?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
kernel32: Pass completion value to ntdll layer.
[wine/wine64.git] / dlls / wined3d / glsl_shader.c
blobb0ca134549a55a9a72db4e39d83344a4c5ea573e
1 /*
2 * GLSL pixel and vertex shader implementation
3 *
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
20 */
21
22 /*
23 * D3D shader asm has swizzles on source parameters, and write masks for
24 * destination parameters. GLSL uses swizzles for both. The result of this is
25 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
26 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
27 * mask for the destination parameter into account.
28 */
29
30 #include "config.h"
31 #include <stdio.h>
32 #include "wined3d_private.h"
33
34 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
35 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
36
37 #define GLINFO_LOCATION (*gl_info)
38
39 typedef struct {
40 char reg_name[50];
41 char mask_str[6];
42 } glsl_dst_param_t;
43
44 typedef struct {
45 char reg_name[50];
46 char param_str[100];
47 } glsl_src_param_t;
48
49 typedef struct {
50 const char *name;
51 DWORD coord_mask;
52 } glsl_sample_function_t;
53
54 /** Prints the GLSL info log which will contain error messages if they exist */
55 void print_glsl_info_log(WineD3D_GL_Info *gl_info, GLhandleARB obj) {
56
57 int infologLength = 0;
58 char *infoLog;
59
60 GL_EXTCALL(glGetObjectParameterivARB(obj,
61 GL_OBJECT_INFO_LOG_LENGTH_ARB,
62 &infologLength));
63
64 /* A size of 1 is just a null-terminated string, so the log should be bigger than
65 * that if there are errors. */
66 if (infologLength > 1)
67 {
68 infoLog = HeapAlloc(GetProcessHeap(), 0, infologLength);
69 GL_EXTCALL(glGetInfoLogARB(obj, infologLength, NULL, infoLog));
70 FIXME("Error received from GLSL shader #%u: %s\n", obj, debugstr_a(infoLog));
71 HeapFree(GetProcessHeap(), 0, infoLog);
72 }
73 }
74
75 /**
76 * Loads (pixel shader) samplers
77 */
78 static void shader_glsl_load_psamplers(
79 WineD3D_GL_Info *gl_info,
80 IWineD3DStateBlock* iface) {
81
82 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
83 GLhandleARB programId = stateBlock->glsl_program->programId;
84 GLhandleARB name_loc;
85 int i;
86 char sampler_name[20];
87
88 for (i = 0; i < MAX_FRAGMENT_SAMPLERS; ++i) {
89 snprintf(sampler_name, sizeof(sampler_name), "Psampler%d", i);
90 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
91 if (name_loc != -1) {
92 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[i];
93 if (mapped_unit != -1 && mapped_unit < GL_LIMITS(fragment_samplers)) {
94 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
95 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
96 checkGLcall("glUniform1iARB");
97 } else {
98 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
99 }
100 }
101 }
102 }
103
104 static void shader_glsl_load_vsamplers(WineD3D_GL_Info *gl_info, IWineD3DStateBlock* iface) {
105 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
106 GLhandleARB programId = stateBlock->glsl_program->programId;
107 GLhandleARB name_loc;
108 char sampler_name[20];
109 int i;
110
111 for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i) {
112 snprintf(sampler_name, sizeof(sampler_name), "Vsampler%d", i);
113 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
114 if (name_loc != -1) {
115 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[MAX_FRAGMENT_SAMPLERS + i];
116 if (mapped_unit != -1 && mapped_unit < GL_LIMITS(combined_samplers)) {
117 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
118 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
119 checkGLcall("glUniform1iARB");
120 } else {
121 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
122 }
123 }
124 }
125 }
126
127 /**
128 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
129 * When constant_list == NULL, it will load all the constants.
130 */
131 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info,
132 unsigned int max_constants, float* constants, GLhandleARB *constant_locations,
133 struct list *constant_list) {
134 constants_entry *constant;
135 local_constant* lconst;
136 GLhandleARB tmp_loc;
137 DWORD i, j, k;
138 DWORD *idx;
139
140 if (TRACE_ON(d3d_shader)) {
141 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
142 idx = constant->idx;
143 j = constant->count;
144 while (j--) {
145 i = *idx++;
146 tmp_loc = constant_locations[i];
147 if (tmp_loc != -1) {
148 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
149 constants[i * 4 + 0], constants[i * 4 + 1],
150 constants[i * 4 + 2], constants[i * 4 + 3]);
151 }
152 }
153 }
154 }
155
156 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
157 if(WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 &&
158 shader_is_pshader_version(This->baseShader.hex_version)) {
159 float lcl_const[4];
160
161 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
162 idx = constant->idx;
163 j = constant->count;
164 while (j--) {
165 i = *idx++;
166 tmp_loc = constant_locations[i];
167 if (tmp_loc != -1) {
168 /* We found this uniform name in the program - go ahead and send the data */
169 k = i * 4;
170 if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
171 else if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
172 else lcl_const[0] = constants[k + 0];
173 if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
174 else if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
175 else lcl_const[1] = constants[k + 1];
176 if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
177 else if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
178 else lcl_const[2] = constants[k + 2];
179 if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
180 else if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
181 else lcl_const[3] = constants[k + 3];
182
183 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, lcl_const));
184 }
185 }
186 }
187 } else {
188 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
189 idx = constant->idx;
190 j = constant->count;
191 while (j--) {
192 i = *idx++;
193 tmp_loc = constant_locations[i];
194 if (tmp_loc != -1) {
195 /* We found this uniform name in the program - go ahead and send the data */
196 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, constants + (i * 4)));
197 }
198 }
199 }
200 }
201 checkGLcall("glUniform4fvARB()");
202
203 /* Load immediate constants */
204 if (TRACE_ON(d3d_shader)) {
205 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
206 tmp_loc = constant_locations[lconst->idx];
207 if (tmp_loc != -1) {
208 GLfloat* values = (GLfloat*)lconst->value;
209 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
210 values[0], values[1], values[2], values[3]);
211 }
212 }
213 }
214 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
215 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
216 tmp_loc = constant_locations[lconst->idx];
217 if (tmp_loc != -1) {
218 /* We found this uniform name in the program - go ahead and send the data */
219 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, (GLfloat*)lconst->value));
220 }
221 }
222 checkGLcall("glUniform4fvARB()");
223 }
224
225 /**
226 * Loads integer constants (aka uniforms) into the currently set GLSL program.
227 * When @constants_set == NULL, it will load all the constants.
228 */
229 static void shader_glsl_load_constantsI(
230 IWineD3DBaseShaderImpl* This,
231 WineD3D_GL_Info *gl_info,
232 GLhandleARB programId,
233 unsigned max_constants,
234 int* constants,
235 BOOL* constants_set) {
236
237 GLhandleARB tmp_loc;
238 int i;
239 char tmp_name[8];
240 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
241 const char* prefix = is_pshader? "PI":"VI";
242 struct list* ptr;
243
244 for (i=0; i<max_constants; ++i) {
245 if (NULL == constants_set || constants_set[i]) {
246
247 TRACE_(d3d_constants)("Loading constants %i: %i, %i, %i, %i\n",
248 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
249
250 /* TODO: Benchmark and see if it would be beneficial to store the
251 * locations of the constants to avoid looking up each time */
252 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
253 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
254 if (tmp_loc != -1) {
255 /* We found this uniform name in the program - go ahead and send the data */
256 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, &constants[i*4]));
257 checkGLcall("glUniform4ivARB");
258 }
259 }
260 }
261
262 /* Load immediate constants */
263 ptr = list_head(&This->baseShader.constantsI);
264 while (ptr) {
265 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
266 unsigned int idx = lconst->idx;
267 GLint* values = (GLint*) lconst->value;
268
269 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
270 values[0], values[1], values[2], values[3]);
271
272 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
273 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
274 if (tmp_loc != -1) {
275 /* We found this uniform name in the program - go ahead and send the data */
276 GL_EXTCALL(glUniform4ivARB(tmp_loc, 1, values));
277 checkGLcall("glUniform4ivARB");
278 }
279 ptr = list_next(&This->baseShader.constantsI, ptr);
280 }
281 }
282
283 /**
284 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
285 * When @constants_set == NULL, it will load all the constants.
286 */
287 static void shader_glsl_load_constantsB(
288 IWineD3DBaseShaderImpl* This,
289 WineD3D_GL_Info *gl_info,
290 GLhandleARB programId,
291 unsigned max_constants,
292 BOOL* constants,
293 BOOL* constants_set) {
294
295 GLhandleARB tmp_loc;
296 int i;
297 char tmp_name[8];
298 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
299 const char* prefix = is_pshader? "PB":"VB";
300 struct list* ptr;
301
302 for (i=0; i<max_constants; ++i) {
303 if (NULL == constants_set || constants_set[i]) {
304
305 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i*4]);
306
307 /* TODO: Benchmark and see if it would be beneficial to store the
308 * locations of the constants to avoid looking up each time */
309 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
310 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
311 if (tmp_loc != -1) {
312 /* We found this uniform name in the program - go ahead and send the data */
313 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i*4]));
314 checkGLcall("glUniform1ivARB");
315 }
316 }
317 }
318
319 /* Load immediate constants */
320 ptr = list_head(&This->baseShader.constantsB);
321 while (ptr) {
322 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
323 unsigned int idx = lconst->idx;
324 GLint* values = (GLint*) lconst->value;
325
326 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
327
328 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
329 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
330 if (tmp_loc != -1) {
331 /* We found this uniform name in the program - go ahead and send the data */
332 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
333 checkGLcall("glUniform1ivARB");
334 }
335 ptr = list_next(&This->baseShader.constantsB, ptr);
336 }
337 }
338
339
340
341 /**
342 * Loads the app-supplied constants into the currently set GLSL program.
