779     BOOL has_bumpmat = FALSE;
780     int i;
781
782     for(i = 0; i < This->numbumpenvmatconsts; i++) {
783         if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
784             has_bumpmat = TRUE;
785             break;
786         }
787     }
788
789     pshader_get_register_name(arg->shader, arg->dst, dst_name);
790     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
791     strcat(dst_name, dst_wmask);
792
793     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
794     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
795
796     if(has_bumpmat) {
797         /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
798         shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
799         shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
800         shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
801         shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
802
803         shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
804     } else {
805         shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
806     }
807 }
808
809 static void pshader_hw_cnd(const SHADER_OPCODE_ARG *arg)
810 {
811     SHADER_BUFFER* buffer = arg->buffer;
812     char dst_wmask[20];
813     char dst_name[50];
814     char src_name[3][50];
815     BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
816     DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
817
818     /* FIXME: support output modifiers */
819
820     /* Handle output register */
821     pshader_get_register_name(arg->shader, arg->dst, dst_name);
822     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
823
824     /* Generate input register names (with modifiers) */
825     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
826     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
827     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
828
829     /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
830     if (arg->reg_maps->shader_version <= WINED3DPS_VERSION(1, 3) && arg->opcode_token & WINED3DSI_COISSUE)
831     {
832         shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
833     } else {
834         shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
835         shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
836                                 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
837     }
838     if (shift != 0)
839         pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
840 }
841
842 static void pshader_hw_cmp(const SHADER_OPCODE_ARG *arg)
843 {
844     SHADER_BUFFER* buffer = arg->buffer;
845     char dst_wmask[20];
846     char dst_name[50];
847     char src_name[3][50];
848     DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
849     BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
850
851     /* FIXME: support output modifiers */
852
853     /* Handle output register */
854     pshader_get_register_name(arg->shader, arg->dst, dst_name);
855     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
856
857     /* Generate input register names (with modifiers) */
858     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
859     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
860     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
861
862     shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
863                    src_name[0], src_name[2], src_name[1]);
864
865     if (shift != 0)
866         pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
867 }
868
869 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
870  * dst = dot2(src0, src1) + src2 */
871 static void pshader_hw_dp2add(const SHADER_OPCODE_ARG *arg)
872 {
873     SHADER_BUFFER* buffer = arg->buffer;
874     char dst_wmask[20];
875     char dst_name[50];
876     char src_name[3][50];
877     DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
878     BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
879
880     pshader_get_register_name(arg->shader, arg->dst, dst_name);
881     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
882
883     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
884     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
885     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
886
887     /* Emulate a DP2 with a DP3 and 0.0 */
888     shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
889     shader_addline(buffer, "MOV TMP.z, 0.0;\n");
890     shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
891     shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
892
893     if (shift != 0)
894         pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
895 }
896
897 /* Map the opcode 1-to-1 to the GL code */
898 static void shader_hw_map2gl(const SHADER_OPCODE_ARG *arg)
899 {
900     CONST SHADER_OPCODE* curOpcode = arg->opcode;
901     SHADER_BUFFER* buffer = arg->buffer;
902     DWORD dst = arg->dst;
903     const DWORD *src = arg->src;
904     const char *instruction;
905     char arguments[256];
906     unsigned int i;
907
908     switch (curOpcode->opcode)
909     {
910         case WINED3DSIO_ABS: instruction = "ABS"; break;
911         case WINED3DSIO_ADD: instruction = "ADD"; break;
912         case WINED3DSIO_CRS: instruction = "XPD"; break;
913         case WINED3DSIO_DP3: instruction = "DP3"; break;
914         case WINED3DSIO_DP4: instruction = "DP4"; break;
915         case WINED3DSIO_DST: instruction = "DST"; break;
916         case WINED3DSIO_EXP: instruction = "EX2"; break;
917         case WINED3DSIO_EXPP: instruction = "EXP"; break;
918         case WINED3DSIO_FRC: instruction = "FRC"; break;
919         case WINED3DSIO_LIT: instruction = "LIT"; break;
920         case WINED3DSIO_LOG: instruction = "LG2"; break;
921         case WINED3DSIO_LOGP: instruction = "LOG"; break;
922         case WINED3DSIO_LRP: instruction = "LRP"; break;
923         case WINED3DSIO_MAD: instruction = "MAD"; break;
924         case WINED3DSIO_MAX: instruction = "MAX"; break;
925         case WINED3DSIO_MIN: instruction = "MIN"; break;
926         case WINED3DSIO_MOV: instruction = "MOV"; break;
927         case WINED3DSIO_MUL: instruction = "MUL"; break;
928         case WINED3DSIO_NOP: instruction = "NOP"; break;
929         case WINED3DSIO_POW: instruction = "POW"; break;
930         case WINED3DSIO_SGE: instruction = "SGE"; break;
931         case WINED3DSIO_SLT: instruction = "SLT"; break;
932         case WINED3DSIO_SUB: instruction = "SUB"; break;
933         default: instruction = "";
934             FIXME("Unhandled opcode %s\n", curOpcode->name);
935             break;
936     }
937
938     if (shader_is_pshader_version(arg->reg_maps->shader_version))
939     {
940         /* Output token related */
941         char output_rname[256];
942         char output_wmask[20];
943         char operands[4][100];
944         BOOL saturate = FALSE;
945         BOOL centroid = FALSE;
946         BOOL partialprecision = FALSE;
947         const char *modifier;
948         DWORD shift;
949
950         if (!curOpcode->num_params)
951         {
952             ERR("Opcode \"%s\" has no parameters\n", curOpcode->name);
953             return;
954         }
955
956         /* Process modifiers */
957         if (dst & WINED3DSP_DSTMOD_MASK)
958         {
959             DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
960
961             saturate = mask & WINED3DSPDM_SATURATE;
962             centroid = mask & WINED3DSPDM_MSAMPCENTROID;
963             partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
964             mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
965             if (mask)
966                 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
967
968             if (centroid)
969                 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
970         }
971         shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
972         modifier = (saturate && !shift) ? "_SAT" : "";
973
974         /* Generate input register names (with modifiers) */
975         for (i = 1; i < curOpcode->num_params; ++i)
976             pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
977
978         /* Handle output register */
979         pshader_get_register_name(arg->shader, dst, output_rname);
980         strcpy(operands[0], output_rname);
981         shader_arb_get_write_mask(arg, dst, output_wmask);
982         strcat(operands[0], output_wmask);
983
984         arguments[0] = '\0';
985         strcat(arguments, operands[0]);
986         for (i = 1; i < curOpcode->num_params; i++)
987         {
988             strcat(arguments, ", ");
989             strcat(arguments, operands[i]);
990         }
991         shader_addline(buffer, "%s%s %s;\n", instruction, modifier, arguments);
992
993         /* A shift requires another line. */
994         if (shift) pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
995     } else {
996         /* Note that vshader_program_add_param() adds spaces. */
997
998         arguments[0] = '\0';
999         if (curOpcode->num_params > 0)
1000         {
1001             vshader_program_add_param(arg, dst, FALSE, arguments);
1002             for (i = 1; i < curOpcode->num_params; ++i)
1003             {
1004                 strcat(arguments, ",");
1005                 vshader_program_add_param(arg, src[i-1], TRUE, arguments);
1006             }
1007         }
1008         shader_addline(buffer, "%s%s;\n", instruction, arguments);
1009     }
1010 }
1011
1012 static void shader_hw_mov(const SHADER_OPCODE_ARG *arg)
1013 {
1014     IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1015
1016     if ((WINED3DSHADER_VERSION_MAJOR(arg->reg_maps->shader_version) == 1
1017             && !shader_is_pshader_version(arg->reg_maps->shader_version)
1018             && shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)
1019             || arg->opcode->opcode == WINED3DSIO_MOVA)
1020     {
1021         SHADER_BUFFER *buffer = arg->buffer;
1022         char src0_param[256];
1023
1024         if (arg->opcode->opcode == WINED3DSIO_MOVA)
1025             FIXME("mova should round\n");
1026
1027         src0_param[0] = '\0';
1028         if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
1029         {
1030             vshader_program_add_param(arg, arg->src[0], TRUE, src0_param);
1031             shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", src0_param);
1032             shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1033         }
1034         else
1035         {
1036             /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1037              * with more than one component. Thus replicate the first source argument over all
1038              * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1039              */
1040             DWORD parm = arg->src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1041             if((arg->src[0] & WINED3DVS_X_W) == WINED3DVS_X_W)
1042                 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1043             else if((arg->src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z)
1044                 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1045             else if((arg->src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y)
1046                 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1047             else if((arg->src[0] & WINED3DVS_X_X) == WINED3DVS_X_X)
1048                 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1049             vshader_program_add_param(arg, parm, TRUE, src0_param);
1050             shader_addline(buffer, "ARL A0.x, %s;\n", src0_param);
1051         }
1052     }
1053     else
1054     {
1055         shader_hw_map2gl(arg);
1056     }
1057 }
1058
1059 static void pshader_hw_texkill(const SHADER_OPCODE_ARG *arg)
1060 {
1061     DWORD shader_version = arg->reg_maps->shader_version;
1062     SHADER_BUFFER* buffer = arg->buffer;
1063     char reg_dest[40];
1064
1065     /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1066      * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1067      */
1068     pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1069
1070     if (shader_version >= WINED3DPS_VERSION(2,0))
1071     {
1072         /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1073         shader_addline(buffer, "KIL %s;\n", reg_dest);
1074     } else {
1075         /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1076          * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1077          */
1078         shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1079         shader_addline(buffer, "MOV TMP.w, one.w;\n");
1080         shader_addline(buffer, "KIL TMP;\n");
1081     }
1082 }
1083
1084 static void pshader_hw_tex(const SHADER_OPCODE_ARG *arg)
1085 {
1086     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1087     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1088
1089     DWORD dst = arg->dst;
1090     const DWORD *src = arg->src;
1091     SHADER_BUFFER* buffer = arg->buffer;
1092     DWORD shader_version = arg->reg_maps->shader_version;
1093     BOOL projected = FALSE, bias = FALSE;
1094
1095     char reg_dest[40];
1096     char reg_coord[40];
1097     DWORD reg_dest_code;
1098     DWORD reg_sampler_code;
1099
1100     /* All versions have a destination register */
1101     reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1102     pshader_get_register_name(arg->shader, dst, reg_dest);
1103
1104     /* 1.0-1.3: Use destination register as coordinate source.
1105        1.4+: Use provided coordinate source register. */
1106    if (shader_version < WINED3DPS_VERSION(1,4))
1107       strcpy(reg_coord, reg_dest);
1108    else
1109       pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1110
1111   /* 1.0-1.4: Use destination register number as texture code.
