?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Release 0.9.39.
[wine/gsoc-2012-control.git] / dlls / wined3d / glsl_shader.c
blobafffa23fbfb858000d91a71bbfb5d854b55bc96f
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 = (char *)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< GL_LIMITS(samplers); ++i) {
89 if (stateBlock->textures[i] != NULL) {
90 snprintf(sampler_name, sizeof(sampler_name), "Psampler%d", i);
91 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
92 if (name_loc != -1) {
93 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[i];
94 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
95 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
96 checkGLcall("glUniform1iARB");
97 }
98 }
99 }
100 }
101
102 /**
103 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
104 * When constant_list == NULL, it will load all the constants.
105 */
106 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info,
107 unsigned int max_constants, float* constants, GLhandleARB *constant_locations,
108 struct list *constant_list) {
109 constants_entry *constant;
110 local_constant* lconst;
111 GLhandleARB tmp_loc;
112 DWORD i, j;
113 DWORD *idx;
114
115 if (TRACE_ON(d3d_shader)) {
116 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
117 idx = constant->idx;
118 j = constant->count;
119 while (j--) {
120 i = *idx++;
121 tmp_loc = constant_locations[i];
122 if (tmp_loc != -1) {
123 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
124 constants[i * 4 + 0], constants[i * 4 + 1],
125 constants[i * 4 + 2], constants[i * 4 + 3]);
126 }
127 }
128 }
129 }
130 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
131 idx = constant->idx;
132 j = constant->count;
133 while (j--) {
134 i = *idx++;
135 tmp_loc = constant_locations[i];
136 if (tmp_loc != -1) {
137 /* We found this uniform name in the program - go ahead and send the data */
138 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, constants + (i * 4)));
139 }
140 }
141 }
142 checkGLcall("glUniform4fvARB()");
143
144 /* Load immediate constants */
145 if (TRACE_ON(d3d_shader)) {
146 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
147 tmp_loc = constant_locations[lconst->idx];
148 if (tmp_loc != -1) {
149 GLfloat* values = (GLfloat*)lconst->value;
150 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
151 values[0], values[1], values[2], values[3]);
152 }
153 }
154 }
155 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
156 tmp_loc = constant_locations[lconst->idx];
157 if (tmp_loc != -1) {
158 /* We found this uniform name in the program - go ahead and send the data */
159 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, (GLfloat*)lconst->value));
160 }
161 }
162 checkGLcall("glUniform4fvARB()");
163 }
164
165 /**
166 * Loads integer constants (aka uniforms) into the currently set GLSL program.
167 * When @constants_set == NULL, it will load all the constants.
168 */
169 static void shader_glsl_load_constantsI(
170 IWineD3DBaseShaderImpl* This,
171 WineD3D_GL_Info *gl_info,
172 GLhandleARB programId,
173 unsigned max_constants,
174 int* constants,
175 BOOL* constants_set) {
176
177 GLhandleARB tmp_loc;
178 int i;
179 char tmp_name[8];
180 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
181 const char* prefix = is_pshader? "PI":"VI";
182 struct list* ptr;
183
184 for (i=0; i<max_constants; ++i) {
185 if (NULL == constants_set || constants_set[i]) {
186
187 TRACE_(d3d_constants)("Loading constants %i: %i, %i, %i, %i\n",
188 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
189
190 /* TODO: Benchmark and see if it would be beneficial to store the
191 * locations of the constants to avoid looking up each time */
192 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
193 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
194 if (tmp_loc != -1) {
195 /* We found this uniform name in the program - go ahead and send the data */
196 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, &constants[i*4]));
197 checkGLcall("glUniform4ivARB");
198 }
199 }
200 }
201
202 /* Load immediate constants */
203 ptr = list_head(&This->baseShader.constantsI);
204 while (ptr) {
205 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
206 unsigned int idx = lconst->idx;
207 GLint* values = (GLint*) lconst->value;
208
209 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
210 values[0], values[1], values[2], values[3]);
211
212 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
213 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
214 if (tmp_loc != -1) {
215 /* We found this uniform name in the program - go ahead and send the data */
216 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, values));
217 checkGLcall("glUniform4ivARB");
218 }
219 ptr = list_next(&This->baseShader.constantsI, ptr);
220 }
221 }
222
223 /**
224 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
225 * When @constants_set == NULL, it will load all the constants.
226 */
227 static void shader_glsl_load_constantsB(
228 IWineD3DBaseShaderImpl* This,
229 WineD3D_GL_Info *gl_info,
230 GLhandleARB programId,
231 unsigned max_constants,
232 BOOL* constants,
233 BOOL* constants_set) {
234
235 GLhandleARB tmp_loc;
236 int i;
237 char tmp_name[8];
238 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
239 const char* prefix = is_pshader? "PB":"VB";
240 struct list* ptr;
241
242 for (i=0; i<max_constants; ++i) {
243 if (NULL == constants_set || constants_set[i]) {
244
245 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i*4]);
246
247 /* TODO: Benchmark and see if it would be beneficial to store the
248 * locations of the constants to avoid looking up each time */
249 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
250 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
251 if (tmp_loc != -1) {
252 /* We found this uniform name in the program - go ahead and send the data */
253 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i*4]));
254 checkGLcall("glUniform1ivARB");
255 }
256 }
257 }
258
259 /* Load immediate constants */
260 ptr = list_head(&This->baseShader.constantsB);
261 while (ptr) {
262 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
263 unsigned int idx = lconst->idx;
264 GLint* values = (GLint*) lconst->value;
265
266 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
267
268 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
269 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
270 if (tmp_loc != -1) {
271 /* We found this uniform name in the program - go ahead and send the data */
272 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
273 checkGLcall("glUniform1ivARB");
274 }
275 ptr = list_next(&This->baseShader.constantsB, ptr);
276 }
277 }
278
279
280
281 /**
282 * Loads the app-supplied constants into the currently set GLSL program.
283 */
284 void shader_glsl_load_constants(
285 IWineD3DDevice* device,
286 char usePixelShader,
287 char useVertexShader) {
288
289 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
290 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
291 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
292
293 GLhandleARB *constant_locations;
294 struct list *constant_list;
295 GLhandleARB programId;
296 GLint pos;
297
298 if (!stateBlock->glsl_program) {
299 /* No GLSL program set - nothing to do. */
300 return;
301 }
302 programId = stateBlock->glsl_program->programId;
303
304 if (useVertexShader) {
305 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
306 GLint pos;
307
308 constant_locations = stateBlock->glsl_program->vuniformF_locations;
309 constant_list = &stateBlock->set_vconstantsF;
310
311 /* Load DirectX 9 float constants/uniforms for vertex shader */
312 shader_glsl_load_constantsF(vshader, gl_info, GL_LIMITS(vshader_constantsF),
313 stateBlock->vertexShaderConstantF, constant_locations, constant_list);
314
315 /* Load DirectX 9 integer constants/uniforms for vertex shader */
316 shader_glsl_load_constantsI(vshader, gl_info, programId, MAX_CONST_I,
317 stateBlock->vertexShaderConstantI,
318 stateBlock->set.vertexShaderConstantsI);
319
320 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
321 shader_glsl_load_constantsB(vshader, gl_info, programId, MAX_CONST_B,
322 stateBlock->vertexShaderConstantB,
323 stateBlock->set.vertexShaderConstantsB);
324
325 /* Upload the position fixup params */
326 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "posFixup"));
327 checkGLcall("glGetUniformLocationARB");
328 GL_EXTCALL(glUniform4fvARB(pos, 1, &deviceImpl->posFixup[0]));
329 checkGLcall("glUniform4fvARB");
330 }
331
332 if (usePixelShader) {
333
334 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
335
336 constant_locations = stateBlock->glsl_program->puniformF_locations;
337 constant_list = &stateBlock->set_pconstantsF;
338
339 /* Load pixel shader samplers */
340 shader_glsl_load_psamplers(gl_info, (IWineD3DStateBlock*) stateBlock);
341
342 /* Load DirectX 9 float constants/uniforms for pixel shader */
343 shader_glsl_load_constantsF(pshader, gl_info, GL_LIMITS(pshader_constantsF),
344 stateBlock->pixelShaderConstantF, constant_locations, constant_list);
345
346 /* Load DirectX 9 integer constants/uniforms for pixel shader */
347 shader_glsl_load_constantsI(pshader, gl_info, programId, MAX_CONST_I,
348 stateBlock->pixelShaderConstantI,
349 stateBlock->set.pixelShaderConstantsI);
350
351 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
352 shader_glsl_load_constantsB(pshader, gl_info, programId, MAX_CONST_B,
353 stateBlock->pixelShaderConstantB,
354 stateBlock->set.pixelShaderConstantsB);
355
356 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
357 * It can't be 0 for a valid texbem instruction.