343 */
344 void shader_glsl_load_constants(
345 IWineD3DDevice* device,
346 char usePixelShader,
347 char useVertexShader) {
348
349 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
350 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
351 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
352
353 GLhandleARB *constant_locations;
354 struct list *constant_list;
355 GLhandleARB programId;
356 GLint pos;
357
358 if (!stateBlock->glsl_program) {
359 /* No GLSL program set - nothing to do. */
360 return;
361 }
362 programId = stateBlock->glsl_program->programId;
363
364 if (useVertexShader) {
365 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
366 GLint pos;
367
368 constant_locations = stateBlock->glsl_program->vuniformF_locations;
369 constant_list = &stateBlock->set_vconstantsF;
370
371 /* Load vertex shader samplers */
372 shader_glsl_load_vsamplers(gl_info, (IWineD3DStateBlock*)stateBlock);
373
374 /* Load DirectX 9 float constants/uniforms for vertex shader */
375 shader_glsl_load_constantsF(vshader, gl_info, GL_LIMITS(vshader_constantsF),
376 stateBlock->vertexShaderConstantF, constant_locations, constant_list);
377
378 /* Load DirectX 9 integer constants/uniforms for vertex shader */
379 shader_glsl_load_constantsI(vshader, gl_info, programId, MAX_CONST_I,
380 stateBlock->vertexShaderConstantI,
381 stateBlock->changed.vertexShaderConstantsI);
382
383 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
384 shader_glsl_load_constantsB(vshader, gl_info, programId, MAX_CONST_B,
385 stateBlock->vertexShaderConstantB,
386 stateBlock->changed.vertexShaderConstantsB);
387
388 /* Upload the position fixup params */
389 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "posFixup"));
390 checkGLcall("glGetUniformLocationARB");
391 GL_EXTCALL(glUniform4fvARB(pos, 1, &deviceImpl->posFixup[0]));
392 checkGLcall("glUniform4fvARB");
393 }
394
395 if (usePixelShader) {
396
397 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
398
399 constant_locations = stateBlock->glsl_program->puniformF_locations;
400 constant_list = &stateBlock->set_pconstantsF;
401
402 /* Load pixel shader samplers */
403 shader_glsl_load_psamplers(gl_info, (IWineD3DStateBlock*) stateBlock);
404
405 /* Load DirectX 9 float constants/uniforms for pixel shader */
406 shader_glsl_load_constantsF(pshader, gl_info, GL_LIMITS(pshader_constantsF),
407 stateBlock->pixelShaderConstantF, constant_locations, constant_list);
408
409 /* Load DirectX 9 integer constants/uniforms for pixel shader */
410 shader_glsl_load_constantsI(pshader, gl_info, programId, MAX_CONST_I,
411 stateBlock->pixelShaderConstantI,
412 stateBlock->changed.pixelShaderConstantsI);
413
414 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
415 shader_glsl_load_constantsB(pshader, gl_info, programId, MAX_CONST_B,
416 stateBlock->pixelShaderConstantB,
417 stateBlock->changed.pixelShaderConstantsB);
418
419 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
420 * It can't be 0 for a valid texbem instruction.
421 */
422 if(((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat != -1) {
423 float *data = (float *) &stateBlock->textureState[(int) ((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
424 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "bumpenvmat"));
425 checkGLcall("glGetUniformLocationARB");
426 GL_EXTCALL(glUniformMatrix2fvARB(pos, 1, 0, data));
427 checkGLcall("glUniformMatrix2fvARB");
428
429 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
430 * is set too, so we can check that in the needsbumpmat check
431 */
432 if(((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams != -1) {
433 int stage = ((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams;
434 GLfloat *scale = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLSCALE];
435 GLfloat *offset = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET];
436
437 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "luminancescale"));
438 checkGLcall("glGetUniformLocationARB");
439 GL_EXTCALL(glUniform1fvARB(pos, 1, scale));
440 checkGLcall("glUniform1fvARB");
441 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "luminanceoffset"));
442 checkGLcall("glGetUniformLocationARB");
443 GL_EXTCALL(glUniform1fvARB(pos, 1, offset));
444 checkGLcall("glUniform1fvARB");
445 }
446 } else if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
447 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
448 float comparison[4];
449 float mul_low[4];
450
451 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
452 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
453 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
454
455 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
456 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
457 } else {
458 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
459 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
460
461 mul_low[0] = 1.0; mul_low[1] = 1.0;
462 mul_low[2] = 1.0; mul_low[3] = 1.0;
463 }
464
465 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_comparison"));
466 checkGLcall("glGetUniformLocationARB");
467 GL_EXTCALL(glUniform4fvARB(pos, 1, comparison));
468 pos = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_mul_low"));
469 checkGLcall("glGetUniformLocationARB");
470 GL_EXTCALL(glUniform4fvARB(pos, 1, mul_low));
471 }
472 }
473 }
474
475 /** Generate the variable & register declarations for the GLSL output target */
476 void shader_generate_glsl_declarations(
477 IWineD3DBaseShader *iface,
478 shader_reg_maps* reg_maps,
479 SHADER_BUFFER* buffer,
480 WineD3D_GL_Info* gl_info) {
481
482 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
483 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
484 int i;
485 unsigned int extra_constants_needed = 0;
486
487 /* There are some minor differences between pixel and vertex shaders */
488 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
489 char prefix = pshader ? 'P' : 'V';
490
491 /* Prototype the subroutines */
492 for (i = 0; i < This->baseShader.limits.label; i++) {
493 if (reg_maps->labels[i])
494 shader_addline(buffer, "void subroutine%lu();\n", i);
495 }
496
497 /* Declare the constants (aka uniforms) */
498 if (This->baseShader.limits.constant_float > 0) {
499 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
500 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
501 shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
502 }
503
504 if (This->baseShader.limits.constant_int > 0)
505 shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
506
507 if (This->baseShader.limits.constant_bool > 0)
508 shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
509
510 if(!pshader) {
511 shader_addline(buffer, "uniform vec4 posFixup;\n");
512 } else {
513 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
514
515 if(reg_maps->bumpmat != -1) {
516 shader_addline(buffer, "uniform mat2 bumpenvmat;\n");
517 if(reg_maps->luminanceparams) {
518 shader_addline(buffer, "uniform float luminancescale;\n");
519 shader_addline(buffer, "uniform float luminanceoffset;\n");
520 extra_constants_needed++;
521 }
522 extra_constants_needed++;
523 }
524
525 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
526 ps_impl->srgb_enabled = 1;
527 if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
528 shader_addline(buffer, "uniform vec4 srgb_mul_low;\n");
529 shader_addline(buffer, "uniform vec4 srgb_comparison;\n");
530 ps_impl->srgb_mode_hardcoded = 0;
531 } else {
532 ps_impl->srgb_mode_hardcoded = 1;
533 shader_addline(buffer, "const vec4 srgb_mul_low = {%f, %f, %f, %f};\n",
534 srgb_mul_low, srgb_mul_low, srgb_mul_low, srgb_mul_low);
535 shader_addline(buffer, "const vec4 srgb_comparison = {%f, %f, %f, %f};\n",
536 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
537 }
538 } else {
539 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
540
541 /* Do not write any srgb fixup into the shader to save shader size and processing time.
