?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Make the code aware of GL_ARB_texture_rectangle.
[wine/gsoc_dplay.git] / dlls / wined3d / glsl_shader.c
blobdd18709b0a47e173f48047a1e92e25ffa3d4e078
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 GLhandleARB programId) {
82
83 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
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, GLhandleARB programId) {
105 IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
106 GLhandleARB name_loc;
107 char sampler_name[20];
108 int i;
109
110 for (i = 0; i < MAX_VERTEX_SAMPLERS; ++i) {
111 snprintf(sampler_name, sizeof(sampler_name), "Vsampler%d", i);
112 name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
113 if (name_loc != -1) {
114 int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[MAX_FRAGMENT_SAMPLERS + i];
115 if (mapped_unit != -1 && mapped_unit < GL_LIMITS(combined_samplers)) {
116 TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
117 GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
118 checkGLcall("glUniform1iARB");
119 } else {
120 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name, mapped_unit);
121 }
122 }
123 }
124 }
125
126 /**
127 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
128 * When constant_list == NULL, it will load all the constants.
129 */
130 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info,
131 unsigned int max_constants, float* constants, GLhandleARB *constant_locations,
132 struct list *constant_list) {
133 constants_entry *constant;
134 local_constant* lconst;
135 GLhandleARB tmp_loc;
136 DWORD i, j, k;
137 DWORD *idx;
138
139 if (TRACE_ON(d3d_shader)) {
140 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
141 idx = constant->idx;
142 j = constant->count;
143 while (j--) {
144 i = *idx++;
145 tmp_loc = constant_locations[i];
146 if (tmp_loc != -1) {
147 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
148 constants[i * 4 + 0], constants[i * 4 + 1],
149 constants[i * 4 + 2], constants[i * 4 + 3]);
150 }
151 }
152 }
153 }
154
155 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
156 if(WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 &&
157 shader_is_pshader_version(This->baseShader.hex_version)) {
158 float lcl_const[4];
159
160 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
161 idx = constant->idx;
162 j = constant->count;
163 while (j--) {
164 i = *idx++;
165 tmp_loc = constant_locations[i];
166 if (tmp_loc != -1) {
167 /* We found this uniform name in the program - go ahead and send the data */
168 k = i * 4;
169 if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
170 else if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
171 else lcl_const[0] = constants[k + 0];
172 if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
173 else if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
174 else lcl_const[1] = constants[k + 1];
175 if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
176 else if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
177 else lcl_const[2] = constants[k + 2];
178 if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
179 else if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
180 else lcl_const[3] = constants[k + 3];
181
182 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, lcl_const));
183 }
184 }
185 }
186 } else {
187 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
188 idx = constant->idx;
189 j = constant->count;
190 while (j--) {
191 i = *idx++;
192 tmp_loc = constant_locations[i];
193 if (tmp_loc != -1) {
194 /* We found this uniform name in the program - go ahead and send the data */
195 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, constants + (i * 4)));
196 }
197 }
198 }
199 }
200 checkGLcall("glUniform4fvARB()");
201
202 if(!This->baseShader.load_local_constsF) {
203 TRACE("No need to load local float constants for this shader\n");
204 return;
205 }
206
207 /* Load immediate constants */
208 if (TRACE_ON(d3d_shader)) {
209 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
210 tmp_loc = constant_locations[lconst->idx];
211 if (tmp_loc != -1) {
212 GLfloat* values = (GLfloat*)lconst->value;
213 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
214 values[0], values[1], values[2], values[3]);
215 }
216 }
217 }
218 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
219 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
220 tmp_loc = constant_locations[lconst->idx];
221 if (tmp_loc != -1) {
222 /* We found this uniform name in the program - go ahead and send the data */
223 GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, (GLfloat*)lconst->value));
224 }
225 }
226 checkGLcall("glUniform4fvARB()");
227 }
228
229 /**
230 * Loads integer constants (aka uniforms) into the currently set GLSL program.
231 * When @constants_set == NULL, it will load all the constants.