358 */
359 if(((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat != -1) {
360 float *data = (float *) &stateBlock->textureState[(int) ((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
361 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "bumpenvmat"));
362 checkGLcall("glGetUniformLocationARB");
363 GL_EXTCALL(glUniformMatrix2fvARB(pos, 1, 0, data));
364 checkGLcall("glUniform4fvARB");
365 }
366 }
367 }
368
369 /** Generate the variable & register declarations for the GLSL output target */
370 void shader_generate_glsl_declarations(
371 IWineD3DBaseShader *iface,
372 shader_reg_maps* reg_maps,
373 SHADER_BUFFER* buffer,
374 WineD3D_GL_Info* gl_info) {
375
376 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
377 int i;
378
379 /* There are some minor differences between pixel and vertex shaders */
380 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
381 char prefix = pshader ? 'P' : 'V';
382
383 /* Prototype the subroutines */
384 for (i = 0; i < This->baseShader.limits.label; i++) {
385 if (reg_maps->labels[i])
386 shader_addline(buffer, "void subroutine%lu();\n", i);
387 }
388
389 /* Declare the constants (aka uniforms) */
390 if (This->baseShader.limits.constant_float > 0) {
391 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
392 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
393 shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
394 }
395
396 if (This->baseShader.limits.constant_int > 0)
397 shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
398
399 if (This->baseShader.limits.constant_bool > 0)
400 shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
401
402 if(!pshader)
403 shader_addline(buffer, "uniform vec4 posFixup;\n");
404 else if(reg_maps->bumpmat != -1)
405 shader_addline(buffer, "uniform mat2 bumpenvmat;\n");
406
407 /* Declare texture samplers */
408 for (i = 0; i < This->baseShader.limits.sampler; i++) {
409 if (reg_maps->samplers[i]) {
410
411 DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
412 switch (stype) {
413
414 case WINED3DSTT_1D:
415 shader_addline(buffer, "uniform sampler1D %csampler%lu;\n", prefix, i);
416 break;
417 case WINED3DSTT_2D:
418 shader_addline(buffer, "uniform sampler2D %csampler%lu;\n", prefix, i);
419 break;
420 case WINED3DSTT_CUBE:
421 shader_addline(buffer, "uniform samplerCube %csampler%lu;\n", prefix, i);
422 break;
423 case WINED3DSTT_VOLUME:
424 shader_addline(buffer, "uniform sampler3D %csampler%lu;\n", prefix, i);
425 break;
426 default:
427 shader_addline(buffer, "uniform unsupported_sampler %csampler%lu;\n", prefix, i);
428 FIXME("Unrecognized sampler type: %#x\n", stype);
429 break;
430 }
431 }
432 }
433
434 /* Declare address variables */
435 for (i = 0; i < This->baseShader.limits.address; i++) {
436 if (reg_maps->address[i])
437 shader_addline(buffer, "ivec4 A%d;\n", i);
438 }
439
440 /* Declare texture coordinate temporaries and initialize them */
441 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
442 if (reg_maps->texcoord[i])
443 shader_addline(buffer, "vec4 T%lu = gl_TexCoord[%lu];\n", i, i);
444 }
445
446 /* Declare input register temporaries */
447 for (i=0; i < This->baseShader.limits.packed_input; i++) {
448 if (reg_maps->packed_input[i])
449 shader_addline(buffer, "vec4 IN%lu;\n", i);
450 }
451
452 /* Declare output register temporaries */
453 for (i = 0; i < This->baseShader.limits.packed_output; i++) {
454 if (reg_maps->packed_output[i])
455 shader_addline(buffer, "vec4 OUT%lu;\n", i);
456 }
457
458 /* Declare temporary variables */
459 for(i = 0; i < This->baseShader.limits.temporary; i++) {
460 if (reg_maps->temporary[i])
461 shader_addline(buffer, "vec4 R%lu;\n", i);
462 }
463
464 /* Declare attributes */
465 for (i = 0; i < This->baseShader.limits.attributes; i++) {
466 if (reg_maps->attributes[i])
467 shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
468 }
469
470 /* Declare loop register aL */
471 if (reg_maps->loop) {
472 shader_addline(buffer, "int aL;\n");
473 shader_addline(buffer, "int tmpInt;\n");
474 }
475
476 /* Temporary variables for matrix operations */
477 shader_addline(buffer, "vec4 tmp0;\n");
478 shader_addline(buffer, "vec4 tmp1;\n");
479
480 /* Start the main program */
481 shader_addline(buffer, "void main() {\n");
482 }
483
484 /*****************************************************************************
485 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
486 *
487 * For more information, see http://wiki.winehq.org/DirectX-Shaders
488 ****************************************************************************/
489
490 /* Prototypes */
491 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
492 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
493
494 /** Used for opcode modifiers - They multiply the result by the specified amount */
495 static const char * const shift_glsl_tab[] = {
496 "", /* 0 (none) */
497 "2.0 * ", /* 1 (x2) */
498 "4.0 * ", /* 2 (x4) */
499 "8.0 * ", /* 3 (x8) */
500 "16.0 * ", /* 4 (x16) */
501 "32.0 * ", /* 5 (x32) */
502 "", /* 6 (x64) */
503 "", /* 7 (x128) */
504 "", /* 8 (d256) */
505 "", /* 9 (d128) */
506 "", /* 10 (d64) */
507 "", /* 11 (d32) */
508 "0.0625 * ", /* 12 (d16) */
509 "0.125 * ", /* 13 (d8) */
510 "0.25 * ", /* 14 (d4) */
511 "0.5 * " /* 15 (d2) */
512 };
513
514 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
515 static void shader_glsl_gen_modifier (
516 const DWORD instr,
517 const char *in_reg,
518 const char *in_regswizzle,
519 char *out_str) {
520
521 out_str[0] = 0;
522
523 if (instr == WINED3DSIO_TEXKILL)
524 return;
525
526 switch (instr & WINED3DSP_SRCMOD_MASK) {
527 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
528 case WINED3DSPSM_DW:
529 case WINED3DSPSM_NONE:
530 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
531 break;
532 case WINED3DSPSM_NEG:
533 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
534 break;
535 case WINED3DSPSM_NOT:
536 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
537 break;
538 case WINED3DSPSM_BIAS:
539 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
540 break;
541 case WINED3DSPSM_BIASNEG:
542 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
543 break;
544 case WINED3DSPSM_SIGN:
545 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
546 break;
547 case WINED3DSPSM_SIGNNEG:
548 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
549 break;
550 case WINED3DSPSM_COMP:
551 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
552 break;
553 case WINED3DSPSM_X2:
554 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
555 break;
556 case WINED3DSPSM_X2NEG:
557 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
558 break;
559 case WINED3DSPSM_ABS:
560 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
561 break;
562 case WINED3DSPSM_ABSNEG:
563 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
564 break;
565 default:
566 FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
567 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
568 }
569 }
570
571 /** Writes the GLSL variable name that corresponds to the register that the
572 * DX opcode parameter is trying to access */
573 static void shader_glsl_get_register_name(
574 const DWORD param,
575 const DWORD addr_token,
576 char* regstr,
577 BOOL* is_color,
578 SHADER_OPCODE_ARG* arg) {
579
580 /* oPos, oFog and oPts in D3D */
581 static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
582
583 DWORD reg = param & WINED3DSP_REGNUM_MASK;
584 DWORD regtype = shader_get_regtype(param);
585 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
586 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
587 WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
588
589 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
590 char tmpStr[50];
591
592 *is_color = FALSE;
593
594 switch (regtype) {
595 case WINED3DSPR_TEMP:
596 sprintf(tmpStr, "R%u", reg);
597 break;
598 case WINED3DSPR_INPUT:
599 if (pshader) {
600 /* Pixel shaders >= 3.0 */
601 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
602 sprintf(tmpStr, "IN%u", reg);
603 else {
604 if (reg==0)
605 strcpy(tmpStr, "gl_Color");
606 else
607 strcpy(tmpStr, "gl_SecondaryColor");
608 }
609 } else {
610 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
611 *is_color = TRUE;
612 sprintf(tmpStr, "attrib%u", reg);
613 }
614 break;
615 case WINED3DSPR_CONST:
616 {
617 const char* prefix = pshader? "PC":"VC";
618
619 /* Relative addressing */
620 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
621
622 /* Relative addressing on shaders 2.0+ have a relative address token,
623 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
624 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
625 glsl_src_param_t rel_param;
626 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
627 sprintf(tmpStr, "%s[%s + %u]", prefix, rel_param.param_str, reg);
628 } else
629 sprintf(tmpStr, "%s[A0.x + %u]", prefix, reg);
630
631 } else
632 sprintf(tmpStr, "%s[%u]", prefix, reg);
633
634 break;
635 }
636 case WINED3DSPR_CONSTINT:
637 if (pshader)
638 sprintf(tmpStr, "PI[%u]", reg);
639 else
640 sprintf(tmpStr, "VI[%u]", reg);
641 break;
642 case WINED3DSPR_CONSTBOOL:
643 if (pshader)
644 sprintf(tmpStr, "PB[%u]", reg);
645 else
646 sprintf(tmpStr, "VB[%u]", reg);
647 break;
648 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
649 if (pshader) {
650 sprintf(tmpStr, "T%u", reg);
651 } else {
652 sprintf(tmpStr, "A%u", reg);
653 }
654 break;
655 case WINED3DSPR_LOOP:
656 sprintf(tmpStr, "aL");
657 break;
658 case WINED3DSPR_SAMPLER:
659 if (pshader)
660 sprintf(tmpStr, "Psampler%u", reg);
661 else
662 sprintf(tmpStr, "Vsampler%u", reg);
663 break;
664 case WINED3DSPR_COLOROUT:
665 if (reg >= GL_LIMITS(buffers)) {
666 WARN("Write to render target %u, only %d supported\n", reg, 4);
