?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push 583922ece91f6a3e6a32c772f3574cd16e6257fa
[wine/hacks.git] / dlls / wined3d / glsl_shader.c
blob8a0603b16ded30e70144f1e605d1bc3cc118eed9
1 /*
2 * GLSL pixel and vertex shader implementation
3 *
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
20 */
21
22 /*
23 * D3D shader asm has swizzles on source parameters, and write masks for
24 * destination parameters. GLSL uses swizzles for both. The result of this is
25 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
26 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
27 * mask for the destination parameter into account.
28 */
29
30 #include "config.h"
31 #include <stdio.h>
32 #include "wined3d_private.h"
33
34 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
35 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
36
37 #define GLINFO_LOCATION (*gl_info)
38
39 typedef struct {
40 char reg_name[50];
41 char mask_str[6];
42 } glsl_dst_param_t;
43
44 typedef struct {
45 char reg_name[50];
46 char param_str[100];
47 } glsl_src_param_t;
48
49 typedef struct {
50 const char *name;
51 DWORD coord_mask;
52 } glsl_sample_function_t;
53
54 /** Prints the GLSL info log which will contain error messages if they exist */
55 void print_glsl_info_log(WineD3D_GL_Info *gl_info, GLhandleARB obj) {
56
57 int infologLength = 0;
58 char *infoLog;
59
60 GL_EXTCALL(glGetObjectParameterivARB(obj,
61 GL_OBJECT_INFO_LOG_LENGTH_ARB,
62 &infologLength));
63
64 /* A size of 1 is just a null-terminated string, so the log should be bigger than
65 * that if there are errors. */
66 if (infologLength > 1)
67 {
68 infoLog = HeapAlloc(GetProcessHeap(), 0, infologLength);
69 GL_EXTCALL(glGetInfoLogARB(obj, infologLength, NULL, infoLog));
70 FIXME("Error received from GLSL shader #%u: %s\n", obj, debugstr_a(infoLog));
71 HeapFree(GetProcessHeap(), 0, infoLog);
72 }
73 }
74
75 /**
76 * Loads (pixel shader) samplers
77 */
78 static void shader_glsl_load_psamplers(
79 WineD3D_GL_Info *gl_info,
80 IWineD3DStateBlock* iface) {
81
82 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
83 GLhandleARB programId = stateBlock->glsl_program->programId;
84 GLhandleARB name_loc;
85 int i;
86 char sampler_name[20];
87
88 for (i = 0; i < MAX_FRAGMENT_SAMPLERS; ++i) {
89 snprintf(sampler_name, sizeof(sampler_name), "Psampler%d", i);
90 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
91 if (name_loc != -1) {
92 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[i];
93 if (mapped_unit != -1 && mapped_unit < GL_LIMITS(fragment_samplers)) {
94 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
95 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
96 checkGLcall("glUniform1iARB");
97 } else {
98 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
99 }
100 }
101 }
102 }
103
104 static void shader_glsl_load_vsamplers(WineD3D_GL_Info *gl_info, IWineD3DStateBlock* iface) {
105 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
106 GLhandleARB programId = stateBlock->glsl_program->programId;
107 GLhandleARB name_loc;
108 char sampler_name[20];
109 int i;
110
111 for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i) {
112 snprintf(sampler_name, sizeof(sampler_name), "Vsampler%d", i);
113 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
114 if (name_loc != -1) {
115 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[MAX_FRAGMENT_SAMPLERS + i];
116 if (mapped_unit != -1 && mapped_unit < GL_LIMITS(combined_samplers)) {
117 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
118 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
119 checkGLcall("glUniform1iARB");
120 } else {
121 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
122 }
123 }
124 }
125 }
126
127 /**
128 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
129 * When constant_list == NULL, it will load all the constants.
130 */
131 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info,
132 unsigned int max_constants, float* constants, GLhandleARB *constant_locations,
133 struct list *constant_list) {
134 constants_entry *constant;
135 local_constant* lconst;
136 GLhandleARB tmp_loc;
137 DWORD i, j, k;
138 DWORD *idx;
139
140 if (TRACE_ON(d3d_shader)) {
141 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
142 idx = constant->idx;
143 j = constant->count;
144 while (j--) {
145 i = *idx++;
146 tmp_loc = constant_locations[i];
147 if (tmp_loc != -1) {
148 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
149 constants[i * 4 + 0], constants[i * 4 + 1],
150 constants[i * 4 + 2], constants[i * 4 + 3]);
151 }
152 }
153 }
154 }
155
156 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
157 if(WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 &&
158 shader_is_pshader_version(This->baseShader.hex_version)) {
159 float lcl_const[4];
160
161 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
162 idx = constant->idx;
163 j = constant->count;
164 while (j--) {
165 i = *idx++;
166 tmp_loc = constant_locations[i];
167 if (tmp_loc != -1) {
168 /* We found this uniform name in the program - go ahead and send the data */
169 k = i * 4;
170 if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
171 else if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
172 else lcl_const[0] = constants[k + 0];
173 if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
174 else if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
175 else lcl_const[1] = constants[k + 1];
176 if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
177 else if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
178 else lcl_const[2] = constants[k + 2];
179 if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
180 else if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
181 else lcl_const[3] = constants[k + 3];
182
183 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, lcl_const));
184 }
185 }
186 }
187 } else {
188 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
189 idx = constant->idx;
190 j = constant->count;
191 while (j--) {
192 i = *idx++;
193 tmp_loc = constant_locations[i];
194 if (tmp_loc != -1) {
195 /* We found this uniform name in the program - go ahead and send the data */
196 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, constants + (i * 4)));
197 }
198 }
199 }
200 }
201 checkGLcall("glUniform4fvARB()");
202
203 /* Load immediate constants */
204 if (TRACE_ON(d3d_shader)) {
205 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
206 tmp_loc = constant_locations[lconst->idx];
207 if (tmp_loc != -1) {
208 GLfloat* values = (GLfloat*)lconst->value;
209 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
210 values[0], values[1], values[2], values[3]);
211 }
212 }
213 }
214 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
215 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
216 tmp_loc = constant_locations[lconst->idx];
217 if (tmp_loc != -1) {
218 /* We found this uniform name in the program - go ahead and send the data */
219 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, (GLfloat*)lconst->value));
220 }
221 }
222 checkGLcall("glUniform4fvARB()");
223 }
224
225 /**
226 * Loads integer constants (aka uniforms) into the currently set GLSL program.
227 * When @constants_set == NULL, it will load all the constants.
228 */
229 static void shader_glsl_load_constantsI(
230 IWineD3DBaseShaderImpl* This,
231 WineD3D_GL_Info *gl_info,
232 GLhandleARB programId,
233 unsigned max_constants,
234 int* constants,
235 BOOL* constants_set) {
236
237 GLhandleARB tmp_loc;
238 int i;
239 char tmp_name[8];
240 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
241 const char* prefix = is_pshader? "PI":"VI";
242 struct list* ptr;
243
244 for (i=0; i<max_constants; ++i) {
245 if (NULL == constants_set || constants_set[i]) {
246
247 TRACE_(d3d_constants)("Loading constants %i: %i, %i, %i, %i\n",
248 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
249
250 /* TODO: Benchmark and see if it would be beneficial to store the
251 * locations of the constants to avoid looking up each time */
252 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
253 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
254 if (tmp_loc != -1) {
255 /* We found this uniform name in the program - go ahead and send the data */
256 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, &constants[i*4]));
257 checkGLcall("glUniform4ivARB");
258 }
259 }
260 }
261
262 /* Load immediate constants */
263 ptr = list_head(&This->baseShader.constantsI);
264 while (ptr) {
265 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
266 unsigned int idx = lconst->idx;
267 GLint* values = (GLint*) lconst->value;
268
269 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
270 values[0], values[1], values[2], values[3]);
271
272 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
273 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
274 if (tmp_loc != -1) {
275 /* We found this uniform name in the program - go ahead and send the data */
276 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, values));
277 checkGLcall("glUniform4ivARB");
278 }
279 ptr = list_next(&This->baseShader.constantsI, ptr);
280 }
281 }
282
283 /**
284 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
285 * When @constants_set == NULL, it will load all the constants.
286 */
287 static void shader_glsl_load_constantsB(
288 IWineD3DBaseShaderImpl* This,
289 WineD3D_GL_Info *gl_info,
290 GLhandleARB programId,
291 unsigned max_constants,
292 BOOL* constants,
293 BOOL* constants_set) {
294
295 GLhandleARB tmp_loc;
296 int i;
297 char tmp_name[8];
298 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
299 const char* prefix = is_pshader? "PB":"VB";
300 struct list* ptr;
301
302 for (i=0; i<max_constants; ++i) {
303 if (NULL == constants_set || constants_set[i]) {
304
305 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i*4]);
306
307 /* TODO: Benchmark and see if it would be beneficial to store the
308 * locations of the constants to avoid looking up each time */
309 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
310 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
311 if (tmp_loc != -1) {
312 /* We found this uniform name in the program - go ahead and send the data */
313 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i*4]));
314 checkGLcall("glUniform1ivARB");
315 }
316 }
317 }
318
319 /* Load immediate constants */
320 ptr = list_head(&This->baseShader.constantsB);
321 while (ptr) {
322 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
323 unsigned int idx = lconst->idx;
324 GLint* values = (GLint*) lconst->value;
325
326 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
327
328 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
329 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
330 if (tmp_loc != -1) {
331 /* We found this uniform name in the program - go ahead and send the data */
332 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
333 checkGLcall("glUniform1ivARB");
334 }
335 ptr = list_next(&This->baseShader.constantsB, ptr);
336 }
337 }
338
339
340
341 /**
342 * Loads the app-supplied constants into the currently set GLSL program.
