wined3d: Get rid of the WINED3DDEVICE_CREATION_PARAMETERS typedef.
[wine] / dlls / wined3d / ati_fragment_shader.c
1 /*
2  * Fixed function pipeline replacement using GL_ATI_fragment_shader
3  *
4  * Copyright 2008 Stefan Dösinger(for CodeWeavers)
5  *
6  * This library is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * This library is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with this library; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
19  */
20
21 #include "config.h"
22
23 #include <math.h>
24 #include <stdio.h>
25
26 #include "wined3d_private.h"
27
28 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
29 WINE_DECLARE_DEBUG_CHANNEL(d3d);
30
31 /* GL locking for state handlers is done by the caller. */
32
33 /* Some private defines, Constant associations, etc.
34  * Env bump matrix and per stage constant should be independent,
35  * a stage that bump maps can't read the per state constant
36  */
37 #define ATI_FFP_CONST_BUMPMAT(i) (GL_CON_0_ATI + i)
38 #define ATI_FFP_CONST_CONSTANT0 GL_CON_0_ATI
39 #define ATI_FFP_CONST_CONSTANT1 GL_CON_1_ATI
40 #define ATI_FFP_CONST_CONSTANT2 GL_CON_2_ATI
41 #define ATI_FFP_CONST_CONSTANT3 GL_CON_3_ATI
42 #define ATI_FFP_CONST_CONSTANT4 GL_CON_4_ATI
43 #define ATI_FFP_CONST_CONSTANT5 GL_CON_5_ATI
44 #define ATI_FFP_CONST_TFACTOR   GL_CON_6_ATI
45
46 /* GL_ATI_fragment_shader specific fixed function pipeline description. "Inherits" from the common one */
47 struct atifs_ffp_desc
48 {
49     struct ffp_frag_desc parent;
50     GLuint shader;
51     unsigned int num_textures_used;
52 };
53
54 struct atifs_private_data
55 {
56     struct wine_rb_tree fragment_shaders; /* A rb-tree to track fragment pipeline replacement shaders */
57 };
58
59 static const char *debug_dstmod(GLuint mod) {
60     switch(mod) {
61         case GL_NONE:               return "GL_NONE";
62         case GL_2X_BIT_ATI:         return "GL_2X_BIT_ATI";
63         case GL_4X_BIT_ATI:         return "GL_4X_BIT_ATI";
64         case GL_8X_BIT_ATI:         return "GL_8X_BIT_ATI";
65         case GL_HALF_BIT_ATI:       return "GL_HALF_BIT_ATI";
66         case GL_QUARTER_BIT_ATI:    return "GL_QUARTER_BIT_ATI";
67         case GL_EIGHTH_BIT_ATI:     return "GL_EIGHTH_BIT_ATI";
68         case GL_SATURATE_BIT_ATI:   return "GL_SATURATE_BIT_ATI";
69         default:                    return "Unexpected modifier\n";
70     }
71 }
72
73 static const char *debug_argmod(GLuint mod) {
74     switch(mod) {
75         case GL_NONE:
76             return "GL_NONE";
77
78         case GL_2X_BIT_ATI:
79             return "GL_2X_BIT_ATI";
80         case GL_COMP_BIT_ATI:
81             return "GL_COMP_BIT_ATI";
82         case GL_NEGATE_BIT_ATI:
83             return "GL_NEGATE_BIT_ATI";
84         case GL_BIAS_BIT_ATI:
85             return "GL_BIAS_BIT_ATI";
86
87         case GL_2X_BIT_ATI | GL_COMP_BIT_ATI:
88             return "GL_2X_BIT_ATI | GL_COMP_BIT_ATI";
89         case GL_2X_BIT_ATI | GL_NEGATE_BIT_ATI:
90             return "GL_2X_BIT_ATI | GL_NEGATE_BIT_ATI";
91         case GL_2X_BIT_ATI | GL_BIAS_BIT_ATI:
92             return "GL_2X_BIT_ATI | GL_BIAS_BIT_ATI";
93         case GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI:
94             return "GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI";
95         case GL_COMP_BIT_ATI | GL_BIAS_BIT_ATI:
96             return "GL_COMP_BIT_ATI | GL_BIAS_BIT_ATI";
97         case GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI:
98             return "GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI";
99
100         case GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI:
101             return "GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI";
102         case GL_2X_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI:
103             return "GL_2X_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI";
104         case GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_BIAS_BIT_ATI:
105             return "GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_BIAS_BIT_ATI";
106         case GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI:
107             return "GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI";
108
109         case GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI:
110             return "GL_2X_BIT_ATI | GL_COMP_BIT_ATI | GL_NEGATE_BIT_ATI | GL_BIAS_BIT_ATI";
111
112         default:
113             return "Unexpected argmod combination\n";
114     }
115 }
116 static const char *debug_register(GLuint reg) {
117     switch(reg) {
118         case GL_REG_0_ATI:                  return "GL_REG_0_ATI";
119         case GL_REG_1_ATI:                  return "GL_REG_1_ATI";
120         case GL_REG_2_ATI:                  return "GL_REG_2_ATI";
121         case GL_REG_3_ATI:                  return "GL_REG_3_ATI";
122         case GL_REG_4_ATI:                  return "GL_REG_4_ATI";
123         case GL_REG_5_ATI:                  return "GL_REG_5_ATI";
124
125         case GL_CON_0_ATI:                  return "GL_CON_0_ATI";
126         case GL_CON_1_ATI:                  return "GL_CON_1_ATI";
127         case GL_CON_2_ATI:                  return "GL_CON_2_ATI";
128         case GL_CON_3_ATI:                  return "GL_CON_3_ATI";
129         case GL_CON_4_ATI:                  return "GL_CON_4_ATI";
130         case GL_CON_5_ATI:                  return "GL_CON_5_ATI";
131         case GL_CON_6_ATI:                  return "GL_CON_6_ATI";
132         case GL_CON_7_ATI:                  return "GL_CON_7_ATI";
133
134         case GL_ZERO:                       return "GL_ZERO";
135         case GL_ONE:                        return "GL_ONE";
136         case GL_PRIMARY_COLOR:              return "GL_PRIMARY_COLOR";
137         case GL_SECONDARY_INTERPOLATOR_ATI: return "GL_SECONDARY_INTERPOLATOR_ATI";
138
139         default:                            return "Unknown register\n";
140     }
141 }
142
143 static const char *debug_swizzle(GLuint swizzle) {
144     switch(swizzle) {
145         case GL_SWIZZLE_STR_ATI:        return "GL_SWIZZLE_STR_ATI";
146         case GL_SWIZZLE_STQ_ATI:        return "GL_SWIZZLE_STQ_ATI";
147         case GL_SWIZZLE_STR_DR_ATI:     return "GL_SWIZZLE_STR_DR_ATI";
148         case GL_SWIZZLE_STQ_DQ_ATI:     return "GL_SWIZZLE_STQ_DQ_ATI";
149         default:                        return "unknown swizzle";
150     }
151 }
152
153 static const char *debug_rep(GLuint rep) {
154     switch(rep) {
155         case GL_NONE:                   return "GL_NONE";
156         case GL_RED:                    return "GL_RED";