667 }
668 if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
669 sprintf(tmpStr, "gl_FragData[%u]", reg);
670 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
671 sprintf(tmpStr, "gl_FragColor");
672 }
673 break;
674 case WINED3DSPR_RASTOUT:
675 sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
676 break;
677 case WINED3DSPR_DEPTHOUT:
678 sprintf(tmpStr, "gl_FragDepth");
679 break;
680 case WINED3DSPR_ATTROUT:
681 if (reg == 0) {
682 sprintf(tmpStr, "gl_FrontColor");
683 } else {
684 sprintf(tmpStr, "gl_FrontSecondaryColor");
685 }
686 break;
687 case WINED3DSPR_TEXCRDOUT:
688 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
689 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
690 sprintf(tmpStr, "OUT%u", reg);
691 else
692 sprintf(tmpStr, "gl_TexCoord[%u]", reg);
693 break;
694 case WINED3DSPR_MISCTYPE:
695 if (reg == 0) {
696 /* vPos */
697 sprintf(tmpStr, "gl_FragCoord");
698 } else {
699 /* gl_FrontFacing could be used for vFace, but note that
700 * gl_FrontFacing is a bool, while vFace is a float for
701 * which the sign determines front/back */
702 FIXME("Unhandled misctype register %d\n", reg);
703 sprintf(tmpStr, "unrecognized_register");
704 }
705 break;
706 default:
707 FIXME("Unhandled register name Type(%d)\n", regtype);
708 sprintf(tmpStr, "unrecognized_register");
709 break;
710 }
711
712 strcat(regstr, tmpStr);
713 }
714
715 /* Get the GLSL write mask for the destination register */
716 static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
717 char *ptr = write_mask;
718 DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
719
720 if (shader_is_scalar(param)) {
721 mask = WINED3DSP_WRITEMASK_0;
722 } else {
723 *ptr++ = '.';
724 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
725 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
726 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
727 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
728 }
729
730 *ptr = '\0';
731
732 return mask;
733 }
734
735 static size_t shader_glsl_get_write_mask_size(DWORD write_mask) {
736 size_t size = 0;
737
738 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
739 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
740 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
741 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
742
743 return size;
744 }
745
746 static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
747 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
748 * but addressed as "rgba". To fix this we need to swap the register's x
749 * and z components. */
750 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
751 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
752 char *ptr = swizzle_str;
753
754 if (!shader_is_scalar(param)) {
755 *ptr++ = '.';
756 /* swizzle bits fields: wwzzyyxx */
757 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
758 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
759 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
760 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
761 }
762
763 *ptr = '\0';
764 }
765
766 /* From a given parameter token, generate the corresponding GLSL string.
767 * Also, return the actual register name and swizzle in case the
768 * caller needs this information as well. */
769 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
770 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
771 BOOL is_color = FALSE;
772 char swizzle_str[6];
773
774 src_param->reg_name[0] = '\0';
775 src_param->param_str[0] = '\0';
776 swizzle_str[0] = '\0';
777
778 shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
779
780 shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
781 shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
782 }
783
784 /* From a given parameter token, generate the corresponding GLSL string.
785 * Also, return the actual register name and swizzle in case the
786 * caller needs this information as well. */
787 static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
788 const DWORD addr_token, glsl_dst_param_t *dst_param) {
789 BOOL is_color = FALSE;
790
791 dst_param->mask_str[0] = '\0';
792 dst_param->reg_name[0] = '\0';
793
794 shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
795 return shader_glsl_get_write_mask(param, dst_param->mask_str);
796 }
797
798 /* Append the destination part of the instruction to the buffer, return the effective write mask */
799 static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
800 glsl_dst_param_t dst_param;
801 DWORD mask;
802 int shift;
803
804 mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
805
806 if(mask) {
807 shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
808 shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
809 }
810
811 return mask;
812 }
813
814 /* Append the destination part of the instruction to the buffer, return the effective write mask */
815 static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
816 return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
817 }
818
819 /** Process GLSL instruction modifiers */
820 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
821
822 DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
823
824 if (arg->opcode->dst_token && mask != 0) {
825 glsl_dst_param_t dst_param;
826
827 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
828
829 if (mask & WINED3DSPDM_SATURATE) {
830 /* _SAT means to clamp the value of the register to between 0 and 1 */
831 shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
832 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
833 }
834 if (mask & WINED3DSPDM_MSAMPCENTROID) {
835 FIXME("_centroid modifier not handled\n");
836 }
837 if (mask & WINED3DSPDM_PARTIALPRECISION) {
838 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
839 }
840 }
841 }
842
843 static inline const char* shader_get_comp_op(
844 const DWORD opcode) {
845
846 DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
847 switch (op) {
848 case COMPARISON_GT: return ">";
849 case COMPARISON_EQ: return "==";
850 case COMPARISON_GE: return ">=";
851 case COMPARISON_LT: return "<";
852 case COMPARISON_NE: return "!=";
853 case COMPARISON_LE: return "<=";
854 default:
855 FIXME("Unrecognized comparison value: %u\n", op);
856 return "(\?\?)";
857 }
858 }
859
860 static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, glsl_sample_function_t *sample_function) {
861 /* Note that there's no such thing as a projected cube texture. */
862 switch(sampler_type) {
863 case WINED3DSTT_1D:
864 sample_function->name = projected ? "texture1DProj" : "texture1D";
865 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
866 break;
867 case WINED3DSTT_2D:
868 sample_function->name = projected ? "texture2DProj" : "texture2D";
869 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
870 break;
871 case WINED3DSTT_CUBE:
872 sample_function->name = "textureCube";
873 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
874 break;
875 case WINED3DSTT_VOLUME:
876 sample_function->name = projected ? "texture3DProj" : "texture3D";
877 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
878 break;
879 default:
880 sample_function->name = "";
881 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
882 break;
883 }
884 }
885
886
887 /*****************************************************************************
888 *
889 * Begin processing individual instruction opcodes
890 *
891 ****************************************************************************/
892
893 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
894 void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
895 CONST SHADER_OPCODE* curOpcode = arg->opcode;
896 SHADER_BUFFER* buffer = arg->buffer;
897 glsl_src_param_t src0_param;
898 glsl_src_param_t src1_param;
899 DWORD write_mask;
900 char op;
901
902 /* Determine the GLSL operator to use based on the opcode */
903 switch (curOpcode->opcode) {
904 case WINED3DSIO_MUL: op = '*'; break;
905 case WINED3DSIO_ADD: op = '+'; break;
906 case WINED3DSIO_SUB: op = '-'; break;
907 default:
908 op = ' ';
909 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
910 break;
911 }
912
913 write_mask = shader_glsl_append_dst(buffer, arg);
914 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
915 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
916 shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
917 }
918
919 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
920 void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
921 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
922 SHADER_BUFFER* buffer = arg->buffer;
923 glsl_src_param_t src0_param;
924 DWORD write_mask;
925
926 write_mask = shader_glsl_append_dst(buffer, arg);
927 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
928
929 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
930 * shader versions WINED3DSIO_MOVA is used for this. */
931 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
932 !shader_is_pshader_version(shader->baseShader.hex_version) &&
933 shader_get_regtype(arg->dst) == WINED3DSPR_ADDR) ||
934 arg->opcode->opcode == WINED3DSIO_MOVA) {
935 /* We need to *round* to the nearest int here. */
936 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
937 if (mask_size > 1) {
938 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);
939 } else {
940 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param.param_str, src0_param.param_str);
941 }
942 } else {
943 shader_addline(buffer, "%s);\n", src0_param.param_str);
944 }
945 }
946
947 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
948 void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
949 CONST SHADER_OPCODE* curOpcode = arg->opcode;
950 SHADER_BUFFER* buffer = arg->buffer;