343 */
344 void shader_glsl_load_constants(
345 IWineD3DDevice* device,
346 char usePixelShader,
347 char useVertexShader) {
348
349 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
350 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
351 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
352
353 GLhandleARB *constant_locations;
354 struct list *constant_list;
355 GLhandleARB programId;
356 GLint pos;
357
358 if (!stateBlock->glsl_program) {
359 /* No GLSL program set - nothing to do. */
360 return;
361 }
362 programId = stateBlock->glsl_program->programId;
363
364 if (useVertexShader) {
365 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
366 GLint pos;
367
368 constant_locations = stateBlock->glsl_program->vuniformF_locations;
369 constant_list = &stateBlock->set_vconstantsF;
370
371 /* Load vertex shader samplers */
372 shader_glsl_load_vsamplers(gl_info, (IWineD3DStateBlock*)stateBlock);
373
374 /* Load DirectX 9 float constants/uniforms for vertex shader */
375 shader_glsl_load_constantsF(vshader, gl_info, GL_LIMITS(vshader_constantsF),
376 stateBlock->vertexShaderConstantF, constant_locations, constant_list);
377
378 /* Load DirectX 9 integer constants/uniforms for vertex shader */
379 shader_glsl_load_constantsI(vshader, gl_info, programId, MAX_CONST_I,
380 stateBlock->vertexShaderConstantI,
381 stateBlock->changed.vertexShaderConstantsI);
382
383 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
384 shader_glsl_load_constantsB(vshader, gl_info, programId, MAX_CONST_B,
385 stateBlock->vertexShaderConstantB,
386 stateBlock->changed.vertexShaderConstantsB);
387
388 /* Upload the position fixup params */
389 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "posFixup"));
390 checkGLcall("glGetUniformLocationARB");
391 GL_EXTCALL(glUniform4fvARB(pos, 1, &deviceImpl->posFixup[0]));
392 checkGLcall("glUniform4fvARB");
393 }
394
395 if (usePixelShader) {
396
397 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
398
399 constant_locations = stateBlock->glsl_program->puniformF_locations;
400 constant_list = &stateBlock->set_pconstantsF;
401
402 /* Load pixel shader samplers */
403 shader_glsl_load_psamplers(gl_info, (IWineD3DStateBlock*) stateBlock);
404
405 /* Load DirectX 9 float constants/uniforms for pixel shader */
406 shader_glsl_load_constantsF(pshader, gl_info, GL_LIMITS(pshader_constantsF),
407 stateBlock->pixelShaderConstantF, constant_locations, constant_list);
408
409 /* Load DirectX 9 integer constants/uniforms for pixel shader */
410 shader_glsl_load_constantsI(pshader, gl_info, programId, MAX_CONST_I,
411 stateBlock->pixelShaderConstantI,
412 stateBlock->changed.pixelShaderConstantsI);
413
414 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
415 shader_glsl_load_constantsB(pshader, gl_info, programId, MAX_CONST_B,
416 stateBlock->pixelShaderConstantB,
417 stateBlock->changed.pixelShaderConstantsB);
418
419 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
420 * It can't be 0 for a valid texbem instruction.
421 */
422 if(((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat != -1) {
423 float *data = (float *) &stateBlock->textureState[(int) ((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
424 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "bumpenvmat"));
425 checkGLcall("glGetUniformLocationARB");
426 GL_EXTCALL(glUniformMatrix2fvARB(pos, 1, 0, data));
427 checkGLcall("glUniformMatrix2fvARB");
428
429 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
430 * is set too, so we can check that in the needsbumpmat check
431 */
432 if(((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams != -1) {
433 int stage = ((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams;
434 GLfloat *scale = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLSCALE];
435 GLfloat *offset = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET];
436
437 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "luminancescale"));
438 checkGLcall("glGetUniformLocationARB");
439 GL_EXTCALL(glUniform1fvARB(pos, 1, scale));
440 checkGLcall("glUniform1fvARB");
441 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "luminanceoffset"));
442 checkGLcall("glGetUniformLocationARB");
443 GL_EXTCALL(glUniform1fvARB(pos, 1, offset));
444 checkGLcall("glUniform1fvARB");
445 }
446 } else if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
447 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
448 float comparison[4];
449 float mul_low[4];
450
451 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
452 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
453 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
454
455 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
456 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
457 } else {
458 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
459 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
460
461 mul_low[0] = 1.0; mul_low[1] = 1.0;
462 mul_low[2] = 1.0; mul_low[3] = 1.0;
463 }
464
465 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_comparison"));
466 checkGLcall("glGetUniformLocationARB");
467 GL_EXTCALL(glUniform4fvARB(pos, 1, comparison));
468 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_mul_low"));
469 checkGLcall("glGetUniformLocationARB");
470 GL_EXTCALL(glUniform4fvARB(pos, 1, mul_low));
471 }
472 if(((IWineD3DPixelShaderImpl *) pshader)->vpos_uniform) {
473 float correction_params[4];
474 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "ycorrection"));
475 checkGLcall("glGetUniformLocationARB");
476 if(deviceImpl->render_offscreen) {
477 correction_params[0] = 0.0;
478 correction_params[1] = 1.0;
479 } else {
480 /* position is window relative, not viewport relative */
481 correction_params[0] = ((IWineD3DSurfaceImpl *) deviceImpl->render_targets[0])->currentDesc.Height;
482 correction_params[1] = -1.0;
483 }
484 GL_EXTCALL(glUniform4fvARB(pos, 1, correction_params));
485 }
486 }
487 }
488
489 /** Generate the variable & register declarations for the GLSL output target */
490 void shader_generate_glsl_declarations(
491 IWineD3DBaseShader *iface,
492 shader_reg_maps* reg_maps,
493 SHADER_BUFFER* buffer,
494 WineD3D_GL_Info* gl_info) {
495
496 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
497 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
498 int i;
499 unsigned int extra_constants_needed = 0;
500
501 /* There are some minor differences between pixel and vertex shaders */
502 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
503 char prefix = pshader ? 'P' : 'V';
504
505 /* Prototype the subroutines */
506 for (i = 0; i < This->baseShader.limits.label; i++) {
507 if (reg_maps->labels[i])
508 shader_addline(buffer, "void subroutine%lu();\n", i);
509 }
510
511 /* Declare the constants (aka uniforms) */
512 if (This->baseShader.limits.constant_float > 0) {
513 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
514 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
515 shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
516 }
517
518 if (This->baseShader.limits.constant_int > 0)
519 shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
520
521 if (This->baseShader.limits.constant_bool > 0)
522 shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
523
524 if(!pshader) {
525 shader_addline(buffer, "uniform vec4 posFixup;\n");
526 } else {
527 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
528
529 if(reg_maps->bumpmat != -1) {
530 shader_addline(buffer, "uniform mat2 bumpenvmat;\n");
531 if(reg_maps->luminanceparams) {
532 shader_addline(buffer, "uniform float luminancescale;\n");
533 shader_addline(buffer, "uniform float luminanceoffset;\n");
534 extra_constants_needed++;
535 }
536 extra_constants_needed++;
537 }
538
539 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
540 ps_impl->srgb_enabled = 1;
541 if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
542 shader_addline(buffer, "uniform vec4 srgb_mul_low;\n");
543 shader_addline(buffer, "uniform vec4 srgb_comparison;\n");
544 ps_impl->srgb_mode_hardcoded = 0;
545 extra_constants_needed++;
546 } else {
547 ps_impl->srgb_mode_hardcoded = 1;
548 shader_addline(buffer, "const vec4 srgb_mul_low = {%f, %f, %f, %f};\n",
549 srgb_mul_low, srgb_mul_low, srgb_mul_low, srgb_mul_low);
550 shader_addline(buffer, "const vec4 srgb_comparison = {%f, %f, %f, %f};\n",
551 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
552 }
553 } else {
554 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
555
556 /* Do not write any srgb fixup into the shader to save shader size and processing time.
557 * As a consequence, we can't toggle srgb write on without recompilation
558 */
559 ps_impl->srgb_enabled = 0;
560 ps_impl->srgb_mode_hardcoded = 1;
561 }
562 if(reg_maps->vpos) {
563 if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
564 shader_addline(buffer, "uniform vec4 ycorrection;\n");
565 ((IWineD3DPixelShaderImpl *) This)->vpos_uniform = 1;
566 extra_constants_needed++;
567 } else {
568 /* This happens because we do not have proper tracking of the constant registers that are
569 * actually used, only the max limit of the shader version
570 */
571 FIXME("Cannot find a free uniform for vpos correction params\n");
572 shader_addline(buffer, "const vec4 ycorrection = {%f, %f, 0.0, 0.0};\n",
573 device->render_offscreen ? 0.0 : ((IWineD3DSurfaceImpl *) device->render_targets[0])->currentDesc.Height,
574 device->render_offscreen ? 1.0 : -1.0);
575 }
576 shader_addline(buffer, "vec4 vpos;\n");
577 }
578 }
579
580 /* Declare texture samplers */
581 for (i = 0; i < This->baseShader.limits.sampler; i++) {
582 if (reg_maps->samplers[i]) {
583
584 DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
585 switch (stype) {
586
587 case WINED3DSTT_1D:
588 shader_addline(buffer, "uniform sampler1D %csampler%lu;\n", prefix, i);
589 break;
590 case WINED3DSTT_2D:
591 shader_addline(buffer, "uniform sampler2D %csampler%lu;\n", prefix, i);
592 break;
593 case WINED3DSTT_CUBE:
594 shader_addline(buffer, "uniform samplerCube %csampler%lu;\n", prefix, i);
595 break;
596 case WINED3DSTT_VOLUME:
597 shader_addline(buffer, "uniform sampler3D %csampler%lu;\n", prefix, i);
598 break;
599 default:
600 shader_addline(buffer, "uniform unsupported_sampler %csampler%lu;\n", prefix, i);
601 FIXME("Unrecognized sampler type: %#x\n", stype);
602 break;
603 }
604 }
605 }
606
607 /* Declare address variables */
608 for (i = 0; i < This->baseShader.limits.address; i++) {
609 if (reg_maps->address[i])
610 shader_addline(buffer, "ivec4 A%d;\n", i);
611 }
612
613 /* Declare texture coordinate temporaries and initialize them */
614 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
615 if (reg_maps->texcoord[i])
616 shader_addline(buffer, "vec4 T%lu = gl_TexCoord[%lu];\n", i, i);