157         case GL_GREEN:                  return "GL_GREEN";
158         case GL_BLUE:                   return "GL_BLUE";
159         default:                        return "unknown argrep";
160     }
161 }
162
163 static const char *debug_op(GLuint op) {
164     switch(op) {
165         case GL_MOV_ATI:                return "GL_MOV_ATI";
166         case GL_ADD_ATI:                return "GL_ADD_ATI";
167         case GL_MUL_ATI:                return "GL_MUL_ATI";
168         case GL_SUB_ATI:                return "GL_SUB_ATI";
169         case GL_DOT3_ATI:               return "GL_DOT3_ATI";
170         case GL_DOT4_ATI:               return "GL_DOT4_ATI";
171         case GL_MAD_ATI:                return "GL_MAD_ATI";
172         case GL_LERP_ATI:               return "GL_LERP_ATI";
173         case GL_CND_ATI:                return "GL_CND_ATI";
174         case GL_CND0_ATI:               return "GL_CND0_ATI";
175         case GL_DOT2_ADD_ATI:           return "GL_DOT2_ADD_ATI";
176         default:                        return "unexpected op";
177     }
178 }
179
180 static const char *debug_mask(GLuint mask) {
181     switch(mask) {
182         case GL_NONE:                           return "GL_NONE";
183         case GL_RED_BIT_ATI:                    return "GL_RED_BIT_ATI";
184         case GL_GREEN_BIT_ATI:                  return "GL_GREEN_BIT_ATI";
185         case GL_BLUE_BIT_ATI:                   return "GL_BLUE_BIT_ATI";
186         case GL_RED_BIT_ATI | GL_GREEN_BIT_ATI: return "GL_RED_BIT_ATI | GL_GREEN_BIT_ATI";
187         case GL_RED_BIT_ATI | GL_BLUE_BIT_ATI:  return "GL_RED_BIT_ATI | GL_BLUE_BIT_ATI";
188         case GL_GREEN_BIT_ATI | GL_BLUE_BIT_ATI:return "GL_GREEN_BIT_ATI | GL_BLUE_BIT_ATI";
189         case GL_RED_BIT_ATI | GL_GREEN_BIT_ATI | GL_BLUE_BIT_ATI:return "GL_RED_BIT_ATI | GL_GREEN_BIT_ATI | GL_BLUE_BIT_ATI";
190         default:                                return "Unexpected writemask";
191     }
192 }
193
194 static void wrap_op1(const struct wined3d_gl_info *gl_info, GLuint op, GLuint dst, GLuint dstMask, GLuint dstMod,
195         GLuint arg1, GLuint arg1Rep, GLuint arg1Mod)
196 {
197     if(dstMask == GL_ALPHA) {
198         TRACE("glAlphaFragmentOp1ATI(%s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst), debug_dstmod(dstMod),
199               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod));
200         GL_EXTCALL(glAlphaFragmentOp1ATI(op, dst, dstMod, arg1, arg1Rep, arg1Mod));
201     } else {
202         TRACE("glColorFragmentOp1ATI(%s, %s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst),
203               debug_mask(dstMask), debug_dstmod(dstMod),
204               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod));
205         GL_EXTCALL(glColorFragmentOp1ATI(op, dst, dstMask, dstMod, arg1, arg1Rep, arg1Mod));
206     }
207 }
208
209 static void wrap_op2(const struct wined3d_gl_info *gl_info, GLuint op, GLuint dst, GLuint dstMask, GLuint dstMod,
210         GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod)
211 {
212     if(dstMask == GL_ALPHA) {
213         TRACE("glAlphaFragmentOp2ATI(%s, %s, %s, %s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst), debug_dstmod(dstMod),
214               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod),
215               debug_register(arg2), debug_rep(arg2Rep), debug_argmod(arg2Mod));
216         GL_EXTCALL(glAlphaFragmentOp2ATI(op, dst, dstMod, arg1, arg1Rep, arg1Mod, arg2, arg2Rep, arg2Mod));
217     } else {
218         TRACE("glColorFragmentOp2ATI(%s, %s, %s, %s, %s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst),
219               debug_mask(dstMask), debug_dstmod(dstMod),
220               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod),
221               debug_register(arg2), debug_rep(arg2Rep), debug_argmod(arg2Mod));
222         GL_EXTCALL(glColorFragmentOp2ATI(op, dst, dstMask, dstMod, arg1, arg1Rep, arg1Mod, arg2, arg2Rep, arg2Mod));
223     }
224 }
225
226 static void wrap_op3(const struct wined3d_gl_info *gl_info, GLuint op, GLuint dst, GLuint dstMask, GLuint dstMod,
227         GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod,
228         GLuint arg3, GLuint arg3Rep, GLuint arg3Mod)
229 {
230     if(dstMask == GL_ALPHA) {
231         /* Leave some free space to fit "GL_NONE, " in to align most alpha and color op lines */
232         TRACE("glAlphaFragmentOp3ATI(%s, %s,          %s, %s, %s, %s, %s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst), debug_dstmod(dstMod),
233               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod),
234               debug_register(arg2), debug_rep(arg2Rep), debug_argmod(arg2Mod),
235               debug_register(arg3), debug_rep(arg3Rep), debug_argmod(arg3Mod));
236         GL_EXTCALL(glAlphaFragmentOp3ATI(op, dst, dstMod,
237                                          arg1, arg1Rep, arg1Mod,
238                                          arg2, arg2Rep, arg2Mod,
239                                          arg3, arg3Rep, arg3Mod));
240     } else {
241         TRACE("glColorFragmentOp3ATI(%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s)\n", debug_op(op), debug_register(dst),
242               debug_mask(dstMask), debug_dstmod(dstMod),
243               debug_register(arg1), debug_rep(arg1Rep), debug_argmod(arg1Mod),
244               debug_register(arg2), debug_rep(arg2Rep), debug_argmod(arg2Mod),
245               debug_register(arg3), debug_rep(arg3Rep), debug_argmod(arg3Mod));
246         GL_EXTCALL(glColorFragmentOp3ATI(op, dst, dstMask, dstMod,
247                                          arg1, arg1Rep, arg1Mod,
248                                          arg2, arg2Rep, arg2Mod,
249                                          arg3, arg3Rep, arg3Mod));
250     }
251 }
252
253 static GLuint register_for_arg(DWORD arg, const struct wined3d_gl_info *gl_info,
254         unsigned int stage, GLuint *mod, GLuint *rep, GLuint tmparg)
255 {
256     GLenum ret;
257
258     if(mod) *mod = GL_NONE;
259     if(arg == ARG_UNUSED)
260     {
261         if (rep) *rep = GL_NONE;
262         return -1; /* This is the marker for unused registers */
263     }
264
265     switch(arg & WINED3DTA_SELECTMASK) {
266         case WINED3DTA_DIFFUSE:
267             ret = GL_PRIMARY_COLOR;
268             break;
269
270         case WINED3DTA_CURRENT:
271             /* Note that using GL_REG_0_ATI for the passed on register is safe because
272              * texture0 is read at stage0, so in the worst case it is read in the
273              * instruction writing to reg0. Afterwards texture0 is not used any longer.