1112      2.0+: Use provided sampler number as texure code. */
1113   if (shader_version < WINED3DPS_VERSION(2,0))
1114      reg_sampler_code = reg_dest_code;
1115   else
1116      reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1117
1118   /* projection flag:
1119    * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1120    * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1121    * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1122    */
1123   if (shader_version < WINED3DPS_VERSION(1,4))
1124   {
1125       DWORD flags = 0;
1126       if(reg_sampler_code < MAX_TEXTURES) {
1127         flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1128       }
1129       if (flags & WINED3DTTFF_PROJECTED) {
1130           projected = TRUE;
1131       }
1132   }
1133   else if (shader_version < WINED3DPS_VERSION(2,0))
1134   {
1135       DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1136       if (src_mod == WINED3DSPSM_DZ) {
1137           projected = TRUE;
1138       } else if(src_mod == WINED3DSPSM_DW) {
1139           projected = TRUE;
1140       }
1141   } else {
1142       if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1143           projected = TRUE;
1144       }
1145       if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1146           bias = TRUE;
1147       }
1148   }
1149   shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1150 }
1151
1152 static void pshader_hw_texcoord(const SHADER_OPCODE_ARG *arg)
1153 {
1154     DWORD dst = arg->dst;
1155     SHADER_BUFFER* buffer = arg->buffer;
1156
1157     char tmp[20];
1158     shader_arb_get_write_mask(arg, dst, tmp);
1159     if (arg->reg_maps->shader_version != WINED3DPS_VERSION(1,4))
1160     {
1161         DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1162         shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1163     } else {
1164         DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1165         char reg_src[40];
1166
1167         pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1168         shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1169    }
1170 }
1171
1172 static void pshader_hw_texreg2ar(const SHADER_OPCODE_ARG *arg)
1173 {
1174      SHADER_BUFFER* buffer = arg->buffer;
1175      IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1176      IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1177      DWORD flags;
1178
1179      DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1180      char dst_str[8];
1181      char src_str[50];
1182
1183      sprintf(dst_str, "T%u", reg1);
1184      pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1185      shader_addline(buffer, "MOV TMP.x, %s.w;\n", src_str);
1186      shader_addline(buffer, "MOV TMP.y, %s.x;\n", src_str);
1187      flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1188      shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1189 }
1190
1191 static void pshader_hw_texreg2gb(const SHADER_OPCODE_ARG *arg)
1192 {
1193      SHADER_BUFFER* buffer = arg->buffer;
1194
1195      DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1196      char dst_str[8];
1197      char src_str[50];
1198
1199      sprintf(dst_str, "T%u", reg1);
1200      pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1201      shader_addline(buffer, "MOV TMP.x, %s.y;\n", src_str);
1202      shader_addline(buffer, "MOV TMP.y, %s.z;\n", src_str);
1203      shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1204 }
1205
1206 static void pshader_hw_texreg2rgb(const SHADER_OPCODE_ARG *arg)
1207 {
1208     SHADER_BUFFER* buffer = arg->buffer;
1209     DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1210     char dst_str[8];
1211     char src_str[50];
1212
1213     sprintf(dst_str, "T%u", reg1);
1214     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1215     shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1216 }
1217
1218 static void pshader_hw_texbem(const SHADER_OPCODE_ARG *arg)
1219 {
1220     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1221     BOOL has_bumpmat = FALSE;
1222     BOOL has_luminance = FALSE;
1223     int i;
1224
1225     DWORD dst = arg->dst;
1226     DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1227     SHADER_BUFFER* buffer = arg->buffer;
1228
1229     char reg_coord[40];
1230     DWORD reg_dest_code;
1231
1232     /* All versions have a destination register */
1233     reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1234     /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1235     pshader_get_register_name(arg->shader, dst, reg_coord);
1236
1237     for(i = 0; i < This->numbumpenvmatconsts; i++) {
1238         if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1239             has_bumpmat = TRUE;
1240             break;
1241         }
1242     }
1243     for(i = 0; i < This->numbumpenvmatconsts; i++) {
1244         if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1245             has_luminance = TRUE;
1246             break;
1247         }
1248     }
1249
1250     if(has_bumpmat) {
1251         /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1252
1253         shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1254         shader_addline(buffer, "DP3 TMP.x, TMP2, T%u;\n", src);
1255         shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1256         shader_addline(buffer, "DP3 TMP.y, TMP2, T%u;\n", src);
1257
1258         /* with projective textures, texbem only divides the static texture coord, not the displacement,
1259          * so we can't let the GL handle this.
1260          */
1261         if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1262               & WINED3DTTFF_PROJECTED) {
1263             shader_addline(buffer, "RCP TMP2.w, %s.w;\n", reg_coord);
1264             shader_addline(buffer, "MUL TMP2.xy, %s, TMP2.w;\n", reg_coord);
1265             shader_addline(buffer, "ADD TMP.xy, TMP, TMP2;\n");
1266         } else {
1267             shader_addline(buffer, "ADD TMP.xy, TMP, %s;\n", reg_coord);
1268         }
1269
1270         shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1271
1272         if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1273             shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1274                            src, reg_dest_code, reg_dest_code);
1275             shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1276         }
1277
1278     } else {
1279         DWORD tf;
1280         if(reg_dest_code < MAX_TEXTURES) {
1281             tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1282         } else {
1283             tf = 0;
1284         }
1285         /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1286         shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1287     }
1288 }
1289
1290 static void pshader_hw_texm3x2pad(const SHADER_OPCODE_ARG *arg)
1291 {
1292     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1293     SHADER_BUFFER* buffer = arg->buffer;
1294     char src0_name[50];
1295
1296     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1297     shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1298 }
1299
1300 static void pshader_hw_texm3x2tex(const SHADER_OPCODE_ARG *arg)
1301 {
1302     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1303     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1304     DWORD flags;
1305     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1306     SHADER_BUFFER* buffer = arg->buffer;
1307     char dst_str[8];
1308     char src0_name[50];
1309
1310     sprintf(dst_str, "T%u", reg);
1311     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1312     shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1313     flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1314     shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1315 }
1316
1317 static void pshader_hw_texm3x3pad(const SHADER_OPCODE_ARG *arg)
1318 {
1319     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1320     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1321     SHADER_BUFFER* buffer = arg->buffer;
1322     SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1323     char src0_name[50];
1324
1325     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1326     shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1327     current_state->texcoord_w[current_state->current_row++] = reg;
1328 }
1329
1330 static void pshader_hw_texm3x3tex(const SHADER_OPCODE_ARG *arg)
1331 {
1332     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1333     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1334     DWORD flags;
1335     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1336     SHADER_BUFFER* buffer = arg->buffer;
1337     SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1338     char dst_str[8];
1339     char src0_name[50];
1340
1341     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1342     shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1343
1344     /* Sample the texture using the calculated coordinates */
1345     sprintf(dst_str, "T%u", reg);
1346     flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1347     shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1348     current_state->current_row = 0;
1349 }
1350
1351 static void pshader_hw_texm3x3vspec(const SHADER_OPCODE_ARG *arg)
1352 {
1353     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1354     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1355     DWORD flags;
1356     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1357     SHADER_BUFFER* buffer = arg->buffer;
1358     SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1359     char dst_str[8];
1360     char src0_name[50];
1361
1362     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1363     shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1364
1365     /* Construct the eye-ray vector from w coordinates */
1366     shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1367     shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1368     shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1369
1370     /* Calculate reflection vector
1371      */
1372     shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1373     /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1374     shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1375     shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1376     shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1377     shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1378     shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1379
1380     /* Sample the texture using the calculated coordinates */
1381     sprintf(dst_str, "T%u", reg);
1382     flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1383     shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1384     current_state->current_row = 0;
1385 }
1386
1387 static void pshader_hw_texm3x3spec(const SHADER_OPCODE_ARG *arg)
1388 {
1389     IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1390     IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1391     DWORD flags;
1392     DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1393     DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1394     SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1395     SHADER_BUFFER* buffer = arg->buffer;
1396     char dst_str[8];
1397     char src0_name[50];
1398
1399     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1400     shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1401
1402     /* Calculate reflection vector.
1403      *
1404      *               dot(N, E)
1405      * TMP.xyz = 2 * --------- * N - E
1406      *               dot(N, N)
1407      *
1408      * Which normalizes the normal vector
1409      */
1410     shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1411     shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1412     shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1413     shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1414     shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1415     shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1416
1417     /* Sample the texture using the calculated coordinates */
1418     sprintf(dst_str, "T%u", reg);
1419     flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1420     shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1421     current_state->current_row = 0;
1422 }
1423
1424 static void pshader_hw_texdepth(const SHADER_OPCODE_ARG *arg)
1425 {
1426     SHADER_BUFFER* buffer = arg->buffer;
1427     char dst_name[50];
1428
1429     /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1430      * which is essentially an input, is the destination register because it is the first
1431      * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1432      * here
1433      */
1434     pshader_get_register_name(arg->shader, arg->dst, dst_name);
1435
1436     /* According to the msdn, the source register(must be r5) is unusable after
1437      * the texdepth instruction, so we're free to modify it
1438      */
1439     shader_addline(buffer, "MIN %s.y, %s.y, one.y;\n", dst_name, dst_name);
1440
1441     /* How to deal with the special case dst_name.g == 0? if r != 0, then
1442      * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1443      * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1444      */
1445     shader_addline(buffer, "RCP %s.y, %s.y;\n", dst_name, dst_name);
1446     shader_addline(buffer, "MUL TMP.x, %s.x, %s.y;\n", dst_name, dst_name);
1447     shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1448     shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1449 }
1450
1451 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1452  * Take a 3-component dot product of the TexCoord[dstreg] and src,
1453  * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1454 static void pshader_hw_texdp3tex(const SHADER_OPCODE_ARG *arg)
1455 {
1456     SHADER_BUFFER* buffer = arg->buffer;
1457     DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1458     char src0[50];
1459     char dst_str[8];
1460
1461     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1462     shader_addline(buffer, "MOV TMP, 0.0;\n");
1463     shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1464
1465     sprintf(dst_str, "T%u", sampler_idx);
1466     shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1467 }
1468
1469 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1470  * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1471 static void pshader_hw_texdp3(const SHADER_OPCODE_ARG *arg)
1472 {
1473     char src0[50];
1474     char dst_str[50];
1475     char dst_mask[6];
1476     DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1477     SHADER_BUFFER* buffer = arg->buffer;
1478
1479     /* Handle output register */
1480     pshader_get_register_name(arg->shader, arg->dst, dst_str);
1481     shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1482
1483     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1484     shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1485
1486     /* TODO: Handle output modifiers */
1487 }
1488
1489 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1490  * Perform the 3rd row of a 3x3 matrix multiply */
1491 static void pshader_hw_texm3x3(const SHADER_OPCODE_ARG *arg)
1492 {
1493     SHADER_BUFFER* buffer = arg->buffer;
1494     char dst_str[50];
1495     char dst_mask[6];
1496     char src0[50];
1497     DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1498
1499     pshader_get_register_name(arg->shader, arg->dst, dst_str);
1500     shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1501
1502     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1503     shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1504     shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1505
1506     /* TODO: Handle output modifiers */
1507 }
1508
1509 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1510  * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1511  * Calculate tmp0.y = TexCoord[dstreg] . src.xyz;  (tmp0.x has already been calculated)
1512  * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1513  */
1514 static void pshader_hw_texm3x2depth(const SHADER_OPCODE_ARG *arg)
1515 {
1516     SHADER_BUFFER* buffer = arg->buffer;
1517     DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1518     char src0[50];
1519
1520     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1521     shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1522
1523     /* How to deal with the special case dst_name.g == 0? if r != 0, then
1524      * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1525      * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1526      */
1527     shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1528     shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1529     shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1530     shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1531 }
1532
1533 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1534     Vertex/Pixel shaders to ARB_vertex_program codes */
1535 static void shader_hw_mnxn(const SHADER_OPCODE_ARG *arg)
1536 {
1537     int i;
1538     int nComponents = 0;
1539     SHADER_OPCODE_ARG tmpArg;
1540     IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1541     const SHADER_OPCODE *opcode_table = shader->baseShader.shader_ins;
1542     DWORD shader_version = arg->reg_maps->shader_version;
1543
1544     memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1545
1546     /* Set constants for the temporary argument */
1547     tmpArg.shader      = arg->shader;
1548     tmpArg.buffer      = arg->buffer;
1549     tmpArg.src[0]      = arg->src[0];
1550     tmpArg.src_addr[0] = arg->src_addr[0];
1551     tmpArg.src_addr[1] = arg->src_addr[1];
1552     tmpArg.reg_maps = arg->reg_maps;
1553
1554     switch(arg->opcode->opcode) {
1555     case WINED3DSIO_M4x4:
1556         nComponents = 4;
1557         tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1558         break;
1559     case WINED3DSIO_M4x3:
1560         nComponents = 3;
1561         tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1562         break;
1563     case WINED3DSIO_M3x4:
1564         nComponents = 4;
1565         tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1566         break;
1567     case WINED3DSIO_M3x3:
1568         nComponents = 3;
1569         tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1570         break;
1571     case WINED3DSIO_M3x2:
1572         nComponents = 2;
1573         tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1574         break;
1575     default:
1576         break;
1577     }
1578
1579     for (i = 0; i < nComponents; i++) {
1580         tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1581         tmpArg.src[1] = arg->src[1]+i;
1582         shader_hw_map2gl(&tmpArg);
1583     }
1584 }
1585
1586 static void vshader_hw_rsq_rcp(const SHADER_OPCODE_ARG *arg)
1587 {
1588     CONST SHADER_OPCODE* curOpcode = arg->opcode;