617 }
618
619 /* Declare input register temporaries */
620 for (i=0; i < This->baseShader.limits.packed_input; i++) {
621 if (reg_maps->packed_input[i])
622 shader_addline(buffer, "vec4 IN%lu;\n", i);
623 }
624
625 /* Declare output register temporaries */
626 for (i = 0; i < This->baseShader.limits.packed_output; i++) {
627 if (reg_maps->packed_output[i])
628 shader_addline(buffer, "vec4 OUT%lu;\n", i);
629 }
630
631 /* Declare temporary variables */
632 for(i = 0; i < This->baseShader.limits.temporary; i++) {
633 if (reg_maps->temporary[i])
634 shader_addline(buffer, "vec4 R%lu;\n", i);
635 }
636
637 /* Declare attributes */
638 for (i = 0; i < This->baseShader.limits.attributes; i++) {
639 if (reg_maps->attributes[i])
640 shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
641 }
642
643 /* Declare loop registers aLx */
644 for (i = 0; i < reg_maps->loop_depth; i++) {
645 shader_addline(buffer, "int aL%u;\n", i);
646 shader_addline(buffer, "int tmpInt%u;\n", i);
647 }
648
649 /* Temporary variables for matrix operations */
650 shader_addline(buffer, "vec4 tmp0;\n");
651 shader_addline(buffer, "vec4 tmp1;\n");
652
653 /* Start the main program */
654 shader_addline(buffer, "void main() {\n");
655 if(pshader && reg_maps->vpos) {
656 shader_addline(buffer, "vpos = vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1) - 0.5;\n");
657 }
658 }
659
660 /*****************************************************************************
661 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
662 *
663 * For more information, see http://wiki.winehq.org/DirectX-Shaders
664 ****************************************************************************/
665
666 /* Prototypes */
667 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
668 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
669
670 /** Used for opcode modifiers - They multiply the result by the specified amount */
671 static const char * const shift_glsl_tab[] = {
672 "", /* 0 (none) */
673 "2.0 * ", /* 1 (x2) */
674 "4.0 * ", /* 2 (x4) */
675 "8.0 * ", /* 3 (x8) */
676 "16.0 * ", /* 4 (x16) */
677 "32.0 * ", /* 5 (x32) */
678 "", /* 6 (x64) */
679 "", /* 7 (x128) */
680 "", /* 8 (d256) */
681 "", /* 9 (d128) */
682 "", /* 10 (d64) */
683 "", /* 11 (d32) */
684 "0.0625 * ", /* 12 (d16) */
685 "0.125 * ", /* 13 (d8) */
686 "0.25 * ", /* 14 (d4) */
687 "0.5 * " /* 15 (d2) */
688 };
689
690 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
691 static void shader_glsl_gen_modifier (
692 const DWORD instr,
693 const char *in_reg,
694 const char *in_regswizzle,
695 char *out_str) {
696
697 out_str[0] = 0;
698
699 if (instr == WINED3DSIO_TEXKILL)
700 return;
701
702 switch (instr & WINED3DSP_SRCMOD_MASK) {
703 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
704 case WINED3DSPSM_DW:
705 case WINED3DSPSM_NONE:
706 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
707 break;
708 case WINED3DSPSM_NEG:
709 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
710 break;
711 case WINED3DSPSM_NOT:
712 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
713 break;
714 case WINED3DSPSM_BIAS:
715 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
716 break;
717 case WINED3DSPSM_BIASNEG:
718 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
719 break;
720 case WINED3DSPSM_SIGN:
721 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
722 break;
723 case WINED3DSPSM_SIGNNEG:
724 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
725 break;
726 case WINED3DSPSM_COMP:
727 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
728 break;
729 case WINED3DSPSM_X2:
730 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
731 break;
732 case WINED3DSPSM_X2NEG:
733 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
734 break;
735 case WINED3DSPSM_ABS:
736 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
737 break;
738 case WINED3DSPSM_ABSNEG:
739 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
740 break;
741 default:
742 FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
743 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
744 }
745 }
746
747 /** Writes the GLSL variable name that corresponds to the register that the
748 * DX opcode parameter is trying to access */
749 static void shader_glsl_get_register_name(
750 const DWORD param,
751 const DWORD addr_token,
752 char* regstr,
753 BOOL* is_color,
754 SHADER_OPCODE_ARG* arg) {
755
756 /* oPos, oFog and oPts in D3D */
757 static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
758
759 DWORD reg = param & WINED3DSP_REGNUM_MASK;
760 DWORD regtype = shader_get_regtype(param);
761 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
762 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
763 WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
764
765 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
766 char tmpStr[50];
767
768 *is_color = FALSE;
769
770 switch (regtype) {
771 case WINED3DSPR_TEMP:
772 sprintf(tmpStr, "R%u", reg);
773 break;
774 case WINED3DSPR_INPUT:
775 if (pshader) {
776 /* Pixel shaders >= 3.0 */
777 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
778 sprintf(tmpStr, "IN%u", reg);
779 else {
780 if (reg==0)
781 strcpy(tmpStr, "gl_Color");
782 else
783 strcpy(tmpStr, "gl_SecondaryColor");
784 }
785 } else {
786 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
787 *is_color = TRUE;
788 sprintf(tmpStr, "attrib%u", reg);
789 }
790 break;
791 case WINED3DSPR_CONST:
792 {
793 const char* prefix = pshader? "PC":"VC";
794
795 /* Relative addressing */
796 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
797
798 /* Relative addressing on shaders 2.0+ have a relative address token,
799 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
800 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
801 glsl_src_param_t rel_param;
802 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
803 sprintf(tmpStr, "%s[%s + %u]", prefix, rel_param.param_str, reg);
804 } else
805 sprintf(tmpStr, "%s[A0.x + %u]", prefix, reg);
806
807 } else
808 sprintf(tmpStr, "%s[%u]", prefix, reg);
809
810 break;
811 }
812 case WINED3DSPR_CONSTINT:
813 if (pshader)
814 sprintf(tmpStr, "PI[%u]", reg);
815 else
816 sprintf(tmpStr, "VI[%u]", reg);
817 break;
818 case WINED3DSPR_CONSTBOOL:
819 if (pshader)
820 sprintf(tmpStr, "PB[%u]", reg);
821 else
822 sprintf(tmpStr, "VB[%u]", reg);
823 break;
824 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
825 if (pshader) {
826 sprintf(tmpStr, "T%u", reg);
827 } else {
828 sprintf(tmpStr, "A%u", reg);
829 }
830 break;
831 case WINED3DSPR_LOOP:
832 sprintf(tmpStr, "aL%u", This->baseShader.cur_loop_regno - 1);
833 break;
834 case WINED3DSPR_SAMPLER:
835 if (pshader)
836 sprintf(tmpStr, "Psampler%u", reg);
837 else
838 sprintf(tmpStr, "Vsampler%u", reg);
839 break;
840 case WINED3DSPR_COLOROUT:
841 if (reg >= GL_LIMITS(buffers)) {
842 WARN("Write to render target %u, only %d supported\n", reg, 4);
843 }
844 if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
845 sprintf(tmpStr, "gl_FragData[%u]", reg);
846 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
847 sprintf(tmpStr, "gl_FragColor");
848 }
849 break;
850 case WINED3DSPR_RASTOUT:
851 sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
852 break;
853 case WINED3DSPR_DEPTHOUT:
854 sprintf(tmpStr, "gl_FragDepth");
855 break;
856 case WINED3DSPR_ATTROUT:
857 if (reg == 0) {
858 sprintf(tmpStr, "gl_FrontColor");
859 } else {
860 sprintf(tmpStr, "gl_FrontSecondaryColor");
861 }
862 break;
863 case WINED3DSPR_TEXCRDOUT:
864 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
865 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
866 sprintf(tmpStr, "OUT%u", reg);
867 else
868 sprintf(tmpStr, "gl_TexCoord[%u]", reg);
869 break;
870 case WINED3DSPR_MISCTYPE:
871 if (reg == 0) {
872 /* vPos */
873 sprintf(tmpStr, "vpos");
874 } else if (reg == 1){
875 /* Note that gl_FrontFacing is a bool, while vFace is
876 * a float for which the sign determines front/back
877 */
878 sprintf(tmpStr, "(gl_FrontFacing ? 1.0 : -1.0)");
879 } else {
880 FIXME("Unhandled misctype register %d\n", reg);
881 sprintf(tmpStr, "unrecognized_register");
882 }
883 break;
884 default:
885 FIXME("Unhandled register name Type(%d)\n", regtype);
886 sprintf(tmpStr, "unrecognized_register");
887 break;
888 }
889
890 strcat(regstr, tmpStr);
891 }
892
893 /* Get the GLSL write mask for the destination register */
894 static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
895 char *ptr = write_mask;
896 DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
897
898 if (shader_is_scalar(param)) {
899 mask = WINED3DSP_WRITEMASK_0;
900 } else {
901 *ptr++ = '.';
902 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
903 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
904 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
905 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
906 }
907
908 *ptr = '\0';
909
910 return mask;
911 }
912
913 static size_t shader_glsl_get_write_mask_size(DWORD write_mask) {
914 size_t size = 0;
915
916 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
917 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
918 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
919 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
920
921 return size;
922 }
923
924 static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
925 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
926 * but addressed as "rgba". To fix this we need to swap the register's x
927 * and z components. */
928 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
929 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
930 char *ptr = swizzle_str;
931
932 if (!shader_is_scalar(param)) {
933 *ptr++ = '.';
934 /* swizzle bits fields: wwzzyyxx */
935 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
936 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
937 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
938 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
939 }
940
941 *ptr = '\0';
942 }
943
944 /* From a given parameter token, generate the corresponding GLSL string.
945 * Also, return the actual register name and swizzle in case the
946 * caller needs this information as well. */
947 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
948 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
949 BOOL is_color = FALSE;
950 char swizzle_str[6];
951
952 src_param->reg_name[0] = '\0';
953 src_param->param_str[0] = '\0';
954 swizzle_str[0] = '\0';
955
956 shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
957
958 shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
959 shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
960 }
961
962 /* From a given parameter token, generate the corresponding GLSL string.