274              * If we're reading from current
275              */
276             ret = stage ? GL_REG_0_ATI : GL_PRIMARY_COLOR;
277             break;
278
279         case WINED3DTA_TEXTURE:
280             ret = GL_REG_0_ATI + stage;
281             break;
282
283         case WINED3DTA_TFACTOR:
284             ret = ATI_FFP_CONST_TFACTOR;
285             break;
286
287         case WINED3DTA_SPECULAR:
288             ret = GL_SECONDARY_INTERPOLATOR_ATI;
289             break;
290
291         case WINED3DTA_TEMP:
292             ret = tmparg;
293             break;
294
295         case WINED3DTA_CONSTANT:
296             FIXME("Unhandled source argument WINED3DTA_TEMP\n");
297             ret = GL_CON_0_ATI;
298             break;
299
300         default:
301             FIXME("Unknown source argument %d\n", arg);
302             ret = GL_ZERO;
303     }
304
305     if(arg & WINED3DTA_COMPLEMENT) {
306         if(mod) *mod |= GL_COMP_BIT_ATI;
307     }
308     if(arg & WINED3DTA_ALPHAREPLICATE) {
309         if(rep) *rep = GL_ALPHA;
310     } else {
311         if(rep) *rep = GL_NONE;
312     }
313     return ret;
314 }
315
316 static GLuint find_tmpreg(const struct texture_stage_op op[MAX_TEXTURES])
317 {
318     int lowest_read = -1;
319     int lowest_write = -1;
320     int i;
321     BOOL tex_used[MAX_TEXTURES];
322
323     memset(tex_used, 0, sizeof(tex_used));
324     for(i = 0; i < MAX_TEXTURES; i++) {
325         if(op[i].cop == WINED3DTOP_DISABLE) {
326             break;
327         }
328
329         if(lowest_read == -1 &&
330           (op[i].carg1 == WINED3DTA_TEMP || op[i].carg2 == WINED3DTA_TEMP || op[i].carg0 == WINED3DTA_TEMP ||
331            op[i].aarg1 == WINED3DTA_TEMP || op[i].aarg2 == WINED3DTA_TEMP || op[i].aarg0 == WINED3DTA_TEMP)) {
332             lowest_read = i;
333         }
334
335         if(lowest_write == -1 && op[i].dst == tempreg) {
336             lowest_write = i;
337         }
338
339         if(op[i].carg1 == WINED3DTA_TEXTURE || op[i].carg2 == WINED3DTA_TEXTURE || op[i].carg0 == WINED3DTA_TEXTURE ||
340            op[i].aarg1 == WINED3DTA_TEXTURE || op[i].aarg2 == WINED3DTA_TEXTURE || op[i].aarg0 == WINED3DTA_TEXTURE) {
341             tex_used[i] = TRUE;
342         }
343     }
344
345     /* Temp reg not read? We don't need it, return GL_NONE */
346     if(lowest_read == -1) return GL_NONE;
347
348     if(lowest_write >= lowest_read) {
349         FIXME("Temp register read before being written\n");
350     }
351
352     if(lowest_write == -1) {
353         /* This needs a test. Maybe we are supposed to return 0.0/0.0/0.0/0.0, or fail drawprim, or whatever */
354         FIXME("Temp register read without being written\n");
355         return GL_REG_1_ATI;
356     } else if(lowest_write >= 1) {
357         /* If we're writing to the temp reg at earliest in stage 1, we can use register 1 for the temp result.
358          * there may be texture data stored in reg 1, but we do not need it any longer since stage 1 already
359          * read it
360          */
361         return GL_REG_1_ATI;
362     } else {
363         /* Search for a free texture register. We have 6 registers available. GL_REG_0_ATI is already used
364          * for the regular result
365          */
366         for(i = 1; i < 6; i++) {
367             if(!tex_used[i]) {
368                 return GL_REG_0_ATI + i;
369             }
370         }
371         /* What to do here? Report it in ValidateDevice? */
372         FIXME("Could not find a register for the temporary register\n");
373         return 0;
374     }
375 }
376
377 static GLuint gen_ati_shader(const struct texture_stage_op op[MAX_TEXTURES], const struct wined3d_gl_info *gl_info)
378 {
379     GLuint ret = GL_EXTCALL(glGenFragmentShadersATI(1));
380     unsigned int stage;
381     GLuint arg0, arg1, arg2, extrarg;
382     GLuint dstmod, argmod0, argmod1, argmod2, argmodextra;
383     GLuint rep0, rep1, rep2;
384     GLuint swizzle;
385     GLuint tmparg = find_tmpreg(op);
386     GLuint dstreg;
387
388     if(!ret) {
389         ERR("Failed to generate a GL_ATI_fragment_shader shader id\n");
390         return 0;
391     }
392     GL_EXTCALL(glBindFragmentShaderATI(ret));
393     checkGLcall("GL_EXTCALL(glBindFragmentShaderATI(ret))");
394
395     TRACE("glBeginFragmentShaderATI()\n");
396     GL_EXTCALL(glBeginFragmentShaderATI());
397     checkGLcall("GL_EXTCALL(glBeginFragmentShaderATI())");
398
399     /* Pass 1: Generate sampling instructions for perturbation maps */
400     for (stage = 0; stage < gl_info->limits.textures; ++stage)
401     {
402         if(op[stage].cop == WINED3DTOP_DISABLE) break;
403         if(op[stage].cop != WINED3DTOP_BUMPENVMAP &&
404            op[stage].cop != WINED3DTOP_BUMPENVMAPLUMINANCE) continue;
405
406         TRACE("glSampleMapATI(GL_REG_%d_ATI, GL_TEXTURE_%d_ARB, GL_SWIZZLE_STR_ATI)\n",
407               stage, stage);
408         GL_EXTCALL(glSampleMapATI(GL_REG_0_ATI + stage,
409                    GL_TEXTURE0_ARB + stage,
410                    GL_SWIZZLE_STR_ATI));
411         if(op[stage + 1].projected == proj_none) {
412             swizzle = GL_SWIZZLE_STR_ATI;
413         } else if(op[stage + 1].projected == proj_count4) {
414             swizzle = GL_SWIZZLE_STQ_DQ_ATI;
415         } else {
416             swizzle = GL_SWIZZLE_STR_DR_ATI;
417         }
418         TRACE("glPassTexCoordATI(GL_REG_%d_ATI, GL_TEXTURE_%d_ARB, %s)\n",
419               stage + 1, stage + 1, debug_swizzle(swizzle));
420         GL_EXTCALL(glPassTexCoordATI(GL_REG_0_ATI + stage + 1,
421                    GL_TEXTURE0_ARB + stage + 1,
422                    swizzle));
423     }
424
425     /* Pass 2: Generate perturbation calculations */
426     for (stage = 0; stage < gl_info->limits.textures; ++stage)
427     {
428         GLuint argmodextra_x, argmodextra_y;
429         struct color_fixup_desc fixup;
430
431         if(op[stage].cop == WINED3DTOP_DISABLE) break;
432         if(op[stage].cop != WINED3DTOP_BUMPENVMAP &&
433            op[stage].cop != WINED3DTOP_BUMPENVMAPLUMINANCE) continue;
434
435         fixup = op[stage].color_fixup;
436         if (fixup.x_source != CHANNEL_SOURCE_X || fixup.y_source != CHANNEL_SOURCE_Y)
437         {
438             FIXME("Swizzles not implemented\n");
439             argmodextra_x = GL_NONE;
440             argmodextra_y = GL_NONE;
441         }
442         else
443         {
444             /* Nice thing, we get the color correction for free :-) */
445             argmodextra_x = fixup.x_sign_fixup ? GL_2X_BIT_ATI | GL_BIAS_BIT_ATI : GL_NONE;
446             argmodextra_y = fixup.y_sign_fixup ? GL_2X_BIT_ATI | GL_BIAS_BIT_ATI : GL_NONE;
447         }
448
449         wrap_op3(gl_info, GL_DOT2_ADD_ATI, GL_REG_0_ATI + stage + 1, GL_RED_BIT_ATI, GL_NONE,
450                  GL_REG_0_ATI + stage, GL_NONE, argmodextra_x,
451                  ATI_FFP_CONST_BUMPMAT(stage), GL_NONE, GL_2X_BIT_ATI | GL_BIAS_BIT_ATI,
452                  GL_REG_0_ATI + stage + 1, GL_RED, GL_NONE);
453
454         /* Don't use GL_DOT2_ADD_ATI here because we cannot configure it to read the blue and alpha
455          * component of the bump matrix. Instead do this with two MADs:
456          *
457          * coord.a = tex.r * bump.b + coord.g
458          * coord.g = tex.g * bump.a + coord.a
459          *
460          * The first instruction writes to alpha so it can be coissued with the above DOT2_ADD.
