?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Fix a typo in shader_glsl_cross.
[wine/multimedia.git] / dlls / wined3d / glsl_shader.c
blobb3ea71ecfbb6c57067cf9d10e1a8bf0412cdede9
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 }
447 }
448 }
449
450 /** Generate the variable & register declarations for the GLSL output target */
451 void shader_generate_glsl_declarations(
452 IWineD3DBaseShader *iface,
453 shader_reg_maps* reg_maps,
454 SHADER_BUFFER* buffer,
455 WineD3D_GL_Info* gl_info) {
456
457 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
458 int i;
459
460 /* There are some minor differences between pixel and vertex shaders */
461 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
462 char prefix = pshader ? 'P' : 'V';
463
464 /* Prototype the subroutines */
465 for (i = 0; i < This->baseShader.limits.label; i++) {
466 if (reg_maps->labels[i])
467 shader_addline(buffer, "void subroutine%lu();\n", i);
468 }
469
470 /* Declare the constants (aka uniforms) */
471 if (This->baseShader.limits.constant_float > 0) {
472 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
473 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
474 shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
475 }
476
477 if (This->baseShader.limits.constant_int > 0)
478 shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
479
480 if (This->baseShader.limits.constant_bool > 0)
481 shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
482
483 if(!pshader)
484 shader_addline(buffer, "uniform vec4 posFixup;\n");
485 else if(reg_maps->bumpmat != -1) {
486 shader_addline(buffer, "uniform mat2 bumpenvmat;\n");
487 if(reg_maps->luminanceparams) {
488 shader_addline(buffer, "uniform float luminancescale;\n");
489 shader_addline(buffer, "uniform float luminanceoffset;\n");
490 }
491 }
492
493 /* Declare texture samplers */
494 for (i = 0; i < This->baseShader.limits.sampler; i++) {
495 if (reg_maps->samplers[i]) {
496
497 DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
498 switch (stype) {
499
500 case WINED3DSTT_1D:
501 shader_addline(buffer, "uniform sampler1D %csampler%lu;\n", prefix, i);
502 break;
503 case WINED3DSTT_2D:
504 shader_addline(buffer, "uniform sampler2D %csampler%lu;\n", prefix, i);
505 break;
506 case WINED3DSTT_CUBE:
507 shader_addline(buffer, "uniform samplerCube %csampler%lu;\n", prefix, i);
508 break;
509 case WINED3DSTT_VOLUME:
510 shader_addline(buffer, "uniform sampler3D %csampler%lu;\n", prefix, i);
511 break;
512 default:
513 shader_addline(buffer, "uniform unsupported_sampler %csampler%lu;\n", prefix, i);
514 FIXME("Unrecognized sampler type: %#x\n", stype);
515 break;
516 }
517 }
518 }
519
520 /* Declare address variables */
521 for (i = 0; i < This->baseShader.limits.address; i++) {
522 if (reg_maps->address[i])
523 shader_addline(buffer, "ivec4 A%d;\n", i);
524 }
525
526 /* Declare texture coordinate temporaries and initialize them */
527 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
528 if (reg_maps->texcoord[i])
529 shader_addline(buffer, "vec4 T%lu = gl_TexCoord[%lu];\n", i, i);
530 }
531
532 /* Declare input register temporaries */
533 for (i=0; i < This->baseShader.limits.packed_input; i++) {
534 if (reg_maps->packed_input[i])
535 shader_addline(buffer, "vec4 IN%lu;\n", i);
536 }
537
538 /* Declare output register temporaries */
539 for (i = 0; i < This->baseShader.limits.packed_output; i++) {
540 if (reg_maps->packed_output[i])
541 shader_addline(buffer, "vec4 OUT%lu;\n", i);
542 }
543
544 /* Declare temporary variables */
545 for(i = 0; i < This->baseShader.limits.temporary; i++) {
546 if (reg_maps->temporary[i])
547 shader_addline(buffer, "vec4 R%lu;\n", i);
548 }
549
550 /* Declare attributes */
551 for (i = 0; i < This->baseShader.limits.attributes; i++) {
552 if (reg_maps->attributes[i])
553 shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
554 }
555
556 /* Declare loop register aL */
557 if (reg_maps->loop) {
558 shader_addline(buffer, "int aL;\n");
559 shader_addline(buffer, "int tmpInt;\n");
560 }
561
562 /* Temporary variables for matrix operations */
563 shader_addline(buffer, "vec4 tmp0;\n");
564 shader_addline(buffer, "vec4 tmp1;\n");
565
566 /* Start the main program */
567 shader_addline(buffer, "void main() {\n");
568 }
569
570 /*****************************************************************************
571 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
572 *
573 * For more information, see http://wiki.winehq.org/DirectX-Shaders
574 ****************************************************************************/
575
576 /* Prototypes */
577 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
578 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
579
580 /** Used for opcode modifiers - They multiply the result by the specified amount */
581 static const char * const shift_glsl_tab[] = {
582 "", /* 0 (none) */
583 "2.0 * ", /* 1 (x2) */
584 "4.0 * ", /* 2 (x4) */
585 "8.0 * ", /* 3 (x8) */
586 "16.0 * ", /* 4 (x16) */
587 "32.0 * ", /* 5 (x32) */
588 "", /* 6 (x64) */
589 "", /* 7 (x128) */
590 "", /* 8 (d256) */
591 "", /* 9 (d128) */
592 "", /* 10 (d64) */
593 "", /* 11 (d32) */
594 "0.0625 * ", /* 12 (d16) */
595 "0.125 * ", /* 13 (d8) */
596 "0.25 * ", /* 14 (d4) */
597 "0.5 * " /* 15 (d2) */
598 };
599
600 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
601 static void shader_glsl_gen_modifier (
602 const DWORD instr,
603 const char *in_reg,
604 const char *in_regswizzle,
605 char *out_str) {
606
607 out_str[0] = 0;
608
609 if (instr == WINED3DSIO_TEXKILL)
610 return;
611
612 switch (instr & WINED3DSP_SRCMOD_MASK) {
613 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
614 case WINED3DSPSM_DW:
615 case WINED3DSPSM_NONE:
616 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
617 break;
618 case WINED3DSPSM_NEG:
619 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
620 break;
621 case WINED3DSPSM_NOT:
622 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
623 break;
624 case WINED3DSPSM_BIAS:
625 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
626 break;
627 case WINED3DSPSM_BIASNEG:
628 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
629 break;
630 case WINED3DSPSM_SIGN:
631 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
632 break;
633 case WINED3DSPSM_SIGNNEG:
634 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
635 break;
636 case WINED3DSPSM_COMP:
637 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
638 break;
639 case WINED3DSPSM_X2:
640 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
641 break;
642 case WINED3DSPSM_X2NEG:
643 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
644 break;
645 case WINED3DSPSM_ABS:
646 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
647 break;
648 case WINED3DSPSM_ABSNEG:
649 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
650 break;
651 default:
652 FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
653 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
654 }
655 }
656
657 /** Writes the GLSL variable name that corresponds to the register that the
658 * DX opcode parameter is trying to access */
659 static void shader_glsl_get_register_name(
660 const DWORD param,
661 const DWORD addr_token,
662 char* regstr,
663 BOOL* is_color,
664 SHADER_OPCODE_ARG* arg) {
665
666 /* oPos, oFog and oPts in D3D */
667 static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
668
669 DWORD reg = param & WINED3DSP_REGNUM_MASK;
670 DWORD regtype = shader_get_regtype(param);
671 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
672 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
673 WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
674
675 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
676 char tmpStr[50];
677
678 *is_color = FALSE;
679
680 switch (regtype) {
681 case WINED3DSPR_TEMP:
682 sprintf(tmpStr, "R%u", reg);
683 break;
684 case WINED3DSPR_INPUT:
685 if (pshader) {
686 /* Pixel shaders >= 3.0 */
687 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
688 sprintf(tmpStr, "IN%u", reg);
689 else {
690 if (reg==0)
691 strcpy(tmpStr, "gl_Color");
692 else
693 strcpy(tmpStr, "gl_SecondaryColor");
694 }
695 } else {
696 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
697 *is_color = TRUE;
698 sprintf(tmpStr, "attrib%u", reg);
699 }
700 break;
701 case WINED3DSPR_CONST:
702 {
703 const char* prefix = pshader? "PC":"VC";
704
705 /* Relative addressing */
706 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
707
708 /* Relative addressing on shaders 2.0+ have a relative address token,
709 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
710 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
711 glsl_src_param_t rel_param;
712 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
713 sprintf(tmpStr, "%s[%s + %u]", prefix, rel_param.param_str, reg);
714 } else
715 sprintf(tmpStr, "%s[A0.x + %u]", prefix, reg);
716
717 } else
718 sprintf(tmpStr, "%s[%u]", prefix, reg);
719
720 break;
721 }
722 case WINED3DSPR_CONSTINT:
723 if (pshader)
724 sprintf(tmpStr, "PI[%u]", reg);
725 else
726 sprintf(tmpStr, "VI[%u]", reg);
727 break;
728 case WINED3DSPR_CONSTBOOL:
729 if (pshader)
730 sprintf(tmpStr, "PB[%u]", reg);
731 else
732 sprintf(tmpStr, "VB[%u]", reg);
733 break;
734 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
735 if (pshader) {
736 sprintf(tmpStr, "T%u", reg);
737 } else {
738 sprintf(tmpStr, "A%u", reg);
739 }
740 break;
741 case WINED3DSPR_LOOP:
742 sprintf(tmpStr, "aL");
743 break;
744 case WINED3DSPR_SAMPLER:
745 if (pshader)
746 sprintf(tmpStr, "Psampler%u", reg);
747 else
748 sprintf(tmpStr, "Vsampler%u", reg);
749 break;
750 case WINED3DSPR_COLOROUT:
751 if (reg >= GL_LIMITS(buffers)) {
752 WARN("Write to render target %u, only %d supported\n", reg, 4);
753 }
754 if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
755 sprintf(tmpStr, "gl_FragData[%u]", reg);
756 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
757 sprintf(tmpStr, "gl_FragColor");
758 }
759 break;
760 case WINED3DSPR_RASTOUT:
761 sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
762 break;
763 case WINED3DSPR_DEPTHOUT:
764 sprintf(tmpStr, "gl_FragDepth");
765 break;
766 case WINED3DSPR_ATTROUT:
767 if (reg == 0) {
768 sprintf(tmpStr, "gl_FrontColor");
769 } else {
770 sprintf(tmpStr, "gl_FrontSecondaryColor");
771 }
772 break;
773 case WINED3DSPR_TEXCRDOUT:
774 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
775 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
776 sprintf(tmpStr, "OUT%u", reg);
777 else
778 sprintf(tmpStr, "gl_TexCoord[%u]", reg);
779 break;
780 case WINED3DSPR_MISCTYPE:
781 if (reg == 0) {
782 /* vPos */
783 sprintf(tmpStr, "gl_FragCoord");
784 } else {
785 /* gl_FrontFacing could be used for vFace, but note that
786 * gl_FrontFacing is a bool, while vFace is a float for
787 * which the sign determines front/back */
788 FIXME("Unhandled misctype register %d\n", reg);
789 sprintf(tmpStr, "unrecognized_register");
790 }
791 break;
792 default:
793 FIXME("Unhandled register name Type(%d)\n", regtype);
794 sprintf(tmpStr, "unrecognized_register");
795 break;
796 }
797
798 strcat(regstr, tmpStr);
799 }
800
801 /* Get the GLSL write mask for the destination register */
802 static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
803 char *ptr = write_mask;
804 DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
805
806 if (shader_is_scalar(param)) {
807 mask = WINED3DSP_WRITEMASK_0;
808 } else {
809 *ptr++ = '.';
810 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
811 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
812 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
813 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
814 }
815
816 *ptr = '\0';
817
818 return mask;
819 }
820
821 static size_t shader_glsl_get_write_mask_size(DWORD write_mask) {
822 size_t size = 0;
823
824 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
825 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
826 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
827 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
828
829 return size;
830 }
831
832 static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
833 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
834 * but addressed as "rgba". To fix this we need to swap the register's x
835 * and z components. */
836 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
837 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
838 char *ptr = swizzle_str;
839
840 if (!shader_is_scalar(param)) {
841 *ptr++ = '.';
842 /* swizzle bits fields: wwzzyyxx */
843 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
844 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
845 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
846 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
847 }
848
849 *ptr = '\0';
850 }
851
852 /* From a given parameter token, generate the corresponding GLSL string.
853 * Also, return the actual register name and swizzle in case the
854 * caller needs this information as well. */
855 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
856 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
857 BOOL is_color = FALSE;
858 char swizzle_str[6];
859
860 src_param->reg_name[0] = '\0';
861 src_param->param_str[0] = '\0';
862 swizzle_str[0] = '\0';
863
864 shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
865
866 shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
867 shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
868 }
869
870 /* From a given parameter token, generate the corresponding GLSL string.