963 * Also, return the actual register name and swizzle in case the
964 * caller needs this information as well. */
965 static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
966 const DWORD addr_token, glsl_dst_param_t *dst_param) {
967 BOOL is_color = FALSE;
968
969 dst_param->mask_str[0] = '\0';
970 dst_param->reg_name[0] = '\0';
971
972 shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
973 return shader_glsl_get_write_mask(param, dst_param->mask_str);
974 }
975
976 /* Append the destination part of the instruction to the buffer, return the effective write mask */
977 static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
978 glsl_dst_param_t dst_param;
979 DWORD mask;
980 int shift;
981
982 mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
983
984 if(mask) {
985 shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
986 shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
987 }
988
989 return mask;
990 }
991
992 /* Append the destination part of the instruction to the buffer, return the effective write mask */
993 static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
994 return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
995 }
996
997 /** Process GLSL instruction modifiers */
998 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
999
1000 DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
1001
1002 if (arg->opcode->dst_token && mask != 0) {
1003 glsl_dst_param_t dst_param;
1004
1005 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
1006
1007 if (mask & WINED3DSPDM_SATURATE) {
1008 /* _SAT means to clamp the value of the register to between 0 and 1 */
1009 shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
1010 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
1011 }
1012 if (mask & WINED3DSPDM_MSAMPCENTROID) {
1013 FIXME("_centroid modifier not handled\n");
1014 }
1015 if (mask & WINED3DSPDM_PARTIALPRECISION) {
1016 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1017 }
1018 }
1019 }
1020
1021 static inline const char* shader_get_comp_op(
1022 const DWORD opcode) {
1023
1024 DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
1025 switch (op) {
1026 case COMPARISON_GT: return ">";
1027 case COMPARISON_EQ: return "==";
1028 case COMPARISON_GE: return ">=";
1029 case COMPARISON_LT: return "<";
1030 case COMPARISON_NE: return "!=";
1031 case COMPARISON_LE: return "<=";
1032 default:
1033 FIXME("Unrecognized comparison value: %u\n", op);
1034 return "(\?\?)";
1035 }
1036 }
1037
1038 static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, glsl_sample_function_t *sample_function) {
1039 /* Note that there's no such thing as a projected cube texture. */
1040 switch(sampler_type) {
1041 case WINED3DSTT_1D:
1042 sample_function->name = projected ? "texture1DProj" : "texture1D";
1043 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
1044 break;
1045 case WINED3DSTT_2D:
1046 sample_function->name = projected ? "texture2DProj" : "texture2D";
1047 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
1048 break;
1049 case WINED3DSTT_CUBE:
1050 sample_function->name = "textureCube";
1051 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1052 break;
1053 case WINED3DSTT_VOLUME:
1054 sample_function->name = projected ? "texture3DProj" : "texture3D";
1055 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1056 break;
1057 default:
1058 sample_function->name = "";
1059 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
1060 break;
1061 }
1062 }
1063
1064 static void shader_glsl_color_correction(SHADER_OPCODE_ARG* arg) {
1065 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1066 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
1067 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
1068 glsl_dst_param_t dst_param;
1069 glsl_dst_param_t dst_param2;
1070 WINED3DFORMAT fmt;
1071 WINED3DFORMAT conversion_group;
1072 IWineD3DBaseTextureImpl *texture;
1073 DWORD mask, mask_size;
1074 UINT i;
1075 BOOL recorded = FALSE;
1076 DWORD sampler_idx;
1077 DWORD hex_version = shader->baseShader.hex_version;
1078
1079 switch(arg->opcode->opcode) {
1080 case WINED3DSIO_TEX:
1081 if (hex_version < WINED3DPS_VERSION(2,0)) {
1082 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1083 } else {
1084 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1085 }
1086 break;
1087
1088 case WINED3DSIO_TEXLDL:
1089 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
1090 return;
1091
1092 case WINED3DSIO_TEXDP3TEX:
1093 case WINED3DSIO_TEXM3x3TEX:
1094 case WINED3DSIO_TEXM3x3SPEC:
1095 case WINED3DSIO_TEXM3x3VSPEC:
1096 case WINED3DSIO_TEXBEM:
1097 case WINED3DSIO_TEXREG2AR:
1098 case WINED3DSIO_TEXREG2GB:
1099 case WINED3DSIO_TEXREG2RGB:
1100 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1101 break;
1102
1103 default:
1104 /* Not a texture sampling instruction, nothing to do */
1105 return;
1106 };
1107
1108 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
1109 if(texture) {
1110 fmt = texture->resource.format;
1111 conversion_group = texture->baseTexture.shader_conversion_group;
1112 } else {
1113 fmt = WINED3DFMT_UNKNOWN;
1114 conversion_group = WINED3DFMT_UNKNOWN;
1115 }
1116
1117 /* before doing anything, record the sampler with the format in the format conversion list,
1118 * but check if it's not there already
1119 */
1120 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
1121 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
1122 recorded = TRUE;
1123 break;
1124 }
1125 }
1126 if(!recorded) {
1127 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
1128 shader->baseShader.num_sampled_samplers++;
1129 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
1130 }
1131
1132 switch(fmt) {
1133 case WINED3DFMT_V8U8:
1134 case WINED3DFMT_V16U16:
1135 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
1136 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
1137 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
1138 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_2, &dst_param);
1139 mask_size = shader_glsl_get_write_mask_size(mask);
1140 if(mask_size >= 3) {
1141 shader_addline(arg->buffer, "%s.%c = 1.0;\n", dst_param.reg_name, dst_param.mask_str[3]);
1142 }
1143 } else {
1144 /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
1145 mask = shader_glsl_add_dst_param(arg, arg->dst,
1146 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1,
1147 &dst_param);
1148 mask_size = shader_glsl_get_write_mask_size(mask);
1149 if(mask_size >= 2) {
1150 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1151 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1152 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1153 } else if(mask_size == 1) {
1154 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str[1],
1155 dst_param.reg_name, dst_param.mask_str[1]);
1156 }
1157 }
1158 break;
1159
1160 case WINED3DFMT_X8L8V8U8:
1161 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1162 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
1163 * and a(X) is always 1.0
1164 */
1165 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1166 mask_size = shader_glsl_get_write_mask_size(mask);
1167 if(mask_size >= 2) {
1168 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1169 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1170 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1171 } else if(mask_size == 1) {
1172 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1173 dst_param.reg_name, dst_param.mask_str[1],
1174 dst_param.reg_name, dst_param.mask_str[1]);
1175 }
1176 }
1177 break;
1178
1179 case WINED3DFMT_L6V5U5:
1180 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1181 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1182 mask_size = shader_glsl_get_write_mask_size(mask);
1183 shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_2, &dst_param2);
1184 if(mask_size >= 3) {
1185 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1186 shader_addline(arg->buffer, "tmp0.g = %s.%c;\n",
1187 dst_param.reg_name, dst_param.mask_str[2]);
1188 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1189 dst_param.reg_name, dst_param.mask_str[2], dst_param.mask_str[1],
1190 dst_param2.reg_name, dst_param.mask_str[1], dst_param.mask_str[3]);
1191 shader_addline(arg->buffer, "%s.%c = tmp0.g;\n", dst_param.reg_name,
1192 dst_param.mask_str[3]);
1193 } else if(mask_size == 2) {
1194 /* This is bad: We have VL, but we need VU */
1195 FIXME("2 components sampled from a converted L6V5U5 texture\n");
1196 } else {
1197 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1198 dst_param.reg_name, dst_param.mask_str[1],
1199 dst_param2.reg_name, dst_param.mask_str[1]);
1200 }
1201 }
1202 break;
1203
1204 case WINED3DFMT_Q8W8V8U8:
1205 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1206 /* Correct the sign in all channels. The writemask just applies as-is, no
1207 * need for checking the mask size
1208 */
1209 shader_glsl_add_dst_param(arg, arg->dst,
1210 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 |
1211 WINED3DSP_WRITEMASK_2 | WINED3DSP_WRITEMASK_3,
1212 &dst_param);
1213 shader_addline(arg->buffer, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str,
1214 dst_param.reg_name, dst_param.mask_str);
1215 }
1216 break;
1217
1218 /* stupid compiler */
1219 default:
1220 break;
1221 }
1222 }
1223
1224 /*****************************************************************************
1225 *
1226 * Begin processing individual instruction opcodes
1227 *
1228 ****************************************************************************/
1229
1230 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1231 void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
1232 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1233 SHADER_BUFFER* buffer = arg->buffer;
1234 glsl_src_param_t src0_param;
1235 glsl_src_param_t src1_param;
1236 DWORD write_mask;
1237 char op;
1238
1239 /* Determine the GLSL operator to use based on the opcode */
1240 switch (curOpcode->opcode) {
1241 case WINED3DSIO_MUL: op = '*'; break;
1242 case WINED3DSIO_ADD: op = '+'; break;
1243 case WINED3DSIO_SUB: op = '-'; break;
1244 default:
1245 op = ' ';
1246 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1247 break;
1248 }
1249
1250 write_mask = shader_glsl_append_dst(buffer, arg);
1251 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1252 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1253 shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
1254 }
1255
1256 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1257 void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
1258 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1259 SHADER_BUFFER* buffer = arg->buffer;
1260 glsl_src_param_t src0_param;
1261 DWORD write_mask;
1262
1263 write_mask = shader_glsl_append_dst(buffer, arg);
1264 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1265
1266 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1267 * shader versions WINED3DSIO_MOVA is used for this. */
1268 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
1269 !shader_is_pshader_version(shader->baseShader.hex_version) &&
1270 shader_get_regtype(arg->dst) == WINED3DSPR_ADDR) ||
1271 arg->opcode->opcode == WINED3DSIO_MOVA) {
1272 /* We need to *round* to the nearest int here. */
1273 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1274 if (mask_size > 1) {
1275 shader_addline(buffer, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n", mask_size, src0_param.param_str, mask_size, src0_param.param_str);
1276 } else {
1277 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param.param_str, src0_param.param_str);
1278 }
1279 } else {
1280 shader_addline(buffer, "%s);\n", src0_param.param_str);
1281 }
1282 }
1283
1284 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1285 void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
1286 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1287 SHADER_BUFFER* buffer = arg->buffer;
1288 glsl_src_param_t src0_param;
1289 glsl_src_param_t src1_param;
1290 DWORD dst_write_mask, src_write_mask;
1291 size_t dst_size = 0;
1292
1293 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1294 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1295
1296 /* dp3 works on vec3, dp4 on vec4 */
1297 if (curOpcode->opcode == WINED3DSIO_DP4) {
1298 src_write_mask = WINED3DSP_WRITEMASK_ALL;
1299 } else {
1300 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1301 }
1302
1303 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
1304 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
1305
1306 if (dst_size > 1) {
1307 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1308 } else {
1309 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
1310 }
1311 }
1312
1313 /* Note that this instruction has some restrictions. The destination write mask
1314 * can't contain the w component, and the source swizzles have to be .xyzw */
1315 void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
1316 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1317 glsl_src_param_t src0_param;
1318 glsl_src_param_t src1_param;
1319 char dst_mask[6];
1320
1321 shader_glsl_get_write_mask(arg->dst, dst_mask);
1322 shader_glsl_append_dst(arg->buffer, arg);
1323 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1324 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1325 shader_addline(arg->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
1326 }
1327
1328 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1329 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1330 * GLSL uses the value as-is. */
1331 void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
1332 SHADER_BUFFER *buffer = arg->buffer;
1333 glsl_src_param_t src0_param;
1334 glsl_src_param_t src1_param;
1335 DWORD dst_write_mask;
1336 size_t dst_size;
1337
1338 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1339 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1340
1341 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1342 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1343
1344 if (dst_size > 1) {
1345 shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1346 } else {
1347 shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
1348 }
1349 }
1350
1351 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1352 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1353 * GLSL uses the value as-is. */
1354 void shader_glsl_log(SHADER_OPCODE_ARG *arg) {
1355 SHADER_BUFFER *buffer = arg->buffer;
1356 glsl_src_param_t src0_param;
1357 DWORD dst_write_mask;
1358 size_t dst_size;
1359
1360 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1361 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1362
1363 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1364
1365 if (dst_size > 1) {
1366 shader_addline(buffer, "vec%d(log2(abs(%s))));\n", dst_size, src0_param.param_str);
1367 } else {
1368 shader_addline(buffer, "log2(abs(%s)));\n", src0_param.param_str);
1369 }
1370 }
1371
1372 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1373 void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