542 * As a consequence, we can't toggle srgb write on without recompilation
543 */
544 ps_impl->srgb_enabled = 0;
545 ps_impl->srgb_mode_hardcoded = 1;
546 }
547 }
548
549 /* Declare texture samplers */
550 for (i = 0; i < This->baseShader.limits.sampler; i++) {
551 if (reg_maps->samplers[i]) {
552
553 DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
554 switch (stype) {
555
556 case WINED3DSTT_1D:
557 shader_addline(buffer, "uniform sampler1D %csampler%lu;\n", prefix, i);
558 break;
559 case WINED3DSTT_2D:
560 shader_addline(buffer, "uniform sampler2D %csampler%lu;\n", prefix, i);
561 break;
562 case WINED3DSTT_CUBE:
563 shader_addline(buffer, "uniform samplerCube %csampler%lu;\n", prefix, i);
564 break;
565 case WINED3DSTT_VOLUME:
566 shader_addline(buffer, "uniform sampler3D %csampler%lu;\n", prefix, i);
567 break;
568 default:
569 shader_addline(buffer, "uniform unsupported_sampler %csampler%lu;\n", prefix, i);
570 FIXME("Unrecognized sampler type: %#x\n", stype);
571 break;
572 }
573 }
574 }
575
576 /* Declare address variables */
577 for (i = 0; i < This->baseShader.limits.address; i++) {
578 if (reg_maps->address[i])
579 shader_addline(buffer, "ivec4 A%d;\n", i);
580 }
581
582 /* Declare texture coordinate temporaries and initialize them */
583 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
584 if (reg_maps->texcoord[i])
585 shader_addline(buffer, "vec4 T%lu = gl_TexCoord[%lu];\n", i, i);
586 }
587
588 /* Declare input register temporaries */
589 for (i=0; i < This->baseShader.limits.packed_input; i++) {
590 if (reg_maps->packed_input[i])
591 shader_addline(buffer, "vec4 IN%lu;\n", i);
592 }
593
594 /* Declare output register temporaries */
595 for (i = 0; i < This->baseShader.limits.packed_output; i++) {
596 if (reg_maps->packed_output[i])
597 shader_addline(buffer, "vec4 OUT%lu;\n", i);
598 }
599
600 /* Declare temporary variables */
601 for(i = 0; i < This->baseShader.limits.temporary; i++) {
602 if (reg_maps->temporary[i])
603 shader_addline(buffer, "vec4 R%lu;\n", i);
604 }
605
606 /* Declare attributes */
607 for (i = 0; i < This->baseShader.limits.attributes; i++) {
608 if (reg_maps->attributes[i])
609 shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
610 }
611
612 /* Declare loop register aL */
613 if (reg_maps->loop) {
614 shader_addline(buffer, "int aL;\n");
615 shader_addline(buffer, "int tmpInt;\n");
616 }
617
618 /* Temporary variables for matrix operations */
619 shader_addline(buffer, "vec4 tmp0;\n");
620 shader_addline(buffer, "vec4 tmp1;\n");
621
622 /* Start the main program */
623 shader_addline(buffer, "void main() {\n");
624 }
625
626 /*****************************************************************************
627 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
628 *
629 * For more information, see http://wiki.winehq.org/DirectX-Shaders
630 ****************************************************************************/
631
632 /* Prototypes */
633 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
634 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
635
636 /** Used for opcode modifiers - They multiply the result by the specified amount */
637 static const char * const shift_glsl_tab[] = {
638 "", /* 0 (none) */
639 "2.0 * ", /* 1 (x2) */
640 "4.0 * ", /* 2 (x4) */
641 "8.0 * ", /* 3 (x8) */
642 "16.0 * ", /* 4 (x16) */
643 "32.0 * ", /* 5 (x32) */
644 "", /* 6 (x64) */
645 "", /* 7 (x128) */
646 "", /* 8 (d256) */
647 "", /* 9 (d128) */
648 "", /* 10 (d64) */
649 "", /* 11 (d32) */
650 "0.0625 * ", /* 12 (d16) */
651 "0.125 * ", /* 13 (d8) */
652 "0.25 * ", /* 14 (d4) */
653 "0.5 * " /* 15 (d2) */
654 };
655
656 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
657 static void shader_glsl_gen_modifier (
658 const DWORD instr,
659 const char *in_reg,
660 const char *in_regswizzle,
661 char *out_str) {
662
663 out_str[0] = 0;
664
665 if (instr == WINED3DSIO_TEXKILL)
666 return;
667
668 switch (instr & WINED3DSP_SRCMOD_MASK) {
669 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
670 case WINED3DSPSM_DW:
671 case WINED3DSPSM_NONE:
672 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
673 break;
674 case WINED3DSPSM_NEG:
675 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
676 break;
677 case WINED3DSPSM_NOT:
678 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
679 break;
680 case WINED3DSPSM_BIAS:
681 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
682 break;
683 case WINED3DSPSM_BIASNEG:
684 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
685 break;
686 case WINED3DSPSM_SIGN:
687 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
688 break;
689 case WINED3DSPSM_SIGNNEG:
690 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
691 break;
692 case WINED3DSPSM_COMP:
693 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
694 break;
695 case WINED3DSPSM_X2:
696 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
697 break;
698 case WINED3DSPSM_X2NEG:
699 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
700 break;
701 case WINED3DSPSM_ABS:
702 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
703 break;
704 case WINED3DSPSM_ABSNEG:
705 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
706 break;
707 default:
708 FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
709 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
710 }
711 }
712
713 /** Writes the GLSL variable name that corresponds to the register that the
714 * DX opcode parameter is trying to access */
715 static void shader_glsl_get_register_name(
716 const DWORD param,
717 const DWORD addr_token,
718 char* regstr,
719 BOOL* is_color,
720 SHADER_OPCODE_ARG* arg) {
721
722 /* oPos, oFog and oPts in D3D */
723 static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
724
725 DWORD reg = param & WINED3DSP_REGNUM_MASK;
726 DWORD regtype = shader_get_regtype(param);
727 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
728 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
729 WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
730
731 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
732 char tmpStr[50];
733
734 *is_color = FALSE;
735
736 switch (regtype) {
737 case WINED3DSPR_TEMP:
738 sprintf(tmpStr, "R%u", reg);
739 break;
740 case WINED3DSPR_INPUT:
741 if (pshader) {
742 /* Pixel shaders >= 3.0 */
743 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
744 sprintf(tmpStr, "IN%u", reg);
745 else {
746 if (reg==0)
747 strcpy(tmpStr, "gl_Color");
748 else
749 strcpy(tmpStr, "gl_SecondaryColor");
750 }
751 } else {
752 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
753 *is_color = TRUE;
754 sprintf(tmpStr, "attrib%u", reg);
755 }
756 break;
757 case WINED3DSPR_CONST:
758 {
759 const char* prefix = pshader? "PC":"VC";
760
761 /* Relative addressing */
762 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
763
764 /* Relative addressing on shaders 2.0+ have a relative address token,
765 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
766 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
767 glsl_src_param_t rel_param;
768 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
769 sprintf(tmpStr, "%s[%s + %u]", prefix, rel_param.param_str, reg);
770 } else
771 sprintf(tmpStr, "%s[A0.x + %u]", prefix, reg);
772
773 } else
774 sprintf(tmpStr, "%s[%u]", prefix, reg);
775
776 break;
777 }
778 case WINED3DSPR_CONSTINT:
779 if (pshader)
780 sprintf(tmpStr, "PI[%u]", reg);
781 else
782 sprintf(tmpStr, "VI[%u]", reg);
783 break;
784 case WINED3DSPR_CONSTBOOL:
785 if (pshader)
786 sprintf(tmpStr, "PB[%u]", reg);
787 else
788 sprintf(tmpStr, "VB[%u]", reg);
789 break;
790 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
791 if (pshader) {
792 sprintf(tmpStr, "T%u", reg);
793 } else {
794 sprintf(tmpStr, "A%u", reg);
795 }
796 break;
797 case WINED3DSPR_LOOP:
798 sprintf(tmpStr, "aL");
799 break;
800 case WINED3DSPR_SAMPLER:
801 if (pshader)
802 sprintf(tmpStr, "Psampler%u", reg);
803 else
804 sprintf(tmpStr, "Vsampler%u", reg);
805 break;
806 case WINED3DSPR_COLOROUT:
807 if (reg >= GL_LIMITS(buffers)) {
808 WARN("Write to render target %u, only %d supported\n", reg, 4);
809 }
810 if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
811 sprintf(tmpStr, "gl_FragData[%u]", reg);
812 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
813 sprintf(tmpStr, "gl_FragColor");
814 }
815 break;
816 case WINED3DSPR_RASTOUT:
817 sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
818 break;
819 case WINED3DSPR_DEPTHOUT:
820 sprintf(tmpStr, "gl_FragDepth");
821 break;
822 case WINED3DSPR_ATTROUT:
823 if (reg == 0) {
824 sprintf(tmpStr, "gl_FrontColor");
825 } else {
826 sprintf(tmpStr, "gl_FrontSecondaryColor");
827 }
828 break;
829 case WINED3DSPR_TEXCRDOUT:
830 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
831 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
832 sprintf(tmpStr, "OUT%u", reg);
833 else
834 sprintf(tmpStr, "gl_TexCoord[%u]", reg);
835 break;
836 case WINED3DSPR_MISCTYPE:
837 if (reg == 0) {
838 /* vPos */
839 sprintf(tmpStr, "gl_FragCoord");
840 } else {
841 /* gl_FrontFacing could be used for vFace, but note that
842 * gl_FrontFacing is a bool, while vFace is a float for
843 * which the sign determines front/back */
844 FIXME("Unhandled misctype register %d\n", reg);
845 sprintf(tmpStr, "unrecognized_register");
846 }
847 break;
848 default:
849 FIXME("Unhandled register name Type(%d)\n", regtype);
850 sprintf(tmpStr, "unrecognized_register");
851 break;
852 }
853
854 strcat(regstr, tmpStr);
855 }
856
857 /* Get the GLSL write mask for the destination register */
858 static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
859 char *ptr = write_mask;
860 DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
861
862 if (shader_is_scalar(param)) {
863 mask = WINED3DSP_WRITEMASK_0;
864 } else {
865 *ptr++ = '.';
866 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
867 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
868 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
869 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
870 }
871
872 *ptr = '\0';
873
874 return mask;
875 }
876
877 static size_t shader_glsl_get_write_mask_size(DWORD write_mask) {
878 size_t size = 0;
879
880 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
881 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
882 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
883 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
884
885 return size;
886 }
887
888 static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
889 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
890 * but addressed as "rgba". To fix this we need to swap the register's x
891 * and z components. */
892 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
893 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
894 char *ptr = swizzle_str;
895
896 if (!shader_is_scalar(param)) {
897 *ptr++ = '.';
898 /* swizzle bits fields: wwzzyyxx */
899 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
900 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
901 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
902 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
903 }
904
905 *ptr = '\0';
906 }
907
908 /* From a given parameter token, generate the corresponding GLSL string.
909 * Also, return the actual register name and swizzle in case the
910 * caller needs this information as well. */
911 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
912 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
913 BOOL is_color = FALSE;
914 char swizzle_str[6];
915
916 src_param->reg_name[0] = '\0';
917 src_param->param_str[0] = '\0';
918 swizzle_str[0] = '\0';
919
920 shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
921
922 shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
923 shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
924 }
925
926 /* From a given parameter token, generate the corresponding GLSL string.