232 */
233 static void shader_glsl_load_constantsI(
234 IWineD3DBaseShaderImpl* This,
235 WineD3D_GL_Info *gl_info,
236 GLhandleARB programId,
237 GLhandleARB locations[MAX_CONST_I],
238 unsigned max_constants,
239 int* constants,
240 BOOL* constants_set) {
241
242 int i;
243 struct list* ptr;
244
245 for (i=0; i<max_constants; ++i) {
246 if (NULL == constants_set || constants_set[i]) {
247
248 TRACE_(d3d_constants)("Loading constants %i: %i, %i, %i, %i\n",
249 i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
250
251 /* We found this uniform name in the program - go ahead and send the data */
252 GL_EXTCALL(glUniform4ivARB(locations[i], 1, &constants[i*4]));
253 checkGLcall("glUniform4ivARB");
254 }
255 }
256
257 /* Load immediate constants */
258 ptr = list_head(&This->baseShader.constantsI);
259 while (ptr) {
260 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
261 unsigned int idx = lconst->idx;
262 GLint* values = (GLint*) lconst->value;
263
264 TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
265 values[0], values[1], values[2], values[3]);
266
267 /* We found this uniform name in the program - go ahead and send the data */
268 GL_EXTCALL(glUniform4ivARB(locations[idx], 1, values));
269 checkGLcall("glUniform4ivARB");
270 ptr = list_next(&This->baseShader.constantsI, ptr);
271 }
272 }
273
274 /**
275 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
276 * When @constants_set == NULL, it will load all the constants.
277 */
278 static void shader_glsl_load_constantsB(
279 IWineD3DBaseShaderImpl* This,
280 WineD3D_GL_Info *gl_info,
281 GLhandleARB programId,
282 unsigned max_constants,
283 BOOL* constants,
284 BOOL* constants_set) {
285
286 GLhandleARB tmp_loc;
287 int i;
288 char tmp_name[8];
289 char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
290 const char* prefix = is_pshader? "PB":"VB";
291 struct list* ptr;
292
293 for (i=0; i<max_constants; ++i) {
294 if (NULL == constants_set || constants_set[i]) {
295
296 TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i]);
297
298 /* TODO: Benchmark and see if it would be beneficial to store the
299 * locations of the constants to avoid looking up each time */
300 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
301 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
302 if (tmp_loc != -1) {
303 /* We found this uniform name in the program - go ahead and send the data */
304 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i]));
305 checkGLcall("glUniform1ivARB");
306 }
307 }
308 }
309
310 /* Load immediate constants */
311 ptr = list_head(&This->baseShader.constantsB);
312 while (ptr) {
313 local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
314 unsigned int idx = lconst->idx;
315 GLint* values = (GLint*) lconst->value;
316
317 TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
318
319 snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, idx);
320 tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
321 if (tmp_loc != -1) {
322 /* We found this uniform name in the program - go ahead and send the data */
323 GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, values));
324 checkGLcall("glUniform1ivARB");
325 }
326 ptr = list_next(&This->baseShader.constantsB, ptr);
327 }
328 }
329
330
331
332 /**
333 * Loads the app-supplied constants into the currently set GLSL program.