1374 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1375 SHADER_BUFFER* buffer = arg->buffer;
1376 glsl_src_param_t src_param;
1377 const char *instruction;
1378 char arguments[256];
1379 DWORD write_mask;
1380 unsigned i;
1381
1382 /* Determine the GLSL function to use based on the opcode */
1383 /* TODO: Possibly make this a table for faster lookups */
1384 switch (curOpcode->opcode) {
1385 case WINED3DSIO_MIN: instruction = "min"; break;
1386 case WINED3DSIO_MAX: instruction = "max"; break;
1387 case WINED3DSIO_ABS: instruction = "abs"; break;
1388 case WINED3DSIO_FRC: instruction = "fract"; break;
1389 case WINED3DSIO_NRM: instruction = "normalize"; break;
1390 case WINED3DSIO_LOGP:
1391 case WINED3DSIO_LOG: instruction = "log2"; break;
1392 case WINED3DSIO_EXP: instruction = "exp2"; break;
1393 case WINED3DSIO_SGN: instruction = "sign"; break;
1394 case WINED3DSIO_DSX: instruction = "dFdx"; break;
1395 case WINED3DSIO_DSY: instruction = "dFdy"; break;
1396 default: instruction = "";
1397 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1398 break;
1399 }
1400
1401 write_mask = shader_glsl_append_dst(buffer, arg);
1402
1403 arguments[0] = '\0';
1404 if (curOpcode->num_params > 0) {
1405 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
1406 strcat(arguments, src_param.param_str);
1407 for (i = 2; i < curOpcode->num_params; ++i) {
1408 strcat(arguments, ", ");
1409 shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
1410 strcat(arguments, src_param.param_str);
1411 }
1412 }
1413
1414 shader_addline(buffer, "%s(%s));\n", instruction, arguments);
1415 }
1416
1417 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1418 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1419 * dst.x = 2^(floor(src))
1420 * dst.y = src - floor(src)
1421 * dst.z = 2^src (partial precision is allowed, but optional)
1422 * dst.w = 1.0;
1423 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1424 * dst = 2^src; (partial precision is allowed, but optional)
1425 */
1426 void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
1427 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1428 glsl_src_param_t src_param;
1429
1430 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1431
1432 if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1433 char dst_mask[6];
1434
1435 shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
1436 shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
1437 shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
1438 shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
1439
1440 shader_glsl_append_dst(arg->buffer, arg);
1441 shader_glsl_get_write_mask(arg->dst, dst_mask);
1442 shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
1443 } else {
1444 DWORD write_mask;
1445 size_t mask_size;
1446
1447 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1448 mask_size = shader_glsl_get_write_mask_size(write_mask);
1449
1450 if (mask_size > 1) {
1451 shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
1452 } else {
1453 shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
1454 }
1455 }
1456 }
1457
1458 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1459 void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
1460 glsl_src_param_t src_param;
1461 DWORD write_mask;
1462 size_t mask_size;
1463
1464 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1465 mask_size = shader_glsl_get_write_mask_size(write_mask);
1466 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1467
1468 if (mask_size > 1) {
1469 shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
1470 } else {
1471 shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
1472 }
1473 }
1474
1475 void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
1476 SHADER_BUFFER* buffer = arg->buffer;
1477 glsl_src_param_t src_param;
1478 DWORD write_mask;
1479 size_t mask_size;
1480
1481 write_mask = shader_glsl_append_dst(buffer, arg);
1482 mask_size = shader_glsl_get_write_mask_size(write_mask);
1483
1484 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1485
1486 if (mask_size > 1) {
1487 shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
1488 } else {
1489 shader_addline(buffer, "inversesqrt(%s));\n", src_param.param_str);
1490 }
1491 }
1492
1493 /** Process signed comparison opcodes in GLSL. */
1494 void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
1495 glsl_src_param_t src0_param;
1496 glsl_src_param_t src1_param;
1497 DWORD write_mask;
1498 size_t mask_size;
1499
1500 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1501 mask_size = shader_glsl_get_write_mask_size(write_mask);
1502 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1503 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1504
1505 if (mask_size > 1) {
1506 const char *compare;
1507
1508 switch(arg->opcode->opcode) {
1509 case WINED3DSIO_SLT: compare = "lessThan"; break;
1510 case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
1511 default: compare = "";
1512 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1513 }
1514
1515 shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
1516 src0_param.param_str, src1_param.param_str);
1517 } else {
1518 const char *compare;
1519
1520 switch(arg->opcode->opcode) {
1521 case WINED3DSIO_SLT: compare = "<"; break;
1522 case WINED3DSIO_SGE: compare = ">="; break;
1523 default: compare = "";
1524 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1525 }
1526
1527 shader_addline(arg->buffer, "(%s %s %s) ? 1.0 : 0.0);\n",
1528 src0_param.param_str, compare, src1_param.param_str);
1529 }
1530 }
1531
1532 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1533 void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
1534 glsl_src_param_t src0_param;
1535 glsl_src_param_t src1_param;
1536 glsl_src_param_t src2_param;
1537 DWORD write_mask, cmp_channel = 0;
1538 unsigned int i, j;
1539 char mask_char[6];
1540 BOOL temp_destination = FALSE;
1541
1542 DWORD src0reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
1543 DWORD src1reg = arg->src[1] & WINED3DSP_REGNUM_MASK;
1544 DWORD src2reg = arg->src[2] & WINED3DSP_REGNUM_MASK;
1545 DWORD src0regtype = shader_get_regtype(arg->src[0]);
1546 DWORD src1regtype = shader_get_regtype(arg->src[1]);
1547 DWORD src2regtype = shader_get_regtype(arg->src[2]);
1548 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1549 DWORD dstregtype = shader_get_regtype(arg->dst);
1550
1551 /* Cycle through all source0 channels */
1552 for (i=0; i<4; i++) {
1553 write_mask = 0;
1554 /* Find the destination channels which use the current source0 channel */
1555 for (j=0; j<4; j++) {
1556 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1557 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1558 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1559 }
1560 }
1561
1562 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1563 * The first lines may overwrite source parameters of the following lines.
1564 * Deal with that by using a temporary destination register if needed
1565 */
1566 if((src0reg == dstreg && src0regtype == dstregtype) ||
1567 (src1reg == dstreg && src1regtype == dstregtype) ||
1568 (src2reg == dstreg && src2regtype == dstregtype)) {
1569
1570 write_mask = shader_glsl_get_write_mask(arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask), mask_char);
1571 if (!write_mask) continue;
1572 shader_addline(arg->buffer, "tmp0%s = (", mask_char);
1573 temp_destination = TRUE;
1574 } else {
1575 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1576 if (!write_mask) continue;
1577 }
1578
1579 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1580 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1581 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1582
1583 shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
1584 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1585 }
1586
1587 if(temp_destination) {
1588 shader_glsl_get_write_mask(arg->dst, mask_char);
1589 shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst);
1590 shader_addline(arg->buffer, "tmp0%s);\n", mask_char);
1591 }
1592
1593 }
1594
1595 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1596 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1597 * the compare is done per component of src0. */
1598 void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
1599 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1600 glsl_src_param_t src0_param;
1601 glsl_src_param_t src1_param;
1602 glsl_src_param_t src2_param;
1603 DWORD write_mask, cmp_channel = 0;
1604 unsigned int i, j;
1605
1606 if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
1607 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1608 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1609 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1610 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1611
1612 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1613 if(arg->opcode_token & WINED3DSI_COISSUE) {
1614 shader_addline(arg->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
1615 } else {
1616 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1617 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1618 }
1619 return;
1620 }
1621 /* Cycle through all source0 channels */
1622 for (i=0; i<4; i++) {
1623 write_mask = 0;
1624 /* Find the destination channels which use the current source0 channel */
1625 for (j=0; j<4; j++) {
1626 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1627 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1628 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1629 }
1630 }
1631 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1632 if (!write_mask) continue;
1633
1634 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1635 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1636 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1637
1638 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1639 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1640 }
1641 }
1642
1643 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1644 void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
1645 glsl_src_param_t src0_param;
1646 glsl_src_param_t src1_param;
1647 glsl_src_param_t src2_param;
1648 DWORD write_mask;
1649
1650 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1651 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1652 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1653 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1654 shader_addline(arg->buffer, "(%s * %s) + %s);\n",
1655 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1656 }
1657
1658 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1659 Vertex shaders to GLSL codes */
1660 void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
1661 int i;
1662 int nComponents = 0;
1663 SHADER_OPCODE_ARG tmpArg;
1664
1665 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1666
1667 /* Set constants for the temporary argument */
1668 tmpArg.shader = arg->shader;
1669 tmpArg.buffer = arg->buffer;
1670 tmpArg.src[0] = arg->src[0];
1671 tmpArg.src_addr[0] = arg->src_addr[0];
1672 tmpArg.src_addr[1] = arg->src_addr[1];
1673 tmpArg.reg_maps = arg->reg_maps;
1674
1675 switch(arg->opcode->opcode) {
1676 case WINED3DSIO_M4x4:
1677 nComponents = 4;
1678 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1679 break;
1680 case WINED3DSIO_M4x3:
1681 nComponents = 3;
1682 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1683 break;
1684 case WINED3DSIO_M3x4:
1685 nComponents = 4;
1686 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1687 break;
1688 case WINED3DSIO_M3x3:
1689 nComponents = 3;
1690 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1691 break;
1692 case WINED3DSIO_M3x2:
1693 nComponents = 2;
1694 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1695 break;
1696 default:
1697 break;
1698 }
1699
1700 for (i = 0; i < nComponents; i++) {
1701 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1702 tmpArg.src[1] = arg->src[1]+i;
1703 shader_glsl_dot(&tmpArg);
1704 }
1705 }
1706
1707 /**
1708 The LRP instruction performs a component-wise linear interpolation
1709 between the second and third operands using the first operand as the
1710 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1711 This is equivalent to mix(src2, src1, src0);
1712 */
1713 void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
1714 glsl_src_param_t src0_param;
1715 glsl_src_param_t src1_param;
1716 glsl_src_param_t src2_param;
1717 DWORD write_mask;
1718
1719 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1720
1721 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1722 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1723 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1724
1725 shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
1726 src2_param.param_str, src1_param.param_str, src0_param.param_str);
1727 }
1728
1729 /** Process the WINED3DSIO_LIT instruction in GLSL:
1730 * dst.x = dst.w = 1.0
1731 * dst.y = (src0.x > 0) ? src0.x
1732 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1733 * where src.w is clamped at +- 128
1734 */
1735 void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
1736 glsl_src_param_t src0_param;
1737 glsl_src_param_t src1_param;
1738 glsl_src_param_t src3_param;
1739 char dst_mask[6];
1740
1741 shader_glsl_append_dst(arg->buffer, arg);
1742 shader_glsl_get_write_mask(arg->dst, dst_mask);
1743
1744 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1745 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
1746 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
1747
1748 shader_addline(arg->buffer, "vec4(1.0, (%s > 0.0 ? %s : 0.0), (%s > 0.0 ? ((%s > 0.0) ? pow(%s, clamp(%s, -128.0, 128.0)) : 0.0) : 0.0), 1.0)%s);\n",
1749 src0_param.param_str, src0_param.param_str, src0_param.param_str, src1_param.param_str, src1_param.param_str, src3_param.param_str, dst_mask);
1750 }
1751
1752 /** Process the WINED3DSIO_DST instruction in GLSL:
1753 * dst.x = 1.0
1754 * dst.y = src0.x * src0.y
1755 * dst.z = src0.z
1756 * dst.w = src1.w
1757 */
1758 void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
1759 glsl_src_param_t src0y_param;
1760 glsl_src_param_t src0z_param;
1761 glsl_src_param_t src1y_param;
1762 glsl_src_param_t src1w_param;
1763 char dst_mask[6];
1764
1765 shader_glsl_append_dst(arg->buffer, arg);
1766 shader_glsl_get_write_mask(arg->dst, dst_mask);
1767
1768 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
1769 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
1770 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
1771 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
1772
1773 shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1774 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
1775 }
1776
1777 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1778 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1779 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1780 *
1781 * dst.x = cos(src0.?)