871 * Also, return the actual register name and swizzle in case the
872 * caller needs this information as well. */
873 static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
874 const DWORD addr_token, glsl_dst_param_t *dst_param) {
875 BOOL is_color = FALSE;
876
877 dst_param->mask_str[0] = '\0';
878 dst_param->reg_name[0] = '\0';
879
880 shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
881 return shader_glsl_get_write_mask(param, dst_param->mask_str);
882 }
883
884 /* Append the destination part of the instruction to the buffer, return the effective write mask */
885 static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
886 glsl_dst_param_t dst_param;
887 DWORD mask;
888 int shift;
889
890 mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
891
892 if(mask) {
893 shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
894 shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
895 }
896
897 return mask;
898 }
899
900 /* Append the destination part of the instruction to the buffer, return the effective write mask */
901 static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
902 return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
903 }
904
905 /** Process GLSL instruction modifiers */
906 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
907
908 DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
909
910 if (arg->opcode->dst_token && mask != 0) {
911 glsl_dst_param_t dst_param;
912
913 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
914
915 if (mask & WINED3DSPDM_SATURATE) {
916 /* _SAT means to clamp the value of the register to between 0 and 1 */
917 shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
918 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
919 }
920 if (mask & WINED3DSPDM_MSAMPCENTROID) {
921 FIXME("_centroid modifier not handled\n");
922 }
923 if (mask & WINED3DSPDM_PARTIALPRECISION) {
924 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
925 }
926 }
927 }
928
929 static inline const char* shader_get_comp_op(
930 const DWORD opcode) {
931
932 DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
933 switch (op) {
934 case COMPARISON_GT: return ">";
935 case COMPARISON_EQ: return "==";
936 case COMPARISON_GE: return ">=";
937 case COMPARISON_LT: return "<";
938 case COMPARISON_NE: return "!=";
939 case COMPARISON_LE: return "<=";
940 default:
941 FIXME("Unrecognized comparison value: %u\n", op);
942 return "(\?\?)";
943 }
944 }
945
946 static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, glsl_sample_function_t *sample_function) {
947 /* Note that there's no such thing as a projected cube texture. */
948 switch(sampler_type) {
949 case WINED3DSTT_1D:
950 sample_function->name = projected ? "texture1DProj" : "texture1D";
951 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
952 break;
953 case WINED3DSTT_2D:
954 sample_function->name = projected ? "texture2DProj" : "texture2D";
955 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
956 break;
957 case WINED3DSTT_CUBE:
958 sample_function->name = "textureCube";
959 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
960 break;
961 case WINED3DSTT_VOLUME:
962 sample_function->name = projected ? "texture3DProj" : "texture3D";
963 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
964 break;
965 default:
966 sample_function->name = "";
967 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
968 break;
969 }
970 }
971
972
973 /*****************************************************************************
974 *
975 * Begin processing individual instruction opcodes
976 *
977 ****************************************************************************/
978
979 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
980 void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
981 CONST SHADER_OPCODE* curOpcode = arg->opcode;
982 SHADER_BUFFER* buffer = arg->buffer;
983 glsl_src_param_t src0_param;
984 glsl_src_param_t src1_param;
985 DWORD write_mask;
986 char op;
987
988 /* Determine the GLSL operator to use based on the opcode */
989 switch (curOpcode->opcode) {
990 case WINED3DSIO_MUL: op = '*'; break;
991 case WINED3DSIO_ADD: op = '+'; break;
992 case WINED3DSIO_SUB: op = '-'; break;
993 default:
994 op = ' ';
995 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
996 break;
997 }
998
999 write_mask = shader_glsl_append_dst(buffer, arg);
1000 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1001 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1002 shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
1003 }
1004
1005 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1006 void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
1007 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1008 SHADER_BUFFER* buffer = arg->buffer;
1009 glsl_src_param_t src0_param;
1010 DWORD write_mask;
1011
1012 write_mask = shader_glsl_append_dst(buffer, arg);
1013 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1014
1015 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1016 * shader versions WINED3DSIO_MOVA is used for this. */
1017 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
1018 !shader_is_pshader_version(shader->baseShader.hex_version) &&
1019 shader_get_regtype(arg->dst) == WINED3DSPR_ADDR) ||
1020 arg->opcode->opcode == WINED3DSIO_MOVA) {
1021 /* We need to *round* to the nearest int here. */
1022 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1023 if (mask_size > 1) {
1024 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);
1025 } else {
1026 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param.param_str, src0_param.param_str);
1027 }
1028 } else {
1029 shader_addline(buffer, "%s);\n", src0_param.param_str);
1030 }
1031 }
1032
1033 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1034 void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
1035 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1036 SHADER_BUFFER* buffer = arg->buffer;
1037 glsl_src_param_t src0_param;
1038 glsl_src_param_t src1_param;
1039 DWORD dst_write_mask, src_write_mask;
1040 size_t dst_size = 0;
1041
1042 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1043 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1044
1045 /* dp3 works on vec3, dp4 on vec4 */
1046 if (curOpcode->opcode == WINED3DSIO_DP4) {
1047 src_write_mask = WINED3DSP_WRITEMASK_ALL;
1048 } else {
1049 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1050 }
1051
1052 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
1053 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
1054
1055 if (dst_size > 1) {
1056 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1057 } else {
1058 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
1059 }
1060 }
1061
1062 /* Note that this instruction has some restrictions. The destination write mask
1063 * can't contain the w component, and the source swizzles have to be .xyzw */
1064 void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
1065 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1066 glsl_src_param_t src0_param;
1067 glsl_src_param_t src1_param;
1068 char dst_mask[6];
1069
1070 shader_glsl_get_write_mask(arg->dst, dst_mask);
1071 shader_glsl_append_dst(arg->buffer, arg);
1072 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1073 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1074 shader_addline(arg->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
1075 }
1076
1077 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1078 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1079 * GLSL uses the value as-is. */
1080 void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
1081 SHADER_BUFFER *buffer = arg->buffer;
1082 glsl_src_param_t src0_param;
1083 glsl_src_param_t src1_param;
1084 DWORD dst_write_mask;
1085 size_t dst_size;
1086
1087 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1088 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1089
1090 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1091 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1092
1093 if (dst_size > 1) {
1094 shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1095 } else {
1096 shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
1097 }
1098 }
1099
1100 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1101 void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
1102 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1103 SHADER_BUFFER* buffer = arg->buffer;
1104 glsl_src_param_t src_param;
1105 const char *instruction;
1106 char arguments[256];
1107 DWORD write_mask;
1108 unsigned i;
1109
1110 /* Determine the GLSL function to use based on the opcode */
1111 /* TODO: Possibly make this a table for faster lookups */
1112 switch (curOpcode->opcode) {
1113 case WINED3DSIO_MIN: instruction = "min"; break;
1114 case WINED3DSIO_MAX: instruction = "max"; break;
1115 case WINED3DSIO_ABS: instruction = "abs"; break;
1116 case WINED3DSIO_FRC: instruction = "fract"; break;
1117 case WINED3DSIO_NRM: instruction = "normalize"; break;
1118 case WINED3DSIO_LOGP:
1119 case WINED3DSIO_LOG: instruction = "log2"; break;
1120 case WINED3DSIO_EXP: instruction = "exp2"; break;
1121 case WINED3DSIO_SGN: instruction = "sign"; break;
1122 case WINED3DSIO_DSX: instruction = "dFdx"; break;
1123 case WINED3DSIO_DSY: instruction = "dFdy"; break;
1124 default: instruction = "";
1125 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1126 break;
1127 }
1128
1129 write_mask = shader_glsl_append_dst(buffer, arg);
1130
1131 arguments[0] = '\0';
1132 if (curOpcode->num_params > 0) {
1133 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
1134 strcat(arguments, src_param.param_str);
1135 for (i = 2; i < curOpcode->num_params; ++i) {
1136 strcat(arguments, ", ");
1137 shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
1138 strcat(arguments, src_param.param_str);
1139 }
1140 }
1141
1142 shader_addline(buffer, "%s(%s));\n", instruction, arguments);
1143 }
1144
1145 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1146 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1147 * dst.x = 2^(floor(src))
1148 * dst.y = src - floor(src)
1149 * dst.z = 2^src (partial precision is allowed, but optional)
1150 * dst.w = 1.0;
1151 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1152 * dst = 2^src; (partial precision is allowed, but optional)
1153 */
1154 void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
1155 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1156 glsl_src_param_t src_param;
1157
1158 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1159
1160 if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1161 char dst_mask[6];
1162
1163 shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
1164 shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
1165 shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