927 * Also, return the actual register name and swizzle in case the
928 * caller needs this information as well. */
929 static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
930 const DWORD addr_token, glsl_dst_param_t *dst_param) {
931 BOOL is_color = FALSE;
932
933 dst_param->mask_str[0] = '\0';
934 dst_param->reg_name[0] = '\0';
935
936 shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
937 return shader_glsl_get_write_mask(param, dst_param->mask_str);
938 }
939
940 /* Append the destination part of the instruction to the buffer, return the effective write mask */
941 static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
942 glsl_dst_param_t dst_param;
943 DWORD mask;
944 int shift;
945
946 mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
947
948 if(mask) {
949 shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
950 shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
951 }
952
953 return mask;
954 }
955
956 /* Append the destination part of the instruction to the buffer, return the effective write mask */
957 static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
958 return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
959 }
960
961 /** Process GLSL instruction modifiers */
962 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
963
964 DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
965
966 if (arg->opcode->dst_token && mask != 0) {
967 glsl_dst_param_t dst_param;
968
969 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
970
971 if (mask & WINED3DSPDM_SATURATE) {
972 /* _SAT means to clamp the value of the register to between 0 and 1 */
973 shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
974 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
975 }
976 if (mask & WINED3DSPDM_MSAMPCENTROID) {
977 FIXME("_centroid modifier not handled\n");
978 }
979 if (mask & WINED3DSPDM_PARTIALPRECISION) {
980 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
981 }
982 }
983 }
984
985 static inline const char* shader_get_comp_op(
986 const DWORD opcode) {
987
988 DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
989 switch (op) {
990 case COMPARISON_GT: return ">";
991 case COMPARISON_EQ: return "==";
992 case COMPARISON_GE: return ">=";
993 case COMPARISON_LT: return "<";
994 case COMPARISON_NE: return "!=";
995 case COMPARISON_LE: return "<=";
996 default:
997 FIXME("Unrecognized comparison value: %u\n", op);
998 return "(\?\?)";
999 }
1000 }
1001
1002 static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, glsl_sample_function_t *sample_function) {
1003 /* Note that there's no such thing as a projected cube texture. */
1004 switch(sampler_type) {
1005 case WINED3DSTT_1D:
1006 sample_function->name = projected ? "texture1DProj" : "texture1D";
1007 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
1008 break;
1009 case WINED3DSTT_2D:
1010 sample_function->name = projected ? "texture2DProj" : "texture2D";
1011 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
1012 break;
1013 case WINED3DSTT_CUBE:
1014 sample_function->name = "textureCube";
1015 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1016 break;
1017 case WINED3DSTT_VOLUME:
1018 sample_function->name = projected ? "texture3DProj" : "texture3D";
1019 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1020 break;
1021 default:
1022 sample_function->name = "";
1023 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
1024 break;
1025 }
1026 }
1027
1028 static void shader_glsl_color_correction(SHADER_OPCODE_ARG* arg) {
1029 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1030 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
1031 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
1032 glsl_dst_param_t dst_param;
1033 glsl_dst_param_t dst_param2;
1034 WINED3DFORMAT fmt;
1035 WINED3DFORMAT conversion_group;
1036 IWineD3DBaseTextureImpl *texture;
1037 DWORD mask, mask_size;
1038 UINT i;
1039 BOOL recorded = FALSE;
1040 DWORD sampler_idx;
1041 DWORD hex_version = shader->baseShader.hex_version;
1042
1043 switch(arg->opcode->opcode) {
1044 case WINED3DSIO_TEX:
1045 if (hex_version < WINED3DPS_VERSION(2,0)) {
1046 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1047 } else {
1048 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1049 }
1050 break;
1051
1052 case WINED3DSIO_TEXLDL:
1053 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
1054 return;
1055
1056 case WINED3DSIO_TEXDP3TEX:
1057 case WINED3DSIO_TEXM3x3TEX:
1058 case WINED3DSIO_TEXM3x3SPEC:
1059 case WINED3DSIO_TEXM3x3VSPEC:
1060 case WINED3DSIO_TEXBEM:
1061 case WINED3DSIO_TEXREG2AR:
1062 case WINED3DSIO_TEXREG2GB:
1063 case WINED3DSIO_TEXREG2RGB:
1064 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1065 break;
1066
1067 default:
1068 /* Not a texture sampling instruction, nothing to do */
1069 return;
1070 };
1071
1072 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
1073 if(texture) {
1074 fmt = texture->resource.format;
1075 conversion_group = texture->baseTexture.shader_conversion_group;
1076 } else {
1077 fmt = WINED3DFMT_UNKNOWN;
1078 conversion_group = WINED3DFMT_UNKNOWN;
1079 }
1080
1081 /* before doing anything, record the sampler with the format in the format conversion list,
1082 * but check if it's not there already
1083 */
1084 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
1085 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
1086 recorded = TRUE;
1087 break;
1088 }
1089 }
1090 if(!recorded) {
1091 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
1092 shader->baseShader.num_sampled_samplers++;
1093 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
1094 }
1095
1096 switch(fmt) {
1097 case WINED3DFMT_V8U8:
1098 case WINED3DFMT_V16U16:
1099 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
1100 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
1101 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
1102 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_2, &dst_param);
1103 mask_size = shader_glsl_get_write_mask_size(mask);
1104 if(mask_size >= 3) {
1105 shader_addline(arg->buffer, "%s.%c = 1.0;\n", dst_param.reg_name, dst_param.mask_str[3]);
1106 }
1107 } else {
1108 /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
1109 mask = shader_glsl_add_dst_param(arg, arg->dst,
1110 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1,
1111 &dst_param);
1112 mask_size = shader_glsl_get_write_mask_size(mask);
1113 if(mask_size >= 2) {
1114 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1115 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1116 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1117 } else if(mask_size == 1) {
1118 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str[1],
1119 dst_param.reg_name, dst_param.mask_str[1]);
1120 }
1121 }
1122 break;
1123
1124 case WINED3DFMT_X8L8V8U8:
1125 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1126 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
1127 * and a(X) is always 1.0
1128 */
1129 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1130 mask_size = shader_glsl_get_write_mask_size(mask);
1131 if(mask_size >= 2) {
1132 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1133 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1134 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1135 } else if(mask_size == 1) {
1136 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1137 dst_param.reg_name, dst_param.mask_str[1],
1138 dst_param.reg_name, dst_param.mask_str[1]);
1139 }
1140 }
1141 break;
1142
1143 case WINED3DFMT_L6V5U5:
1144 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1145 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1146 mask_size = shader_glsl_get_write_mask_size(mask);
1147 shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_2, &dst_param2);
1148 if(mask_size >= 3) {
1149 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1150 shader_addline(arg->buffer, "tmp0.g = %s.%c;\n",
1151 dst_param.reg_name, dst_param.mask_str[2]);
1152 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1153 dst_param.reg_name, dst_param.mask_str[2], dst_param.mask_str[1],
1154 dst_param2.reg_name, dst_param.mask_str[1], dst_param.mask_str[3]);
1155 shader_addline(arg->buffer, "%s.%c = tmp0.g;\n", dst_param.reg_name,
1156 dst_param.mask_str[3]);
1157 } else if(mask_size == 2) {
1158 /* This is bad: We have VL, but we need VU */
1159 FIXME("2 components sampled from a converted L6V5U5 texture\n");
1160 } else {
1161 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1162 dst_param.reg_name, dst_param.mask_str[1],
1163 dst_param2.reg_name, dst_param.mask_str[1]);
1164 }
1165 }
1166 break;
1167
1168 case WINED3DFMT_Q8W8V8U8:
1169 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1170 /* Correct the sign in all channels. The writemask just applies as-is, no
1171 * need for checking the mask size
1172 */
1173 shader_glsl_add_dst_param(arg, arg->dst,
1174 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 |
1175 WINED3DSP_WRITEMASK_2 | WINED3DSP_WRITEMASK_3,
1176 &dst_param);
1177 shader_addline(arg->buffer, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str,
1178 dst_param.reg_name, dst_param.mask_str);
1179 }
1180 break;
1181
1182 /* stupid compiler */
1183 default:
1184 break;
1185 }
1186 }
1187
1188 /*****************************************************************************
1189 *
1190 * Begin processing individual instruction opcodes
1191 *
1192 ****************************************************************************/
1193
1194 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1195 void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
1196 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1197 SHADER_BUFFER* buffer = arg->buffer;
1198 glsl_src_param_t src0_param;
1199 glsl_src_param_t src1_param;
1200 DWORD write_mask;
1201 char op;
1202
1203 /* Determine the GLSL operator to use based on the opcode */
1204 switch (curOpcode->opcode) {
1205 case WINED3DSIO_MUL: op = '*'; break;
1206 case WINED3DSIO_ADD: op = '+'; break;
1207 case WINED3DSIO_SUB: op = '-'; break;
1208 default:
1209 op = ' ';
1210 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1211 break;
1212 }
1213
1214 write_mask = shader_glsl_append_dst(buffer, arg);
1215 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1216 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1217 shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
1218 }
1219
1220 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1221 void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
1222 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1223 SHADER_BUFFER* buffer = arg->buffer;
1224 glsl_src_param_t src0_param;
1225 DWORD write_mask;
1226
1227 write_mask = shader_glsl_append_dst(buffer, arg);
1228 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1229
1230 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1231 * shader versions WINED3DSIO_MOVA is used for this. */
1232 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
1233 !shader_is_pshader_version(shader->baseShader.hex_version) &&
1234 shader_get_regtype(arg->dst) == WINED3DSPR_ADDR) ||
1235 arg->opcode->opcode == WINED3DSIO_MOVA) {
1236 /* We need to *round* to the nearest int here. */
1237 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1238 if (mask_size > 1) {
1239 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);
1240 } else {
1241 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param.param_str, src0_param.param_str);
1242 }
1243 } else {
1244 shader_addline(buffer, "%s);\n", src0_param.param_str);
1245 }
1246 }
1247
1248 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1249 void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
1250 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1251 SHADER_BUFFER* buffer = arg->buffer;
1252 glsl_src_param_t src0_param;
1253 glsl_src_param_t src1_param;
1254 DWORD dst_write_mask, src_write_mask;
1255 size_t dst_size = 0;
1256
1257 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1258 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1259
1260 /* dp3 works on vec3, dp4 on vec4 */
1261 if (curOpcode->opcode == WINED3DSIO_DP4) {
1262 src_write_mask = WINED3DSP_WRITEMASK_ALL;
1263 } else {
1264 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1265 }
1266
1267 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
1268 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
1269
1270 if (dst_size > 1) {
1271 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1272 } else {
1273 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
1274 }
1275 }
1276
1277 /* Note that this instruction has some restrictions. The destination write mask
1278 * can't contain the w component, and the source swizzles have to be .xyzw */
1279 void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
1280 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1281 glsl_src_param_t src0_param;
1282 glsl_src_param_t src1_param;
1283 char dst_mask[6];
1284
1285 shader_glsl_get_write_mask(arg->dst, dst_mask);
1286 shader_glsl_append_dst(arg->buffer, arg);
1287 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1288 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1289 shader_addline(arg->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
1290 }
1291
1292 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1293 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1294 * GLSL uses the value as-is. */
1295 void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
1296 SHADER_BUFFER *buffer = arg->buffer;
1297 glsl_src_param_t src0_param;
1298 glsl_src_param_t src1_param;
1299 DWORD dst_write_mask;
1300 size_t dst_size;
1301
1302 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1303 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1304