461          * coord.a is unused. If the perturbed texture is projected, this was already handled
462          * in the glPassTexCoordATI above.
463          */
464         wrap_op3(gl_info, GL_MAD_ATI, GL_REG_0_ATI + stage + 1, GL_ALPHA, GL_NONE,
465                  GL_REG_0_ATI + stage, GL_RED, argmodextra_y,
466                  ATI_FFP_CONST_BUMPMAT(stage), GL_BLUE, GL_NONE,
467                  GL_REG_0_ATI + stage + 1, GL_GREEN, GL_NONE);
468         wrap_op3(gl_info, GL_MAD_ATI, GL_REG_0_ATI + stage + 1, GL_GREEN_BIT_ATI, GL_NONE,
469                  GL_REG_0_ATI + stage, GL_GREEN, argmodextra_y,
470                  ATI_FFP_CONST_BUMPMAT(stage), GL_ALPHA, GL_NONE,
471                  GL_REG_0_ATI + stage + 1, GL_ALPHA, GL_NONE);
472     }
473
474     /* Pass 3: Generate sampling instructions for regular textures */
475     for (stage = 0; stage < gl_info->limits.textures; ++stage)
476     {
477         if(op[stage].cop == WINED3DTOP_DISABLE) {
478             break;
479         }
480
481         if(op[stage].projected == proj_none) {
482             swizzle = GL_SWIZZLE_STR_ATI;
483         } else if(op[stage].projected == proj_count3) {
484             swizzle = GL_SWIZZLE_STR_DR_ATI;
485         } else {
486             swizzle = GL_SWIZZLE_STQ_DQ_ATI;
487         }
488
489         if((op[stage].carg0 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
490            (op[stage].carg1 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
491            (op[stage].carg2 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
492            (op[stage].aarg0 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
493            (op[stage].aarg1 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
494            (op[stage].aarg2 & WINED3DTA_SELECTMASK) == WINED3DTA_TEXTURE ||
495             op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) {
496
497             if(stage > 0 &&
498                (op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
499                 op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
500                 TRACE("glSampleMapATI(GL_REG_%d_ATI, GL_REG_%d_ATI, GL_SWIZZLE_STR_ATI)\n",
501                       stage, stage);
502                 GL_EXTCALL(glSampleMapATI(GL_REG_0_ATI + stage,
503                            GL_REG_0_ATI + stage,
504                            GL_SWIZZLE_STR_ATI));
505             } else {
506                 TRACE("glSampleMapATI(GL_REG_%d_ATI, GL_TEXTURE_%d_ARB, %s)\n",
507                     stage, stage, debug_swizzle(swizzle));
508                 GL_EXTCALL(glSampleMapATI(GL_REG_0_ATI + stage,
509                                         GL_TEXTURE0_ARB + stage,
510                                         swizzle));
511             }
512         }
513     }
514
515     /* Pass 4: Generate the arithmetic instructions */
516     for (stage = 0; stage < MAX_TEXTURES; ++stage)
517     {
518         if (op[stage].cop == WINED3DTOP_DISABLE)
519         {
520             if (!stage)
521             {
522                 /* Handle complete texture disabling gracefully */
523                 wrap_op1(gl_info, GL_MOV_ATI, GL_REG_0_ATI, GL_NONE, GL_NONE,
524                          GL_PRIMARY_COLOR, GL_NONE, GL_NONE);
525                 wrap_op1(gl_info, GL_MOV_ATI, GL_REG_0_ATI, GL_ALPHA, GL_NONE,
526                          GL_PRIMARY_COLOR, GL_NONE, GL_NONE);
527             }
528             break;
529         }
530
531         if(op[stage].dst == tempreg) {
532             /* If we're writing to D3DTA_TEMP, but never reading from it we don't have to write there in the first place.
533              * skip the entire stage, this saves some GPU time
534              */
535             if(tmparg == GL_NONE) continue;
536
537             dstreg = tmparg;
538         } else {
539             dstreg = GL_REG_0_ATI;
540         }
541
542         arg0 = register_for_arg(op[stage].carg0, gl_info, stage, &argmod0, &rep0, tmparg);
543         arg1 = register_for_arg(op[stage].carg1, gl_info, stage, &argmod1, &rep1, tmparg);
544         arg2 = register_for_arg(op[stage].carg2, gl_info, stage, &argmod2, &rep2, tmparg);
545         dstmod = GL_NONE;
546         argmodextra = GL_NONE;
547         extrarg = GL_NONE;
548
549         switch(op[stage].cop) {
550             case WINED3DTOP_SELECTARG2:
551                 arg1 = arg2;
552                 argmod1 = argmod2;
553                 rep1 = rep2;
554                 /* fall through */
555             case WINED3DTOP_SELECTARG1:
556                 wrap_op1(gl_info, GL_MOV_ATI, dstreg, GL_NONE, GL_NONE,
557                          arg1, rep1, argmod1);
558                 break;
559
560             case WINED3DTOP_MODULATE4X:
561                 if(dstmod == GL_NONE) dstmod = GL_4X_BIT_ATI;
562                 /* fall through */
563             case WINED3DTOP_MODULATE2X:
564                 if(dstmod == GL_NONE) dstmod = GL_2X_BIT_ATI;
565                 dstmod |= GL_SATURATE_BIT_ATI;
566                 /* fall through */
567             case WINED3DTOP_MODULATE:
568                 wrap_op2(gl_info, GL_MUL_ATI, dstreg, GL_NONE, dstmod,
569                          arg1, rep1, argmod1,
570                          arg2, rep2, argmod2);
571                 break;
572
573             case WINED3DTOP_ADDSIGNED2X:
574                 dstmod = GL_2X_BIT_ATI;
575                 /* fall through */
576             case WINED3DTOP_ADDSIGNED:
577                 argmodextra = GL_BIAS_BIT_ATI;
578                 /* fall through */
579             case WINED3DTOP_ADD:
580                 dstmod |= GL_SATURATE_BIT_ATI;
581                 wrap_op2(gl_info, GL_ADD_ATI, GL_REG_0_ATI, GL_NONE, dstmod,
582                          arg1, rep1, argmod1,
583                          arg2, rep2, argmodextra | argmod2);
584                 break;
585
586             case WINED3DTOP_SUBTRACT:
587                 dstmod |= GL_SATURATE_BIT_ATI;
588                 wrap_op2(gl_info, GL_SUB_ATI, dstreg, GL_NONE, dstmod,
589                          arg1, rep1, argmod1,
590                          arg2, rep2, argmod2);
591                 break;
592
593             case WINED3DTOP_ADDSMOOTH:
594                 argmodextra = argmod1 & GL_COMP_BIT_ATI ? argmod1 & ~GL_COMP_BIT_ATI : argmod1 | GL_COMP_BIT_ATI;
595                 /* Dst = arg1 + * arg2(1 -arg 1)
596                  *     = arg2 * (1 - arg1) + arg1
597                  */
598                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_NONE, GL_SATURATE_BIT_ATI,
599                          arg2, rep2, argmod2,
600                          arg1, rep1, argmodextra,
601                          arg1, rep1, argmod1);
602                 break;
603
604             case WINED3DTOP_BLENDCURRENTALPHA:
605                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_CURRENT, gl_info, stage, NULL, NULL, -1);
606                 /* fall through */
607             case WINED3DTOP_BLENDFACTORALPHA:
608                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_TFACTOR, gl_info, stage, NULL, NULL, -1);
609                 /* fall through */
610             case WINED3DTOP_BLENDTEXTUREALPHA:
611                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_TEXTURE, gl_info, stage, NULL, NULL, -1);
612                 /* fall through */
613             case WINED3DTOP_BLENDDIFFUSEALPHA:
614                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_DIFFUSE, gl_info, stage, NULL, NULL, -1);
615                 wrap_op3(gl_info, GL_LERP_ATI, dstreg, GL_NONE, GL_NONE,
616                          extrarg, GL_ALPHA, GL_NONE,
617                          arg1, rep1, argmod1,
618                          arg2, rep2, argmod2);
619                 break;
620
621             case WINED3DTOP_BLENDTEXTUREALPHAPM:
622                 arg0 = register_for_arg(WINED3DTA_TEXTURE, gl_info, stage, NULL, NULL, -1);
623                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_NONE, GL_NONE,
624                          arg2, rep2,  argmod2,
625                          arg0, GL_ALPHA, GL_COMP_BIT_ATI,