1166 shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
1167
1168 shader_glsl_append_dst(arg->buffer, arg);
1169 shader_glsl_get_write_mask(arg->dst, dst_mask);
1170 shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
1171 } else {
1172 DWORD write_mask;
1173 size_t mask_size;
1174
1175 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1176 mask_size = shader_glsl_get_write_mask_size(write_mask);
1177
1178 if (mask_size > 1) {
1179 shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
1180 } else {
1181 shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
1182 }
1183 }
1184 }
1185
1186 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1187 void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
1188 glsl_src_param_t src_param;
1189 DWORD write_mask;
1190 size_t mask_size;
1191
1192 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1193 mask_size = shader_glsl_get_write_mask_size(write_mask);
1194 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1195
1196 if (mask_size > 1) {
1197 shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
1198 } else {
1199 shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
1200 }
1201 }
1202
1203 void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
1204 SHADER_BUFFER* buffer = arg->buffer;
1205 glsl_src_param_t src_param;
1206 DWORD write_mask;
1207 size_t mask_size;
1208
1209 write_mask = shader_glsl_append_dst(buffer, arg);
1210 mask_size = shader_glsl_get_write_mask_size(write_mask);
1211
1212 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1213
1214 if (mask_size > 1) {
1215 shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
1216 } else {
1217 shader_addline(buffer, "inversesqrt(%s));\n", src_param.param_str);
1218 }
1219 }
1220
1221 /** Process signed comparison opcodes in GLSL. */
1222 void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
1223 glsl_src_param_t src0_param;
1224 glsl_src_param_t src1_param;
1225 DWORD write_mask;
1226 size_t mask_size;
1227
1228 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1229 mask_size = shader_glsl_get_write_mask_size(write_mask);
1230 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1231 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1232
1233 if (mask_size > 1) {
1234 const char *compare;
1235
1236 switch(arg->opcode->opcode) {
1237 case WINED3DSIO_SLT: compare = "lessThan"; break;
1238 case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
1239 default: compare = "";
1240 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1241 }
1242
1243 shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
1244 src0_param.param_str, src1_param.param_str);
1245 } else {
1246 const char *compare;
1247
1248 switch(arg->opcode->opcode) {
1249 case WINED3DSIO_SLT: compare = "<"; break;
1250 case WINED3DSIO_SGE: compare = ">="; break;
1251 default: compare = "";
1252 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1253 }
1254
1255 shader_addline(arg->buffer, "(%s %s %s) ? 1.0 : 0.0);\n",
1256 src0_param.param_str, compare, src1_param.param_str);
1257 }
1258 }
1259
1260 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1261 void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
1262 glsl_src_param_t src0_param;
1263 glsl_src_param_t src1_param;
1264 glsl_src_param_t src2_param;
1265 DWORD write_mask, cmp_channel = 0;
1266 unsigned int i, j;
1267
1268 /* Cycle through all source0 channels */
1269 for (i=0; i<4; i++) {
1270 write_mask = 0;
1271 /* Find the destination channels which use the current source0 channel */
1272 for (j=0; j<4; j++) {
1273 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1274 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1275 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1276 }
1277 }
1278 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1279 if (!write_mask) continue;
1280
1281 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1282 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1283 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1284
1285 shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
1286 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1287 }
1288 }
1289
1290 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1291 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1292 * the compare is done per component of src0. */
1293 void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
1294 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1295 glsl_src_param_t src0_param;
1296 glsl_src_param_t src1_param;
1297 glsl_src_param_t src2_param;
1298 DWORD write_mask, cmp_channel = 0;
1299 unsigned int i, j;
1300
1301 if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
1302 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1303 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1304 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1305 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1306
1307 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1308 if(arg->opcode_token & WINED3DSI_COISSUE) {
1309 shader_addline(arg->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
1310 } else {
1311 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1312 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1313 }
1314 return;
1315 }
1316 /* Cycle through all source0 channels */
1317 for (i=0; i<4; i++) {
1318 write_mask = 0;
1319 /* Find the destination channels which use the current source0 channel */
1320 for (j=0; j<4; j++) {
1321 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1322 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1323 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1324 }
1325 }
1326 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1327 if (!write_mask) continue;
1328
1329 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1330 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1331 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1332
1333 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1334 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1335 }
1336 }
1337
1338 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1339 void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
1340 glsl_src_param_t src0_param;
1341 glsl_src_param_t src1_param;
1342 glsl_src_param_t src2_param;
1343 DWORD write_mask;
1344
1345 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1346 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1347 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1348 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1349 shader_addline(arg->buffer, "(%s * %s) + %s);\n",
1350 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1351 }
1352
1353 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1354 Vertex shaders to GLSL codes */
1355 void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
1356 int i;
1357 int nComponents = 0;
1358 SHADER_OPCODE_ARG tmpArg;
1359
1360 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1361
1362 /* Set constants for the temporary argument */
1363 tmpArg.shader = arg->shader;
1364 tmpArg.buffer = arg->buffer;
1365 tmpArg.src[0] = arg->src[0];
1366 tmpArg.src_addr[0] = arg->src_addr[0];
1367 tmpArg.src_addr[1] = arg->src_addr[1];
1368 tmpArg.reg_maps = arg->reg_maps;
1369
1370 switch(arg->opcode->opcode) {
1371 case WINED3DSIO_M4x4:
1372 nComponents = 4;
1373 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1374 break;
1375 case WINED3DSIO_M4x3:
1376 nComponents = 3;
1377 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1378 break;
1379 case WINED3DSIO_M3x4:
1380 nComponents = 4;
1381 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1382 break;
1383 case WINED3DSIO_M3x3:
1384 nComponents = 3;
1385 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1386 break;
1387 case WINED3DSIO_M3x2:
1388 nComponents = 2;
1389 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1390 break;
1391 default:
1392 break;
1393 }
1394
1395 for (i = 0; i < nComponents; i++) {
1396 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1397 tmpArg.src[1] = arg->src[1]+i;
1398 shader_glsl_dot(&tmpArg);
1399 }
1400 }
1401
1402 /**
1403 The LRP instruction performs a component-wise linear interpolation
1404 between the second and third operands using the first operand as the
1405 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1406 This is equivalent to mix(src2, src1, src0);
1407 */
1408 void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
1409 glsl_src_param_t src0_param;
1410 glsl_src_param_t src1_param;
1411 glsl_src_param_t src2_param;
1412 DWORD write_mask;
1413
1414 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1415
1416 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1417 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1418 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1419
1420 shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
1421 src2_param.param_str, src1_param.param_str, src0_param.param_str);
1422 }
1423
1424 /** Process the WINED3DSIO_LIT instruction in GLSL:
1425 * dst.x = dst.w = 1.0
1426 * dst.y = (src0.x > 0) ? src0.x
1427 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1428 * where src.w is clamped at +- 128
1429 */
1430 void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
1431 glsl_src_param_t src0_param;
1432 glsl_src_param_t src1_param;
1433 glsl_src_param_t src3_param;
1434 char dst_mask[6];
1435
1436 shader_glsl_append_dst(arg->buffer, arg);
1437 shader_glsl_get_write_mask(arg->dst, dst_mask);
1438
1439 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1440 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
1441 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
1442
1443 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",
1444 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);
1445 }
1446
1447 /** Process the WINED3DSIO_DST instruction in GLSL:
1448 * dst.x = 1.0
1449 * dst.y = src0.x * src0.y
1450 * dst.z = src0.z
1451 * dst.w = src1.w
1452 */
1453 void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
1454 glsl_src_param_t src0y_param;
1455 glsl_src_param_t src0z_param;
1456 glsl_src_param_t src1y_param;
1457 glsl_src_param_t src1w_param;
1458 char dst_mask[6];
1459
1460 shader_glsl_append_dst(arg->buffer, arg);
1461 shader_glsl_get_write_mask(arg->dst, dst_mask);
1462
1463 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
1464 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
1465 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
1466 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
1467
1468 shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1469 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
1470 }
1471
1472 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1473 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1474 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1475 *
1476 * dst.x = cos(src0.?)