951 glsl_src_param_t src0_param;
952 glsl_src_param_t src1_param;
953 DWORD dst_write_mask, src_write_mask;
954 size_t dst_size = 0;
955
956 dst_write_mask = shader_glsl_append_dst(buffer, arg);
957 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
958
959 /* dp3 works on vec3, dp4 on vec4 */
960 if (curOpcode->opcode == WINED3DSIO_DP4) {
961 src_write_mask = WINED3DSP_WRITEMASK_ALL;
962 } else {
963 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
964 }
965
966 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
967 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
968
969 if (dst_size > 1) {
970 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
971 } else {
972 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
973 }
974 }
975
976 /* Note that this instruction has some restrictions. The destination write mask
977 * can't contain the w component, and the source swizzles have to be .xyzw */
978 void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
979 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
980 glsl_src_param_t src0_param;
981 glsl_src_param_t src1_param;
982 char dst_mask[6];
983
984 shader_glsl_get_write_mask(arg->dst, dst_mask);
985 shader_glsl_append_dst(arg->buffer, arg);
986 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
987 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
988 shader_addline(arg->buffer, "cross(%s, %s).%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
989 }
990
991 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
992 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
993 * GLSL uses the value as-is. */
994 void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
995 SHADER_BUFFER *buffer = arg->buffer;
996 glsl_src_param_t src0_param;
997 glsl_src_param_t src1_param;
998 DWORD dst_write_mask;
999 size_t dst_size;
1000
1001 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1002 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1003
1004 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1005 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1006
1007 if (dst_size > 1) {
1008 shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1009 } else {
1010 shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
1011 }
1012 }
1013
1014 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1015 void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
1016 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1017 SHADER_BUFFER* buffer = arg->buffer;
1018 glsl_src_param_t src_param;
1019 const char *instruction;
1020 char arguments[256];
1021 DWORD write_mask;
1022 unsigned i;
1023
1024 /* Determine the GLSL function to use based on the opcode */
1025 /* TODO: Possibly make this a table for faster lookups */
1026 switch (curOpcode->opcode) {
1027 case WINED3DSIO_MIN: instruction = "min"; break;
1028 case WINED3DSIO_MAX: instruction = "max"; break;
1029 case WINED3DSIO_RSQ: instruction = "inversesqrt"; break;
1030 case WINED3DSIO_ABS: instruction = "abs"; break;
1031 case WINED3DSIO_FRC: instruction = "fract"; break;
1032 case WINED3DSIO_NRM: instruction = "normalize"; break;
1033 case WINED3DSIO_LOGP:
1034 case WINED3DSIO_LOG: instruction = "log2"; break;
1035 case WINED3DSIO_EXP: instruction = "exp2"; break;
1036 case WINED3DSIO_SGN: instruction = "sign"; break;
1037 default: instruction = "";
1038 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1039 break;
1040 }
1041
1042 write_mask = shader_glsl_append_dst(buffer, arg);
1043
1044 arguments[0] = '\0';
1045 if (curOpcode->num_params > 0) {
1046 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
1047 strcat(arguments, src_param.param_str);
1048 for (i = 2; i < curOpcode->num_params; ++i) {
1049 strcat(arguments, ", ");
1050 shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
1051 strcat(arguments, src_param.param_str);
1052 }
1053 }
1054
1055 shader_addline(buffer, "%s(%s));\n", instruction, arguments);
1056 }
1057
1058 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1059 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1060 * dst.x = 2^(floor(src))
1061 * dst.y = src - floor(src)
1062 * dst.z = 2^src (partial precision is allowed, but optional)
1063 * dst.w = 1.0;
1064 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1065 * dst = 2^src; (partial precision is allowed, but optional)
1066 */
1067 void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
1068 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1069 glsl_src_param_t src_param;
1070
1071 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1072
1073 if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1074 char dst_mask[6];
1075
1076 shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
1077 shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
1078 shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
1079 shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
1080
1081 shader_glsl_append_dst(arg->buffer, arg);
1082 shader_glsl_get_write_mask(arg->dst, dst_mask);
1083 shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
1084 } else {
1085 DWORD write_mask;
1086 size_t mask_size;
1087
1088 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1089 mask_size = shader_glsl_get_write_mask_size(write_mask);
1090
1091 if (mask_size > 1) {
1092 shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
1093 } else {
1094 shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
1095 }
1096 }
1097 }
1098
1099 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1100 void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
1101 glsl_src_param_t src_param;
1102 DWORD write_mask;
1103 size_t mask_size;
1104
1105 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1106 mask_size = shader_glsl_get_write_mask_size(write_mask);
1107 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1108
1109 if (mask_size > 1) {
1110 shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
1111 } else {
1112 shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
1113 }
1114 }
1115
1116 /** Process signed comparison opcodes in GLSL. */
1117 void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
1118 glsl_src_param_t src0_param;
1119 glsl_src_param_t src1_param;
1120 DWORD write_mask;
1121 size_t mask_size;
1122
1123 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1124 mask_size = shader_glsl_get_write_mask_size(write_mask);
1125 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1126 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1127
1128 if (mask_size > 1) {
1129 const char *compare;
1130
1131 switch(arg->opcode->opcode) {
1132 case WINED3DSIO_SLT: compare = "lessThan"; break;
1133 case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
1134 default: compare = "";
1135 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1136 }
1137
1138 shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
1139 src0_param.param_str, src1_param.param_str);
1140 } else {
1141 const char *compare;
1142
1143 switch(arg->opcode->opcode) {
1144 case WINED3DSIO_SLT: compare = "<"; break;
1145 case WINED3DSIO_SGE: compare = ">="; break;
1146 default: compare = "";
1147 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1148 }
1149
1150 shader_addline(arg->buffer, "(%s %s %s) ? 1.0 : 0.0);\n",
1151 src0_param.param_str, compare, src1_param.param_str);
1152 }
1153 }
1154
1155 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1156 void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
1157 glsl_src_param_t src0_param;
1158 glsl_src_param_t src1_param;
1159 glsl_src_param_t src2_param;
1160 DWORD write_mask, cmp_channel = 0;
1161 unsigned int i, j;
1162
1163 /* Cycle through all source0 channels */
1164 for (i=0; i<4; i++) {
1165 write_mask = 0;
1166 /* Find the destination channels which use the current source0 channel */
1167 for (j=0; j<4; j++) {
1168 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1169 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1170 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1171 }
1172 }
1173 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1174 if (!write_mask) continue;
1175
1176 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1177 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1178 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1179
1180 shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
1181 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1182 }
1183 }
1184
1185 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1186 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1187 * the compare is done per component of src0. */
1188 void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
1189 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1190 glsl_src_param_t src0_param;
1191 glsl_src_param_t src1_param;
1192 glsl_src_param_t src2_param;
1193 DWORD write_mask, cmp_channel = 0;
1194 unsigned int i, j;
1195
1196 if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
1197 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1198 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1199 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1200 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1201 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1202 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1203 return;
1204 }
1205 /* Cycle through all source0 channels */
1206 for (i=0; i<4; i++) {
1207 write_mask = 0;
1208 /* Find the destination channels which use the current source0 channel */
1209 for (j=0; j<4; j++) {
1210 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1211 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1212 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1213 }
1214 }
1215 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1216 if (!write_mask) continue;