1589     SHADER_BUFFER* buffer = arg->buffer;
1590     DWORD dst = arg->dst;
1591     DWORD src = arg->src[0];
1592     DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1593     const char *instruction;
1594
1595     char tmpLine[256];
1596
1597     switch(curOpcode->opcode)
1598     {
1599         case WINED3DSIO_RSQ: instruction = "RSQ"; break;
1600         case WINED3DSIO_RCP: instruction = "RCP"; break;
1601         default: instruction = "";
1602             FIXME("Unhandled opcode %s\n", curOpcode->name);
1603             break;
1604     }
1605
1606     strcpy(tmpLine, instruction);
1607     vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1608     strcat(tmpLine, ",");
1609     vshader_program_add_param(arg, src, TRUE, tmpLine);
1610     if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1611         /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1612          * .w is used
1613          */
1614         strcat(tmpLine, ".w");
1615     }
1616
1617     shader_addline(buffer, "%s;\n", tmpLine);
1618 }
1619
1620 static void shader_hw_nrm(const SHADER_OPCODE_ARG *arg)
1621 {
1622     SHADER_BUFFER* buffer = arg->buffer;
1623     char dst_name[50];
1624     char src_name[50];
1625     char dst_wmask[20];
1626     DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1627     BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1628
1629     pshader_get_register_name(arg->shader, arg->dst, dst_name);
1630     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1631
1632     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1633     shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1634     shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1635     /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1636     shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1637                    src_name);
1638
1639     if (shift != 0)
1640         pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1641 }
1642
1643 static void shader_hw_sincos(const SHADER_OPCODE_ARG *arg)
1644 {
1645     /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1646      * must contain fixed constants. So we need a separate function to filter those constants and
1647      * can't use map2gl
1648      */
1649     SHADER_BUFFER* buffer = arg->buffer;
1650     char dst_name[50];
1651     char src_name[50];
1652     char dst_wmask[20];
1653     DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1654     BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1655
1656     pshader_get_register_name(arg->shader, arg->dst, dst_name);
1657     shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1658
1659     pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1660     shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1661                    src_name);
1662
1663     if (shift != 0)
1664         pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1665
1666 }
1667
1668 static GLuint create_arb_blt_vertex_program(const WineD3D_GL_Info *gl_info)
1669 {
1670     GLuint program_id = 0;
1671     const char *blt_vprogram =
1672         "!!ARBvp1.0\n"
1673         "PARAM c[1] = { { 1, 0.5 } };\n"
1674         "MOV result.position, vertex.position;\n"
1675         "MOV result.color, c[0].x;\n"
1676         "MOV result.texcoord[0], vertex.texcoord[0];\n"
1677         "END\n";
1678
1679     GL_EXTCALL(glGenProgramsARB(1, &program_id));
1680     GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1681     GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1682
1683     if (glGetError() == GL_INVALID_OPERATION) {
1684         GLint pos;
1685         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1686         FIXME("Vertex program error at position %d: %s\n", pos,
1687             debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1688     }
1689
1690     return program_id;
1691 }
1692
1693 static GLuint create_arb_blt_fragment_program(const WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1694 {
1695     GLuint program_id = 0;
1696     static const char * const blt_fprograms[tex_type_count] =
1697     {
1698         /* tex_1d */
1699         NULL,
1700         /* tex_2d */
1701         "!!ARBfp1.0\n"
1702         "TEMP R0;\n"
1703         "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1704         "MOV result.depth.z, R0.x;\n"
1705         "END\n",
1706         /* tex_3d */
1707         NULL,
1708         /* tex_cube */
1709         "!!ARBfp1.0\n"
1710         "TEMP R0;\n"
1711         "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1712         "MOV result.depth.z, R0.x;\n"
1713         "END\n",
1714         /* tex_rect */
1715         "!!ARBfp1.0\n"
1716         "TEMP R0;\n"
1717         "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1718         "MOV result.depth.z, R0.x;\n"
1719         "END\n",
1720     };
1721
1722     if (!blt_fprograms[tex_type])
1723     {
1724         FIXME("tex_type %#x not supported\n", tex_type);
1725         tex_type = tex_2d;
1726     }
1727
1728     GL_EXTCALL(glGenProgramsARB(1, &program_id));
1729     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1730     GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1731
1732     if (glGetError() == GL_INVALID_OPERATION) {
1733         GLint pos;
1734         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1735         FIXME("Fragment program error at position %d: %s\n", pos,
1736             debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1737     }
1738
1739     return program_id;
1740 }
1741
1742 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1743     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1744     struct shader_arb_priv *priv = This->shader_priv;
1745     const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1746
1747     if (useVS) {
1748         struct vs_compile_args compile_args;
1749
1750         TRACE("Using vertex shader\n");
1751         find_vs_compile_args((IWineD3DVertexShaderImpl *) This->stateBlock->vertexShader, This->stateBlock, &compile_args);
1752         priv->current_vprogram_id = find_gl_vshader((IWineD3DVertexShaderImpl *) This->stateBlock->vertexShader, &compile_args);
1753
1754         /* Bind the vertex program */
1755         GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1756         checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
1757
1758         /* Enable OpenGL vertex programs */
1759         glEnable(GL_VERTEX_PROGRAM_ARB);
1760         checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1761         TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1762     } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1763         priv->current_vprogram_id = 0;
1764         glDisable(GL_VERTEX_PROGRAM_ARB);
1765         checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1766     }
1767
1768     if (usePS) {
1769         struct ps_compile_args compile_args;
1770         TRACE("Using pixel shader\n");
1771         find_ps_compile_args((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader, This->stateBlock, &compile_args);
1772         priv->current_fprogram_id = find_gl_pshader((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader,
1773                                                     &compile_args);
1774
1775         /* Bind the fragment program */
1776         GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1777         checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
1778
1779         if(!priv->use_arbfp_fixed_func) {
1780             /* Enable OpenGL fragment programs */
1781             glEnable(GL_FRAGMENT_PROGRAM_ARB);
1782             checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1783         }
1784         TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1785     } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1786         /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1787          * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1788          * replacement shader
1789          */
1790         glDisable(GL_FRAGMENT_PROGRAM_ARB);
1791         checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1792         priv->current_fprogram_id = 0;
1793     }
1794 }
1795
1796 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1797     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1798     struct shader_arb_priv *priv = This->shader_priv;
1799     GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1800     const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1801
1802     if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1803     GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1804     glEnable(GL_VERTEX_PROGRAM_ARB);
1805
1806     if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1807     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1808     glEnable(GL_FRAGMENT_PROGRAM_ARB);
1809 }
1810
1811 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1812     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1813     struct shader_arb_priv *priv = This->shader_priv;
1814     const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1815
1816     if (priv->current_vprogram_id) {
1817         GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1818         checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1819
1820         glEnable(GL_VERTEX_PROGRAM_ARB);
1821         checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1822
1823         TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1824     } else {
1825         glDisable(GL_VERTEX_PROGRAM_ARB);
1826         checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1827     }
1828
1829     if (priv->current_fprogram_id) {
1830         GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1831         checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1832
1833         glEnable(GL_FRAGMENT_PROGRAM_ARB);
1834         checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1835
1836         TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1837     } else {
1838         glDisable(GL_FRAGMENT_PROGRAM_ARB);
1839         checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1840     }
1841 }
1842
1843 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1844     IWineD3DBaseShaderImpl *baseShader = (IWineD3DBaseShaderImpl *) iface;
1845     const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)baseShader->baseShader.device)->adapter->gl_info;
1846
1847     if (shader_is_pshader_version(baseShader->baseShader.reg_maps.shader_version))
1848     {
1849         IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
1850         UINT i;
1851
1852         ENTER_GL();
1853         for(i = 0; i < This->num_gl_shaders; i++) {
1854             GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1855             checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1856         }
1857         LEAVE_GL();
1858         HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1859         This->gl_shaders = NULL;
1860         This->num_gl_shaders = 0;
1861         This->shader_array_size = 0;
1862     } else {
1863         IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *) iface;
1864         UINT i;
1865
1866         ENTER_GL();
1867         for(i = 0; i < This->num_gl_shaders; i++) {
1868             GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1869             checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1870         }
1871         LEAVE_GL();
1872         HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1873         This->gl_shaders = NULL;
1874         This->num_gl_shaders = 0;
1875         This->shader_array_size = 0;
1876     }
1877 }
1878
1879 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1880     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1881     This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1882     return WINED3D_OK;
1883 }
1884
1885 static void shader_arb_free(IWineD3DDevice *iface) {
1886     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1887     const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1888     struct shader_arb_priv *priv = This->shader_priv;
1889     int i;
1890
1891     if(priv->depth_blt_vprogram_id) {
1892         GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1893     }
1894     for (i = 0; i < tex_type_count; ++i) {
1895         if (priv->depth_blt_fprogram_id[i]) {
1896             GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
1897         }
1898     }
1899
1900     HeapFree(GetProcessHeap(), 0, This->shader_priv);
1901 }
1902
1903 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1904     return TRUE;
1905 }
1906
1907 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1908                                       const char *tmp2, const char *tmp3, const char *tmp4) {
1909     /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1910
1911     /* Calculate the > 0.0031308 case */
1912     shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1913     shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1914     shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1915     shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1916     shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1917     /* Calculate the < case */
1918     shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1919     /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1920     shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1921     shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1922     /* Store the components > 0.0031308 in the destination */
1923     shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1924     /* Add the components that are < 0.0031308 */
1925     shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1926     /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1927 }
1928
1929 static GLuint shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer, const struct ps_compile_args *args) {
1930     IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1931     const shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1932     CONST DWORD *function = This->baseShader.function;
1933     DWORD shader_version = reg_maps->shader_version;
1934     const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1935     const local_constant *lconst;
1936     GLuint retval;
1937     const char *fragcolor;
1938
1939     /*  Create the hw ARB shader */
1940     shader_addline(buffer, "!!ARBfp1.0\n");
1941
1942     if (shader_version < WINED3DPS_VERSION(3,0)) {
1943         switch(args->fog) {
1944             case FOG_OFF:
1945                 break;
1946             case FOG_LINEAR:
1947                 shader_addline(buffer, "OPTION ARB_fog_linear;\n");
1948                 break;
1949             case FOG_EXP:
1950                 shader_addline(buffer, "OPTION ARB_fog_exp;\n");
1951                 break;
1952             case FOG_EXP2:
1953                 shader_addline(buffer, "OPTION ARB_fog_exp2;\n");
1954                 break;
1955         }
1956     }
1957
1958     shader_addline(buffer, "TEMP TMP;\n");     /* Used in matrix ops */
1959     shader_addline(buffer, "TEMP TMP2;\n");    /* Used in matrix ops */
1960     shader_addline(buffer, "TEMP TA;\n");      /* Used for modifiers */
1961     shader_addline(buffer, "TEMP TB;\n");      /* Used for modifiers */
1962     shader_addline(buffer, "TEMP TC;\n");      /* Used for modifiers */
1963     shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1964     shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1965     shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1966
1967     if (shader_version < WINED3DPS_VERSION(2,0)) {
1968         fragcolor = "R0";
1969     } else {
1970         shader_addline(buffer, "TEMP TMP_COLOR;\n");
1971         fragcolor = "TMP_COLOR";
1972     }
1973
1974     /* Base Declarations */
1975     shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1976
1977     /* Base Shader Body */
1978     shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1979
1980     if(args->srgb_correction) {
1981         arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
1982     }
1983     shader_addline(buffer, "MOV result.color, %s;\n", fragcolor);
1984     shader_addline(buffer, "END\n");
1985
1986     /* TODO: change to resource.glObjectHandle or something like that */
1987     GL_EXTCALL(glGenProgramsARB(1, &retval));
1988
1989     TRACE("Creating a hw pixel shader, prg=%d\n", retval);
1990     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, retval));
1991
1992     TRACE("Created hw pixel shader, prg=%d\n", retval);
1993     /* Create the program and check for errors */
1994     GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1995                buffer->bsize, buffer->buffer));
1996
1997     if (glGetError() == GL_INVALID_OPERATION) {
1998         GLint errPos;
1999         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2000         FIXME("HW PixelShader Error at position %d: %s\n",
2001               errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2002         retval = 0;
2003     }
2004
2005     /* Load immediate constants */
2006     if(!This->baseShader.load_local_constsF) {
2007         LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2008             const float *value = (const float *)lconst->value;
2009             GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2010             checkGLcall("glProgramLocalParameter4fvARB");
2011         }
2012     }
2013
2014     return retval;
2015 }
2016
2017 static GLuint shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer, const struct vs_compile_args *args) {
2018     IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2019     const shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
2020     CONST DWORD *function = This->baseShader.function;
2021     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2022     const WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2023     const local_constant *lconst;
2024     GLuint ret;
2025
2026     /*  Create the hw ARB shader */
2027     shader_addline(buffer, "!!ARBvp1.0\n");
2028     shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2029
2030     /* Mesa supports only 95 constants */
2031     if (GL_VEND(MESA) || GL_VEND(WINE))
2032         This->baseShader.limits.constant_float =
2033                 min(95, This->baseShader.limits.constant_float);
2034
2035     shader_addline(buffer, "TEMP TMP;\n");
2036
2037     /* Base Declarations */
2038     shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2039
2040     /* We need a constant to fixup the final position */
2041     shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2042
2043     /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2044      * for output parameters. D3D in theory does not do that either, but some applications depend on a
2045      * proper initialization of the secondary color, and programs using the fixed function pipeline without
2046      * a replacement shader depend on the texcoord.w being set properly.