1477 * dst.y = sin(src0.?)
1478 * dst.z = dst.z
1479 * dst.w = dst.w
1480 */
1481 void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
1482 glsl_src_param_t src0_param;
1483 DWORD write_mask;
1484
1485 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1486 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1487
1488 switch (write_mask) {
1489 case WINED3DSP_WRITEMASK_0:
1490 shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
1491 break;
1492
1493 case WINED3DSP_WRITEMASK_1:
1494 shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
1495 break;
1496
1497 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
1498 shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
1499 break;
1500
1501 default:
1502 ERR("Write mask should be .x, .y or .xy\n");
1503 break;
1504 }
1505 }
1506
1507 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1508 * Start a for() loop where src1.y is the initial value of aL,
1509 * increment aL by src1.z for a total of src1.x iterations.
1510 * Need to use a temporary variable for this operation.
1511 */
1512 /* FIXME: I don't think nested loops will work correctly this way. */
1513 void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
1514 glsl_src_param_t src1_param;
1515
1516 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
1517
1518 shader_addline(arg->buffer, "for (tmpInt = 0, aL = %s.y; tmpInt < %s.x; tmpInt++, aL += %s.z) {\n",
1519 src1_param.reg_name, src1_param.reg_name, src1_param.reg_name);
1520 }
1521
1522 void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
1523 shader_addline(arg->buffer, "}\n");
1524 }
1525
1526 void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
1527 glsl_src_param_t src0_param;
1528
1529 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1530 shader_addline(arg->buffer, "for (tmpInt = 0; tmpInt < %s; tmpInt++) {\n", src0_param.param_str);
1531 }
1532
1533 void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
1534 glsl_src_param_t src0_param;
1535
1536 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1537 shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
1538 }
1539
1540 void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
1541 glsl_src_param_t src0_param;
1542 glsl_src_param_t src1_param;
1543
1544 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1545 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1546
1547 shader_addline(arg->buffer, "if (%s %s %s) {\n",
1548 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1549 }
1550
1551 void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
1552 shader_addline(arg->buffer, "} else {\n");
1553 }
1554
1555 void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
1556 shader_addline(arg->buffer, "break;\n");
1557 }
1558
1559 /* FIXME: According to MSDN the compare is done per component. */
1560 void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
1561 glsl_src_param_t src0_param;
1562 glsl_src_param_t src1_param;
1563
1564 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1565 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1566
1567 shader_addline(arg->buffer, "if (%s %s %s) break;\n",
1568 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1569 }
1570
1571 void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
1572
1573 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1574 shader_addline(arg->buffer, "}\n");
1575 shader_addline(arg->buffer, "void subroutine%lu () {\n", snum);
1576 }
1577
1578 void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
1579 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1580 shader_addline(arg->buffer, "subroutine%lu();\n", snum);
1581 }
1582
1583 void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
1584 glsl_src_param_t src1_param;
1585
1586 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1587 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1588 shader_addline(arg->buffer, "if (%s) subroutine%lu();\n", src1_param.param_str, snum);
1589 }
1590
1591 /*********************************************
1592 * Pixel Shader Specific Code begins here
1593 ********************************************/
1594 void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
1595 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1596 DWORD hex_version = This->baseShader.hex_version;
1597 char dst_swizzle[6];
1598 glsl_sample_function_t sample_function;
1599 DWORD sampler_type;
1600 DWORD sampler_idx;
1601 BOOL projected;
1602 DWORD mask = 0;
1603
1604 /* All versions have a destination register */
1605 shader_glsl_append_dst(arg->buffer, arg);
1606
1607 /* 1.0-1.4: Use destination register as sampler source.
1608 * 2.0+: Use provided sampler source. */
1609 if (hex_version < WINED3DPS_VERSION(1,4)) {
1610 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1611 DWORD flags;
1612
1613 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1614 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1615
1616 if (flags & WINED3DTTFF_PROJECTED) {
1617 projected = TRUE;
1618 switch (flags & ~WINED3DTTFF_PROJECTED) {
1619 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
1620 case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
1621 case WINED3DTTFF_COUNT3: mask = WINED3DSP_WRITEMASK_2; break;
1622 case WINED3DTTFF_COUNT4:
1623 case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
1624 }
1625 } else {
1626 projected = FALSE;
1627 }
1628 } else if (hex_version < WINED3DPS_VERSION(2,0)) {
1629 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1630 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1631
1632 if (src_mod == WINED3DSPSM_DZ) {
1633 projected = TRUE;
1634 mask = WINED3DSP_WRITEMASK_2;
1635 } else if (src_mod == WINED3DSPSM_DW) {
1636 projected = TRUE;
1637 mask = WINED3DSP_WRITEMASK_3;
1638 } else {
1639 projected = FALSE;
1640 }
1641 } else {
1642 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1643 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1644 /* ps 2.0 texldp instruction always divides by the fourth component. */
1645 projected = TRUE;
1646 mask = WINED3DSP_WRITEMASK_3;
1647 } else {
1648 projected = FALSE;
1649 }
1650 }
1651
1652 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1653 shader_glsl_get_sample_function(sampler_type, projected, &sample_function);
1654 mask |= sample_function.coord_mask;
1655
1656 if (hex_version < WINED3DPS_VERSION(2,0)) {
1657 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
1658 } else {
1659 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1660 }
1661
1662 /* 1.0-1.3: Use destination register as coordinate source.