1217
1218 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1219 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1220 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1221
1222 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1223 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1224 }
1225 }
1226
1227 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1228 void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
1229 glsl_src_param_t src0_param;
1230 glsl_src_param_t src1_param;
1231 glsl_src_param_t src2_param;
1232 DWORD write_mask;
1233
1234 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1235 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1236 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1237 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1238 shader_addline(arg->buffer, "(%s * %s) + %s);\n",
1239 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1240 }
1241
1242 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1243 Vertex shaders to GLSL codes */
1244 void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
1245 int i;
1246 int nComponents = 0;
1247 SHADER_OPCODE_ARG tmpArg;
1248
1249 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1250
1251 /* Set constants for the temporary argument */
1252 tmpArg.shader = arg->shader;
1253 tmpArg.buffer = arg->buffer;
1254 tmpArg.src[0] = arg->src[0];
1255 tmpArg.src_addr[0] = arg->src_addr[0];
1256 tmpArg.src_addr[1] = arg->src_addr[1];
1257 tmpArg.reg_maps = arg->reg_maps;
1258
1259 switch(arg->opcode->opcode) {
1260 case WINED3DSIO_M4x4:
1261 nComponents = 4;
1262 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1263 break;
1264 case WINED3DSIO_M4x3:
1265 nComponents = 3;
1266 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1267 break;
1268 case WINED3DSIO_M3x4:
1269 nComponents = 4;
1270 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1271 break;
1272 case WINED3DSIO_M3x3:
1273 nComponents = 3;
1274 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1275 break;
1276 case WINED3DSIO_M3x2:
1277 nComponents = 2;
1278 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1279 break;
1280 default:
1281 break;
1282 }
1283
1284 for (i = 0; i < nComponents; i++) {
1285 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1286 tmpArg.src[1] = arg->src[1]+i;
1287 shader_glsl_dot(&tmpArg);
1288 }
1289 }
1290
1291 /**
1292 The LRP instruction performs a component-wise linear interpolation
1293 between the second and third operands using the first operand as the
1294 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1295 This is equivalent to mix(src2, src1, src0);
1296 */
1297 void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
1298 glsl_src_param_t src0_param;
1299 glsl_src_param_t src1_param;
1300 glsl_src_param_t src2_param;
1301 DWORD write_mask;
1302
1303 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1304
1305 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1306 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1307 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1308
1309 shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
1310 src2_param.param_str, src1_param.param_str, src0_param.param_str);
1311 }
1312
1313 /** Process the WINED3DSIO_LIT instruction in GLSL:
1314 * dst.x = dst.w = 1.0
1315 * dst.y = (src0.x > 0) ? src0.x
1316 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1317 * where src.w is clamped at +- 128
1318 */
1319 void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
1320 glsl_src_param_t src0_param;
1321 glsl_src_param_t src1_param;
1322 glsl_src_param_t src3_param;
1323 char dst_mask[6];
1324
1325 shader_glsl_append_dst(arg->buffer, arg);
1326 shader_glsl_get_write_mask(arg->dst, dst_mask);
1327
1328 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1329 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
1330 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
1331
1332 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",
1333 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);
1334 }
1335
1336 /** Process the WINED3DSIO_DST instruction in GLSL:
1337 * dst.x = 1.0
1338 * dst.y = src0.x * src0.y
1339 * dst.z = src0.z
1340 * dst.w = src1.w
1341 */
1342 void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
1343 glsl_src_param_t src0y_param;
1344 glsl_src_param_t src0z_param;
1345 glsl_src_param_t src1y_param;
1346 glsl_src_param_t src1w_param;
1347 char dst_mask[6];
1348
1349 shader_glsl_append_dst(arg->buffer, arg);
1350 shader_glsl_get_write_mask(arg->dst, dst_mask);
1351
1352 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
1353 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
1354 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
1355 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
1356
1357 shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1358 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
1359 }
1360
1361 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1362 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1363 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1364 *
1365 * dst.x = cos(src0.?)
1366 * dst.y = sin(src0.?)
1367 * dst.z = dst.z
1368 * dst.w = dst.w
1369 */
1370 void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
1371 glsl_src_param_t src0_param;
1372 DWORD write_mask;
1373
1374 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1375 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1376
1377 switch (write_mask) {
1378 case WINED3DSP_WRITEMASK_0:
1379 shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
1380 break;
1381
1382 case WINED3DSP_WRITEMASK_1:
1383 shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
1384 break;
1385
1386 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
1387 shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
1388 break;
1389
1390 default:
1391 ERR("Write mask should be .x, .y or .xy\n");
1392 break;
1393 }
1394 }
1395
1396 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1397 * Start a for() loop where src1.y is the initial value of aL,
1398 * increment aL by src1.z for a total of src1.x iterations.
1399 * Need to use a temporary variable for this operation.
1400 */
1401 /* FIXME: I don't think nested loops will work correctly this way. */
1402 void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
1403 glsl_src_param_t src1_param;
1404
1405 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
1406
1407 shader_addline(arg->buffer, "for (tmpInt = 0, aL = %s.y; tmpInt < %s.x; tmpInt++, aL += %s.z) {\n",
1408 src1_param.reg_name, src1_param.reg_name, src1_param.reg_name);
1409 }
1410
1411 void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
1412 shader_addline(arg->buffer, "}\n");
1413 }
1414
1415 void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
1416 glsl_src_param_t src0_param;
1417
1418 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1419 shader_addline(arg->buffer, "for (tmpInt = 0; tmpInt < %s; tmpInt++) {\n", src0_param.param_str);
1420 }
1421
1422 void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
1423 glsl_src_param_t src0_param;
1424
1425 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1426 shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
1427 }
1428
1429 void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
1430 glsl_src_param_t src0_param;
1431 glsl_src_param_t src1_param;
1432
1433 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1434 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1435
1436 shader_addline(arg->buffer, "if (%s %s %s) {\n",
1437 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1438 }
1439
1440 void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
1441 shader_addline(arg->buffer, "} else {\n");
1442 }
1443
1444 void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
1445 shader_addline(arg->buffer, "break;\n");
1446 }
1447
1448 /* FIXME: According to MSDN the compare is done per component. */
1449 void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
1450 glsl_src_param_t src0_param;
1451 glsl_src_param_t src1_param;
1452
1453 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1454 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1455
1456 shader_addline(arg->buffer, "if (%s %s %s) break;\n",
1457 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1458 }
1459
1460 void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
1461
1462 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1463 shader_addline(arg->buffer, "}\n");
1464 shader_addline(arg->buffer, "void subroutine%lu () {\n", snum);
1465 }
1466
1467 void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
1468 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1469 shader_addline(arg->buffer, "subroutine%lu();\n", snum);
1470 }
1471
1472 void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
1473 glsl_src_param_t src1_param;
1474
1475 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1476 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1477 shader_addline(arg->buffer, "if (%s) subroutine%lu();\n", src1_param.param_str, snum);
1478 }
1479
1480 /*********************************************
1481 * Pixel Shader Specific Code begins here
1482 ********************************************/
1483 void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
1484 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1485 DWORD hex_version = This->baseShader.hex_version;
1486 char dst_swizzle[6];
1487 glsl_sample_function_t sample_function;
1488 DWORD sampler_type;
1489 DWORD sampler_idx;
1490 BOOL projected;
1491 DWORD mask = 0;
1492
1493 /* All versions have a destination register */
1494 shader_glsl_append_dst(arg->buffer, arg);
1495
1496 /* 1.0-1.4: Use destination register as sampler source.