2047      *
2048      * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2049      * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2050      * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2051      * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2052      * this can eat a number of instructions, so skip it unless this cap is set as well
2053      */
2054     if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2055         shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2056
2057         if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2058             int i;
2059             for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2060                 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2061                 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2062                     shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2063                 }
2064             }
2065         }
2066     }
2067
2068     /* Base Shader Body */
2069     shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2070
2071     /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
2072      * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
2073      * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
2074      * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
2075      */
2076     if(args->fog_src == VS_FOG_Z) {
2077         shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2078     } else if (!reg_maps->fog) {
2079         shader_addline(buffer, "MOV result.fogcoord, 0.0;\n");
2080     }
2081
2082     /* Write the final position.
2083      *
2084      * OpenGL coordinates specify the center of the pixel while d3d coords specify
2085      * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2086      * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2087      * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2088      */
2089     shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2090     shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2091     shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2092
2093     /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2094      * and the glsl equivalent
2095      */
2096     shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2097
2098     shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2099
2100     shader_addline(buffer, "END\n");
2101
2102     /* TODO: change to resource.glObjectHandle or something like that */
2103     GL_EXTCALL(glGenProgramsARB(1, &ret));
2104
2105     TRACE("Creating a hw vertex shader, prg=%d\n", ret);
2106     GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, ret));
2107
2108     TRACE("Created hw vertex shader, prg=%d\n", ret);
2109     /* Create the program and check for errors */
2110     GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2111                buffer->bsize, buffer->buffer));
2112
2113     if (glGetError() == GL_INVALID_OPERATION) {
2114         GLint errPos;
2115         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2116         FIXME("HW VertexShader Error at position %d: %s\n",
2117               errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2118         ret = -1;
2119     } else {
2120         /* Load immediate constants */
2121         if(!This->baseShader.load_local_constsF) {
2122             LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2123                 const float *value = (const float *)lconst->value;
2124                 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2125             }
2126         }
2127     }
2128     return ret;
2129 }
2130
2131 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
2132 {
2133     /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2134      * then overwrite the shader specific ones
2135      */
2136     none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2137
2138     if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2139         pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2140         TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2141         pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2142     }
2143
2144     if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2145         pCaps->PixelShaderVersion    = WINED3DPS_VERSION(1,4);
2146         pCaps->PixelShader1xMaxValue = 8.0;
2147         TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2148     }
2149 }
2150
2151 static BOOL shader_arb_color_fixup_supported(struct color_fixup_desc fixup)
2152 {
2153     if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
2154     {
2155         TRACE("Checking support for color_fixup:\n");
2156         dump_color_fixup_desc(fixup);
2157     }
2158
2159     /* We support everything except YUV conversions. */
2160     if (!is_yuv_fixup(fixup))
2161     {
2162         TRACE("[OK]\n");
2163         return TRUE;
2164     }
2165
2166     TRACE("[FAILED]\n");
2167     return FALSE;
2168 }
2169
2170 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2171 {
2172     /* WINED3DSIH_ABS           */ shader_hw_map2gl,
2173     /* WINED3DSIH_ADD           */ shader_hw_map2gl,
2174     /* WINED3DSIH_BEM           */ pshader_hw_bem,
2175     /* WINED3DSIH_BREAK         */ NULL,
2176     /* WINED3DSIH_BREAKC        */ NULL,
2177     /* WINED3DSIH_BREAKP        */ NULL,
2178     /* WINED3DSIH_CALL          */ NULL,
2179     /* WINED3DSIH_CALLNZ        */ NULL,
2180     /* WINED3DSIH_CMP           */ pshader_hw_cmp,
2181     /* WINED3DSIH_CND           */ pshader_hw_cnd,
2182     /* WINED3DSIH_CRS           */ shader_hw_map2gl,
2183     /* WINED3DSIH_DCL           */ NULL,
2184     /* WINED3DSIH_DEF           */ NULL,
2185     /* WINED3DSIH_DEFB          */ NULL,
2186     /* WINED3DSIH_DEFI          */ NULL,
2187     /* WINED3DSIH_DP2ADD        */ pshader_hw_dp2add,
2188     /* WINED3DSIH_DP3           */ shader_hw_map2gl,
2189     /* WINED3DSIH_DP4           */ shader_hw_map2gl,
2190     /* WINED3DSIH_DST           */ shader_hw_map2gl,
2191     /* WINED3DSIH_DSX           */ NULL,
2192     /* WINED3DSIH_DSY           */ NULL,
2193     /* WINED3DSIH_ELSE          */ NULL,
2194     /* WINED3DSIH_ENDIF         */ NULL,
2195     /* WINED3DSIH_ENDLOOP       */ NULL,
2196     /* WINED3DSIH_ENDREP        */ NULL,
2197     /* WINED3DSIH_EXP           */ shader_hw_map2gl,
2198     /* WINED3DSIH_EXPP          */ shader_hw_map2gl,
2199     /* WINED3DSIH_FRC           */ shader_hw_map2gl,
2200     /* WINED3DSIH_IF            */ NULL,
2201     /* WINED3DSIH_IFC           */ NULL,
2202     /* WINED3DSIH_LABEL         */ NULL,
2203     /* WINED3DSIH_LIT           */ shader_hw_map2gl,
2204     /* WINED3DSIH_LOG           */ shader_hw_map2gl,
2205     /* WINED3DSIH_LOGP          */ shader_hw_map2gl,
2206     /* WINED3DSIH_LOOP          */ NULL,
2207     /* WINED3DSIH_LRP           */ shader_hw_map2gl,
2208     /* WINED3DSIH_M3x2          */ shader_hw_mnxn,
2209     /* WINED3DSIH_M3x3          */ shader_hw_mnxn,
2210     /* WINED3DSIH_M3x4          */ shader_hw_mnxn,
2211     /* WINED3DSIH_M4x3          */ shader_hw_mnxn,
2212     /* WINED3DSIH_M4x4          */ shader_hw_mnxn,
2213     /* WINED3DSIH_MAD           */ shader_hw_map2gl,
2214     /* WINED3DSIH_MAX           */ shader_hw_map2gl,
2215     /* WINED3DSIH_MIN           */ shader_hw_map2gl,
2216     /* WINED3DSIH_MOV           */ shader_hw_mov,
2217     /* WINED3DSIH_MOVA          */ shader_hw_mov,
2218     /* WINED3DSIH_MUL           */ shader_hw_map2gl,
2219     /* WINED3DSIH_NOP           */ shader_hw_map2gl,
2220     /* WINED3DSIH_NRM           */ shader_hw_nrm,
2221     /* WINED3DSIH_PHASE         */ NULL,
2222     /* WINED3DSIH_POW           */ shader_hw_map2gl,
2223     /* WINED3DSIH_RCP           */ vshader_hw_rsq_rcp,
2224     /* WINED3DSIH_REP           */ NULL,
2225     /* WINED3DSIH_RET           */ NULL,
2226     /* WINED3DSIH_RSQ           */ vshader_hw_rsq_rcp,
2227     /* WINED3DSIH_SETP          */ NULL,
2228     /* WINED3DSIH_SGE           */ shader_hw_map2gl,
2229     /* WINED3DSIH_SGN           */ NULL,
2230     /* WINED3DSIH_SINCOS        */ shader_hw_sincos,
2231     /* WINED3DSIH_SLT           */ shader_hw_map2gl,
2232     /* WINED3DSIH_SUB           */ shader_hw_map2gl,
2233     /* WINED3DSIH_TEX           */ pshader_hw_tex,
2234     /* WINED3DSIH_TEXBEM        */ pshader_hw_texbem,
2235     /* WINED3DSIH_TEXBEML       */ pshader_hw_texbem,
2236     /* WINED3DSIH_TEXCOORD      */ pshader_hw_texcoord,
2237     /* WINED3DSIH_TEXDEPTH      */ pshader_hw_texdepth,
2238     /* WINED3DSIH_TEXDP3        */ pshader_hw_texdp3,
2239     /* WINED3DSIH_TEXDP3TEX     */ pshader_hw_texdp3tex,
2240     /* WINED3DSIH_TEXKILL       */ pshader_hw_texkill,
2241     /* WINED3DSIH_TEXLDD        */ NULL,
2242     /* WINED3DSIH_TEXLDL        */ NULL,
2243     /* WINED3DSIH_TEXM3x2DEPTH  */ pshader_hw_texm3x2depth,
2244     /* WINED3DSIH_TEXM3x2PAD    */ pshader_hw_texm3x2pad,
2245     /* WINED3DSIH_TEXM3x2TEX    */ pshader_hw_texm3x2tex,
2246     /* WINED3DSIH_TEXM3x3       */ pshader_hw_texm3x3,
2247     /* WINED3DSIH_TEXM3x3DIFF   */ NULL,
2248     /* WINED3DSIH_TEXM3x3PAD    */ pshader_hw_texm3x3pad,
2249     /* WINED3DSIH_TEXM3x3SPEC   */ pshader_hw_texm3x3spec,
2250     /* WINED3DSIH_TEXM3x3TEX    */ pshader_hw_texm3x3tex,
2251     /* WINED3DSIH_TEXM3x3VSPEC  */ pshader_hw_texm3x3vspec,
2252     /* WINED3DSIH_TEXREG2AR     */ pshader_hw_texreg2ar,
2253     /* WINED3DSIH_TEXREG2GB     */ pshader_hw_texreg2gb,
2254     /* WINED3DSIH_TEXREG2RGB    */ pshader_hw_texreg2rgb,
2255 };
2256
2257 const shader_backend_t arb_program_shader_backend = {
2258     shader_arb_instruction_handler_table,
2259     shader_arb_select,
2260     shader_arb_select_depth_blt,
2261     shader_arb_deselect_depth_blt,
2262     shader_arb_update_float_vertex_constants,
2263     shader_arb_update_float_pixel_constants,
2264     shader_arb_load_constants,
2265     shader_arb_color_correction,
2266     shader_arb_destroy,
2267     shader_arb_alloc,
2268     shader_arb_free,
2269     shader_arb_dirty_const,
2270     shader_arb_generate_pshader,
2271     shader_arb_generate_vshader,
2272     shader_arb_get_caps,
2273     shader_arb_color_fixup_supported,
2274 };
2275
2276 /* ARB_fragment_program fixed function pipeline replacement definitions */
2277 #define ARB_FFP_CONST_TFACTOR           0
2278 #define ARB_FFP_CONST_SPECULAR_ENABLE   ((ARB_FFP_CONST_TFACTOR) + 1)
2279 #define ARB_FFP_CONST_CONSTANT(i)       ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2280 #define ARB_FFP_CONST_BUMPMAT(i)        ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2281 #define ARB_FFP_CONST_LUMINANCE(i)      ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2282
2283 struct arbfp_ffp_desc
2284 {
2285     struct ffp_frag_desc parent;
2286     GLuint shader;
2287     unsigned int num_textures_used;
2288 };
2289
2290 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2291     if(enable) {
2292         glEnable(GL_FRAGMENT_PROGRAM_ARB);
2293         checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2294     } else {
2295         glDisable(GL_FRAGMENT_PROGRAM_ARB);
2296         checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2297     }
2298 }
2299
2300 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2301     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2302     struct shader_arb_priv *priv;
2303     /* Share private data between the shader backend and the pipeline replacement, if both
2304      * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2305      * if no pixel shader is bound or not
2306      */
2307     if(This->shader_backend == &arb_program_shader_backend) {
2308         This->fragment_priv = This->shader_priv;
2309     } else {
2310         This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2311         if(!This->fragment_priv) return E_OUTOFMEMORY;
2312     }
2313     priv = This->fragment_priv;
2314     priv->fragment_shaders = hash_table_create(ffp_frag_program_key_hash, ffp_frag_program_key_compare);
2315     priv->use_arbfp_fixed_func = TRUE;
2316     return WINED3D_OK;
2317 }
2318
2319 static void arbfp_free_ffpshader(void *value, void *gli) {
2320     const WineD3D_GL_Info *gl_info = gli;
2321     struct arbfp_ffp_desc *entry_arb = value;
2322
2323     ENTER_GL();
2324     GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2325     checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2326     HeapFree(GetProcessHeap(), 0, entry_arb);
2327     LEAVE_GL();
2328 }
2329
2330 static void arbfp_free(IWineD3DDevice *iface) {
2331     IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2332     struct shader_arb_priv *priv = This->fragment_priv;
2333
2334     hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2335     priv->use_arbfp_fixed_func = FALSE;
2336
2337     if(This->shader_backend != &arb_program_shader_backend) {
2338         HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2339     }
2340 }
2341
2342 static void arbfp_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct fragment_caps *caps)
2343 {
2344     caps->TextureOpCaps =  WINED3DTEXOPCAPS_DISABLE                     |
2345                            WINED3DTEXOPCAPS_SELECTARG1                  |
2346                            WINED3DTEXOPCAPS_SELECTARG2                  |
2347                            WINED3DTEXOPCAPS_MODULATE4X                  |
2348                            WINED3DTEXOPCAPS_MODULATE2X                  |
2349                            WINED3DTEXOPCAPS_MODULATE                    |
2350                            WINED3DTEXOPCAPS_ADDSIGNED2X                 |
2351                            WINED3DTEXOPCAPS_ADDSIGNED                   |
2352                            WINED3DTEXOPCAPS_ADD                         |
2353                            WINED3DTEXOPCAPS_SUBTRACT                    |
2354                            WINED3DTEXOPCAPS_ADDSMOOTH                   |
2355                            WINED3DTEXOPCAPS_BLENDCURRENTALPHA           |
2356                            WINED3DTEXOPCAPS_BLENDFACTORALPHA            |
2357                            WINED3DTEXOPCAPS_BLENDTEXTUREALPHA           |
2358                            WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA           |
2359                            WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM         |
2360                            WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR      |
2361                            WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA      |
2362                            WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA   |
2363                            WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR   |