1663 1.4+: Use provided coordinate source register. */
1664 if (hex_version < WINED3DPS_VERSION(1,4)) {
1665 char coord_mask[6];
1666 shader_glsl_get_write_mask(mask, coord_mask);
1667 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
1668 sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
1669 } else {
1670 glsl_src_param_t coord_param;
1671 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
1672 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1673 glsl_src_param_t bias;
1674 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &bias);
1675
1676 shader_addline(arg->buffer, "%s(Psampler%u, %s, %s)%s);\n",
1677 sample_function.name, sampler_idx, coord_param.param_str,
1678 bias.param_str, dst_swizzle);
1679 } else {
1680 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
1681 sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
1682 }
1683 }
1684 }
1685
1686 void shader_glsl_texldl(SHADER_OPCODE_ARG* arg) {
1687 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*)arg->shader;
1688 glsl_sample_function_t sample_function;
1689 glsl_src_param_t coord_param, lod_param;
1690 char dst_swizzle[6];
1691 DWORD sampler_type;
1692 DWORD sampler_idx;
1693
1694 shader_glsl_append_dst(arg->buffer, arg);
1695 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1696
1697 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1698 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1699 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1700 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &coord_param);
1701
1702 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &lod_param);
1703
1704 if (shader_is_pshader_version(This->baseShader.hex_version)) {
1705 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
1706 * However, they seem to work just fine in fragment shaders as well. */
1707 WARN("Using %sLod in fragment shader.\n", sample_function.name);
1708 shader_addline(arg->buffer, "%sLod(Psampler%u, %s, %s)%s);\n",
1709 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
1710 } else {
1711 shader_addline(arg->buffer, "%sLod(Vsampler%u, %s, %s)%s);\n",
1712 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
1713 }
1714 }
1715
1716 void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
1717
1718 /* FIXME: Make this work for more than just 2D textures */
1719
1720 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1721 SHADER_BUFFER* buffer = arg->buffer;
1722 DWORD hex_version = This->baseShader.hex_version;
1723 DWORD write_mask;
1724 char dst_mask[6];
1725
1726 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1727 shader_glsl_get_write_mask(write_mask, dst_mask);
1728
1729 if (hex_version != WINED3DPS_VERSION(1,4)) {
1730 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1731 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
1732 } else {
1733 DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
1734 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1735 char dst_swizzle[6];
1736
1737 shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
1738
1739 if (src_mod == WINED3DSPSM_DZ) {
1740 glsl_src_param_t div_param;
1741 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1742 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
1743
1744 if (mask_size > 1) {
1745 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1746 } else {
1747 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1748 }
1749 } else if (src_mod == WINED3DSPSM_DW) {
1750 glsl_src_param_t div_param;
1751 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1752 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
1753
1754 if (mask_size > 1) {
1755 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1756 } else {
1757 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1758 }
1759 } else {
1760 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
1761 }
1762 }
1763 }
1764
1765 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
1766 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1767 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1768 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
1769 glsl_src_param_t src0_param;
1770 char dst_mask[6];
1771 glsl_sample_function_t sample_function;
1772 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1773 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1774 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1775
1776 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1777
1778 shader_glsl_append_dst(arg->buffer, arg);
1779 shader_glsl_get_write_mask(arg->dst, dst_mask);
1780
1781 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
1782 * scalar, and projected sampling would require 4
1783 */
1784 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1785
1786 switch(count_bits(sample_function.coord_mask)) {
1787 case 1:
1788 shader_addline(arg->buffer, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
1789 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
1790 break;
1791
1792 case 2:
1793 shader_addline(arg->buffer, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
1794 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
1795 break;
1796
1797 case 3:
1798 shader_addline(arg->buffer, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
1799 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
1800 break;
1801 default:
1802 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function.coord_mask));
1803 }
1804 }
1805
1806 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
1807 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1808 void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
1809 glsl_src_param_t src0_param;
1810 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1811 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1812 DWORD dst_mask;
1813 size_t mask_size;
1814
1815 dst_mask = shader_glsl_append_dst(arg->buffer, arg);
1816 mask_size = shader_glsl_get_write_mask_size(dst_mask);
1817 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1818
1819 if (mask_size > 1) {
1820 shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
1821 } else {
1822 shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
1823 }
1824 }
1825
1826 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
1827 * Calculate the depth as dst.x / dst.y */
1828 void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
1829 glsl_dst_param_t dst_param;
1830
1831 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
1832
1833 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
1834 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
1835 * this doesn't always work, so clamp the results manually. Wether or not the x value is clamped at 1
1836 * too is irrelevant, since if x = 0, any y value < 1.0(and > 1.0 is not allowed) results in a result
1837 * >= 1.0 or < 0.0
1838 */
1839 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);
1840 }
1841
1842 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
1843 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1844 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1845 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1846 */
1847 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1848 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1849 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1850 glsl_src_param_t src0_param;
1851
1852 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1853
1854 shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
1855 shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
1856 }
1857
1858 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
1859 * Calculate the 1st of a 2-row matrix multiplication. */
1860 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1861 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1862 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1863 SHADER_BUFFER* buffer = arg->buffer;
1864 glsl_src_param_t src0_param;
1865
1866 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1867 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1868 }
1869
1870 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
1871 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
1872 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1873
1874 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1875 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1876 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1877 SHADER_BUFFER* buffer = arg->buffer;
1878 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1879 glsl_src_param_t src0_param;
1880
1881 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1882 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
1883 current_state->texcoord_w[current_state->current_row++] = reg;
1884 }
1885
1886 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1887 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1888 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1889 SHADER_BUFFER* buffer = arg->buffer;
1890 glsl_src_param_t src0_param;
1891 char dst_mask[6];
1892
1893 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1894 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1895
1896 shader_glsl_append_dst(buffer, arg);
1897 shader_glsl_get_write_mask(arg->dst, dst_mask);
1898
1899 /* Sample the texture using the calculated coordinates */
1900 shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
1901 }
1902
1903 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
1904 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
1905 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1906 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1907 glsl_src_param_t src0_param;
1908 char dst_mask[6];
1909 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1910 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1911 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1912 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1913 glsl_sample_function_t sample_function;
1914
1915 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1916 shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1917
1918 shader_glsl_append_dst(arg->buffer, arg);
1919 shader_glsl_get_write_mask(arg->dst, dst_mask);
1920 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1921
1922 /* Sample the texture using the calculated coordinates */
1923 shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
1924
1925 current_state->current_row = 0;
1926 }
1927
1928 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
1929 * Perform the 3rd row of a 3x3 matrix multiply */
1930 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
1931 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1932 glsl_src_param_t src0_param;
1933 char dst_mask[6];
1934 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1935 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1936 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1937
1938 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1939
1940 shader_glsl_append_dst(arg->buffer, arg);
1941 shader_glsl_get_write_mask(arg->dst, dst_mask);
1942 shader_addline(arg->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
1943
1944 current_state->current_row = 0;
1945 }
1946
1947 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
1948 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
1949 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1950
1951 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1952 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1953 glsl_src_param_t src0_param;
1954 glsl_src_param_t src1_param;
1955 char dst_mask[6];
1956 SHADER_BUFFER* buffer = arg->buffer;
1957 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1958 DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1959 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1960 glsl_sample_function_t sample_function;
1961
1962 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1963 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1964
1965 /* Perform the last matrix multiply operation */
1966 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1967 /* Reflection calculation */
1968 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
1969
1970 shader_glsl_append_dst(buffer, arg);
1971 shader_glsl_get_write_mask(arg->dst, dst_mask);
1972 shader_glsl_get_sample_function(stype, FALSE, &sample_function);
1973
1974 /* Sample the texture */
1975 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
1976
1977 current_state->current_row = 0;
1978 }
1979
1980 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
1981 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
1982 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1983
1984 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1985 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1986 SHADER_BUFFER* buffer = arg->buffer;
1987 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1988 glsl_src_param_t src0_param;
1989 char dst_mask[6];
1990 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1991 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1992 glsl_sample_function_t sample_function;
1993
1994 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1995
1996 /* Perform the last matrix multiply operation */
1997 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
1998
1999 /* Construct the eye-ray vector from w coordinates */
2000 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2001 current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
2002 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2003
2004 shader_glsl_append_dst(buffer, arg);
2005 shader_glsl_get_write_mask(arg->dst, dst_mask);
2006 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2007
2008 /* Sample the texture using the calculated coordinates */
2009 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2010
2011 current_state->current_row = 0;
2012 }
2013
2014 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2015 * Apply a fake bump map transform.
2016 * texbem is pshader <= 1.3 only, this saves a few version checks
2017 */
2018 void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
2019 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2020 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2021 char dst_swizzle[6];
2022 glsl_sample_function_t sample_function;
2023 glsl_src_param_t coord_param;
2024 DWORD sampler_type;
2025 DWORD sampler_idx;
2026 DWORD mask;
2027 DWORD flags;
2028 char coord_mask[6];
2029
2030 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2031 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
2032
2033 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2034 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2035 mask = sample_function.coord_mask;
2036
2037 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
2038
2039 shader_glsl_get_write_mask(mask, coord_mask);
2040
2041 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2042 * so we can't let the GL handle this.