1497 * 2.0+: Use provided sampler source. */
1498 if (hex_version < WINED3DPS_VERSION(1,4)) {
1499 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1500 DWORD flags;
1501
1502 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1503 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1504
1505 if (flags & WINED3DTTFF_PROJECTED) {
1506 projected = TRUE;
1507 switch (flags & ~WINED3DTTFF_PROJECTED) {
1508 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
1509 case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
1510 case WINED3DTTFF_COUNT3: mask = WINED3DSP_WRITEMASK_2; break;
1511 case WINED3DTTFF_COUNT4:
1512 case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
1513 }
1514 } else {
1515 projected = FALSE;
1516 }
1517 } else if (hex_version < WINED3DPS_VERSION(2,0)) {
1518 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1519 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1520
1521 if (src_mod == WINED3DSPSM_DZ) {
1522 projected = TRUE;
1523 mask = WINED3DSP_WRITEMASK_2;
1524 } else if (src_mod == WINED3DSPSM_DW) {
1525 projected = TRUE;
1526 mask = WINED3DSP_WRITEMASK_3;
1527 } else {
1528 projected = FALSE;
1529 }
1530 } else {
1531 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1532 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1533 /* ps 2.0 texldp instruction always divides by the fourth component. */
1534 projected = TRUE;
1535 mask = WINED3DSP_WRITEMASK_3;
1536 } else {
1537 projected = FALSE;
1538 }
1539 }
1540
1541 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1542 shader_glsl_get_sample_function(sampler_type, projected, &sample_function);
1543 mask |= sample_function.coord_mask;
1544
1545 if (hex_version < WINED3DPS_VERSION(2,0)) {
1546 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
1547 } else {
1548 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1549 }
1550
1551 /* 1.0-1.3: Use destination register as coordinate source.
1552 1.4+: Use provided coordinate source register. */
1553 if (hex_version < WINED3DPS_VERSION(1,4)) {
1554 char coord_mask[6];
1555 shader_glsl_get_write_mask(mask, coord_mask);
1556 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
1557 sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
1558 } else {
1559 glsl_src_param_t coord_param;
1560 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
1561 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
1562 sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
1563 }
1564 }
1565
1566 void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
1567
1568 /* FIXME: Make this work for more than just 2D textures */
1569
1570 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1571 SHADER_BUFFER* buffer = arg->buffer;
1572 DWORD hex_version = This->baseShader.hex_version;
1573 DWORD write_mask;
1574 char dst_mask[6];
1575
1576 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1577 shader_glsl_get_write_mask(write_mask, dst_mask);
1578
1579 if (hex_version != WINED3DPS_VERSION(1,4)) {
1580 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1581 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
1582 } else {
1583 DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
1584 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1585 char dst_swizzle[6];
1586
1587 shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
1588
1589 if (src_mod == WINED3DSPSM_DZ) {
1590 glsl_src_param_t div_param;
1591 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1592 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
1593
1594 if (mask_size > 1) {
1595 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1596 } else {
1597 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1598 }
1599 } else if (src_mod == WINED3DSPSM_DW) {
1600 glsl_src_param_t div_param;
1601 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1602 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
1603
1604 if (mask_size > 1) {
1605 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1606 } else {
1607 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1608 }
1609 } else {
1610 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
1611 }
1612 }
1613 }
1614
1615 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
1616 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1617 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1618 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
1619 glsl_src_param_t src0_param;
1620 char dst_mask[6];
1621 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1622 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1623
1624 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1625
1626 shader_glsl_append_dst(arg->buffer, arg);
1627 shader_glsl_get_write_mask(arg->dst, dst_mask);
1628 shader_addline(arg->buffer, "texture1D(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
1629 sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
1630 }
1631
1632 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
1633 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1634 void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
1635 glsl_src_param_t src0_param;
1636 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1637 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1638 DWORD dst_mask;
1639 size_t mask_size;
1640
1641 dst_mask = shader_glsl_append_dst(arg->buffer, arg);
1642 mask_size = shader_glsl_get_write_mask_size(dst_mask);
1643 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1644
1645 if (mask_size > 1) {
1646 shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
1647 } else {
1648 shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
1649 }
1650 }
1651
1652 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
1653 * Calculate the depth as dst.x / dst.y */
1654 void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
1655 glsl_dst_param_t dst_param;
1656
1657 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
1658
1659 shader_addline(arg->buffer, "gl_FragDepth = %s.x / %s.y;\n", dst_param.reg_name, dst_param.reg_name);
1660 }
1661
1662 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
1663 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1664 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1665 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1666 */
1667 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1668 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1669 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1670 glsl_src_param_t src0_param;
1671
1672 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1673
1674 shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
1675 shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
1676 }
1677
1678 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
1679 * Calculate the 1st of a 2-row matrix multiplication. */
1680 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1681 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1682 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1683 SHADER_BUFFER* buffer = arg->buffer;
1684 glsl_src_param_t src0_param;
1685
1686 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1687 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1688 }
1689
1690 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
1691 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
1692 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1693
1694 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1695 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1696 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1697 SHADER_BUFFER* buffer = arg->buffer;
1698 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1699 glsl_src_param_t src0_param;
1700
1701 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1702 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
1703 current_state->texcoord_w[current_state->current_row++] = reg;
1704 }
1705
1706 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1707 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1708 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1709 SHADER_BUFFER* buffer = arg->buffer;
1710 glsl_src_param_t src0_param;
1711 char dst_mask[6];
1712
1713 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1714 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1715
1716 shader_glsl_append_dst(buffer, arg);
1717 shader_glsl_get_write_mask(arg->dst, dst_mask);
1718
1719 /* Sample the texture using the calculated coordinates */
1720 shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
1721 }
1722
1723 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
1724 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
1725 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1726 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1727 glsl_src_param_t src0_param;
1728 char dst_mask[6];
1729 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1730 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1731 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1732 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1733 glsl_sample_function_t sample_function;
1734
1735 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1736 shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1737
1738 shader_glsl_append_dst(arg->buffer, arg);
1739 shader_glsl_get_write_mask(arg->dst, dst_mask);
1740 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1741
1742 /* Sample the texture using the calculated coordinates */
1743 shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
1744
1745 current_state->current_row = 0;
1746 }
1747
1748 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
1749 * Perform the 3rd row of a 3x3 matrix multiply */
1750 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
1751 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1752 glsl_src_param_t src0_param;
1753 char dst_mask[6];
1754 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1755 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1756 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1757
1758 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1759
1760 shader_glsl_append_dst(arg->buffer, arg);
1761 shader_glsl_get_write_mask(arg->dst, dst_mask);
1762 shader_addline(arg->buffer, "vec4(tmp.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
1763
1764 current_state->current_row = 0;
1765 }
1766
1767 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
1768 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
1769 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1770
1771 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1772 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1773 glsl_src_param_t src0_param;
1774 glsl_src_param_t src1_param;
1775 char dst_mask[6];
1776 SHADER_BUFFER* buffer = arg->buffer;
1777 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1778 DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1779 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1780 glsl_sample_function_t sample_function;
1781
1782 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1783 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1784
1785 /* Perform the last matrix multiply operation */
1786 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
1787
1788 /* Calculate reflection vector, 2*(tmp0.src1)*tmp0-src1
1789 * This is equivalent to reflect(-src1, tmp0); */
1790 shader_addline(buffer, "tmp0.xyz = reflect(-(%s), tmp0.xyz);\n", src1_param.param_str);
1791
1792 shader_glsl_append_dst(buffer, arg);
1793 shader_glsl_get_write_mask(arg->dst, dst_mask);
1794 shader_glsl_get_sample_function(stype, FALSE, &sample_function);
1795
1796 /* Sample the texture */
1797 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
1798
1799 current_state->current_row = 0;
1800 }
1801
1802 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
1803 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
1804 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1805
1806 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
1807 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1808 SHADER_BUFFER* buffer = arg->buffer;
1809 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
1810 glsl_src_param_t src0_param;
1811 char dst_mask[6];
1812 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1813 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
1814 glsl_sample_function_t sample_function;
1815
1816 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1817
1818 /* Perform the last matrix multiply operation */
1819 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
1820
1821 /* Construct the eye-ray vector from w coordinates */
1822 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
1823 current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
1824
1825 /* Calculate reflection vector (Assume normal is normalized): RF = 2*(tmp0.tmp1)*tmp0-tmp1
1826 * This is equivalent to reflect(-tmp1, tmp0); */
1827 shader_addline(buffer, "tmp0.xyz = reflect(-tmp1.xyz, tmp0.xyz);\n");
1828
1829 shader_glsl_append_dst(buffer, arg);
1830 shader_glsl_get_write_mask(arg->dst, dst_mask);
1831 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1832
1833 /* Sample the texture using the calculated coordinates */
1834 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
1835
1836 current_state->current_row = 0;
1837 }
1838
1839 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
1840 * Apply a fake bump map transform.