1305 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1306 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1307
1308 if (dst_size > 1) {
1309 shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1310 } else {
1311 shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
1312 }
1313 }
1314
1315 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1316 void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
1317 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1318 SHADER_BUFFER* buffer = arg->buffer;
1319 glsl_src_param_t src_param;
1320 const char *instruction;
1321 char arguments[256];
1322 DWORD write_mask;
1323 unsigned i;
1324
1325 /* Determine the GLSL function to use based on the opcode */
1326 /* TODO: Possibly make this a table for faster lookups */
1327 switch (curOpcode->opcode) {
1328 case WINED3DSIO_MIN: instruction = "min"; break;
1329 case WINED3DSIO_MAX: instruction = "max"; break;
1330 case WINED3DSIO_ABS: instruction = "abs"; break;
1331 case WINED3DSIO_FRC: instruction = "fract"; break;
1332 case WINED3DSIO_NRM: instruction = "normalize"; break;
1333 case WINED3DSIO_LOGP:
1334 case WINED3DSIO_LOG: instruction = "log2"; break;
1335 case WINED3DSIO_EXP: instruction = "exp2"; break;
1336 case WINED3DSIO_SGN: instruction = "sign"; break;
1337 case WINED3DSIO_DSX: instruction = "dFdx"; break;
1338 case WINED3DSIO_DSY: instruction = "dFdy"; break;
1339 default: instruction = "";
1340 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1341 break;
1342 }
1343
1344 write_mask = shader_glsl_append_dst(buffer, arg);
1345
1346 arguments[0] = '\0';
1347 if (curOpcode->num_params > 0) {
1348 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
1349 strcat(arguments, src_param.param_str);
1350 for (i = 2; i < curOpcode->num_params; ++i) {
1351 strcat(arguments, ", ");
1352 shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
1353 strcat(arguments, src_param.param_str);
1354 }
1355 }
1356
1357 shader_addline(buffer, "%s(%s));\n", instruction, arguments);
1358 }
1359
1360 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1361 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1362 * dst.x = 2^(floor(src))
1363 * dst.y = src - floor(src)
1364 * dst.z = 2^src (partial precision is allowed, but optional)
1365 * dst.w = 1.0;
1366 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1367 * dst = 2^src; (partial precision is allowed, but optional)
1368 */
1369 void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
1370 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1371 glsl_src_param_t src_param;
1372
1373 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1374
1375 if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1376 char dst_mask[6];
1377
1378 shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
1379 shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
1380 shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
1381 shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
1382
1383 shader_glsl_append_dst(arg->buffer, arg);
1384 shader_glsl_get_write_mask(arg->dst, dst_mask);
1385 shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
1386 } else {
1387 DWORD write_mask;
1388 size_t mask_size;
1389
1390 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1391 mask_size = shader_glsl_get_write_mask_size(write_mask);
1392
1393 if (mask_size > 1) {
1394 shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
1395 } else {
1396 shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
1397 }
1398 }
1399 }
1400
1401 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1402 void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
1403 glsl_src_param_t src_param;
1404 DWORD write_mask;
1405 size_t mask_size;
1406
1407 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1408 mask_size = shader_glsl_get_write_mask_size(write_mask);
1409 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1410
1411 if (mask_size > 1) {
1412 shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
1413 } else {
1414 shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
1415 }
1416 }
1417
1418 void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
1419 SHADER_BUFFER* buffer = arg->buffer;
1420 glsl_src_param_t src_param;
1421 DWORD write_mask;
1422 size_t mask_size;
1423
1424 write_mask = shader_glsl_append_dst(buffer, arg);
1425 mask_size = shader_glsl_get_write_mask_size(write_mask);
1426
1427 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1428
1429 if (mask_size > 1) {
1430 shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
1431 } else {
1432 shader_addline(buffer, "inversesqrt(%s));\n", src_param.param_str);
1433 }
1434 }
1435
1436 /** Process signed comparison opcodes in GLSL. */
1437 void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
1438 glsl_src_param_t src0_param;
1439 glsl_src_param_t src1_param;
1440 DWORD write_mask;
1441 size_t mask_size;
1442
1443 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1444 mask_size = shader_glsl_get_write_mask_size(write_mask);
1445 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1446 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1447
1448 if (mask_size > 1) {
1449 const char *compare;
1450
1451 switch(arg->opcode->opcode) {
1452 case WINED3DSIO_SLT: compare = "lessThan"; break;
1453 case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
1454 default: compare = "";
1455 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1456 }
1457
1458 shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
1459 src0_param.param_str, src1_param.param_str);
1460 } else {
1461 const char *compare;
1462
1463 switch(arg->opcode->opcode) {
1464 case WINED3DSIO_SLT: compare = "<"; break;
1465 case WINED3DSIO_SGE: compare = ">="; break;
1466 default: compare = "";
1467 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1468 }
1469
1470 shader_addline(arg->buffer, "(%s %s %s) ? 1.0 : 0.0);\n",
1471 src0_param.param_str, compare, src1_param.param_str);
1472 }
1473 }
1474
1475 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1476 void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
1477 glsl_src_param_t src0_param;
1478 glsl_src_param_t src1_param;
1479 glsl_src_param_t src2_param;
1480 DWORD write_mask, cmp_channel = 0;
1481 unsigned int i, j;
1482
1483 /* Cycle through all source0 channels */
1484 for (i=0; i<4; i++) {
1485 write_mask = 0;
1486 /* Find the destination channels which use the current source0 channel */
1487 for (j=0; j<4; j++) {
1488 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1489 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1490 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1491 }
1492 }
1493 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1494 if (!write_mask) continue;
1495
1496 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1497 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1498 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1499
1500 shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
1501 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1502 }
1503 }
1504
1505 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1506 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1507 * the compare is done per component of src0. */
1508 void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
1509 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1510 glsl_src_param_t src0_param;
1511 glsl_src_param_t src1_param;
1512 glsl_src_param_t src2_param;
1513 DWORD write_mask, cmp_channel = 0;
1514 unsigned int i, j;
1515
1516 if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
1517 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1518 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1519 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1520 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1521
1522 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1523 if(arg->opcode_token & WINED3DSI_COISSUE) {
1524 shader_addline(arg->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
1525 } else {
1526 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1527 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1528 }
1529 return;
1530 }
1531 /* Cycle through all source0 channels */
1532 for (i=0; i<4; i++) {
1533 write_mask = 0;
1534 /* Find the destination channels which use the current source0 channel */
1535 for (j=0; j<4; j++) {
1536 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1537 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1538 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1539 }
1540 }
1541 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1542 if (!write_mask) continue;
1543
1544 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1545 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1546 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1547
1548 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1549 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1550 }
1551 }
1552
1553 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1554 void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
1555 glsl_src_param_t src0_param;
1556 glsl_src_param_t src1_param;
1557 glsl_src_param_t src2_param;
1558 DWORD write_mask;
1559
1560 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1561 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1562 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1563 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1564 shader_addline(arg->buffer, "(%s * %s) + %s);\n",
1565 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1566 }
1567
1568 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1569 Vertex shaders to GLSL codes */
1570 void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
1571 int i;
1572 int nComponents = 0;
1573 SHADER_OPCODE_ARG tmpArg;
1574
1575 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1576
1577 /* Set constants for the temporary argument */
1578 tmpArg.shader = arg->shader;
1579 tmpArg.buffer = arg->buffer;
1580 tmpArg.src[0] = arg->src[0];
1581 tmpArg.src_addr[0] = arg->src_addr[0];
1582 tmpArg.src_addr[1] = arg->src_addr[1];
1583 tmpArg.reg_maps = arg->reg_maps;
1584
1585 switch(arg->opcode->opcode) {
1586 case WINED3DSIO_M4x4:
1587 nComponents = 4;
1588 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1589 break;
1590 case WINED3DSIO_M4x3:
1591 nComponents = 3;
1592 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1593 break;
1594 case WINED3DSIO_M3x4:
1595 nComponents = 4;
1596 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1597 break;
1598 case WINED3DSIO_M3x3:
1599 nComponents = 3;
1600 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1601 break;
1602 case WINED3DSIO_M3x2:
1603 nComponents = 2;
1604 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1605 break;
1606 default:
1607 break;
1608 }
1609
1610 for (i = 0; i < nComponents; i++) {
1611 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1612 tmpArg.src[1] = arg->src[1]+i;
1613 shader_glsl_dot(&tmpArg);
1614 }
1615 }
1616
1617 /**
1618 The LRP instruction performs a component-wise linear interpolation
1619 between the second and third operands using the first operand as the
1620 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1621 This is equivalent to mix(src2, src1, src0);
1622 */
1623 void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
1624 glsl_src_param_t src0_param;
1625 glsl_src_param_t src1_param;
1626 glsl_src_param_t src2_param;
1627 DWORD write_mask;
1628
1629 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1630
1631 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1632 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1633 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1634
1635 shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
1636 src2_param.param_str, src1_param.param_str, src0_param.param_str);
1637 }
1638
1639 /** Process the WINED3DSIO_LIT instruction in GLSL:
1640 * dst.x = dst.w = 1.0
1641 * dst.y = (src0.x > 0) ? src0.x
1642 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1643 * where src.w is clamped at +- 128
1644 */
1645 void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
1646 glsl_src_param_t src0_param;
1647 glsl_src_param_t src1_param;
1648 glsl_src_param_t src3_param;
1649 char dst_mask[6];
1650
1651 shader_glsl_append_dst(arg->buffer, arg);
1652 shader_glsl_get_write_mask(arg->dst, dst_mask);
1653
1654 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1655 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
1656 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
1657
1658 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",
1659 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);
1660 }
1661
1662 /** Process the WINED3DSIO_DST instruction in GLSL:
1663 * dst.x = 1.0
1664 * dst.y = src0.x * src0.y
1665 * dst.z = src0.z
1666 * dst.w = src1.w
1667 */
1668 void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
1669 glsl_src_param_t src0y_param;
1670 glsl_src_param_t src0z_param;
1671 glsl_src_param_t src1y_param;
1672 glsl_src_param_t src1w_param;
1673 char dst_mask[6];
1674
1675 shader_glsl_append_dst(arg->buffer, arg);
1676 shader_glsl_get_write_mask(arg->dst, dst_mask);
1677
1678 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
1679 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
1680 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
1681 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
1682
1683 shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1684 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
1685 }
1686
1687 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1688 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1689 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1690 *
1691 * dst.x = cos(src0.?)