2043 */
2044 if (flags & WINED3DTTFF_PROJECTED) {
2045 DWORD div_mask=0;
2046 char coord_div_mask[3];
2047 switch (flags & ~WINED3DTTFF_PROJECTED) {
2048 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2049 case WINED3DTTFF_COUNT2: div_mask = WINED3DSP_WRITEMASK_1; break;
2050 case WINED3DTTFF_COUNT3: div_mask = WINED3DSP_WRITEMASK_2; break;
2051 case WINED3DTTFF_COUNT4:
2052 case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
2053 }
2054 shader_glsl_get_write_mask(div_mask, coord_div_mask);
2055 shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
2056 }
2057
2058 shader_glsl_append_dst(arg->buffer, arg);
2059 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
2060 if(arg->opcode->opcode == WINED3DSIO_TEXBEML) {
2061 glsl_src_param_t luminance_param;
2062 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &luminance_param);
2063 shader_addline(arg->buffer, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )*(%s * luminancescale + luminanceoffset))%s);\n",
2064 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask,
2065 luminance_param.param_str, dst_swizzle);
2066 } else {
2067 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )%s);\n",
2068 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask, dst_swizzle);
2069 }
2070 }
2071
2072 void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
2073 glsl_src_param_t src0_param, src1_param;
2074
2075 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
2076 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
2077
2078 shader_glsl_append_dst(arg->buffer, arg);
2079 shader_addline(arg->buffer, "%s + bumpenvmat * %s);\n",
2080 src0_param.param_str, src1_param.param_str);
2081 }
2082
2083 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2084 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2085 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
2086 glsl_src_param_t src0_param;
2087 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2088 char dst_mask[6];
2089
2090 shader_glsl_append_dst(arg->buffer, arg);
2091 shader_glsl_get_write_mask(arg->dst, dst_mask);
2092 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2093
2094 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2095 }
2096
2097 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2098 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2099 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
2100 glsl_src_param_t src0_param;
2101 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2102 char dst_mask[6];
2103
2104 shader_glsl_append_dst(arg->buffer, arg);
2105 shader_glsl_get_write_mask(arg->dst, dst_mask);
2106 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2107
2108 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2109 }
2110
2111 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2112 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2113 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
2114 glsl_src_param_t src0_param;
2115 char dst_mask[6];
2116 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2117 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2118 glsl_sample_function_t sample_function;
2119
2120 shader_glsl_append_dst(arg->buffer, arg);
2121 shader_glsl_get_write_mask(arg->dst, dst_mask);
2122 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2123 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
2124
2125 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
2126 }
2127
2128 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2129 * If any of the first 3 components are < 0, discard this pixel */
2130 void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
2131 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2132 DWORD hex_version = This->baseShader.hex_version;
2133 glsl_dst_param_t dst_param;
2134
2135 /* The argument is a destination parameter, and no writemasks are allowed */
2136 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2137 if((hex_version >= WINED3DPS_VERSION(2,0))) {
2138 /* 2.0 shaders compare all 4 components in texkill */
2139 shader_addline(arg->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
2140 } else {
2141 /* 1.X shaders only compare the first 3 components, propably due to the nature of the texkill
2142 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2143 * 4 components are defined, only the first 3 are used
2144 */
2145 shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
2146 }
2147 }
2148
2149 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2150 * dst = dot2(src0, src1) + src2 */
2151 void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
2152 glsl_src_param_t src0_param;
2153 glsl_src_param_t src1_param;
2154 glsl_src_param_t src2_param;
2155 DWORD write_mask;
2156 size_t mask_size;
2157
2158 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2159 mask_size = shader_glsl_get_write_mask_size(write_mask);
2160
2161 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
2162 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
2163 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
2164
2165 shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
2166 }
2167
2168 void pshader_glsl_input_pack(
2169 SHADER_BUFFER* buffer,
2170 semantic* semantics_in) {
2171
2172 unsigned int i;
2173
2174 for (i = 0; i < MAX_REG_INPUT; i++) {
2175
2176 DWORD usage_token = semantics_in[i].usage;
2177 DWORD register_token = semantics_in[i].reg;
2178 DWORD usage, usage_idx;
2179 char reg_mask[6];
2180
2181 /* Uninitialized */
2182 if (!usage_token) continue;
2183 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2184 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2185 shader_glsl_get_write_mask(register_token, reg_mask);
2186
2187 switch(usage) {
2188
2189 case WINED3DDECLUSAGE_COLOR:
2190 if (usage_idx == 0)
2191 shader_addline(buffer, "IN%u%s = vec4(gl_Color)%s;\n",
2192 i, reg_mask, reg_mask);
2193 else if (usage_idx == 1)
2194 shader_addline(buffer, "IN%u%s = vec4(gl_SecondaryColor)%s;\n",
2195 i, reg_mask, reg_mask);
2196 else
2197 shader_addline(buffer, "IN%u%s = vec4(unsupported_color_input)%s;\n",
2198 i, reg_mask, reg_mask);
2199 break;
2200
2201 case WINED3DDECLUSAGE_TEXCOORD:
2202 shader_addline(buffer, "IN%u%s = vec4(gl_TexCoord[%u])%s;\n",
2203 i, reg_mask, usage_idx, reg_mask );
2204 break;
2205
2206 case WINED3DDECLUSAGE_FOG:
2207 shader_addline(buffer, "IN%u%s = vec4(gl_FogFragCoord)%s;\n",
2208 i, reg_mask, reg_mask);
2209 break;
2210
2211 default:
2212 shader_addline(buffer, "IN%u%s = vec4(unsupported_input)%s;\n",
2213 i, reg_mask, reg_mask);
2214 }
2215 }
2216 }
2217
2218 /*********************************************
2219 * Vertex Shader Specific Code begins here
2220 ********************************************/
2221
2222 void vshader_glsl_output_unpack(
2223 SHADER_BUFFER* buffer,
2224 semantic* semantics_out) {
2225
2226 unsigned int i;
2227
2228 for (i = 0; i < MAX_REG_OUTPUT; i++) {
2229
2230 DWORD usage_token = semantics_out[i].usage;
2231 DWORD register_token = semantics_out[i].reg;
2232 DWORD usage, usage_idx;
2233 char reg_mask[6];
2234
2235 /* Uninitialized */
2236 if (!usage_token) continue;
2237
2238 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2239 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2240 shader_glsl_get_write_mask(register_token, reg_mask);
2241
2242 switch(usage) {
2243
2244 case WINED3DDECLUSAGE_COLOR:
2245 if (usage_idx == 0)
2246 shader_addline(buffer, "gl_FrontColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2247 else if (usage_idx == 1)
2248 shader_addline(buffer, "gl_FrontSecondaryColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2249 else
2250 shader_addline(buffer, "unsupported_color_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2251 break;
2252
2253 case WINED3DDECLUSAGE_POSITION:
2254 shader_addline(buffer, "gl_Position%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2255 break;
2256
2257 case WINED3DDECLUSAGE_TEXCOORD:
2258 shader_addline(buffer, "gl_TexCoord[%u]%s = OUT%u%s;\n",
2259 usage_idx, reg_mask, i, reg_mask);
2260 break;
2261
2262 case WINED3DDECLUSAGE_PSIZE:
2263 shader_addline(buffer, "gl_PointSize = OUT%u.x;\n", i);
2264 break;
2265
2266 case WINED3DDECLUSAGE_FOG:
2267 shader_addline(buffer, "gl_FogFragCoord = OUT%u%s;\n", i, reg_mask);
2268 break;
2269
2270 default:
2271 shader_addline(buffer, "unsupported_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2272 }
2273 }
2274 }
2275
2276 static void add_glsl_program_entry(IWineD3DDeviceImpl *device, struct glsl_shader_prog_link *entry) {
2277 glsl_program_key_t *key;
2278
2279 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2280 key->vshader = entry->vshader;
2281 key->pshader = entry->pshader;
2282
2283 hash_table_put(device->glsl_program_lookup, key, entry);
2284 }
2285
2286 static struct glsl_shader_prog_link *get_glsl_program_entry(IWineD3DDeviceImpl *device,
2287 GLhandleARB vshader, GLhandleARB pshader) {
2288 glsl_program_key_t key;
2289
2290 key.vshader = vshader;
2291 key.pshader = pshader;
2292
2293 return (struct glsl_shader_prog_link *)hash_table_get(device->glsl_program_lookup, &key);
2294 }
2295
2296 void delete_glsl_program_entry(IWineD3DDevice *iface, struct glsl_shader_prog_link *entry) {
2297 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2298 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2299 glsl_program_key_t *key;
2300
2301 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2302 key->vshader = entry->vshader;
2303 key->pshader = entry->pshader;
2304 hash_table_remove(This->glsl_program_lookup, key);
2305
2306 GL_EXTCALL(glDeleteObjectARB(entry->programId));
2307 if (entry->vshader) list_remove(&entry->vshader_entry);
2308 if (entry->pshader) list_remove(&entry->pshader_entry);
2309 HeapFree(GetProcessHeap(), 0, entry->vuniformF_locations);
2310 HeapFree(GetProcessHeap(), 0, entry->puniformF_locations);
2311 HeapFree(GetProcessHeap(), 0, entry);
2312 }
2313
2314 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
2315 * It sets the programId on the current StateBlock (because it should be called
2316 * inside of the DrawPrimitive() part of the render loop).