2364                            WINED3DTEXOPCAPS_DOTPRODUCT3                 |
2365                            WINED3DTEXOPCAPS_MULTIPLYADD                 |
2366                            WINED3DTEXOPCAPS_LERP                        |
2367                            WINED3DTEXOPCAPS_BUMPENVMAP                  |
2368                            WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2369
2370     /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2371
2372     caps->MaxTextureBlendStages   = 8;
2373     caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2374
2375     caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2376 }
2377 #undef GLINFO_LOCATION
2378
2379 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2380 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2381     float col[4];
2382     IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2383
2384     /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2385      * application provided constants
2386      */
2387     if(device->shader_backend == &arb_program_shader_backend) {
2388         if (use_ps(stateblock)) return;
2389
2390         device = stateblock->wineD3DDevice;
2391         device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2392         device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2393     }
2394
2395     D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2396     GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2397     checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2398
2399 }
2400
2401 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2402     float col[4];
2403     IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2404
2405     /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2406      * application provided constants
2407      */
2408     if(device->shader_backend == &arb_program_shader_backend) {
2409         if (use_ps(stateblock)) return;
2410
2411         device = stateblock->wineD3DDevice;
2412         device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2413         device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2414     }
2415
2416     if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2417         /* The specular color has no alpha */
2418         col[0] = 1.0; col[1] = 1.0;
2419         col[2] = 1.0; col[3] = 0.0;
2420     } else {
2421         col[0] = 0.0; col[1] = 0.0;
2422         col[2] = 0.0; col[3] = 0.0;
2423     }
2424     GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2425     checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2426 }
2427
2428 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2429     DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE + 1);
2430     IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2431     float mat[2][2];
2432
2433     if (use_ps(stateblock))
2434     {
2435         if(stage != 0 &&
2436            ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2437             /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2438              * anyway
2439              */
2440             if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2441                 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2442             }
2443         }
2444
2445         if(device->shader_backend == &arb_program_shader_backend) {
2446             /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2447             return;
2448         }
2449     } else if(device->shader_backend == &arb_program_shader_backend) {
2450         device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2451         device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2452     }
2453
2454     mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2455     mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2456     mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2457     mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2458
2459     GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2460     checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2461 }
2462
2463 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2464     DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE + 1);
2465     IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2466     float param[4];
2467
2468     if (use_ps(stateblock))
2469     {
2470         if(stage != 0 &&
2471            ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2472             /* The pixel shader has to know the luminance offset. Do a constants update if it
2473              * isn't scheduled anyway
2474              */
2475             if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2476                 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2477             }
2478         }
2479
2480         if(device->shader_backend == &arb_program_shader_backend) {
2481             /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2482             return;
2483         }
2484     } else if(device->shader_backend == &arb_program_shader_backend) {
2485         device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2486         device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2487     }
2488
2489     param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2490     param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2491     param[2] = 0.0;
2492     param[3] = 0.0;
2493
2494     GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2495     checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2496 }
2497
2498 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2499     const char *ret;
2500
2501     if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2502
2503     switch(arg & WINED3DTA_SELECTMASK) {
2504         case WINED3DTA_DIFFUSE:
2505             ret = "fragment.color.primary"; break;
2506
2507         case WINED3DTA_CURRENT:
2508             if(stage == 0) ret = "fragment.color.primary";
2509             else ret = "ret";
2510             break;
2511
2512         case WINED3DTA_TEXTURE:
2513             switch(stage) {
2514                 case 0: ret = "tex0"; break;
2515                 case 1: ret = "tex1"; break;
2516                 case 2: ret = "tex2"; break;
2517                 case 3: ret = "tex3"; break;
2518                 case 4: ret = "tex4"; break;
2519                 case 5: ret = "tex5"; break;
2520                 case 6: ret = "tex6"; break;
2521                 case 7: ret = "tex7"; break;
2522                 default: ret = "unknown texture";
2523             }
2524             break;
2525
2526         case WINED3DTA_TFACTOR:
2527             ret = "tfactor"; break;
2528
2529         case WINED3DTA_SPECULAR:
2530             ret = "fragment.color.secondary"; break;
2531
2532         case WINED3DTA_TEMP:
2533             ret = "tempreg"; break;
2534
2535         case WINED3DTA_CONSTANT:
2536             FIXME("Implement perstage constants\n");
2537             switch(stage) {
2538                 case 0: ret = "const0"; break;
2539                 case 1: ret = "const1"; break;
2540                 case 2: ret = "const2"; break;
2541                 case 3: ret = "const3"; break;
2542                 case 4: ret = "const4"; break;
2543                 case 5: ret = "const5"; break;
2544                 case 6: ret = "const6"; break;
2545                 case 7: ret = "const7"; break;
2546                 default: ret = "unknown constant";
2547             }
2548             break;
2549
2550         default:
2551             return "unknown";
2552     }
2553
2554     if(arg & WINED3DTA_COMPLEMENT) {
2555         shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2556         if(argnum == 0) ret = "arg0";
2557         if(argnum == 1) ret = "arg1";
2558         if(argnum == 2) ret = "arg2";
2559     }
2560     if(arg & WINED3DTA_ALPHAREPLICATE) {
2561         shader_addline(buffer, "MOV arg%u, %s.w;\n", argnum, ret);
2562         if(argnum == 0) ret = "arg0";
2563         if(argnum == 1) ret = "arg1";
2564         if(argnum == 2) ret = "arg2";
2565     }
2566     return ret;
2567 }
2568
2569 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2570                           DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2571     const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2572     unsigned int mul = 1;
2573     BOOL mul_final_dest = FALSE;
2574
2575     if(color && alpha) dstmask = "";
2576     else if(color) dstmask = ".xyz";
2577     else dstmask = ".w";
2578
2579     if(dst == tempreg) dstreg = "tempreg";
2580     else dstreg = "ret";
2581
2582     arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2583     arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2584     arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2585
2586     switch(op) {
2587         case WINED3DTOP_DISABLE:
2588             if(stage == 0) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2589             break;
2590
2591         case WINED3DTOP_SELECTARG2:
2592             arg1 = arg2;
2593         case WINED3DTOP_SELECTARG1:
2594             shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2595             break;
2596
2597         case WINED3DTOP_MODULATE4X:
2598             mul = 2;
2599         case WINED3DTOP_MODULATE2X:
2600             mul *= 2;
2601             if(strcmp(dstreg, "result.color") == 0) {
2602                 dstreg = "ret";
2603                 mul_final_dest = TRUE;
2604             }
2605         case WINED3DTOP_MODULATE:
2606             shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2607             break;
2608
2609         case WINED3DTOP_ADDSIGNED2X:
2610             mul = 2;
2611             if(strcmp(dstreg, "result.color") == 0) {
2612                 dstreg = "ret";
2613                 mul_final_dest = TRUE;
2614             }
2615         case WINED3DTOP_ADDSIGNED:
2616             shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2617             arg2 = "arg2";
2618         case WINED3DTOP_ADD:
2619             shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2620             break;
2621
2622         case WINED3DTOP_SUBTRACT:
2623             shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2624             break;
2625
2626         case WINED3DTOP_ADDSMOOTH:
2627             shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2628             shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2629             break;
2630
2631         case WINED3DTOP_BLENDCURRENTALPHA:
2632             arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2633             shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2634             break;
2635         case WINED3DTOP_BLENDFACTORALPHA:
2636             arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2637             shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2638             break;
2639         case WINED3DTOP_BLENDTEXTUREALPHA:
2640             arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2641             shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2642             break;
2643         case WINED3DTOP_BLENDDIFFUSEALPHA:
2644             arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2645             shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2646             break;
2647
2648         case WINED3DTOP_BLENDTEXTUREALPHAPM:
2649             arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2650             shader_addline(buffer, "SUB arg0.w, const.x, %s.w;\n", arg0);
2651             shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg, dstmask, arg2, arg1);
2652             break;
2653
2654         /* D3DTOP_PREMODULATE ???? */
2655
2656         case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2657             shader_addline(buffer, "SUB arg0.w, const.x, %s;\n", arg1);
2658             shader_addline(buffer, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2659             break;
2660         case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2661             shader_addline(buffer, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2662             break;
2663         case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2664             shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2665             shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg, dstmask, arg2, arg1);
2666             break;
2667         case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2668             shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg, dstmask, arg1, arg2, arg1);
2669             break;
2670
2671         case WINED3DTOP_DOTPRODUCT3:
2672             mul = 4;
2673             if(strcmp(dstreg, "result.color") == 0) {
2674                 dstreg = "ret";
2675                 mul_final_dest = TRUE;
2676             }
2677             shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2678             shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2679             shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2680             break;
2681
2682         case WINED3DTOP_MULTIPLYADD:
2683             shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2684             break;
2685
2686         case WINED3DTOP_LERP:
2687             /* The msdn is not quite right here */
2688             shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2689             break;
2690
2691         case WINED3DTOP_BUMPENVMAP:
2692         case WINED3DTOP_BUMPENVMAPLUMINANCE:
2693             /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2694             break;
2695
2696         default:
2697             FIXME("Unhandled texture op %08x\n", op);
2698     }
2699
2700     if(mul == 2) {
2701         shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2702     } else if(mul == 4) {
2703         shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2704     }
2705 }
2706
2707 /* The stateblock is passed for GLINFO_LOCATION */
2708 static GLuint gen_arbfp_ffp_shader(const struct ffp_frag_settings *settings, IWineD3DStateBlockImpl *stateblock)
2709 {
2710     unsigned int stage;
2711     SHADER_BUFFER buffer;
2712     BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2713     BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2714     BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2715     const char *textype;
2716     const char *instr, *sat;
2717     char colorcor_dst[8];
2718     GLuint ret;
2719     DWORD arg0, arg1, arg2;
2720     BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2721     BOOL op_equal;
2722     const char *final_combiner_src = "ret";
2723
2724     /* Find out which textures are read */
2725     for(stage = 0; stage < MAX_TEXTURES; stage++) {