1692 * dst.y = sin(src0.?)
1693 * dst.z = dst.z
1694 * dst.w = dst.w
1695 */
1696 void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
1697 glsl_src_param_t src0_param;
1698 DWORD write_mask;
1699
1700 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1701 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1702
1703 switch (write_mask) {
1704 case WINED3DSP_WRITEMASK_0:
1705 shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
1706 break;
1707
1708 case WINED3DSP_WRITEMASK_1:
1709 shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
1710 break;
1711
1712 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
1713 shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
1714 break;
1715
1716 default:
1717 ERR("Write mask should be .x, .y or .xy\n");
1718 break;
1719 }
1720 }
1721
1722 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1723 * Start a for() loop where src1.y is the initial value of aL,
1724 * increment aL by src1.z for a total of src1.x iterations.
1725 * Need to use a temporary variable for this operation.
1726 */
1727 /* FIXME: I don't think nested loops will work correctly this way. */
1728 void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
1729 glsl_src_param_t src1_param;
1730
1731 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
1732
1733 shader_addline(arg->buffer, "for (tmpInt = 0, aL = %s.y; tmpInt < %s.x; tmpInt++, aL += %s.z) {\n",
1734 src1_param.reg_name, src1_param.reg_name, src1_param.reg_name);
1735 }
1736
1737 void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
1738 shader_addline(arg->buffer, "}\n");
1739 }
1740
1741 void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
1742 glsl_src_param_t src0_param;
1743
1744 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1745 shader_addline(arg->buffer, "for (tmpInt = 0; tmpInt < %s; tmpInt++) {\n", src0_param.param_str);
1746 }
1747
1748 void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
1749 glsl_src_param_t src0_param;
1750
1751 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1752 shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
1753 }
1754
1755 void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
1756 glsl_src_param_t src0_param;
1757 glsl_src_param_t src1_param;
1758
1759 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1760 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1761
1762 shader_addline(arg->buffer, "if (%s %s %s) {\n",
1763 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1764 }
1765
1766 void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
1767 shader_addline(arg->buffer, "} else {\n");
1768 }
1769
1770 void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
1771 shader_addline(arg->buffer, "break;\n");
1772 }
1773
1774 /* FIXME: According to MSDN the compare is done per component. */
1775 void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
1776 glsl_src_param_t src0_param;
1777 glsl_src_param_t src1_param;
1778
1779 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1780 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1781
1782 shader_addline(arg->buffer, "if (%s %s %s) break;\n",
1783 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1784 }
1785
1786 void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
1787
1788 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1789 shader_addline(arg->buffer, "}\n");
1790 shader_addline(arg->buffer, "void subroutine%lu () {\n", snum);
1791 }
1792
1793 void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
1794 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1795 shader_addline(arg->buffer, "subroutine%lu();\n", snum);
1796 }
1797
1798 void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
1799 glsl_src_param_t src1_param;
1800
1801 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
1802 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1803 shader_addline(arg->buffer, "if (%s) subroutine%lu();\n", src1_param.param_str, snum);
1804 }
1805
1806 /*********************************************
1807 * Pixel Shader Specific Code begins here
1808 ********************************************/
1809 void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
1810 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1811 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1812 DWORD hex_version = This->baseShader.hex_version;
1813 char dst_swizzle[6];
1814 glsl_sample_function_t sample_function;
1815 DWORD sampler_type;
1816 DWORD sampler_idx;
1817 BOOL projected;
1818 DWORD mask = 0;
1819
1820 /* All versions have a destination register */
1821 shader_glsl_append_dst(arg->buffer, arg);
1822
1823 /* 1.0-1.4: Use destination register as sampler source.
1824 * 2.0+: Use provided sampler source. */
1825 if (hex_version < WINED3DPS_VERSION(1,4)) {
1826 DWORD flags;
1827
1828 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1829 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1830
1831 if (flags & WINED3DTTFF_PROJECTED) {
1832 projected = TRUE;
1833 switch (flags & ~WINED3DTTFF_PROJECTED) {
1834 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
1835 case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
1836 case WINED3DTTFF_COUNT3: mask = WINED3DSP_WRITEMASK_2; break;
1837 case WINED3DTTFF_COUNT4:
1838 case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
1839 }
1840 } else {
1841 projected = FALSE;
1842 }
1843 } else if (hex_version < WINED3DPS_VERSION(2,0)) {
1844 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1845 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1846
1847 if (src_mod == WINED3DSPSM_DZ) {
1848 projected = TRUE;
1849 mask = WINED3DSP_WRITEMASK_2;
1850 } else if (src_mod == WINED3DSPSM_DW) {
1851 projected = TRUE;
1852 mask = WINED3DSP_WRITEMASK_3;
1853 } else {
1854 projected = FALSE;
1855 }
1856 } else {
1857 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1858 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1859 /* ps 2.0 texldp instruction always divides by the fourth component. */
1860 projected = TRUE;
1861 mask = WINED3DSP_WRITEMASK_3;
1862 } else {
1863 projected = FALSE;
1864 }
1865 }
1866
1867 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1868 shader_glsl_get_sample_function(sampler_type, projected, &sample_function);
1869 mask |= sample_function.coord_mask;
1870
1871 if (hex_version < WINED3DPS_VERSION(2,0)) {
1872 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
1873 } else {
1874 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1875 }
1876
1877 /* 1.0-1.3: Use destination register as coordinate source.
1878 1.4+: Use provided coordinate source register. */
1879 if (hex_version < WINED3DPS_VERSION(1,4)) {
1880 char coord_mask[6];
1881 shader_glsl_get_write_mask(mask, coord_mask);
1882 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
1883 sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
1884 } else {
1885 glsl_src_param_t coord_param;
1886 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
1887 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1888 glsl_src_param_t bias;
1889 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &bias);
1890
1891 shader_addline(arg->buffer, "%s(Psampler%u, %s, %s)%s);\n",
1892 sample_function.name, sampler_idx, coord_param.param_str,
1893 bias.param_str, dst_swizzle);
1894 } else {
1895 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
1896 sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
1897 }
1898 }
1899 }
1900
1901 void shader_glsl_texldl(SHADER_OPCODE_ARG* arg) {
1902 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*)arg->shader;
1903 glsl_sample_function_t sample_function;
1904 glsl_src_param_t coord_param, lod_param;
1905 char dst_swizzle[6];
1906 DWORD sampler_type;
1907 DWORD sampler_idx;
1908
1909 shader_glsl_append_dst(arg->buffer, arg);
1910 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
1911
1912 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1913 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1914 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
1915 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &coord_param);
1916
1917 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &lod_param);
1918
1919 if (shader_is_pshader_version(This->baseShader.hex_version)) {
1920 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
1921 * However, they seem to work just fine in fragment shaders as well. */
1922 WARN("Using %sLod in fragment shader.\n", sample_function.name);
1923 shader_addline(arg->buffer, "%sLod(Psampler%u, %s, %s)%s);\n",
1924 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
1925 } else {
1926 shader_addline(arg->buffer, "%sLod(Vsampler%u, %s, %s)%s);\n",
1927 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
1928 }
1929 }
1930
1931 void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
1932
1933 /* FIXME: Make this work for more than just 2D textures */
1934
1935 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1936 SHADER_BUFFER* buffer = arg->buffer;
1937 DWORD hex_version = This->baseShader.hex_version;
1938 DWORD write_mask;
1939 char dst_mask[6];
1940
1941 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1942 shader_glsl_get_write_mask(write_mask, dst_mask);
1943
1944 if (hex_version != WINED3DPS_VERSION(1,4)) {
1945 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1946 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
1947 } else {
1948 DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
1949 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1950 char dst_swizzle[6];
1951
1952 shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
1953
1954 if (src_mod == WINED3DSPSM_DZ) {
1955 glsl_src_param_t div_param;
1956 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1957 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
1958
1959 if (mask_size > 1) {
1960 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1961 } else {
1962 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1963 }
1964 } else if (src_mod == WINED3DSPSM_DW) {
1965 glsl_src_param_t div_param;
1966 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1967 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
1968
1969 if (mask_size > 1) {
1970 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
1971 } else {
1972 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
1973 }
1974 } else {
1975 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
1976 }
1977 }
1978 }
1979
1980 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
1981 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1982 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1983 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
1984 glsl_src_param_t src0_param;
1985 char dst_mask[6];
1986 glsl_sample_function_t sample_function;
1987 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1988 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1989 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
1990
1991 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1992
1993 shader_glsl_append_dst(arg->buffer, arg);
1994 shader_glsl_get_write_mask(arg->dst, dst_mask);
1995
1996 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
1997 * scalar, and projected sampling would require 4
1998 */
1999 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2000
2001 switch(count_bits(sample_function.coord_mask)) {
2002 case 1:
2003 shader_addline(arg->buffer, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2004 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2005 break;
2006
2007 case 2:
2008 shader_addline(arg->buffer, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2009 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2010 break;
2011
2012 case 3:
2013 shader_addline(arg->buffer, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2014 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2015 break;
2016 default:
2017 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function.coord_mask));
2018 }
2019 }
2020
2021 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2022 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2023 void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
2024 glsl_src_param_t src0_param;
2025 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2026 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2027 DWORD dst_mask;
2028 size_t mask_size;
2029
2030 dst_mask = shader_glsl_append_dst(arg->buffer, arg);
2031 mask_size = shader_glsl_get_write_mask_size(dst_mask);
2032 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2033
2034 if (mask_size > 1) {
2035 shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
2036 } else {
2037 shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