626                          arg1, rep1,  argmod1);
627                 break;
628
629             /* D3DTOP_PREMODULATE ???? */
630
631             case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
632                 argmodextra = argmod1 & GL_COMP_BIT_ATI ? argmod1 & ~GL_COMP_BIT_ATI : argmod1 | GL_COMP_BIT_ATI;
633                 /* fall through */
634             case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
635                 if(!argmodextra) argmodextra = argmod1;
636                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_NONE, GL_SATURATE_BIT_ATI,
637                          arg2, rep2,  argmod2,
638                          arg1, GL_ALPHA, argmodextra,
639                          arg1, rep1,  argmod1);
640                 break;
641
642             case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
643                 argmodextra = argmod1 & GL_COMP_BIT_ATI ? argmod1 & ~GL_COMP_BIT_ATI : argmod1 | GL_COMP_BIT_ATI;
644                 /* fall through */
645             case WINED3DTOP_MODULATECOLOR_ADDALPHA:
646                 if(!argmodextra) argmodextra = argmod1;
647                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_NONE, GL_SATURATE_BIT_ATI,
648                          arg2, rep2,  argmod2,
649                          arg1, rep1,  argmodextra,
650                          arg1, GL_ALPHA, argmod1);
651                 break;
652
653             case WINED3DTOP_DOTPRODUCT3:
654                 wrap_op2(gl_info, GL_DOT3_ATI, dstreg, GL_NONE, GL_4X_BIT_ATI | GL_SATURATE_BIT_ATI,
655                          arg1, rep1, argmod1 | GL_BIAS_BIT_ATI,
656                          arg2, rep2, argmod2 | GL_BIAS_BIT_ATI);
657                 break;
658
659             case WINED3DTOP_MULTIPLYADD:
660                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_NONE, GL_SATURATE_BIT_ATI,
661                          arg1, rep1, argmod1,
662                          arg2, rep2, argmod2,
663                          arg0, rep0, argmod0);
664                 break;
665
666             case WINED3DTOP_LERP:
667                 wrap_op3(gl_info, GL_LERP_ATI, dstreg, GL_NONE, GL_NONE,
668                          arg0, rep0, argmod0,
669                          arg1, rep1, argmod1,
670                          arg2, rep2, argmod2);
671                 break;
672
673             case WINED3DTOP_BUMPENVMAP:
674             case WINED3DTOP_BUMPENVMAPLUMINANCE:
675                 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
676                 break;
677
678             default: FIXME("Unhandled color operation %d on stage %d\n", op[stage].cop, stage);
679         }
680
681         arg0 = register_for_arg(op[stage].aarg0, gl_info, stage, &argmod0, NULL, tmparg);
682         arg1 = register_for_arg(op[stage].aarg1, gl_info, stage, &argmod1, NULL, tmparg);
683         arg2 = register_for_arg(op[stage].aarg2, gl_info, stage, &argmod2, NULL, tmparg);
684         dstmod = GL_NONE;
685         argmodextra = GL_NONE;
686         extrarg = GL_NONE;
687
688         switch(op[stage].aop) {
689             case WINED3DTOP_DISABLE:
690                 /* Get the primary color to the output if on stage 0, otherwise leave register 0 untouched */
691                 if (!stage)
692                 {
693                     wrap_op1(gl_info, GL_MOV_ATI, GL_REG_0_ATI, GL_ALPHA, GL_NONE,
694                              GL_PRIMARY_COLOR, GL_NONE, GL_NONE);
695                 }
696                 break;
697
698             case WINED3DTOP_SELECTARG2:
699                 arg1 = arg2;
700                 argmod1 = argmod2;
701                 /* fall through */
702             case WINED3DTOP_SELECTARG1:
703                 wrap_op1(gl_info, GL_MOV_ATI, dstreg, GL_ALPHA, GL_NONE,
704                          arg1, GL_NONE, argmod1);
705                 break;
706
707             case WINED3DTOP_MODULATE4X:
708                 if(dstmod == GL_NONE) dstmod = GL_4X_BIT_ATI;
709                 /* fall through */
710             case WINED3DTOP_MODULATE2X:
711                 if(dstmod == GL_NONE) dstmod = GL_2X_BIT_ATI;
712                 dstmod |= GL_SATURATE_BIT_ATI;
713                 /* fall through */
714             case WINED3DTOP_MODULATE:
715                 wrap_op2(gl_info, GL_MUL_ATI, dstreg, GL_ALPHA, dstmod,
716                          arg1, GL_NONE, argmod1,
717                          arg2, GL_NONE, argmod2);
718                 break;
719
720             case WINED3DTOP_ADDSIGNED2X:
721                 dstmod = GL_2X_BIT_ATI;
722                 /* fall through */
723             case WINED3DTOP_ADDSIGNED:
724                 argmodextra = GL_BIAS_BIT_ATI;
725                 /* fall through */
726             case WINED3DTOP_ADD:
727                 dstmod |= GL_SATURATE_BIT_ATI;
728                 wrap_op2(gl_info, GL_ADD_ATI, dstreg, GL_ALPHA, dstmod,
729                          arg1, GL_NONE, argmod1,
730                          arg2, GL_NONE, argmodextra | argmod2);
731                 break;
732
733             case WINED3DTOP_SUBTRACT:
734                 dstmod |= GL_SATURATE_BIT_ATI;
735                 wrap_op2(gl_info, GL_SUB_ATI, dstreg, GL_ALPHA, dstmod,
736                          arg1, GL_NONE, argmod1,
737                          arg2, GL_NONE, argmod2);
738                 break;
739
740             case WINED3DTOP_ADDSMOOTH:
741                 argmodextra = argmod1 & GL_COMP_BIT_ATI ? argmod1 & ~GL_COMP_BIT_ATI : argmod1 | GL_COMP_BIT_ATI;
742                 /* Dst = arg1 + * arg2(1 -arg 1)
743                  *     = arg2 * (1 - arg1) + arg1
744                  */
745                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_ALPHA, GL_SATURATE_BIT_ATI,
746                          arg2, GL_NONE, argmod2,
747                          arg1, GL_NONE, argmodextra,
748                          arg1, GL_NONE, argmod1);
749                 break;
750
751             case WINED3DTOP_BLENDCURRENTALPHA:
752                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_CURRENT, gl_info, stage, NULL, NULL, -1);
753                 /* fall through */
754             case WINED3DTOP_BLENDFACTORALPHA:
755                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_TFACTOR, gl_info, stage, NULL, NULL, -1);
756                 /* fall through */
757             case WINED3DTOP_BLENDTEXTUREALPHA:
758                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_TEXTURE, gl_info, stage, NULL, NULL, -1);
759                 /* fall through */
760             case WINED3DTOP_BLENDDIFFUSEALPHA:
761                 if(extrarg == GL_NONE) extrarg = register_for_arg(WINED3DTA_DIFFUSE, gl_info, stage, NULL, NULL, -1);
762                 wrap_op3(gl_info, GL_LERP_ATI, dstreg, GL_ALPHA, GL_NONE,
763                          extrarg, GL_ALPHA, GL_NONE,
764                          arg1, GL_NONE, argmod1,
765                          arg2, GL_NONE, argmod2);
766                 break;
767
768             case WINED3DTOP_BLENDTEXTUREALPHAPM:
769                 arg0 = register_for_arg(WINED3DTA_TEXTURE, gl_info, stage, NULL, NULL, -1);
770                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_ALPHA, GL_NONE,
771                          arg2, GL_NONE,  argmod2,
772                          arg0, GL_ALPHA, GL_COMP_BIT_ATI,
773                          arg1, GL_NONE,  argmod1);
774                 break;
775
776             /* D3DTOP_PREMODULATE ???? */
777
778             case WINED3DTOP_DOTPRODUCT3:
779                 wrap_op2(gl_info, GL_DOT3_ATI, dstreg, GL_ALPHA, GL_4X_BIT_ATI | GL_SATURATE_BIT_ATI,
780                          arg1, GL_NONE, argmod1 | GL_BIAS_BIT_ATI,
781                          arg2, GL_NONE, argmod2 | GL_BIAS_BIT_ATI);
782                 break;
783
784             case WINED3DTOP_MULTIPLYADD:
785                 wrap_op3(gl_info, GL_MAD_ATI, dstreg, GL_ALPHA, GL_SATURATE_BIT_ATI,
786                          arg1, GL_NONE, argmod1,
787                          arg2, GL_NONE, argmod2,
788                          arg0, GL_NONE, argmod0);