1841 * texbem is pshader <= 1.3 only, this saves a few version checks
1842 */
1843 void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
1844 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1845 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1846 char dst_swizzle[6];
1847 glsl_sample_function_t sample_function;
1848 glsl_src_param_t coord_param;
1849 DWORD sampler_type;
1850 DWORD sampler_idx;
1851 DWORD mask;
1852 DWORD flags;
1853 char coord_mask[6];
1854
1855 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1856 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1857
1858 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1859 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1860 mask = sample_function.coord_mask;
1861
1862 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
1863
1864 shader_glsl_get_write_mask(mask, coord_mask);
1865
1866 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1867 * so we can't let the GL handle this.
1868 */
1869 if (flags & WINED3DTTFF_PROJECTED) {
1870 DWORD div_mask=0;
1871 char coord_div_mask[3];
1872 switch (flags & ~WINED3DTTFF_PROJECTED) {
1873 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
1874 case WINED3DTTFF_COUNT2: div_mask = WINED3DSP_WRITEMASK_1; break;
1875 case WINED3DTTFF_COUNT3: div_mask = WINED3DSP_WRITEMASK_2; break;
1876 case WINED3DTTFF_COUNT4:
1877 case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
1878 }
1879 shader_glsl_get_write_mask(div_mask, coord_div_mask);
1880 shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
1881 }
1882
1883 shader_glsl_append_dst(arg->buffer, arg);
1884 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
1885 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )%s);\n",
1886 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask, dst_swizzle);
1887 }
1888
1889 void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
1890 glsl_src_param_t src0_param, src1_param;
1891
1892 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
1893 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
1894
1895 shader_glsl_append_dst(arg->buffer, arg);
1896 shader_addline(arg->buffer, "%s + bumpenvmat * %s);\n",
1897 src0_param.param_str, src1_param.param_str);
1898 }
1899
1900 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
1901 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
1902 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
1903 glsl_src_param_t src0_param;
1904 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1905 char dst_mask[6];
1906
1907 shader_glsl_append_dst(arg->buffer, arg);
1908 shader_glsl_get_write_mask(arg->dst, dst_mask);
1909 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
1910
1911 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
1912 }
1913
1914 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
1915 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
1916 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
1917 glsl_src_param_t src0_param;
1918 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1919 char dst_mask[6];
1920
1921 shader_glsl_append_dst(arg->buffer, arg);
1922 shader_glsl_get_write_mask(arg->dst, dst_mask);
1923 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
1924
1925 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
1926 }
1927
1928 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
1929 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
1930 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1931 glsl_src_param_t src0_param;
1932 char dst_mask[6];
1933 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1934 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1935 glsl_sample_function_t sample_function;
1936
1937 shader_glsl_append_dst(arg->buffer, arg);
1938 shader_glsl_get_write_mask(arg->dst, dst_mask);
1939 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1940 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
1941
1942 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
1943 }
1944
1945 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
1946 * If any of the first 3 components are < 0, discard this pixel */
1947 void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
1948 glsl_dst_param_t dst_param;
1949
1950 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
1951 shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
1952 }
1953
1954 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
1955 * dst = dot2(src0, src1) + src2 */
1956 void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
1957 glsl_src_param_t src0_param;
1958 glsl_src_param_t src1_param;
1959 glsl_src_param_t src2_param;
1960 DWORD write_mask;
1961 size_t mask_size;
1962
1963 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1964 mask_size = shader_glsl_get_write_mask_size(write_mask);
1965
1966 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
1967 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
1968 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
1969
1970 shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
1971 }
1972
1973 void pshader_glsl_input_pack(
1974 SHADER_BUFFER* buffer,
1975 semantic* semantics_in) {
1976
1977 unsigned int i;
1978
1979 for (i = 0; i < MAX_REG_INPUT; i++) {
1980
1981 DWORD usage_token = semantics_in[i].usage;
1982 DWORD register_token = semantics_in[i].reg;
1983 DWORD usage, usage_idx;
1984 char reg_mask[6];
1985
1986 /* Uninitialized */
1987 if (!usage_token) continue;
1988 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
1989 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
1990 shader_glsl_get_write_mask(register_token, reg_mask);
1991
1992 switch(usage) {
1993
1994 case WINED3DDECLUSAGE_COLOR:
1995 if (usage_idx == 0)
1996 shader_addline(buffer, "IN%u%s = vec4(gl_Color)%s;\n",
1997 i, reg_mask, reg_mask);
1998 else if (usage_idx == 1)
1999 shader_addline(buffer, "IN%u%s = vec4(gl_SecondaryColor)%s;\n",
2000 i, reg_mask, reg_mask);
2001 else
2002 shader_addline(buffer, "IN%u%s = vec4(unsupported_color_input)%s;\n",
2003 i, reg_mask, reg_mask);
2004 break;
2005
2006 case WINED3DDECLUSAGE_TEXCOORD:
2007 shader_addline(buffer, "IN%u%s = vec4(gl_TexCoord[%u])%s;\n",
2008 i, reg_mask, usage_idx, reg_mask );
2009 break;
2010
2011 case WINED3DDECLUSAGE_FOG:
2012 shader_addline(buffer, "IN%u%s = vec4(gl_FogFragCoord)%s;\n",
2013 i, reg_mask, reg_mask);
2014 break;
2015
2016 default:
2017 shader_addline(buffer, "IN%u%s = vec4(unsupported_input)%s;\n",
2018 i, reg_mask, reg_mask);
2019 }
2020 }
2021 }
2022
2023 /*********************************************
2024 * Vertex Shader Specific Code begins here
2025 ********************************************/
2026
2027 void vshader_glsl_output_unpack(
2028 SHADER_BUFFER* buffer,
2029 semantic* semantics_out) {
2030
2031 unsigned int i;
2032
2033 for (i = 0; i < MAX_REG_OUTPUT; i++) {
2034
2035 DWORD usage_token = semantics_out[i].usage;
2036 DWORD register_token = semantics_out[i].reg;
2037 DWORD usage, usage_idx;
2038 char reg_mask[6];
2039
2040 /* Uninitialized */
2041 if (!usage_token) continue;
2042
2043 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2044 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2045 shader_glsl_get_write_mask(register_token, reg_mask);
2046
2047 switch(usage) {
2048
2049 case WINED3DDECLUSAGE_COLOR:
2050 if (usage_idx == 0)
2051 shader_addline(buffer, "gl_FrontColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2052 else if (usage_idx == 1)
2053 shader_addline(buffer, "gl_FrontSecondaryColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2054 else
2055 shader_addline(buffer, "unsupported_color_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2056 break;
2057
2058 case WINED3DDECLUSAGE_POSITION:
2059 shader_addline(buffer, "gl_Position%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2060 break;
2061
2062 case WINED3DDECLUSAGE_TEXCOORD:
2063 shader_addline(buffer, "gl_TexCoord[%u]%s = OUT%u%s;\n",
2064 usage_idx, reg_mask, i, reg_mask);
2065 break;
2066
2067 case WINED3DDECLUSAGE_PSIZE:
2068 shader_addline(buffer, "gl_PointSize = OUT%u.x;\n", i);
2069 break;
2070
2071 case WINED3DDECLUSAGE_FOG:
2072 shader_addline(buffer, "gl_FogFragCoord = OUT%u%s;\n", i, reg_mask);
2073 break;
2074
2075 default:
2076 shader_addline(buffer, "unsupported_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2077 }