2726         if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2727         arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2728         arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2729         arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2730         if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2731         if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2732         if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2733
2734         if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2735         if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2736         if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2737             bump_used[stage] = TRUE;
2738             tex_read[stage] = TRUE;
2739         }
2740         if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2741             bump_used[stage] = TRUE;
2742             tex_read[stage] = TRUE;
2743             luminance_used[stage] = TRUE;
2744         } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2745             tfactor_used = TRUE;
2746         }
2747
2748         if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2749             tfactor_used = TRUE;
2750         }
2751
2752         if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2753         if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2754             tempreg_used = TRUE;
2755         }
2756
2757         if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2758         arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2759         arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2760         arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2761         if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2762         if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2763         if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2764
2765         if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2766             tempreg_used = TRUE;
2767         }
2768         if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2769             tfactor_used = TRUE;
2770         }
2771     }
2772
2773     /* Shader header */
2774     shader_buffer_init(&buffer);
2775
2776     shader_addline(&buffer, "!!ARBfp1.0\n");
2777
2778     switch(settings->fog) {
2779         case FOG_OFF:                                                         break;
2780         case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2781         case FOG_EXP:    shader_addline(&buffer, "OPTION ARB_fog_exp;\n");    break;
2782         case FOG_EXP2:   shader_addline(&buffer, "OPTION ARB_fog_exp2;\n");   break;
2783         default: FIXME("Unexpected fog setting %d\n", settings->fog);
2784     }
2785
2786     shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2787     shader_addline(&buffer, "TEMP TMP;\n");
2788     shader_addline(&buffer, "TEMP ret;\n");
2789     if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2790     shader_addline(&buffer, "TEMP arg0;\n");
2791     shader_addline(&buffer, "TEMP arg1;\n");
2792     shader_addline(&buffer, "TEMP arg2;\n");
2793     for(stage = 0; stage < MAX_TEXTURES; stage++) {
2794         if(!tex_read[stage]) continue;
2795         shader_addline(&buffer, "TEMP tex%u;\n", stage);
2796         if(!bump_used[stage]) continue;
2797         shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2798         if(!luminance_used[stage]) continue;
2799         shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2800     }
2801     if(tfactor_used) {
2802         shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2803     }
2804         shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2805
2806     if(settings->sRGB_write) {
2807         shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2808                        srgb_mul_low, srgb_mul_low, srgb_mul_low);
2809         shader_addline(&buffer, "PARAM srgb_comparison =  {%f, %f, %f, %f};\n",
2810                        srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2811         shader_addline(&buffer, "PARAM srgb_pow =  {%f, %f, %f, 1.0};\n",
2812                        srgb_pow, srgb_pow, srgb_pow);
2813         shader_addline(&buffer, "PARAM srgb_mul_hi =  {%f, %f, %f, 1.0};\n",
2814                        srgb_mul_high, srgb_mul_high, srgb_mul_high);
2815         shader_addline(&buffer, "PARAM srgb_sub_hi =  {%f, %f, %f, 0.0};\n",
2816                        srgb_sub_high, srgb_sub_high, srgb_sub_high);
2817     }
2818
2819     /* Generate texture sampling instructions) */
2820     for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2821         if(!tex_read[stage]) continue;
2822
2823         switch(settings->op[stage].tex_type) {
2824             case tex_1d:                    textype = "1D";     break;
2825             case tex_2d:                    textype = "2D";     break;
2826             case tex_3d:                    textype = "3D";     break;
2827             case tex_cube:                  textype = "CUBE";   break;
2828             case tex_rect:                  textype = "RECT";   break;
2829             default: textype = "unexpected_textype";   break;
2830         }
2831
2832         if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2833            settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2834             sat = "";
2835         } else {
2836             sat = "_SAT";
2837         }
2838
2839         if(settings->op[stage].projected == proj_none) {
2840             instr = "TEX";
2841         } else if(settings->op[stage].projected == proj_count4 ||
2842                   settings->op[stage].projected == proj_count3) {
2843             instr = "TXP";
2844         } else {
2845             FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2846             instr = "TXP";
2847         }
2848
2849         if(stage > 0 &&
2850            (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2851             settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2852             shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2853             shader_addline(&buffer, "DP3 ret.x, arg1, tex%u;\n", stage - 1);
2854             shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2855             shader_addline(&buffer, "DP3 ret.y, arg1, tex%u;\n", stage - 1);
2856
2857             /* with projective textures, texbem only divides the static texture coord, not the displacement,
2858              * so multiply the displacement with the dividing parameter before passing it to TXP
2859              */
2860             if (settings->op[stage].projected != proj_none) {
2861                 if(settings->op[stage].projected == proj_count4) {
2862                     shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].w;\n", stage);
2863                     shader_addline(&buffer, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage, stage);
2864                 } else {
2865                     shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].z;\n", stage);
2866                     shader_addline(&buffer, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage, stage);
2867                 }
2868             } else {
2869                 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2870             }
2871
2872             shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2873                            instr, sat, stage, stage, textype);
2874             if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2875                 shader_addline(&buffer, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
2876                                stage - 1, stage - 1, stage - 1);
2877                 shader_addline(&buffer, "MUL tex%u, tex%u, ret.x;\n", stage, stage);
2878             }
2879         } else if(settings->op[stage].projected == proj_count3) {
2880             shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2881             shader_addline(&buffer, "MOV ret.w, ret.z;\n");
2882             shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2883                             instr, sat, stage, stage, textype);
2884         } else {
2885             shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2886                             instr, sat, stage, stage, stage, textype);
2887         }
2888
2889         sprintf(colorcor_dst, "tex%u", stage);
2890         gen_color_correction(&buffer, colorcor_dst, WINED3DSP_WRITEMASK_ALL, "const.x", "const.y",
2891                 settings->op[stage].color_fixup);
2892     }
2893
2894     /* Generate the main shader */
2895     for(stage = 0; stage < MAX_TEXTURES; stage++) {
2896         if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2897             if(stage == 0) {
2898                 final_combiner_src = "fragment.color.primary";
2899             }
2900             break;
2901         }
2902
2903         if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2904            settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2905             op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2906         } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2907                   settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2908             op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2909         } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2910                   settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2911             op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2912         } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2913                   settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2914             op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2915         } else {
2916             op_equal = settings->op[stage].aop   == settings->op[stage].cop &&
2917                        settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2918                        settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2919                        settings->op[stage].carg2 == settings->op[stage].aarg2;
2920         }
2921
2922         if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2923             gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2924                           settings->op[stage].cop, settings->op[stage].carg0,
2925                           settings->op[stage].carg1, settings->op[stage].carg2);
2926             if(stage == 0) {
2927                 shader_addline(&buffer, "MOV ret.w, fragment.color.primary.w;\n");
2928             }
2929         } else if(op_equal) {
2930             gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2931                           settings->op[stage].cop, settings->op[stage].carg0,
2932                           settings->op[stage].carg1, settings->op[stage].carg2);
2933         } else {
2934             gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2935                           settings->op[stage].cop, settings->op[stage].carg0,
2936                           settings->op[stage].carg1, settings->op[stage].carg2);
2937             gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
2938                           settings->op[stage].aop, settings->op[stage].aarg0,
2939                           settings->op[stage].aarg1, settings->op[stage].aarg2);
2940         }
2941     }
2942
2943     if(settings->sRGB_write) {
2944         shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2945         arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2946         shader_addline(&buffer, "MOV result.color.w, ret.w;\n");
2947     } else {
2948         shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2949     }
2950
2951     /* Footer */
2952     shader_addline(&buffer, "END\n");
2953
2954     /* Generate the shader */
2955     GL_EXTCALL(glGenProgramsARB(1, &ret));
2956     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2957     GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2958
2959     if (glGetError() == GL_INVALID_OPERATION) {
2960         GLint pos;
2961         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2962         FIXME("Fragment program error at position %d: %s\n", pos,
2963               debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2964     }
2965     shader_buffer_free(&buffer);
2966     return ret;
2967 }
2968
2969 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2970     IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2971     struct shader_arb_priv *priv = device->fragment_priv;
2972     BOOL use_pshader = use_ps(stateblock);
2973     BOOL use_vshader = use_vs(stateblock);
2974     struct ffp_frag_settings settings;
2975     const struct arbfp_ffp_desc *desc;
2976     unsigned int i;
2977
2978     if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2979         if(!use_pshader && device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2980             /* Reload fixed function constants since they collide with the pixel shader constants */
2981             for(i = 0; i < MAX_TEXTURES; i++) {
2982                 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2983             }
2984             state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2985             state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2986         } else if(use_pshader && !isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2987             device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2988         }
2989         return;
2990     }
2991
2992     if(!use_pshader) {
2993         /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2994         gen_ffp_frag_op(stateblock, &settings, FALSE);
2995         desc = (const struct arbfp_ffp_desc *)find_ffp_frag_shader(priv->fragment_shaders, &settings);
2996         if(!desc) {
2997             struct arbfp_ffp_desc *new_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc));
2998             if (!new_desc)
2999             {
3000                 ERR("Out of memory\n");
3001                 return;
3002             }
3003             new_desc->num_textures_used = 0;
3004             for(i = 0; i < GL_LIMITS(texture_stages); i++) {
3005                 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
3006                 new_desc->num_textures_used = i;
3007             }
3008
3009             memcpy(&new_desc->parent.settings, &settings, sizeof(settings));
3010             new_desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
3011             add_ffp_frag_shader(priv->fragment_shaders, &new_desc->parent);
3012             TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc);
3013             desc = new_desc;
3014         }
3015
3016         /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3017          * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3018          * deactivate it.