1782 * dst.y = sin(src0.?)
1783 * dst.z = dst.z
1784 * dst.w = dst.w
1785 */
1786 void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
1787 glsl_src_param_t src0_param;
1788 DWORD write_mask;
1789
1790 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1791 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1792
1793 switch (write_mask) {
1794 case WINED3DSP_WRITEMASK_0:
1795 shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
1796 break;
1797
1798 case WINED3DSP_WRITEMASK_1:
1799 shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
1800 break;
1801
1802 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
1803 shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
1804 break;
1805
1806 default:
1807 ERR("Write mask should be .x, .y or .xy\n");
1808 break;
1809 }
1810 }
1811
1812 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1813 * Start a for() loop where src1.y is the initial value of aL,
1814 * increment aL by src1.z for a total of src1.x iterations.
1815 * Need to use a temporary variable for this operation.
1816 */
1817 /* FIXME: I don't think nested loops will work correctly this way. */
1818 void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
1819 glsl_src_param_t src1_param;
1820 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1821
1822 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
1823
1824 shader_addline(arg->buffer, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
1825 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno,
1826 src1_param.reg_name, shader->baseShader.cur_loop_depth, src1_param.reg_name,
1827 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno, src1_param.reg_name);
1828
1829 shader->baseShader.cur_loop_depth++;
1830 shader->baseShader.cur_loop_regno++;
1831 }
1832
1833 void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
1834 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1835
1836 shader_addline(arg->buffer, "}\n");
1837
1838 if(arg->opcode->opcode == WINED3DSIO_ENDLOOP) {
1839 shader->baseShader.cur_loop_depth--;
1840 shader->baseShader.cur_loop_regno--;
1841 }
1842 if(arg->opcode->opcode == WINED3DSIO_ENDREP) {
1843 shader->baseShader.cur_loop_depth--;
1844 }
1845 }
1846
1847 void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
1848 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1849 glsl_src_param_t src0_param;
1850
1851 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1852 shader_addline(arg->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
1853 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_depth,
1854 src0_param.param_str, shader->baseShader.cur_loop_depth);
1855 shader->baseShader.cur_loop_depth++;
1856 }
1857
1858 void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
1859 glsl_src_param_t src0_param;
1860
1861 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1862 shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
1863 }
1864
1865 void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
1866 glsl_src_param_t src0_param;
1867 glsl_src_param_t src1_param;
1868
1869 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1870 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1871
1872 shader_addline(arg->buffer, "if (%s %s %s) {\n",
1873 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1874 }
1875
1876 void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
1877 shader_addline(arg->buffer, "} else {\n");
1878 }
1879
1880 void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
1881 shader_addline(arg->buffer, "break;\n");
1882 }
1883
1884 /* FIXME: According to MSDN the compare is done per component. */
1885 void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
1886 glsl_src_param_t src0_param;
1887 glsl_src_param_t src1_param;
1888
1889 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1890 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1891
1892 shader_addline(arg->buffer, "if (%s %s %s) break;\n",
1893 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1894 }
1895
1896 void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
1897
1898 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1899 shader_addline(arg->buffer, "}\n");
1900 shader_addline(arg->buffer, "void subroutine%lu () {\n", snum);
1901 }
1902
1903 void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
1904 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1905 shader_addline(arg->buffer, "subroutine%lu();\n", snum);
1906 }
1907
1908 void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
1909 glsl_src_param_t src1_param;
1910
1911 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1912 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1913 shader_addline(arg->buffer, "if (%s) subroutine%lu();\n", src1_param.param_str, snum);
1914 }
1915
1916 /*********************************************
1917 * Pixel Shader Specific Code begins here
1918 ********************************************/
1919 void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
1920 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1921 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1922 DWORD hex_version = This->baseShader.hex_version;
1923 char dst_swizzle[6];
1924 glsl_sample_function_t sample_function;
1925 DWORD sampler_type;
1926 DWORD sampler_idx;
1927 BOOL projected;
1928 DWORD mask = 0;
1929
1930 /* All versions have a destination register */
1931 shader_glsl_append_dst(arg->buffer, arg);
1932
1933 /* 1.0-1.4: Use destination register as sampler source.
1934 * 2.0+: Use provided sampler source. */
1935 if (hex_version < WINED3DPS_VERSION(1,4)) {
1936 DWORD flags;
1937
1938 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1939 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1940
1941 if (flags & WINED3DTTFF_PROJECTED) {
1942 projected = TRUE;
1943 switch (flags & ~WINED3DTTFF_PROJECTED) {
1944 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
1945 case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
1946 case WINED3DTTFF_COUNT3: mask = WINED3DSP_WRITEMASK_2; break;
1947 case WINED3DTTFF_COUNT4:
1948 case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
1949 }
1950 } else {
1951 projected = FALSE;
1952 }
1953 } else if (hex_version < WINED3DPS_VERSION(2,0)) {
1954 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1955 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1956
1957 if (src_mod == WINED3DSPSM_DZ) {
1958 projected = TRUE;
1959 mask = WINED3DSP_WRITEMASK_2;
1960 } else if (src_mod == WINED3DSPSM_DW) {
1961 projected = TRUE;
1962 mask = WINED3DSP_WRITEMASK_3;
1963 } else {
1964 projected = FALSE;
1965 }
1966 } else {
1967 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1968 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1969 /* ps 2.0 texldp instruction always divides by the fourth component. */
1970 projected = TRUE;
1971 mask = WINED3DSP_WRITEMASK_3;
1972 } else {
1973 projected = FALSE;
1974 }
1975 }
1976
1977 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1978 shader_glsl_get_sample_function(sampler_type, projected, &sample_function);
1979 mask |= sample_function.coord_mask;
1980
1981 if (hex_version < WINED3DPS_VERSION(2,0)) {
1982 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
1983 } else {
1984 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1985 }
1986
1987 /* 1.0-1.3: Use destination register as coordinate source.