2038 }
2039 }
2040
2041 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2042 * Calculate the depth as dst.x / dst.y */
2043 void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
2044 glsl_dst_param_t dst_param;
2045
2046 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2047
2048 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2049 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2050 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2051 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2052 * >= 1.0 or < 0.0
2053 */
2054 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);
2055 }
2056
2057 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2058 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2059 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2060 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2061 */
2062 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
2063 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2064 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2065 glsl_src_param_t src0_param;
2066
2067 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2068
2069 shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
2070 shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2071 }
2072
2073 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2074 * Calculate the 1st of a 2-row matrix multiplication. */
2075 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
2076 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2077 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2078 SHADER_BUFFER* buffer = arg->buffer;
2079 glsl_src_param_t src0_param;
2080
2081 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2082 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2083 }
2084
2085 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2086 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2087 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
2088
2089 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2090 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2091 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2092 SHADER_BUFFER* buffer = arg->buffer;
2093 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2094 glsl_src_param_t src0_param;
2095
2096 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2097 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
2098 current_state->texcoord_w[current_state->current_row++] = reg;
2099 }
2100
2101 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
2102 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2103 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2104 SHADER_BUFFER* buffer = arg->buffer;
2105 glsl_src_param_t src0_param;
2106 char dst_mask[6];
2107
2108 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2109 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2110
2111 shader_glsl_append_dst(buffer, arg);
2112 shader_glsl_get_write_mask(arg->dst, dst_mask);
2113
2114 /* Sample the texture using the calculated coordinates */
2115 shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
2116 }
2117
2118 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2119 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2120 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
2121 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2122 glsl_src_param_t src0_param;
2123 char dst_mask[6];
2124 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2125 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2126 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2127 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2128 glsl_sample_function_t sample_function;
2129
2130 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2131 shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2132
2133 shader_glsl_append_dst(arg->buffer, arg);
2134 shader_glsl_get_write_mask(arg->dst, dst_mask);
2135 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2136
2137 /* Sample the texture using the calculated coordinates */
2138 shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2139
2140 current_state->current_row = 0;
2141 }
2142
2143 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2144 * Perform the 3rd row of a 3x3 matrix multiply */
2145 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
2146 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2147 glsl_src_param_t src0_param;
2148 char dst_mask[6];
2149 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2150 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2151 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2152
2153 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2154
2155 shader_glsl_append_dst(arg->buffer, arg);
2156 shader_glsl_get_write_mask(arg->dst, dst_mask);
2157 shader_addline(arg->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
2158
2159 current_state->current_row = 0;
2160 }
2161
2162 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2163 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2164 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
2165
2166 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2167 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2168 glsl_src_param_t src0_param;
2169 glsl_src_param_t src1_param;
2170 char dst_mask[6];
2171 SHADER_BUFFER* buffer = arg->buffer;
2172 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2173 DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2174 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2175 glsl_sample_function_t sample_function;
2176
2177 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2178 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
2179
2180 /* Perform the last matrix multiply operation */
2181 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2182 /* Reflection calculation */
2183 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
2184
2185 shader_glsl_append_dst(buffer, arg);
2186 shader_glsl_get_write_mask(arg->dst, dst_mask);
2187 shader_glsl_get_sample_function(stype, FALSE, &sample_function);
2188
2189 /* Sample the texture */
2190 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2191
2192 current_state->current_row = 0;
2193 }
2194
2195 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2196 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2197 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
2198
2199 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2200 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2201 SHADER_BUFFER* buffer = arg->buffer;
2202 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2203 glsl_src_param_t src0_param;
2204 char dst_mask[6];
2205 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2206 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2207 glsl_sample_function_t sample_function;
2208
2209 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2210
2211 /* Perform the last matrix multiply operation */
2212 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
2213
2214 /* Construct the eye-ray vector from w coordinates */
2215 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2216 current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
2217 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2218
2219 shader_glsl_append_dst(buffer, arg);
2220 shader_glsl_get_write_mask(arg->dst, dst_mask);
2221 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2222
2223 /* Sample the texture using the calculated coordinates */
2224 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2225
2226 current_state->current_row = 0;
2227 }
2228
2229 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2230 * Apply a fake bump map transform.
2231 * texbem is pshader <= 1.3 only, this saves a few version checks
2232 */
2233 void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
2234 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2235 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2236 char dst_swizzle[6];
2237 glsl_sample_function_t sample_function;
2238 glsl_src_param_t coord_param;
2239 DWORD sampler_type;
2240 DWORD sampler_idx;
2241 DWORD mask;
2242 DWORD flags;
2243 char coord_mask[6];
2244
2245 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2246 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
2247
2248 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2249 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2250 mask = sample_function.coord_mask;
2251
2252 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
2253
2254 shader_glsl_get_write_mask(mask, coord_mask);
2255
2256 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2257 * so we can't let the GL handle this.
2258 */
2259 if (flags & WINED3DTTFF_PROJECTED) {
2260 DWORD div_mask=0;
2261 char coord_div_mask[3];
2262 switch (flags & ~WINED3DTTFF_PROJECTED) {
2263 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2264 case WINED3DTTFF_COUNT2: div_mask = WINED3DSP_WRITEMASK_1; break;
2265 case WINED3DTTFF_COUNT3: div_mask = WINED3DSP_WRITEMASK_2; break;
2266 case WINED3DTTFF_COUNT4:
2267 case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
2268 }
2269 shader_glsl_get_write_mask(div_mask, coord_div_mask);
2270 shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
2271 }
2272
2273 shader_glsl_append_dst(arg->buffer, arg);
2274 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
2275 if(arg->opcode->opcode == WINED3DSIO_TEXBEML) {
2276 glsl_src_param_t luminance_param;
2277 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &luminance_param);
2278 shader_addline(arg->buffer, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )*(%s * luminancescale + luminanceoffset))%s);\n",
2279 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask,
2280 luminance_param.param_str, dst_swizzle);
2281 } else {
2282 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )%s);\n",
2283 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask, dst_swizzle);
2284 }
2285 }
2286
2287 void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
2288 glsl_src_param_t src0_param, src1_param;
2289
2290 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
2291 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
2292
2293 shader_glsl_append_dst(arg->buffer, arg);
2294 shader_addline(arg->buffer, "%s + bumpenvmat * %s);\n",
2295 src0_param.param_str, src1_param.param_str);
2296 }
2297
2298 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2299 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2300 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
2301
2302 glsl_src_param_t src0_param;
2303 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2304 char dst_mask[6];
2305
2306 shader_glsl_append_dst(arg->buffer, arg);
2307 shader_glsl_get_write_mask(arg->dst, dst_mask);
2308 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2309
2310 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2311 }
2312
2313 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2314 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2315 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
2316 glsl_src_param_t src0_param;
2317 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2318 char dst_mask[6];
2319
2320 shader_glsl_append_dst(arg->buffer, arg);
2321 shader_glsl_get_write_mask(arg->dst, dst_mask);
2322 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2323
2324 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2325 }
2326
2327 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2328 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2329 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
2330 glsl_src_param_t src0_param;
2331 char dst_mask[6];
2332 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2333 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2334 glsl_sample_function_t sample_function;
2335
2336 shader_glsl_append_dst(arg->buffer, arg);
2337 shader_glsl_get_write_mask(arg->dst, dst_mask);
2338 shader_glsl_get_sample_function(sampler_type, FALSE, &sample_function);
2339 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
2340
2341 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
2342 }
2343
2344 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2345 * If any of the first 3 components are < 0, discard this pixel */
2346 void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
2347 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2348 DWORD hex_version = This->baseShader.hex_version;
2349 glsl_dst_param_t dst_param;
2350
2351 /* The argument is a destination parameter, and no writemasks are allowed */
2352 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2353 if((hex_version >= WINED3DPS_VERSION(2,0))) {