789                 break;
790
791             case WINED3DTOP_LERP:
792                 wrap_op3(gl_info, GL_LERP_ATI, dstreg, GL_ALPHA, GL_SATURATE_BIT_ATI,
793                          arg1, GL_NONE, argmod1,
794                          arg2, GL_NONE, argmod2,
795                          arg0, GL_NONE, argmod0);
796                 break;
797
798             case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
799             case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
800             case WINED3DTOP_MODULATECOLOR_ADDALPHA:
801             case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
802             case WINED3DTOP_BUMPENVMAP:
803             case WINED3DTOP_BUMPENVMAPLUMINANCE:
804                 ERR("Application uses an invalid alpha operation\n");
805                 break;
806
807             default: FIXME("Unhandled alpha operation %d on stage %d\n", op[stage].aop, stage);
808         }
809     }
810
811     TRACE("glEndFragmentShaderATI()\n");
812     GL_EXTCALL(glEndFragmentShaderATI());
813     checkGLcall("GL_EXTCALL(glEndFragmentShaderATI())");
814     return ret;
815 }
816
817 static void set_tex_op_atifs(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
818 {
819     const struct wined3d_device *device = context->swapchain->device;
820     const struct wined3d_gl_info *gl_info = context->gl_info;
821     const struct atifs_ffp_desc *desc;
822     struct ffp_frag_settings settings;
823     struct atifs_private_data *priv = device->fragment_priv;
824     DWORD mapped_stage;
825     unsigned int i;
826
827     gen_ffp_frag_op(device, state, &settings, TRUE);
828     desc = (const struct atifs_ffp_desc *)find_ffp_frag_shader(&priv->fragment_shaders, &settings);
829     if(!desc) {
830         struct atifs_ffp_desc *new_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc));
831         if (!new_desc)
832         {
833             ERR("Out of memory\n");
834             return;
835         }
836         new_desc->num_textures_used = 0;
837         for (i = 0; i < gl_info->limits.texture_stages; ++i)
838         {
839             if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
840             new_desc->num_textures_used = i;
841         }
842
843         memcpy(&new_desc->parent.settings, &settings, sizeof(settings));
844         new_desc->shader = gen_ati_shader(settings.op, gl_info);
845         add_ffp_frag_shader(&priv->fragment_shaders, &new_desc->parent);
846         TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc);
847         desc = new_desc;
848     }
849
850     /* GL_ATI_fragment_shader depends on the GL_TEXTURE_xD enable settings. Update the texture stages
851      * used by this shader
852      */
853     for (i = 0; i < desc->num_textures_used; ++i)
854     {
855         mapped_stage = device->texUnitMap[i];
856         if (mapped_stage != WINED3D_UNMAPPED_STAGE)
857         {
858             context_active_texture(context, gl_info, mapped_stage);
859             texture_activate_dimensions(state->textures[i], gl_info);
860         }
861     }
862
863     GL_EXTCALL(glBindFragmentShaderATI(desc->shader));
864 }
865
866 static void state_texfactor_atifs(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
867 {
868     const struct wined3d_gl_info *gl_info = context->gl_info;
869     float col[4];
870     D3DCOLORTOGLFLOAT4(state->render_states[WINED3DRS_TEXTUREFACTOR], col);
871
872     GL_EXTCALL(glSetFragmentShaderConstantATI(ATI_FFP_CONST_TFACTOR, col));
873     checkGLcall("glSetFragmentShaderConstantATI(ATI_FFP_CONST_TFACTOR, col)");
874 }
875
876 static void set_bumpmat(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
877 {
878     DWORD stage = (state_id - STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE + 1);
879     const struct wined3d_gl_info *gl_info = context->gl_info;
880     float mat[2][2];
881
882     mat[0][0] = *((float *)&state->texture_states[stage][WINED3DTSS_BUMPENVMAT00]);
883     mat[1][0] = *((float *)&state->texture_states[stage][WINED3DTSS_BUMPENVMAT01]);
884     mat[0][1] = *((float *)&state->texture_states[stage][WINED3DTSS_BUMPENVMAT10]);
885     mat[1][1] = *((float *)&state->texture_states[stage][WINED3DTSS_BUMPENVMAT11]);
886     /* GL_ATI_fragment_shader allows only constants from 0.0 to 1.0, but the bumpmat
887      * constants can be in any range. While they should stay between [-1.0 and 1.0] because
888      * Shader Model 1.x pixel shaders are clamped to that range negative values are used occasionally,
889      * for example by our d3d9 test. So to get negative values scale -1;1 to 0;1 and undo that in the
890      * shader(it is free). This might potentially reduce precision. However, if the hardware does
891      * support proper floats it shouldn't, and if it doesn't we can't get anything better anyway
892      */
893     mat[0][0] = (mat[0][0] + 1.0f) * 0.5f;
894     mat[1][0] = (mat[1][0] + 1.0f) * 0.5f;
895     mat[0][1] = (mat[0][1] + 1.0f) * 0.5f;
896     mat[1][1] = (mat[1][1] + 1.0f) * 0.5f;
897     GL_EXTCALL(glSetFragmentShaderConstantATI(ATI_FFP_CONST_BUMPMAT(stage), (float *) mat));
898     checkGLcall("glSetFragmentShaderConstantATI(ATI_FFP_CONST_BUMPMAT(stage), mat)");
899 }
900
901 static void textransform(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
902 {
903     if (!isStateDirty(context, STATE_PIXELSHADER))
904         set_tex_op_atifs(context, state, state_id);
905 }
906
907 static void atifs_apply_pixelshader(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
908 {
909     const struct wined3d_device *device = context->swapchain->device;
910     BOOL use_vshader = use_vs(state);
911
912     context->last_was_pshader = use_ps(state);
913     /* The ATIFS code does not support pixel shaders currently, but we have to
914      * provide a state handler to call shader_select to select a vertex shader
915      * if one is applied because the vertex shader state may defer calling the
916      * shader backend if the pshader state is dirty.
917      *
918      * In theory the application should not be able to mark the pixel shader
919      * dirty because it cannot create a shader, and thus has no way to set the
920      * state to something != NULL. However, a different pipeline part may link
921      * a different state to its pixelshader handler, thus a pshader state
922      * exists and can be dirtified. Also the pshader is always dirtified at
923      * startup, and blitting disables all shaders and dirtifies all shader
924      * states. If atifs can deal with this it keeps the rest of the code
925      * simpler. */
926     if (!isStateDirty(context, context->state_table[STATE_VSHADER].representative))
927     {
928         device->shader_backend->shader_select(context, FALSE, use_vshader);
929
930         if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && use_vshader)
931             context_apply_state(context, state, STATE_VERTEXSHADERCONSTANT);
932     }
933 }
934
935 static void atifs_srgbwriteenable(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id)
936 {
937     if (state->render_states[WINED3DRS_SRGBWRITEENABLE])
938         WARN("sRGB writes are not supported by this fragment pipe.\n");
939 }
940
941 static const struct StateEntryTemplate atifs_fragmentstate_template[] = {
942     {STATE_RENDER(WINED3DRS_TEXTUREFACTOR),               { STATE_RENDER(WINED3DRS_TEXTUREFACTOR),              state_texfactor_atifs   }, WINED3D_GL_EXT_NONE             },
943     {STATE_RENDER(WINED3DRS_FOGCOLOR),                    { STATE_RENDER(WINED3DRS_FOGCOLOR),                   state_fogcolor          }, WINED3D_GL_EXT_NONE             },
944     {STATE_RENDER(WINED3DRS_FOGDENSITY),                  { STATE_RENDER(WINED3DRS_FOGDENSITY),                 state_fogdensity        }, WINED3D_GL_EXT_NONE             },
945     {STATE_RENDER(WINED3DRS_FOGENABLE),                   { STATE_RENDER(WINED3DRS_FOGENABLE),                  state_fog_fragpart      }, WINED3D_GL_EXT_NONE             },
946     {STATE_RENDER(WINED3DRS_FOGTABLEMODE),                { STATE_RENDER(WINED3DRS_FOGENABLE),                  NULL                    }, WINED3D_GL_EXT_NONE             },