1988 1.4+: Use provided coordinate source register. */
1989 if (hex_version < WINED3DPS_VERSION(1,4)) {
1990 char coord_mask[6];
1991 shader_glsl_get_write_mask(mask, coord_mask);
1992 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
1993 sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
1994 } else {
1995 glsl_src_param_t coord_param;
1996 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
1997 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1998 glsl_src_param_t bias;
1999 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &bias);
2000
2001 shader_addline(arg->buffer, "%s(Psampler%u, %s, %s)%s);\n",
2002 sample_function.name, sampler_idx, coord_param.param_str,
2003 bias.param_str, dst_swizzle);
2004 } else {
2005 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
2006 sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
2007 }
2008 }
2009 }
2010
2011 void shader_glsl_texldl(SHADER_OPCODE_ARG* arg) {
2012 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*)arg->shader;
2013 glsl_sample_function_t sample_function;
2014 glsl_src_param_t coord_param, lod_param;
2015 char dst_swizzle[6];
2016 DWORD sampler_type;
2017 DWORD sampler_idx;
2018
2019 shader_glsl_append_dst(arg->buffer, arg);
2020 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
2021
2022 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
2023 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2024 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2025 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &coord_param);
2026
2027 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &lod_param);
2028
2029 if (shader_is_pshader_version(This->baseShader.hex_version)) {
2030 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2031 * However, they seem to work just fine in fragment shaders as well. */
2032 WARN("Using %sLod in fragment shader.\n", sample_function.name);
2033 shader_addline(arg->buffer, "%sLod(Psampler%u, %s, %s)%s);\n",
2034 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
2035 } else {
2036 shader_addline(arg->buffer, "%sLod(Vsampler%u, %s, %s)%s);\n",
2037 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
2038 }
2039 }
2040
2041 void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
2042
2043 /* FIXME: Make this work for more than just 2D textures */
2044
2045 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2046 SHADER_BUFFER* buffer = arg->buffer;
2047 DWORD hex_version = This->baseShader.hex_version;
2048 DWORD write_mask;
2049 char dst_mask[6];
2050
2051 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2052 shader_glsl_get_write_mask(write_mask, dst_mask);
2053
2054 if (hex_version != WINED3DPS_VERSION(1,4)) {
2055 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2056 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
2057 } else {
2058 DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
2059 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
2060 char dst_swizzle[6];
2061
2062 shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
2063
2064 if (src_mod == WINED3DSPSM_DZ) {
2065 glsl_src_param_t div_param;
2066 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
2067 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
2068
2069 if (mask_size > 1) {
2070 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
2071 } else {
2072 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
2073 }
2074 } else if (src_mod == WINED3DSPSM_DW) {
2075 glsl_src_param_t div_param;
2076 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
2077 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
2078
2079 if (mask_size > 1) {
2080 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
2081 } else {
2082 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
2083 }
2084 } else {
2085 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
2086 }
2087 }
2088 }
2089
2090 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2091 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2092 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2093 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
2094 glsl_src_param_t src0_param;
2095 char dst_mask[6];
2096 glsl_sample_function_t sample_function;
2097 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2098 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2099 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2100
2101 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2102
2103 shader_glsl_append_dst(arg->buffer, arg);
2104 shader_glsl_get_write_mask(arg->dst, dst_mask);
2105
2106 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2107 * scalar, and projected sampling would require 4
2108 */
2109 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2110
2111 switch(count_bits(sample_function.coord_mask)) {
2112 case 1:
2113 shader_addline(arg->buffer, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2114 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2115 break;
2116
2117 case 2:
2118 shader_addline(arg->buffer, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2119 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2120 break;
2121
2122 case 3:
2123 shader_addline(arg->buffer, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2124 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2125 break;
2126 default:
2127 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function.coord_mask));
2128 }
2129 }
2130
2131 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2132 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2133 void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
2134 glsl_src_param_t src0_param;
2135 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2136 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2137 DWORD dst_mask;
2138 size_t mask_size;
2139
2140 dst_mask = shader_glsl_append_dst(arg->buffer, arg);
2141 mask_size = shader_glsl_get_write_mask_size(dst_mask);
2142 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2143
2144 if (mask_size > 1) {
2145 shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
2146 } else {
2147 shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
2148 }
2149 }
2150
2151 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2152 * Calculate the depth as dst.x / dst.y */
2153 void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
2154 glsl_dst_param_t dst_param;
2155
2156 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2157
2158 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2159 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2160 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2161 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2162 * >= 1.0 or < 0.0
2163 */
2164 shader_addline(arg->buffer, "gl_FragDepth = (%s.y == 0.0) ? 1.0 : clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param.reg_name, dst_param.reg_name, dst_param.reg_name);
2165 }
2166
2167 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2168 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2169 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2170 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2171 */
2172 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
2173 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2174 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2175 glsl_src_param_t src0_param;
2176
2177 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2178
2179 shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
2180 shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2181 }
2182
2183 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2184 * Calculate the 1st of a 2-row matrix multiplication. */
2185 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
2186 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2187 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2188 SHADER_BUFFER* buffer = arg->buffer;
2189 glsl_src_param_t src0_param;
2190
2191 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2192 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2193 }
2194
2195 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2196 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2197 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
2198
2199 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2200 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2201 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2202 SHADER_BUFFER* buffer = arg->buffer;
2203 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2204 glsl_src_param_t src0_param;
2205
2206 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2207 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
2208 current_state->texcoord_w[current_state->current_row++] = reg;
2209 }
2210
2211 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
2212 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2213 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2214 SHADER_BUFFER* buffer = arg->buffer;
2215 glsl_src_param_t src0_param;
2216 char dst_mask[6];
2217
2218 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2219 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2220
2221 shader_glsl_append_dst(buffer, arg);
2222 shader_glsl_get_write_mask(arg->dst, dst_mask);
2223
2224 /* Sample the texture using the calculated coordinates */
2225 shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
2226 }
2227
2228 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2229 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2230 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
2231 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2232 glsl_src_param_t src0_param;
2233 char dst_mask[6];
2234 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2235 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2236 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2237 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2238 glsl_sample_function_t sample_function;
2239
2240 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2241 shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2242
2243 shader_glsl_append_dst(arg->buffer, arg);
2244 shader_glsl_get_write_mask(arg->dst, dst_mask);
2245 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2246
2247 /* Sample the texture using the calculated coordinates */
2248 shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2249
2250 current_state->current_row = 0;
2251 }
2252
2253 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2254 * Perform the 3rd row of a 3x3 matrix multiply */
2255 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
2256 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2257 glsl_src_param_t src0_param;
2258 char dst_mask[6];
2259 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2260 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2261 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2262
2263 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2264
2265 shader_glsl_append_dst(arg->buffer, arg);
2266 shader_glsl_get_write_mask(arg->dst, dst_mask);
2267 shader_addline(arg->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
2268
2269 current_state->current_row = 0;
2270 }
2271
2272 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2273 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2274 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
2275
2276 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2277 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2278 glsl_src_param_t src0_param;
2279 glsl_src_param_t src1_param;
2280 char dst_mask[6];
2281 SHADER_BUFFER* buffer = arg->buffer;
2282 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2283 DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2284 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2285 glsl_sample_function_t sample_function;
2286
2287 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2288 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
2289
2290 /* Perform the last matrix multiply operation */
2291 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2292 /* Reflection calculation */
2293 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
2294
2295 shader_glsl_append_dst(buffer, arg);
2296 shader_glsl_get_write_mask(arg->dst, dst_mask);
2297 shader_glsl_get_sample_function(stype, FALSE, &sample_function);
2298
2299 /* Sample the texture */
2300 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2301
2302 current_state->current_row = 0;
2303 }
2304
2305 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2306 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2307 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
2308
2309 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2310 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2311 SHADER_BUFFER* buffer = arg->buffer;
2312 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2313 glsl_src_param_t src0_param;
2314 char dst_mask[6];
2315 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2316 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2317 glsl_sample_function_t sample_function;
2318
2319 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2320
2321 /* Perform the last matrix multiply operation */
2322 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
2323
2324 /* Construct the eye-ray vector from w coordinates */
2325 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2326 current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
2327 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2328
2329 shader_glsl_append_dst(buffer, arg);
2330 shader_glsl_get_write_mask(arg->dst, dst_mask);
2331 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2332
2333 /* Sample the texture using the calculated coordinates */
2334 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2335
2336 current_state->current_row = 0;
2337 }
2338
2339 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2340 * Apply a fake bump map transform.
2341 * texbem is pshader <= 1.3 only, this saves a few version checks
2342 */
2343 void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
2344 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2345 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2346 char dst_swizzle[6];
2347 glsl_sample_function_t sample_function;
2348 glsl_src_param_t coord_param;
2349 DWORD sampler_type;
2350 DWORD sampler_idx;
2351 DWORD mask;
2352 DWORD flags;
2353 char coord_mask[6];
2354
2355 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2356 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
2357
2358 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2359 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2360 mask = sample_function.coord_mask;
2361
2362 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
2363
2364 shader_glsl_get_write_mask(mask, coord_mask);
2365
2366 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2367 * so we can't let the GL handle this.
2368 */
2369 if (flags & WINED3DTTFF_PROJECTED) {
2370 DWORD div_mask=0;
2371 char coord_div_mask[3];
2372 switch (flags & ~WINED3DTTFF_PROJECTED) {
2373 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2374 case WINED3DTTFF_COUNT2: div_mask = WINED3DSP_WRITEMASK_1; break;
2375 case WINED3DTTFF_COUNT3: div_mask = WINED3DSP_WRITEMASK_2; break;
2376 case WINED3DTTFF_COUNT4:
2377 case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
2378 }
2379 shader_glsl_get_write_mask(div_mask, coord_div_mask);
2380 shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
2381 }
2382
2383 shader_glsl_append_dst(arg->buffer, arg);
2384 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
2385 if(arg->opcode->opcode == WINED3DSIO_TEXBEML) {
2386 glsl_src_param_t luminance_param;
2387 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &luminance_param);