3019          */
3020         GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
3021         checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3022         priv->current_fprogram_id = desc->shader;
3023
3024         if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3025             /* Reload fixed function constants since they collide with the pixel shader constants */
3026             for(i = 0; i < MAX_TEXTURES; i++) {
3027                 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3028             }
3029             state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3030             state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3031         }
3032     }
3033
3034     /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3035      * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3036      * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3037      * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3038      * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3039      *
3040      * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3041      * to be compiled before activating them(needs some cleanups in the shader backend interface)
3042      */
3043     if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
3044         device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
3045
3046         if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
3047             device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
3048         }
3049     }
3050     if(use_pshader) {
3051         device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
3052     }
3053 }
3054
3055 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3056  * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3057  * state table, so we need to handle that with a forwarding function. The other invisible side effect
3058  * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3059  * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3060  */
3061 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3062     enum fogsource new_source;
3063
3064     if(!isStateDirty(context, STATE_PIXELSHADER)) {
3065         fragment_prog_arbfp(state, stateblock, context);
3066     }
3067
3068     if(!stateblock->renderState[WINED3DRS_FOGENABLE]) return;
3069
3070     if(stateblock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
3071         if(use_vs(stateblock)) {
3072             new_source = FOGSOURCE_VS;
3073         } else {
3074             if(stateblock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || context->last_was_rhw) {
3075                 new_source = FOGSOURCE_COORD;
3076             } else {
3077                 new_source = FOGSOURCE_FFP;
3078             }
3079         }
3080     } else {
3081         new_source = FOGSOURCE_FFP;
3082     }
3083     if(new_source != context->fog_source) {
3084         context->fog_source = new_source;
3085         state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART), stateblock, context);
3086     }
3087 }
3088
3089 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3090     if(!isStateDirty(context, STATE_PIXELSHADER)) {
3091         fragment_prog_arbfp(state, stateblock, context);
3092     }
3093 }
3094
3095 #undef GLINFO_LOCATION
3096
3097 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3098     {STATE_RENDER(WINED3DRS_TEXTUREFACTOR),               { STATE_RENDER(WINED3DRS_TEXTUREFACTOR),              state_texfactor_arbfp   }, 0                               },
3099     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3100     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3101     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3102     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3103     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3104     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3105     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3106     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3107     {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3108     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3109     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3110     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3111     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3112     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3113     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3114     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3115     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3116     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3117     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3118     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3119     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3120     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3121     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3122     {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3123     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3124     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3125     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3126     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3127     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3128     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3129     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3130     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3131     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3132     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3133     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3134     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3135     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3136     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3137     {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3138     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3139     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3140     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3141     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3142     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3143     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3144     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3145     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3146     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3147     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3148     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3149     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3150     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3151     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3152     {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3153     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3154     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3155     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3156     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3157     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3158     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3159     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3160     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3161     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3162     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3163     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3164     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3165     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3166     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3167     {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3168     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3169     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3170     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3171     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3172     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3173     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3174     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3175     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3176     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3177     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3178     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3179     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3180     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3181     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3182     {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3183     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3184     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3185     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3186     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3187     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3188     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3189     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3190     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3191     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3192     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3193     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3194     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3195     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3196     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3197     {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3198     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3199     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3200     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3201     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3202     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3203     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3204     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3205     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3206     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3207     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3208     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP),           { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3209     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3210     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3211     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3212     {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG),         { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3213     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3214     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3215     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3216     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     set_bumpmat_arbfp       }, 0                               },
3217     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE),     { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3218     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET),    { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE),    tex_bumpenvlum_arbfp    }, 0                               },
3219     { STATE_SAMPLER(0),                                   { STATE_SAMPLER(0),                                   sampler_texdim          }, 0                               },
3220     { STATE_SAMPLER(1),                                   { STATE_SAMPLER(1),                                   sampler_texdim          }, 0                               },
3221     { STATE_SAMPLER(2),                                   { STATE_SAMPLER(2),                                   sampler_texdim          }, 0                               },
3222     { STATE_SAMPLER(3),                                   { STATE_SAMPLER(3),                                   sampler_texdim          }, 0                               },
3223     { STATE_SAMPLER(4),                                   { STATE_SAMPLER(4),                                   sampler_texdim          }, 0                               },
3224     { STATE_SAMPLER(5),                                   { STATE_SAMPLER(5),                                   sampler_texdim          }, 0                               },
3225     { STATE_SAMPLER(6),                                   { STATE_SAMPLER(6),                                   sampler_texdim          }, 0                               },
3226     { STATE_SAMPLER(7),                                   { STATE_SAMPLER(7),                                   sampler_texdim          }, 0                               },
3227     { STATE_PIXELSHADER,                                  { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3228     { STATE_RENDER(WINED3DRS_FOGENABLE),                  { STATE_RENDER(WINED3DRS_FOGENABLE),                  state_arbfp_fog         }, 0                               },
3229     { STATE_RENDER(WINED3DRS_FOGTABLEMODE),               { STATE_RENDER(WINED3DRS_FOGENABLE),                  state_arbfp_fog         }, 0                               },
3230     { STATE_RENDER(WINED3DRS_FOGVERTEXMODE),              { STATE_RENDER(WINED3DRS_FOGENABLE),                  state_arbfp_fog         }, 0                               },
3231     { STATE_RENDER(WINED3DRS_FOGSTART),                   { STATE_RENDER(WINED3DRS_FOGSTART),                   state_fogstartend       }, 0                               },
3232     { STATE_RENDER(WINED3DRS_FOGEND),                     { STATE_RENDER(WINED3DRS_FOGSTART),                   state_fogstartend       }, 0                               },
3233     { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE),            { STATE_PIXELSHADER,                                  fragment_prog_arbfp     }, 0                               },
3234     { STATE_RENDER(WINED3DRS_FOGCOLOR),                   { STATE_RENDER(WINED3DRS_FOGCOLOR),                   state_fogcolor      }, 0                               },
3235     { STATE_RENDER(WINED3DRS_FOGDENSITY),                 { STATE_RENDER(WINED3DRS_FOGDENSITY),                 state_fogdensity    }, 0                               },
3236     {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3237     {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3238     {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3239     {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3240     {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3241     {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3242     {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3243     {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, 0                               },
3244     { STATE_RENDER(WINED3DRS_SPECULARENABLE),             { STATE_RENDER(WINED3DRS_SPECULARENABLE),             state_arb_specularenable}, 0                               },
3245     {0 /* Terminate */,                                   { 0,                                                  0                       }, 0                               },
3246 };
3247
3248 const struct fragment_pipeline arbfp_fragment_pipeline = {
3249     arbfp_enable,
3250     arbfp_get_caps,
3251     arbfp_alloc,
3252     arbfp_free,
3253     shader_arb_color_fixup_supported,
3254     arbfp_fragmentstate_template,
3255     TRUE /* We can disable projected textures */
3256 };
3257
3258 #define GLINFO_LOCATION device->adapter->gl_info
3259
3260 struct arbfp_blit_priv {
3261     GLenum yuy2_rect_shader, yuy2_2d_shader;
3262     GLenum uyvy_rect_shader, uyvy_2d_shader;
3263     GLenum yv12_rect_shader, yv12_2d_shader;
3264 };
3265
3266 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3267     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3268     device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3269     if(!device->blit_priv) {
3270         ERR("Out of memory\n");
3271         return E_OUTOFMEMORY;
3272     }
3273     return WINED3D_OK;
3274 }
3275 static void arbfp_blit_free(IWineD3DDevice *iface) {
3276     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3277     struct arbfp_blit_priv *priv = device->blit_priv;
3278
3279     ENTER_GL();
3280     GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3281     GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3282     GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3283     GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3284     GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3285     GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3286     checkGLcall("Delete yuv programs\n");
3287     LEAVE_GL();
3288 }
3289
3290 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, enum yuv_fixup yuv_fixup, GLenum textype, char *luminance)
3291 {
3292     char chroma;
3293     const char *tex, *texinstr;
3294
3295     if (yuv_fixup == YUV_FIXUP_UYVY) {
3296         chroma = 'x';
3297         *luminance = 'w';
3298     } else {
3299         chroma = 'w';
3300         *luminance = 'x';
3301     }
3302     switch(textype) {
3303         case GL_TEXTURE_2D:             tex = "2D";     texinstr = "TXP"; break;
3304         case GL_TEXTURE_RECTANGLE_ARB:  tex = "RECT";   texinstr = "TEX"; break;
3305         default:
3306             /* This is more tricky than just replacing the texture type - we have to navigate
3307              * properly in the texture to find the correct chroma values
3308              */
3309             FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3310             return FALSE;
3311     }
3312
3313     /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3314      * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3315      * filtering when we sample the texture.
3316      *
3317      * These are the rules for reading the chroma:
3318      *
3319      * Even pixel: Cr
3320      * Even pixel: U
3321      * Odd pixel: V
3322      *
3323      * So we have to get the sampling x position in non-normalized coordinates in integers
3324      */
3325     if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3326         shader_addline(buffer, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
3327         shader_addline(buffer, "MOV texcrd.w, size.x;\n");
3328     } else {
3329         shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3330     }
3331     /* We must not allow filtering between pixel x and x+1, this would mix U and V
3332      * Vertical filtering is ok. However, bear in mind that the pixel center is at
3333      * 0.5, so add 0.5.
3334      */
3335     shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3336     shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3337
3338     /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3339      * even and odd pixels respectively
3340      */
3341     shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3342     shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3343
3344     /* Sample Pixel 1 */
3345     shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3346
3347     /* Put the value into either of the chroma values */
3348     shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3349     shader_addline(buffer, "MUL chroma.x, luminance.%c, temp.x;\n", chroma);
3350     shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3351     shader_addline(buffer, "MUL chroma.y, luminance.%c, temp.x;\n", chroma);
3352
3353     /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3354      * the pixel right to the current one. Otherwise, sample the left pixel.