947     {STATE_RENDER(WINED3DRS_FOGVERTEXMODE),               { STATE_RENDER(WINED3DRS_FOGENABLE),                  NULL                    }, WINED3D_GL_EXT_NONE             },
948     {STATE_RENDER(WINED3DRS_FOGSTART),                    { STATE_RENDER(WINED3DRS_FOGSTART),                   state_fogstartend       }, WINED3D_GL_EXT_NONE             },
949     {STATE_RENDER(WINED3DRS_FOGEND),                      { STATE_RENDER(WINED3DRS_FOGSTART),                   NULL                    }, WINED3D_GL_EXT_NONE             },
950     {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE),             { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE),            atifs_srgbwriteenable   }, WINED3D_GL_EXT_NONE             },
951     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          set_tex_op_atifs        }, WINED3D_GL_EXT_NONE             },
952     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
953     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
954     {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
955     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
956     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
957     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
958     {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
959     {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
960     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
961     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
962     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
963     {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
964     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
965     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
966     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
967     {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
968     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
969     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
970     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
971     {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
972     {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
973     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
974     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
975     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
976     {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
977     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
978     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
979     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
980     {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
981     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
982     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
983     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
984     {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
985     {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
986     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
987     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
988     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
989     {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
990     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
991     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
992     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
993     {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
994     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
995     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
996     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
997     {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
998     {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
999     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
1000     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1001     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1002     {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1003     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1004     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1005     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1006     {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1007     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1008     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1009     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1010     {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1011     {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1012     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
1013     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1014     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1015     {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1016     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1017     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1018     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1019     {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1020     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1021     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1022     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1023     {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1024     {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1025     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
1026     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1027     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1028     {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1029     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1030     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1031     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1032     {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1033     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1034     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1035     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1036     {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1037     {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1038     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
1039     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1040     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1041     {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1042     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1043     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1044     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1045     {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1046     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP),           { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1047     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1048     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1049     {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1050     {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG),         { STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP),          NULL                    }, WINED3D_GL_EXT_NONE             },
1051     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     set_bumpmat             }, WINED3D_GL_EXT_NONE             },
1052     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1053     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1054     {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11),      { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00),     NULL                    }, WINED3D_GL_EXT_NONE             },