2078 }
2079 }
2080
2081 static void add_glsl_program_entry(IWineD3DDeviceImpl *device, struct glsl_shader_prog_link *entry) {
2082 glsl_program_key_t *key;
2083
2084 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2085 key->vshader = entry->vshader;
2086 key->pshader = entry->pshader;
2087
2088 hash_table_put(device->glsl_program_lookup, key, entry);
2089 }
2090
2091 static struct glsl_shader_prog_link *get_glsl_program_entry(IWineD3DDeviceImpl *device,
2092 GLhandleARB vshader, GLhandleARB pshader) {
2093 glsl_program_key_t key;
2094
2095 key.vshader = vshader;
2096 key.pshader = pshader;
2097
2098 return (struct glsl_shader_prog_link *)hash_table_get(device->glsl_program_lookup, &key);
2099 }
2100
2101 void delete_glsl_program_entry(IWineD3DDevice *iface, struct glsl_shader_prog_link *entry) {
2102 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2103 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2104 glsl_program_key_t *key;
2105
2106 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2107 key->vshader = entry->vshader;
2108 key->pshader = entry->pshader;
2109 hash_table_remove(This->glsl_program_lookup, key);
2110
2111 GL_EXTCALL(glDeleteObjectARB(entry->programId));
2112 if (entry->vshader) list_remove(&entry->vshader_entry);
2113 if (entry->pshader) list_remove(&entry->pshader_entry);
2114 HeapFree(GetProcessHeap(), 0, entry->vuniformF_locations);
2115 HeapFree(GetProcessHeap(), 0, entry->puniformF_locations);
2116 HeapFree(GetProcessHeap(), 0, entry);
2117 }
2118
2119 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
2120 * It sets the programId on the current StateBlock (because it should be called
2121 * inside of the DrawPrimitive() part of the render loop).
2122 *
2123 * If a program for the given combination does not exist, create one, and store
2124 * the program in the hash table. If it creates a program, it will link the
2125 * given objects, too.
2126 */
2127 static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
2128 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2129 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2130 IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
2131 IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
2132 struct glsl_shader_prog_link *entry = NULL;
2133 GLhandleARB programId = 0;
2134 int i;
2135 char glsl_name[8];
2136
2137 GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
2138 GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
2139 entry = get_glsl_program_entry(This, vshader_id, pshader_id);
2140 if (entry) {
2141 This->stateBlock->glsl_program = entry;
2142 return;
2143 }
2144
2145 /* If we get to this point, then no matching program exists, so we create one */
2146 programId = GL_EXTCALL(glCreateProgramObjectARB());
2147 TRACE("Created new GLSL shader program %u\n", programId);
2148
2149 /* Create the entry */
2150 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
2151 entry->programId = programId;
2152 entry->vshader = vshader_id;
2153 entry->pshader = pshader_id;
2154 /* Add the hash table entry */
2155 add_glsl_program_entry(This, entry);
2156
2157 /* Set the current program */
2158 This->stateBlock->glsl_program = entry;
2159
2160 /* Attach GLSL vshader */
2161 if (vshader_id) {
2162 int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
2163 char tmp_name[10];
2164
2165 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id, programId);
2166 GL_EXTCALL(glAttachObjectARB(programId, vshader_id));
2167 checkGLcall("glAttachObjectARB");
2168
2169 /* Bind vertex attributes to a corresponding index number to match
2170 * the same index numbers as ARB_vertex_programs (makes loading
2171 * vertex attributes simpler). With this method, we can use the
2172 * exact same code to load the attributes later for both ARB and
2173 * GLSL shaders.
2174 *
2175 * We have to do this here because we need to know the Program ID
2176 * in order to make the bindings work, and it has to be done prior
2177 * to linking the GLSL program. */
2178 for (i = 0; i < max_attribs; ++i) {
2179 snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
2180 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
2181 }
2182 checkGLcall("glBindAttribLocationARB");
2183
2184 list_add_head(&((IWineD3DBaseShaderImpl *)vshader)->baseShader.linked_programs, &entry->vshader_entry);
2185 }
2186
2187 /* Attach GLSL pshader */
2188 if (pshader_id) {
2189 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
2190 GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
2191 checkGLcall("glAttachObjectARB");
2192
2193 list_add_head(&((IWineD3DBaseShaderImpl *)pshader)->baseShader.linked_programs, &entry->pshader_entry);
2194 }
2195
2196 /* Link the program */
2197 TRACE("Linking GLSL shader program %u\n", programId);
2198 GL_EXTCALL(glLinkProgramARB(programId));
2199 print_glsl_info_log(&GLINFO_LOCATION, programId);
2200
2201 entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
2202 for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
2203 snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
2204 entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2205 }
2206 entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
2207 for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
2208 snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
2209 entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2210 }
2211 }
2212
2213 static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
2214 GLhandleARB program_id;
2215 GLhandleARB vshader_id, pshader_id;
2216 const char *blt_vshader[] = {
2217 "void main(void)\n"
2218 "{\n"
2219 " gl_Position = gl_Vertex;\n"
2220 " gl_FrontColor = vec4(1.0);\n"
2221 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
2222 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
2223 "}\n"
2224 };
2225
2226 const char *blt_pshader[] = {
2227 "uniform sampler2D sampler;\n"
2228 "void main(void)\n"
2229 "{\n"
2230 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
2231 "}\n"
2232 };
2233
2234 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
2235 GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
2236 GL_EXTCALL(glCompileShaderARB(vshader_id));
2237
2238 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
2239 GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
2240 GL_EXTCALL(glCompileShaderARB(pshader_id));
2241
2242 program_id = GL_EXTCALL(glCreateProgramObjectARB());
2243 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
2244 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
2245 GL_EXTCALL(glLinkProgramARB(program_id));
2246
2247 print_glsl_info_log(&GLINFO_LOCATION, program_id);
2248
2249 return program_id;
2250 }
2251
2252 static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
2253 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2254 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2255 GLhandleARB program_id = 0;
2256
2257 if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
2258 else This->stateBlock->glsl_program = NULL;
2259
2260 program_id = This->stateBlock->glsl_program ? This->stateBlock->glsl_program->programId : 0;
2261 if (program_id) TRACE("Using GLSL program %u\n", program_id);
2262 GL_EXTCALL(glUseProgramObjectARB(program_id));
2263 checkGLcall("glUseProgramObjectARB");
2264 }
2265
2266 static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
2267 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2268 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2269 static GLhandleARB program_id = 0;
2270 static GLhandleARB loc = -1;
2271
2272 if (!program_id) {
2273 program_id = create_glsl_blt_shader(gl_info);
2274 loc = GL_EXTCALL(glGetUniformLocationARB(program_id, "sampler"));
2275 }
2276
2277 GL_EXTCALL(glUseProgramObjectARB(program_id));
2278 GL_EXTCALL(glUniform1iARB(loc, 0));
2279 }
2280
2281 static void shader_glsl_cleanup(IWineD3DDevice *iface) {
2282 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2283 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2284 GL_EXTCALL(glUseProgramObjectARB(0));
2285 }
2286
2287 const shader_backend_t glsl_shader_backend = {
2288 &shader_glsl_select,
2289 &shader_glsl_select_depth_blt,
2290 &shader_glsl_load_constants,
2291 &shader_glsl_cleanup
2292 };
GSOC 2012 project to enhance the control panel