2354 /* 2.0 shaders compare all 4 components in texkill */
2355 shader_addline(arg->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
2356 } else {
2357 /* 1.X shaders only compare the first 3 components, propably due to the nature of the texkill
2358 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2359 * 4 components are defined, only the first 3 are used
2360 */
2361 shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
2362 }
2363 }
2364
2365 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2366 * dst = dot2(src0, src1) + src2 */
2367 void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
2368 glsl_src_param_t src0_param;
2369 glsl_src_param_t src1_param;
2370 glsl_src_param_t src2_param;
2371 DWORD write_mask;
2372 size_t mask_size;
2373
2374 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2375 mask_size = shader_glsl_get_write_mask_size(write_mask);
2376
2377 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
2378 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
2379 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
2380
2381 shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
2382 }
2383
2384 void pshader_glsl_input_pack(
2385 SHADER_BUFFER* buffer,
2386 semantic* semantics_in) {
2387
2388 unsigned int i;
2389
2390 for (i = 0; i < MAX_REG_INPUT; i++) {
2391
2392 DWORD usage_token = semantics_in[i].usage;
2393 DWORD register_token = semantics_in[i].reg;
2394 DWORD usage, usage_idx;
2395 char reg_mask[6];
2396
2397 /* Uninitialized */
2398 if (!usage_token) continue;
2399 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2400 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2401 shader_glsl_get_write_mask(register_token, reg_mask);
2402
2403 switch(usage) {
2404
2405 case WINED3DDECLUSAGE_COLOR:
2406 if (usage_idx == 0)
2407 shader_addline(buffer, "IN%u%s = vec4(gl_Color)%s;\n",
2408 i, reg_mask, reg_mask);
2409 else if (usage_idx == 1)
2410 shader_addline(buffer, "IN%u%s = vec4(gl_SecondaryColor)%s;\n",
2411 i, reg_mask, reg_mask);
2412 else
2413 shader_addline(buffer, "IN%u%s = vec4(unsupported_color_input)%s;\n",
2414 i, reg_mask, reg_mask);
2415 break;
2416
2417 case WINED3DDECLUSAGE_TEXCOORD:
2418 shader_addline(buffer, "IN%u%s = vec4(gl_TexCoord[%u])%s;\n",
2419 i, reg_mask, usage_idx, reg_mask );
2420 break;
2421
2422 case WINED3DDECLUSAGE_FOG:
2423 shader_addline(buffer, "IN%u%s = vec4(gl_FogFragCoord)%s;\n",
2424 i, reg_mask, reg_mask);
2425 break;
2426
2427 default:
2428 shader_addline(buffer, "IN%u%s = vec4(unsupported_input)%s;\n",
2429 i, reg_mask, reg_mask);
2430 }
2431 }
2432 }
2433
2434 /*********************************************
2435 * Vertex Shader Specific Code begins here
2436 ********************************************/
2437
2438 void vshader_glsl_output_unpack(
2439 SHADER_BUFFER* buffer,
2440 semantic* semantics_out) {
2441
2442 unsigned int i;
2443
2444 for (i = 0; i < MAX_REG_OUTPUT; i++) {
2445
2446 DWORD usage_token = semantics_out[i].usage;
2447 DWORD register_token = semantics_out[i].reg;
2448 DWORD usage, usage_idx;
2449 char reg_mask[6];
2450
2451 /* Uninitialized */
2452 if (!usage_token) continue;
2453
2454 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2455 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2456 shader_glsl_get_write_mask(register_token, reg_mask);
2457
2458 switch(usage) {
2459
2460 case WINED3DDECLUSAGE_COLOR:
2461 if (usage_idx == 0)
2462 shader_addline(buffer, "gl_FrontColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2463 else if (usage_idx == 1)
2464 shader_addline(buffer, "gl_FrontSecondaryColor%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2465 else
2466 shader_addline(buffer, "unsupported_color_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2467 break;
2468
2469 case WINED3DDECLUSAGE_POSITION:
2470 shader_addline(buffer, "gl_Position%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2471 break;
2472
2473 case WINED3DDECLUSAGE_TEXCOORD:
2474 shader_addline(buffer, "gl_TexCoord[%u]%s = OUT%u%s;\n",
2475 usage_idx, reg_mask, i, reg_mask);
2476 break;
2477
2478 case WINED3DDECLUSAGE_PSIZE:
2479 shader_addline(buffer, "gl_PointSize = OUT%u.x;\n", i);
2480 break;
2481
2482 case WINED3DDECLUSAGE_FOG:
2483 shader_addline(buffer, "gl_FogFragCoord = OUT%u%s;\n", i, reg_mask);
2484 break;
2485
2486 default:
2487 shader_addline(buffer, "unsupported_output%s = OUT%u%s;\n", reg_mask, i, reg_mask);
2488 }
2489 }
2490 }
2491
2492 static void add_glsl_program_entry(IWineD3DDeviceImpl *device, struct glsl_shader_prog_link *entry) {
2493 glsl_program_key_t *key;
2494
2495 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2496 key->vshader = entry->vshader;
2497 key->pshader = entry->pshader;
2498
2499 hash_table_put(device->glsl_program_lookup, key, entry);
2500 }
2501
2502 static struct glsl_shader_prog_link *get_glsl_program_entry(IWineD3DDeviceImpl *device,
2503 GLhandleARB vshader, GLhandleARB pshader) {
2504 glsl_program_key_t key;
2505
2506 key.vshader = vshader;
2507 key.pshader = pshader;
2508
2509 return (struct glsl_shader_prog_link *)hash_table_get(device->glsl_program_lookup, &key);
2510 }
2511
2512 void delete_glsl_program_entry(IWineD3DDevice *iface, struct glsl_shader_prog_link *entry) {
2513 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2514 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2515 glsl_program_key_t *key;
2516
2517 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2518 key->vshader = entry->vshader;
2519 key->pshader = entry->pshader;
2520 hash_table_remove(This->glsl_program_lookup, key);
2521
2522 GL_EXTCALL(glDeleteObjectARB(entry->programId));
2523 if (entry->vshader) list_remove(&entry->vshader_entry);
2524 if (entry->pshader) list_remove(&entry->pshader_entry);
2525 HeapFree(GetProcessHeap(), 0, entry->vuniformF_locations);
2526 HeapFree(GetProcessHeap(), 0, entry->puniformF_locations);
2527 HeapFree(GetProcessHeap(), 0, entry);
2528 }
2529
2530 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
2531 * It sets the programId on the current StateBlock (because it should be called
2532 * inside of the DrawPrimitive() part of the render loop).
2533 *
2534 * If a program for the given combination does not exist, create one, and store
2535 * the program in the hash table. If it creates a program, it will link the
2536 * given objects, too.
2537 */
2538 static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
2539 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2540 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2541 IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
2542 IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
2543 struct glsl_shader_prog_link *entry = NULL;
2544 GLhandleARB programId = 0;
2545 int i;
2546 char glsl_name[8];
2547
2548 GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
2549 GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
2550 entry = get_glsl_program_entry(This, vshader_id, pshader_id);
2551 if (entry) {
2552 This->stateBlock->glsl_program = entry;
2553 return;
2554 }
2555
2556 /* If we get to this point, then no matching program exists, so we create one */
2557 programId = GL_EXTCALL(glCreateProgramObjectARB());
2558 TRACE("Created new GLSL shader program %u\n", programId);
2559
2560 /* Create the entry */
2561 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
2562 entry->programId = programId;
2563 entry->vshader = vshader_id;
2564 entry->pshader = pshader_id;
2565 /* Add the hash table entry */
2566 add_glsl_program_entry(This, entry);
2567
2568 /* Set the current program */
2569 This->stateBlock->glsl_program = entry;
2570
2571 /* Attach GLSL vshader */
2572 if (vshader_id) {
2573 int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
2574 char tmp_name[10];
2575
2576 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id, programId);
2577 GL_EXTCALL(glAttachObjectARB(programId, vshader_id));
2578 checkGLcall("glAttachObjectARB");
2579
2580 /* Bind vertex attributes to a corresponding index number to match
2581 * the same index numbers as ARB_vertex_programs (makes loading
2582 * vertex attributes simpler). With this method, we can use the
2583 * exact same code to load the attributes later for both ARB and
2584 * GLSL shaders.
2585 *
2586 * We have to do this here because we need to know the Program ID
2587 * in order to make the bindings work, and it has to be done prior
2588 * to linking the GLSL program. */
2589 for (i = 0; i < max_attribs; ++i) {
2590 snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
2591 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
2592 }
2593 checkGLcall("glBindAttribLocationARB");
2594
2595 list_add_head(&((IWineD3DBaseShaderImpl *)vshader)->baseShader.linked_programs, &entry->vshader_entry);
2596 }
2597
2598 /* Attach GLSL pshader */
2599 if (pshader_id) {
2600 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
2601 GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
2602 checkGLcall("glAttachObjectARB");
2603
2604 list_add_head(&((IWineD3DBaseShaderImpl *)pshader)->baseShader.linked_programs, &entry->pshader_entry);
2605 }
2606
2607 /* Link the program */
2608 TRACE("Linking GLSL shader program %u\n", programId);
2609 GL_EXTCALL(glLinkProgramARB(programId));
2610 print_glsl_info_log(&GLINFO_LOCATION, programId);
2611
2612 entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
2613 for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
2614 snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
2615 entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2616 }
2617 entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
2618 for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
2619 snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
2620 entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
2621 }
2622 }
2623
2624 static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
2625 GLhandleARB program_id;
2626 GLhandleARB vshader_id, pshader_id;
2627 const char *blt_vshader[] = {
2628 "void main(void)\n"
2629 "{\n"
2630 " gl_Position = gl_Vertex;\n"
2631 " gl_FrontColor = vec4(1.0);\n"
2632 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
2633 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
2634 "}\n"
2635 };
2636
2637 const char *blt_pshader[] = {
2638 "uniform sampler2D sampler;\n"
2639 "void main(void)\n"
2640 "{\n"
2641 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
2642 "}\n"
2643 };
2644
2645 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
2646 GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
2647 GL_EXTCALL(glCompileShaderARB(vshader_id));
2648
2649 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
2650 GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
2651 GL_EXTCALL(glCompileShaderARB(pshader_id));
2652
2653 program_id = GL_EXTCALL(glCreateProgramObjectARB());
2654 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
2655 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
2656 GL_EXTCALL(glLinkProgramARB(program_id));
2657
2658 print_glsl_info_log(&GLINFO_LOCATION, program_id);
2659
2660 return program_id;
2661 }
2662
2663 static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
2664 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2665 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2666 GLhandleARB program_id = 0;
2667
2668 if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
2669 else This->stateBlock->glsl_program = NULL;
2670
2671 program_id = This->stateBlock->glsl_program ? This->stateBlock->glsl_program->programId : 0;
2672 if (program_id) TRACE("Using GLSL program %u\n", program_id);
2673 GL_EXTCALL(glUseProgramObjectARB(program_id));
2674 checkGLcall("glUseProgramObjectARB");
2675 }
2676
2677 static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
2678 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2679 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2680 static GLhandleARB program_id = 0;
2681 static GLhandleARB loc = -1;
2682
2683 if (!program_id) {
2684 program_id = create_glsl_blt_shader(gl_info);
2685 loc = GL_EXTCALL(glGetUniformLocationARB(program_id, "sampler"));
2686 }
2687
2688 GL_EXTCALL(glUseProgramObjectARB(program_id));
2689 GL_EXTCALL(glUniform1iARB(loc, 0));
2690 }
2691
2692 static void shader_glsl_cleanup(IWineD3DDevice *iface) {
2693 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2694 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2695 GL_EXTCALL(glUseProgramObjectARB(0));
2696 }
2697
2698 const shader_backend_t glsl_shader_backend = {
2699 &shader_glsl_select,
2700 &shader_glsl_select_depth_blt,
2701 &shader_glsl_load_constants,
2702 &shader_glsl_cleanup,
2703 &shader_glsl_color_correction
2704 };