1055     { STATE_SAMPLER(0),                                   { STATE_SAMPLER(0),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1056     { STATE_SAMPLER(1),                                   { STATE_SAMPLER(1),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1057     { STATE_SAMPLER(2),                                   { STATE_SAMPLER(2),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1058     { STATE_SAMPLER(3),                                   { STATE_SAMPLER(3),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1059     { STATE_SAMPLER(4),                                   { STATE_SAMPLER(4),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1060     { STATE_SAMPLER(5),                                   { STATE_SAMPLER(5),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1061     { STATE_SAMPLER(6),                                   { STATE_SAMPLER(6),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1062     { STATE_SAMPLER(7),                                   { STATE_SAMPLER(7),                                   sampler_texdim          }, WINED3D_GL_EXT_NONE             },
1063     {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1064     {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1065     {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1066     {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1067     {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1068     {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1069     {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1070     {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform      }, WINED3D_GL_EXT_NONE             },
1071     {STATE_PIXELSHADER,                                   { STATE_PIXELSHADER,                                  atifs_apply_pixelshader }, WINED3D_GL_EXT_NONE             },
1072     {0 /* Terminate */,                                   { 0,                                                  0                       }, WINED3D_GL_EXT_NONE             },
1073 };
1074
1075 /* Context activation and GL locking are done by the caller. */
1076 static void atifs_enable(BOOL enable)
1077 {
1078     if(enable) {
1079         glEnable(GL_FRAGMENT_SHADER_ATI);
1080         checkGLcall("glEnable(GL_FRAGMENT_SHADER_ATI)");
1081     } else {
1082         glDisable(GL_FRAGMENT_SHADER_ATI);
1083         checkGLcall("glDisable(GL_FRAGMENT_SHADER_ATI)");
1084     }
1085 }
1086
1087 static void atifs_get_caps(const struct wined3d_gl_info *gl_info, struct fragment_caps *caps)
1088 {
1089     caps->PrimitiveMiscCaps = WINED3DPMISCCAPS_TSSARGTEMP;
1090     caps->TextureOpCaps =  WINED3DTEXOPCAPS_DISABLE                     |
1091                            WINED3DTEXOPCAPS_SELECTARG1                  |
1092                            WINED3DTEXOPCAPS_SELECTARG2                  |
1093                            WINED3DTEXOPCAPS_MODULATE4X                  |
1094                            WINED3DTEXOPCAPS_MODULATE2X                  |
1095                            WINED3DTEXOPCAPS_MODULATE                    |
1096                            WINED3DTEXOPCAPS_ADDSIGNED2X                 |
1097                            WINED3DTEXOPCAPS_ADDSIGNED                   |
1098                            WINED3DTEXOPCAPS_ADD                         |
1099                            WINED3DTEXOPCAPS_SUBTRACT                    |
1100                            WINED3DTEXOPCAPS_ADDSMOOTH                   |
1101                            WINED3DTEXOPCAPS_BLENDCURRENTALPHA           |
1102                            WINED3DTEXOPCAPS_BLENDFACTORALPHA            |
1103                            WINED3DTEXOPCAPS_BLENDTEXTUREALPHA           |
1104                            WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA           |
1105                            WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM         |
1106                            WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR      |
1107                            WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA      |
1108                            WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA   |
1109                            WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR   |
1110                            WINED3DTEXOPCAPS_DOTPRODUCT3                 |
1111                            WINED3DTEXOPCAPS_MULTIPLYADD                 |
1112                            WINED3DTEXOPCAPS_LERP                        |
1113                            WINED3DTEXOPCAPS_BUMPENVMAP;
1114
1115     /* TODO: Implement WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
1116     and WINED3DTEXOPCAPS_PREMODULATE */
1117
1118     /* GL_ATI_fragment_shader always supports 6 textures, which was the limit on r200 cards
1119      * which this extension is exclusively focused on(later cards have GL_ARB_fragment_program).
1120      * If the current card has more than 8 fixed function textures in OpenGL's regular fixed
1121      * function pipeline then the ATI_fragment_shader backend imposes a stricter limit. This
1122      * shouldn't be too hard since Nvidia cards have a limit of 4 textures with the default ffp
1123      * pipeline, and almost all games are happy with that. We can however support up to 8
1124      * texture stages because we have a 2nd pass limit of 8 instructions, and per stage we use
1125      * only 1 instruction.
1126      *
1127      * The proper fix for this is not to use GL_ATI_fragment_shader on cards newer than the
1128      * r200 series and use an ARB or GLSL shader instead
1129      */
1130     caps->MaxTextureBlendStages   = 8;
1131     caps->MaxSimultaneousTextures = 6;
1132 }
1133
1134 static HRESULT atifs_alloc(struct wined3d_device *device)
1135 {
1136     struct atifs_private_data *priv;
1137
1138     device->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct atifs_private_data));
1139     if (!device->fragment_priv)
1140     {
1141         ERR("Out of memory\n");
1142         return E_OUTOFMEMORY;
1143     }
1144     priv = device->fragment_priv;
1145     if (wine_rb_init(&priv->fragment_shaders, &wined3d_ffp_frag_program_rb_functions) == -1)
1146     {
1147         ERR("Failed to initialize rbtree.\n");
1148         HeapFree(GetProcessHeap(), 0, device->fragment_priv);
1149         return E_OUTOFMEMORY;
1150     }
1151     return WINED3D_OK;
1152 }
1153
1154 /* Context activation is done by the caller. */
1155 static void atifs_free_ffpshader(struct wine_rb_entry *entry, void *context)
1156 {
1157     struct wined3d_device *device = context;
1158     const struct wined3d_gl_info *gl_info = &device->adapter->gl_info;
1159     struct atifs_ffp_desc *entry_ati = WINE_RB_ENTRY_VALUE(entry, struct atifs_ffp_desc, parent.entry);
1160
1161     ENTER_GL();
1162     GL_EXTCALL(glDeleteFragmentShaderATI(entry_ati->shader));
1163     checkGLcall("glDeleteFragmentShaderATI(entry->shader)");
1164     HeapFree(GetProcessHeap(), 0, entry_ati);
1165     LEAVE_GL();
1166 }
1167
1168 /* Context activation is done by the caller. */
1169 static void atifs_free(struct wined3d_device *device)
1170 {
1171     struct atifs_private_data *priv = device->fragment_priv;
1172
1173     wine_rb_destroy(&priv->fragment_shaders, atifs_free_ffpshader, device);
1174
1175     HeapFree(GetProcessHeap(), 0, priv);
1176     device->fragment_priv = NULL;
1177 }
1178
1179 static BOOL atifs_color_fixup_supported(struct color_fixup_desc fixup)
1180 {
1181     if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
1182     {
1183         TRACE("Checking support for fixup:\n");
1184         dump_color_fixup_desc(fixup);
1185     }
1186
1187     /* We only support sign fixup of the first two channels. */
1188     if (fixup.x_source == CHANNEL_SOURCE_X && fixup.y_source == CHANNEL_SOURCE_Y
1189             && fixup.z_source == CHANNEL_SOURCE_Z && fixup.w_source == CHANNEL_SOURCE_W
1190             && !fixup.z_sign_fixup && !fixup.w_sign_fixup)
1191     {
1192         TRACE("[OK]\n");
1193         return TRUE;
1194     }
1195
1196     TRACE("[FAILED]\n");
1197     return FALSE;
1198 }
1199
1200 const struct fragment_pipeline atifs_fragment_pipeline = {
1201     atifs_enable,
1202     atifs_get_caps,
1203     atifs_alloc,
1204     atifs_free,
1205     atifs_color_fixup_supported,
1206     atifs_fragmentstate_template,
1207     TRUE /* We can disable projected textures */
1208 };