3355      * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3356      */
3357     shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3358     shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3359     shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3360
3361     /* Put the value into the other chroma */
3362     shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3363     shader_addline(buffer, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma);
3364     shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3365     shader_addline(buffer, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma);
3366
3367     /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3368      * the current one and lerp the two U and V values
3369      */
3370
3371     /* This gives the correctly filtered luminance value */
3372     shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3373
3374     return TRUE;
3375 }
3376
3377 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, GLenum textype, char *luminance)
3378 {
3379     const char *tex;
3380
3381     switch(textype) {
3382         case GL_TEXTURE_2D:             tex = "2D";     break;
3383         case GL_TEXTURE_RECTANGLE_ARB:  tex = "RECT";   break;
3384         default:
3385             FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3386             return FALSE;
3387     }
3388
3389     /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3390      * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3391      * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3392      * pitch of the luminance plane, the packing into the gl texture is a bit
3393      * unfortunate. If the whole texture is interpreted as luminance data it looks
3394      * approximately like this:
3395      *
3396      *        +----------------------------------+----
3397      *        |                                  |
3398      *        |                                  |
3399      *        |                                  |
3400      *        |                                  |
3401      *        |                                  |   2
3402      *        |            LUMINANCE             |   -
3403      *        |                                  |   3
3404      *        |                                  |
3405      *        |                                  |
3406      *        |                                  |
3407      *        |                                  |
3408      *        +----------------+-----------------+----
3409      *        |                |                 |
3410      *        |  U even rows   |  U odd rows     |
3411      *        |                |                 |   1
3412      *        +----------------+------------------   -
3413      *        |                |                 |   3
3414      *        |  V even rows   |  V odd rows     |
3415      *        |                |                 |
3416      *        +----------------+-----------------+----
3417      *        |                |                 |
3418      *        |     0.5        |       0.5       |
3419      *
3420      * So it appears as if there are 4 chroma images, but in fact the odd rows
3421      * in the chroma images are in the same row as the even ones. So its is
3422      * kinda tricky to read
3423      *
3424      * When reading from rectangle textures, keep in mind that the input y coordinates
3425      * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3426      */
3427     shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3428                    2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3429
3430     shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3431     /* the chroma planes have only half the width */
3432     shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3433
3434     /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3435      * the coordinate. Also read the right side of the image when reading odd lines
3436      *
3437      * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3438      * bleeding
3439      */
3440     if(textype == GL_TEXTURE_2D) {
3441
3442         shader_addline(buffer, "RCP chroma.w, size.y;\n");
3443
3444         shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3445
3446         shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3447         shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3448
3449         /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3450         shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3451         shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3452         shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3453         shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3454
3455         /* clamp, keep the half pixel origin in mind */
3456         shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3457         shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3458         shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3459         shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3460     } else {
3461         /* Read from [size - size+size/4] */
3462         shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3463         shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3464
3465         /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3466         shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3467         shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3468         shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3469         shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3470         shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3471
3472         /* Make sure to read exactly from the pixel center */
3473         shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3474         shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3475
3476         /* Clamp */
3477         shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3478         shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3479         shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3480         shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3481         shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3482     }
3483     /* Read the texture, put the result into the output register */
3484     shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3485     shader_addline(buffer, "MOV chroma.x, temp.w;\n");
3486
3487     /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3488      * No need to clamp because we're just reusing the already clamped value from above
3489      */
3490     if(textype == GL_TEXTURE_2D) {
3491         shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3492     } else {
3493         shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3494     }
3495     shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3496     shader_addline(buffer, "MOV chroma.y, temp.w;\n");
3497
3498     /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3499      * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3500      * values due to filtering
3501      */
3502     shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3503     if(textype == GL_TEXTURE_2D) {
3504         /* Multiply the y coordinate by 2/3 and clamp it */
3505         shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3506         shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3507         shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3508         shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3509     } else {
3510         /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
3511          * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3512          * is bigger
3513          */
3514         shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3515         shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3516         shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3517     }
3518     *luminance = 'a';
3519
3520     return TRUE;
3521 }
3522
3523 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, enum yuv_fixup yuv_fixup, GLenum textype)
3524 {
3525     GLenum shader;
3526     SHADER_BUFFER buffer;
3527     char luminance_component;
3528     struct arbfp_blit_priv *priv = device->blit_priv;
3529
3530     /* Shader header */
3531     shader_buffer_init(&buffer);
3532
3533     ENTER_GL();
3534     GL_EXTCALL(glGenProgramsARB(1, &shader));
3535     checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3536     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3537     checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3538     LEAVE_GL();
3539     if(!shader) {
3540         shader_buffer_free(&buffer);
3541         return 0;
3542     }
3543
3544     /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3545      * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3546      * two chroma(U and V) values. Each macropixel has two luminance values, one for
3547      * each single pixel it contains, and one U and one V value shared between both
3548      * pixels.
3549      *
3550      * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3551      * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3552      * take the format into account when generating the read swizzles
3553      *
3554      * Reading the Y value is straightforward - just sample the texture. The hardware
3555      * takes care of filtering in the horizontal and vertical direction.
3556      *
3557      * Reading the U and V values is harder. We have to avoid filtering horizontally,
3558      * because that would mix the U and V values of one pixel or two adjacent pixels.
3559      * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3560      * regardless of the filtering setting. Vertical filtering works automatically
3561      * though - the U and V values of two rows are mixed nicely.
3562      *
3563      * Appart of avoiding filtering issues, the code has to know which value it just
3564      * read, and where it can find the other one. To determine this, it checks if
3565      * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3566      *
3567      * Handling horizontal filtering of U and V values requires reading a 2nd pair
3568      * of pixels, extracting U and V and mixing them. This is not implemented yet.
3569      *
3570      * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3571      * with width / 2. This way one read gives all 3 values, finding U and V is easy
3572      * in an unfiltered situation. Finding the luminance on the other hand requires
3573      * finding out if it is an odd or even pixel. The real drawback of this approach
3574      * is filtering. This would have to be emulated completely in the shader, reading
3575      * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3576      * vertically. Beyond that it would require adjustments to the texture handling
3577      * code to deal with the width scaling
3578      */
3579     shader_addline(&buffer, "!!ARBfp1.0\n");
3580     shader_addline(&buffer, "TEMP luminance;\n");
3581     shader_addline(&buffer, "TEMP temp;\n");
3582     shader_addline(&buffer, "TEMP chroma;\n");
3583     shader_addline(&buffer, "TEMP texcrd;\n");
3584     shader_addline(&buffer, "TEMP texcrd2;\n");
3585     shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3586     shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3587     shader_addline(&buffer, "PARAM size = program.local[0];\n");
3588
3589     switch (yuv_fixup)
3590     {
3591         case YUV_FIXUP_UYVY:
3592         case YUV_FIXUP_YUY2:
3593             if (!gen_planar_yuv_read(&buffer, yuv_fixup, textype, &luminance_component))
3594             {
3595                 shader_buffer_free(&buffer);
3596                 return 0;
3597             }
3598             break;
3599
3600         case YUV_FIXUP_YV12:
3601             if (!gen_yv12_read(&buffer, textype, &luminance_component))
3602             {
3603                 shader_buffer_free(&buffer);
3604                 return 0;
3605             }
3606             break;
3607
3608         default:
3609             FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3610             shader_buffer_free(&buffer);
3611             return 0;
3612     }
3613
3614     /* Calculate the final result. Formula is taken from
3615      * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3616      * ranges from -0.5 to 0.5
3617      */
3618     shader_addline(&buffer, "SUB chroma.xy, chroma, coef.y;\n");
3619
3620     shader_addline(&buffer, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component);
3621     shader_addline(&buffer, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component);
3622     shader_addline(&buffer, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
3623     shader_addline(&buffer, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component);
3624     shader_addline(&buffer, "END\n");
3625
3626     ENTER_GL();
3627     GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3628
3629     if (glGetError() == GL_INVALID_OPERATION) {
3630         GLint pos;
3631         glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3632         FIXME("Fragment program error at position %d: %s\n", pos,
3633               debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3634     }
3635     shader_buffer_free(&buffer);
3636     LEAVE_GL();
3637
3638     switch (yuv_fixup)
3639     {
3640         case YUV_FIXUP_YUY2:
3641             if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yuy2_rect_shader = shader;
3642             else priv->yuy2_2d_shader = shader;
3643             break;
3644
3645         case YUV_FIXUP_UYVY:
3646             if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->uyvy_rect_shader = shader;
3647             else priv->uyvy_2d_shader = shader;
3648             break;
3649
3650         case YUV_FIXUP_YV12:
3651             if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yv12_rect_shader = shader;
3652             else priv->yv12_2d_shader = shader;
3653             break;
3654     }
3655
3656     return shader;
3657 }
3658
3659 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3660     GLenum shader;
3661     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3662     float size[4] = {width, height, 1, 1};
3663     struct arbfp_blit_priv *priv = device->blit_priv;
3664     const struct GlPixelFormatDesc *glDesc;
3665     enum yuv_fixup yuv_fixup;
3666
3667     getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3668
3669     if (!is_yuv_fixup(glDesc->color_fixup))
3670     {
3671         TRACE("Fixup:\n");
3672         dump_color_fixup_desc(glDesc->color_fixup);
3673         /* Don't bother setting up a shader for unconverted formats */
3674         ENTER_GL();
3675         glEnable(textype);
3676         checkGLcall("glEnable(textype)");
3677         LEAVE_GL();
3678         return WINED3D_OK;
3679     }
3680
3681     yuv_fixup = get_yuv_fixup(glDesc->color_fixup);
3682
3683     switch(yuv_fixup)
3684     {
3685         case YUV_FIXUP_YUY2:
3686             shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yuy2_rect_shader : priv->yuy2_2d_shader;
3687             break;
3688
3689         case YUV_FIXUP_UYVY:
3690             shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->uyvy_rect_shader : priv->uyvy_2d_shader;
3691             break;
3692
3693         case YUV_FIXUP_YV12:
3694             shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yv12_rect_shader : priv->yv12_2d_shader;
3695             break;
3696
3697         default:
3698             FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup);
3699             ENTER_GL();
3700             glEnable(textype);
3701             checkGLcall("glEnable(textype)");
3702             LEAVE_GL();
3703             return E_NOTIMPL;
3704     }
3705
3706     if (!shader) shader = gen_yuv_shader(device, yuv_fixup, textype);
3707
3708     ENTER_GL();
3709     glEnable(GL_FRAGMENT_PROGRAM_ARB);
3710     checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3711     GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3712     checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3713     GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3714     checkGLcall("glProgramLocalParameter4fvARB");
3715     LEAVE_GL();
3716
3717     return WINED3D_OK;
3718 }
3719
3720 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3721     IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3722
3723     ENTER_GL();
3724     glDisable(GL_FRAGMENT_PROGRAM_ARB);
3725     checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3726     glDisable(GL_TEXTURE_2D);
3727     checkGLcall("glDisable(GL_TEXTURE_2D)");
3728     if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3729         glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3730         checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3731     }
3732     if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3733         glDisable(GL_TEXTURE_RECTANGLE_ARB);
3734         checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3735     }
3736     LEAVE_GL();
3737 }
3738
3739 static BOOL arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup)
3740 {
3741     enum yuv_fixup yuv_fixup;
3742
3743     if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
3744     {
3745         TRACE("Checking support for fixup:\n");
3746         dump_color_fixup_desc(fixup);
3747     }
3748
3749     if (is_identity_fixup(fixup))
3750     {
3751         TRACE("[OK]\n");
3752         return TRUE;
3753     }
3754
3755     /* We only support YUV conversions. */
3756     if (!is_yuv_fixup(fixup))
3757     {
3758         TRACE("[FAILED]\n");
3759         return FALSE;
3760     }
3761
3762     yuv_fixup = get_yuv_fixup(fixup);
3763     switch(yuv_fixup)
3764     {
3765         case YUV_FIXUP_YUY2:
3766         case YUV_FIXUP_UYVY:
3767         case YUV_FIXUP_YV12:
3768             TRACE("[OK]\n");
3769             return TRUE;
3770
3771         default:
3772             FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3773             TRACE("[FAILED]\n");
3774             return FALSE;
3775     }
3776 }
3777
3778 const struct blit_shader arbfp_blit = {
3779     arbfp_blit_alloc,
3780     arbfp_blit_free,
3781     arbfp_blit_set,
3782     arbfp_blit_unset,
3783     arbfp_blit_color_fixup_supported,
3784 };
3785
3786 #undef GLINFO_LOCATION