2388 shader_addline(arg->buffer, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )*(%s * luminancescale + luminanceoffset))%s);\n",
2389 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask,
2390 luminance_param.param_str, dst_swizzle);
2391 } else {
2392 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )%s);\n",
2393 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask, dst_swizzle);
2394 }
2395 }
2396
2397 void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
2398 glsl_src_param_t src0_param, src1_param;
2399
2400 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
2401 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
2402
2403 shader_glsl_append_dst(arg->buffer, arg);
2404 shader_addline(arg->buffer, "%s + bumpenvmat * %s);\n",
2405 src0_param.param_str, src1_param.param_str);
2406 }
2407
2408 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2409 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2410 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
2411
2412 glsl_src_param_t src0_param;
2413 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2414 char dst_mask[6];
2415
2416 shader_glsl_append_dst(arg->buffer, arg);
2417 shader_glsl_get_write_mask(arg->dst, dst_mask);
2418 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2419
2420 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2421 }
2422
2423 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2424 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2425 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
2426 glsl_src_param_t src0_param;
2427 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2428 char dst_mask[6];
2429
2430 shader_glsl_append_dst(arg->buffer, arg);
2431 shader_glsl_get_write_mask(arg->dst, dst_mask);
2432 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2433
2434 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2435 }
2436
2437 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2438 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2439 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
2440 glsl_src_param_t src0_param;
2441 char dst_mask[6];
2442 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2443 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2444 glsl_sample_function_t sample_function;
2445
2446 shader_glsl_append_dst(arg->buffer, arg);
2447 shader_glsl_get_write_mask(arg->dst, dst_mask);
2448 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2449 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
2450
2451 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
2452 }
2453
2454 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2455 * If any of the first 3 components are < 0, discard this pixel */
2456 void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
2457 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2458 DWORD hex_version = This->baseShader.hex_version;
2459 glsl_dst_param_t dst_param;
2460
2461 /* The argument is a destination parameter, and no writemasks are allowed */
2462 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2463 if((hex_version >= WINED3DPS_VERSION(2,0))) {
2464 /* 2.0 shaders compare all 4 components in texkill */
2465 shader_addline(arg->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
2466 } else {
2467 /* 1.X shaders only compare the first 3 components, propably due to the nature of the texkill
2468 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2469 * 4 components are defined, only the first 3 are used
2470 */
2471 shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
2472 }
2473 }
2474
2475 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2476 * dst = dot2(src0, src1) + src2 */
2477 void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
2478 glsl_src_param_t src0_param;
2479 glsl_src_param_t src1_param;
2480 glsl_src_param_t src2_param;
2481 DWORD write_mask;
2482 size_t mask_size;
2483
2484 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2485 mask_size = shader_glsl_get_write_mask_size(write_mask);
2486
2487 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
2488 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
2489 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
2490
2491 shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
2492 }
2493
2494 void pshader_glsl_input_pack(
2495 SHADER_BUFFER* buffer,
2496 semantic* semantics_in) {
2497
2498 unsigned int i;
2499
2500 for (i = 0; i < MAX_REG_INPUT; i++) {
2501
2502 DWORD usage_token = semantics_in[i].usage;
2503 DWORD register_token = semantics_in[i].reg;
2504 DWORD usage, usage_idx;
2505 char reg_mask[6];
2506
2507 /* Uninitialized */
2508 if (!usage_token) continue;
2509 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2510 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2511 shader_glsl_get_write_mask(register_token, reg_mask);
2512
2513 switch(usage) {
2514
2515 case WINED3DDECLUSAGE_COLOR:
2516 if (usage_idx == 0)
2517 shader_addline(buffer, "IN%u%s = vec4(gl_Color)%s;\n",
2518 i, reg_mask, reg_mask);
2519 else if (usage_idx == 1)
2520 shader_addline(buffer, "IN%u%s = vec4(gl_SecondaryColor)%s;\n",
2521 i, reg_mask, reg_mask);
2522 else
2523 shader_addline(buffer, "IN%u%s = vec4(unsupported_color_input)%s;\n",
2524 i, reg_mask, reg_mask);
2525 break;
2526
2527 case WINED3DDECLUSAGE_TEXCOORD:
2528 shader_addline(buffer, "IN%u%s = vec4(gl_TexCoord[%u])%s;\n",
2529 i, reg_mask, usage_idx, reg_mask );
2530 break;
2531
2532 case WINED3DDECLUSAGE_FOG:
2533 shader_addline(buffer, "IN%u%s = vec4(gl_FogFragCoord)%s;\n",
2534 i, reg_mask, reg_mask);
2535 break;
2536
2537 default:
2538 shader_addline(buffer, "IN%u%s = vec4(unsupported_input)%s;\n",
2539 i, reg_mask, reg_mask);
2540 }
2541 }
2542 }
2543
2544 /*********************************************
2545 * Vertex Shader Specific Code begins here
2546 ********************************************/
2547
2548 void vshader_glsl_output_unpack(
2549 SHADER_BUFFER* buffer,
2550 semantic* semantics_out) {
2551
2552 unsigned int i;
2553
2554 for (i = 0; i < MAX_REG_OUTPUT; i++) {
2555
2556 DWORD usage_token = semantics_out[i].usage;
2557 DWORD register_token = semantics_out[i].reg;
2558 DWORD usage, usage_idx;
2559 char reg_mask[6];
2560
2561 /* Uninitialized */
2562 if (!usage_token) continue;
2563
2564 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2565 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2566 shader_glsl_get_write_mask(register_token, reg_mask);
2567
2568 switch(usage) {
2569
2570 case WINED3DDECLUSAGE_COLOR:
2571 if (usage_idx == 0)
2572 shader_addline(buffer, "gl_FrontColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2573 else if (usage_idx == 1)
2574 shader_addline(buffer, "gl_FrontSecondaryColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2575 else
2576 shader_addline(buffer, "unsupported_color_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2577 break;
2578
2579 case WINED3DDECLUSAGE_POSITION:
2580 shader_addline(buffer, "gl_Position%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2581 break;
2582
2583 case WINED3DDECLUSAGE_TEXCOORD:
2584 shader_addline(buffer, "gl_TexCoord[%u]%s = OUT%u%s;\n",
2585 usage_idx, reg_mask, i, reg_mask);
2586 break;
2587
2588 case WINED3DDECLUSAGE_PSIZE:
2589 shader_addline(buffer, "gl_PointSize = OUT%u.x;\n", i);
2590 break;
2591
2592 case WINED3DDECLUSAGE_FOG:
2593 shader_addline(buffer, "gl_FogFragCoord = OUT%u%s;\n", i, reg_mask);
2594 break;
2595
2596 default:
2597 shader_addline(buffer, "unsupported_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2598 }
2599 }
2600 }
2601
2602 static void add_glsl_program_entry(IWineD3DDeviceImpl *device, struct glsl_shader_prog_link *entry) {
2603 glsl_program_key_t *key;
2604
2605 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2606 key->vshader = entry->vshader;
2607 key->pshader = entry->pshader;
2608
2609 hash_table_put(device->glsl_program_lookup, key, entry);
2610 }
2611
2612 static struct glsl_shader_prog_link *get_glsl_program_entry(IWineD3DDeviceImpl *device,
2613 GLhandleARB vshader, GLhandleARB pshader) {
2614 glsl_program_key_t key;
2615
2616 key.vshader = vshader;
2617 key.pshader = pshader;
2618
2619 return (struct glsl_shader_prog_link *)hash_table_get(device->glsl_program_lookup, &key);
2620 }
2621
2622 void delete_glsl_program_entry(IWineD3DDevice *iface, struct glsl_shader_prog_link *entry) {
2623 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2624 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2625 glsl_program_key_t *key;
2626
2627 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2628 key->vshader = entry->vshader;
2629 key->pshader = entry->pshader;
2630 hash_table_remove(This->glsl_program_lookup, key);
2631
2632 GL_EXTCALL(glDeleteObjectARB(entry->programId));
2633 if (entry->vshader) list_remove(&entry->vshader_entry);
2634 if (entry->pshader) list_remove(&entry->pshader_entry);
2635 HeapFree(GetProcessHeap(), 0, entry->vuniformF_locations);
2636 HeapFree(GetProcessHeap(), 0, entry->puniformF_locations);
2637 HeapFree(GetProcessHeap(), 0, entry);
2638 }
2639
2640 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
2641 * It sets the programId on the current StateBlock (because it should be called
2642 * inside of the DrawPrimitive() part of the render loop).
2643 *
2644 * If a program for the given combination does not exist, create one, and store
2645 * the program in the hash table. If it creates a program, it will link the
2646 * given objects, too.
2647 */
2648 static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
2649 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2650 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2651 IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
2652 IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
2653 struct glsl_shader_prog_link *entry = NULL;
2654 GLhandleARB programId = 0;
2655 int i;
2656 char glsl_name[8];
2657
2658 GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
2659 GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
2660 entry = get_glsl_program_entry(This, vshader_id, pshader_id);
2661 if (entry) {
2662 This->stateBlock->glsl_program = entry;
2663 return;
2664 }
2665
2666 /* If we get to this point, then no matching program exists, so we create one */
2667 programId = GL_EXTCALL(glCreateProgramObjectARB());
2668 TRACE("Created new GLSL shader program %u\n", programId);
2669
2670 /* Create the entry */
2671 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
2672 entry->programId = programId;
2673 entry->vshader = vshader_id;
2674 entry->pshader = pshader_id;
2675 /* Add the hash table entry */
2676 add_glsl_program_entry(This, entry);
2677
2678 /* Set the current program */
2679 This->stateBlock->glsl_program = entry;
2680
2681 /* Attach GLSL vshader */
2682 if (vshader_id) {
2683 int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
2684 char tmp_name[10];
2685
2686 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id, programId);
2687 GL_EXTCALL(glAttachObjectARB(programId, vshader_id));
2688 checkGLcall("glAttachObjectARB");
2689
2690 /* Bind vertex attributes to a corresponding index number to match
2691 * the same index numbers as ARB_vertex_programs (makes loading
2692 * vertex attributes simpler). With this method, we can use the
2693 * exact same code to load the attributes later for both ARB and
2694 * GLSL shaders.
2695 *
2696 * We have to do this here because we need to know the Program ID
2697 * in order to make the bindings work, and it has to be done prior
2698 * to linking the GLSL program. */
2699 for (i = 0; i < max_attribs; ++i) {
2700 snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
2701 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
2702 }
2703 checkGLcall("glBindAttribLocationARB");
2704
2705 list_add_head(&((IWineD3DBaseShaderImpl *)vshader)->baseShader.linked_programs, &entry->vshader_entry);
2706 }
2707
2708 /* Attach GLSL pshader */
2709 if (pshader_id) {
2710 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
2711 GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
2712 checkGLcall("glAttachObjectARB");
2713
2714 list_add_head(&((IWineD3DBaseShaderImpl *)pshader)->baseShader.linked_programs, &entry->pshader_entry);
2715 }
2716
2717 /* Link the program */
2718 TRACE("Linking GLSL shader program %u\n", programId);
2719 GL_EXTCALL(glLinkProgramARB(programId));
2720 print_glsl_info_log(&GLINFO_LOCATION, programId);
2721
2722 entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
2723 for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
2724 snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
2725 entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2726 }
2727 entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
2728 for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
2729 snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
2730 entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2731 }
2732 }
2733
2734 static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
2735 GLhandleARB program_id;
2736 GLhandleARB vshader_id, pshader_id;
2737 const char *blt_vshader[] = {
2738 "void main(void)\n"
2739 "{\n"
2740 " gl_Position = gl_Vertex;\n"
2741 " gl_FrontColor = vec4(1.0);\n"
2742 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
2743 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
2744 "}\n"
2745 };
2746
2747 const char *blt_pshader[] = {
2748 "uniform sampler2D sampler;\n"
2749 "void main(void)\n"
2750 "{\n"
2751 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
2752 "}\n"
2753 };
2754
2755 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
2756 GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
2757 GL_EXTCALL(glCompileShaderARB(vshader_id));
2758
2759 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
2760 GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
2761 GL_EXTCALL(glCompileShaderARB(pshader_id));
2762
2763 program_id = GL_EXTCALL(glCreateProgramObjectARB());
2764 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
2765 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
2766 GL_EXTCALL(glLinkProgramARB(program_id));
2767
2768 print_glsl_info_log(&GLINFO_LOCATION, program_id);
2769
2770 return program_id;
2771 }
2772
2773 static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
2774 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2775 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2776 GLhandleARB program_id = 0;
2777
2778 if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
2779 else This->stateBlock->glsl_program = NULL;
2780
2781 program_id = This->stateBlock->glsl_program ? This->stateBlock->glsl_program->programId : 0;
2782 if (program_id) TRACE("Using GLSL program %u\n", program_id);
2783 GL_EXTCALL(glUseProgramObjectARB(program_id));
2784 checkGLcall("glUseProgramObjectARB");
2785 }
2786
2787 static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
2788 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2789 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2790 static GLhandleARB program_id = 0;
2791 static GLhandleARB loc = -1;
2792
2793 if (!program_id) {
2794 program_id = create_glsl_blt_shader(gl_info);
2795 loc = GL_EXTCALL(glGetUniformLocationARB(program_id, "sampler"));
2796 }
2797
2798 GL_EXTCALL(glUseProgramObjectARB(program_id));
2799 GL_EXTCALL(glUniform1iARB(loc, 0));
2800 }
2801
2802 static void shader_glsl_cleanup(IWineD3DDevice *iface) {
2803 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2804 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2805 GL_EXTCALL(glUseProgramObjectARB(0));
2806 }
2807
2808 const shader_backend_t glsl_shader_backend = {
2809 &shader_glsl_select,
2810 &shader_glsl_select_depth_blt,
2811 &shader_glsl_load_constants,
2812 &shader_glsl_cleanup,
2813 &shader_glsl_color_correction
2814 };
useful hacks and other patches not (yet) suitable for mainline