2317 *
2318 * If a program for the given combination does not exist, create one, and store
2319 * the program in the hash table. If it creates a program, it will link the
2320 * given objects, too.
2321 */
2322 static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
2323 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2324 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2325 IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
2326 IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
2327 struct glsl_shader_prog_link *entry = NULL;
2328 GLhandleARB programId = 0;
2329 int i;
2330 char glsl_name[8];
2331
2332 GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
2333 GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
2334 entry = get_glsl_program_entry(This, vshader_id, pshader_id);
2335 if (entry) {
2336 This->stateBlock->glsl_program = entry;
2337 return;
2338 }
2339
2340 /* If we get to this point, then no matching program exists, so we create one */
2341 programId = GL_EXTCALL(glCreateProgramObjectARB());
2342 TRACE("Created new GLSL shader program %u\n", programId);
2343
2344 /* Create the entry */
2345 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
2346 entry->programId = programId;
2347 entry->vshader = vshader_id;
2348 entry->pshader = pshader_id;
2349 /* Add the hash table entry */
2350 add_glsl_program_entry(This, entry);
2351
2352 /* Set the current program */
2353 This->stateBlock->glsl_program = entry;
2354
2355 /* Attach GLSL vshader */
2356 if (vshader_id) {
2357 int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
2358 char tmp_name[10];
2359
2360 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id, programId);
2361 GL_EXTCALL(glAttachObjectARB(programId, vshader_id));
2362 checkGLcall("glAttachObjectARB");
2363
2364 /* Bind vertex attributes to a corresponding index number to match
2365 * the same index numbers as ARB_vertex_programs (makes loading
2366 * vertex attributes simpler). With this method, we can use the
2367 * exact same code to load the attributes later for both ARB and
2368 * GLSL shaders.
2369 *
2370 * We have to do this here because we need to know the Program ID
2371 * in order to make the bindings work, and it has to be done prior
2372 * to linking the GLSL program. */
2373 for (i = 0; i < max_attribs; ++i) {
2374 snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
2375 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
2376 }
2377 checkGLcall("glBindAttribLocationARB");
2378
2379 list_add_head(&((IWineD3DBaseShaderImpl *)vshader)->baseShader.linked_programs, &entry->vshader_entry);
2380 }
2381
2382 /* Attach GLSL pshader */
2383 if (pshader_id) {
2384 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
2385 GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
2386 checkGLcall("glAttachObjectARB");
2387
2388 list_add_head(&((IWineD3DBaseShaderImpl *)pshader)->baseShader.linked_programs, &entry->pshader_entry);
2389 }
2390
2391 /* Link the program */
2392 TRACE("Linking GLSL shader program %u\n", programId);
2393 GL_EXTCALL(glLinkProgramARB(programId));
2394 print_glsl_info_log(&GLINFO_LOCATION, programId);
2395
2396 entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
2397 for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
2398 snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
2399 entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2400 }
2401 entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
2402 for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
2403 snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
2404 entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2405 }
2406 }
2407
2408 static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
2409 GLhandleARB program_id;
2410 GLhandleARB vshader_id, pshader_id;
2411 const char *blt_vshader[] = {
2412 "void main(void)\n"
2413 "{\n"
2414 " gl_Position = gl_Vertex;\n"
2415 " gl_FrontColor = vec4(1.0);\n"
2416 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
2417 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
2418 "}\n"
2419 };
2420
2421 const char *blt_pshader[] = {
2422 "uniform sampler2D sampler;\n"
2423 "void main(void)\n"
2424 "{\n"
2425 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
2426 "}\n"
2427 };
2428
2429 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
2430 GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
2431 GL_EXTCALL(glCompileShaderARB(vshader_id));
2432
2433 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
2434 GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
2435 GL_EXTCALL(glCompileShaderARB(pshader_id));
2436
2437 program_id = GL_EXTCALL(glCreateProgramObjectARB());
2438 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
2439 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
2440 GL_EXTCALL(glLinkProgramARB(program_id));
2441
2442 print_glsl_info_log(&GLINFO_LOCATION, program_id);
2443
2444 return program_id;
2445 }
2446
2447 static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
2448 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2449 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2450 GLhandleARB program_id = 0;
2451
2452 if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
2453 else This->stateBlock->glsl_program = NULL;
2454
2455 program_id = This->stateBlock->glsl_program ? This->stateBlock->glsl_program->programId : 0;
2456 if (program_id) TRACE("Using GLSL program %u\n", program_id);
2457 GL_EXTCALL(glUseProgramObjectARB(program_id));
2458 checkGLcall("glUseProgramObjectARB");
2459 }
2460
2461 static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
2462 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2463 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2464 static GLhandleARB program_id = 0;
2465 static GLhandleARB loc = -1;
2466
2467 if (!program_id) {
2468 program_id = create_glsl_blt_shader(gl_info);
2469 loc = GL_EXTCALL(glGetUniformLocationARB(program_id, "sampler"));
2470 }
2471
2472 GL_EXTCALL(glUseProgramObjectARB(program_id));
2473 GL_EXTCALL(glUniform1iARB(loc, 0));
2474 }
2475
2476 static void shader_glsl_cleanup(IWineD3DDevice *iface) {
2477 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2478 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2479 GL_EXTCALL(glUseProgramObjectARB(0));
2480 }
2481
2482 const shader_backend_t glsl_shader_backend = {
2483 &shader_glsl_select,
2484 &shader_glsl_select_depth_blt,
2485 &shader_glsl_load_constants,
2486 &shader_glsl_cleanup
2487 };
Maarten's multimedia branch