334 */
335 void shader_glsl_load_constants(
336 IWineD3DDevice* device,
337 char usePixelShader,
338 char useVertexShader) {
339
340 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
341 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
342 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
343
344 GLhandleARB *constant_locations;
345 struct list *constant_list;
346 GLhandleARB programId;
347 struct glsl_shader_prog_link *prog = stateBlock->glsl_program;
348
349 if (!prog) {
350 /* No GLSL program set - nothing to do. */
351 return;
352 }
353 programId = prog->programId;
354
355 if (useVertexShader) {
356 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
357
358 constant_locations = prog->vuniformF_locations;
359 constant_list = &stateBlock->set_vconstantsF;
360
361 /* Load DirectX 9 float constants/uniforms for vertex shader */
362 shader_glsl_load_constantsF(vshader, gl_info, GL_LIMITS(vshader_constantsF),
363 stateBlock->vertexShaderConstantF, constant_locations, constant_list);
364
365 /* Load DirectX 9 integer constants/uniforms for vertex shader */
366 shader_glsl_load_constantsI(vshader, gl_info, programId,
367 prog->vuniformI_locations, MAX_CONST_I,
368 stateBlock->vertexShaderConstantI,
369 stateBlock->changed.vertexShaderConstantsI);
370
371 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
372 shader_glsl_load_constantsB(vshader, gl_info, programId, MAX_CONST_B,
373 stateBlock->vertexShaderConstantB,
374 stateBlock->changed.vertexShaderConstantsB);
375
376 /* Upload the position fixup params */
377 GL_EXTCALL(glUniform4fvARB(prog->posFixup_location, 1, &deviceImpl->posFixup[0]));
378 checkGLcall("glUniform4fvARB");
379 }
380
381 if (usePixelShader) {
382
383 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
384
385 constant_locations = prog->puniformF_locations;
386 constant_list = &stateBlock->set_pconstantsF;
387
388 /* Load DirectX 9 float constants/uniforms for pixel shader */
389 shader_glsl_load_constantsF(pshader, gl_info, GL_LIMITS(pshader_constantsF),
390 stateBlock->pixelShaderConstantF, constant_locations, constant_list);
391
392 /* Load DirectX 9 integer constants/uniforms for pixel shader */
393 shader_glsl_load_constantsI(pshader, gl_info, programId,
394 prog->puniformI_locations, MAX_CONST_I,
395 stateBlock->pixelShaderConstantI,
396 stateBlock->changed.pixelShaderConstantsI);
397
398 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
399 shader_glsl_load_constantsB(pshader, gl_info, programId, MAX_CONST_B,
400 stateBlock->pixelShaderConstantB,
401 stateBlock->changed.pixelShaderConstantsB);
402
403 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
404 * It can't be 0 for a valid texbem instruction.
405 */
406 if(((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat != -1) {
407 float *data = (float *) &stateBlock->textureState[(int) ((IWineD3DPixelShaderImpl *) pshader)->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
408 GL_EXTCALL(glUniformMatrix2fvARB(prog->bumpenvmat_location, 1, 0, data));
409 checkGLcall("glUniformMatrix2fvARB");
410
411 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
412 * is set too, so we can check that in the needsbumpmat check
413 */
414 if(((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams != -1) {
415 int stage = ((IWineD3DPixelShaderImpl *) pshader)->baseShader.reg_maps.luminanceparams;
416 GLfloat *scale = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLSCALE];
417 GLfloat *offset = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET];
418
419 GL_EXTCALL(glUniform1fvARB(prog->luminancescale_location, 1, scale));
420 checkGLcall("glUniform1fvARB");
421 GL_EXTCALL(glUniform1fvARB(prog->luminanceoffset_location, 1, offset));
422 checkGLcall("glUniform1fvARB");
423 }
424 } else if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
425 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
426 float comparison[4];
427 float mul_low[4];
428
429 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
430 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
431 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
432
433 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
434 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
435 } else {
436 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
437 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
438
439 mul_low[0] = 1.0; mul_low[1] = 1.0;
440 mul_low[2] = 1.0; mul_low[3] = 1.0;
441 }
442
443 GL_EXTCALL(glUniform4fvARB(prog->srgb_comparison_location, 1, comparison));
444 GL_EXTCALL(glUniform4fvARB(prog->srgb_mul_low_location, 1, mul_low));
445 }
446 if(((IWineD3DPixelShaderImpl *) pshader)->vpos_uniform) {
447 float correction_params[4];
448 if(deviceImpl->render_offscreen) {
449 correction_params[0] = 0.0;
450 correction_params[1] = 1.0;
451 } else {
452 /* position is window relative, not viewport relative */
453 correction_params[0] = ((IWineD3DSurfaceImpl *) deviceImpl->render_targets[0])->currentDesc.Height;
454 correction_params[1] = -1.0;
455 }
456 GL_EXTCALL(glUniform4fvARB(prog->ycorrection_location, 1, correction_params));
457 }
458 }
459 }
460
461 /** Generate the variable & register declarations for the GLSL output target */
462 void shader_generate_glsl_declarations(
463 IWineD3DBaseShader *iface,
464 shader_reg_maps* reg_maps,
465 SHADER_BUFFER* buffer,
466 WineD3D_GL_Info* gl_info) {
467
468 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
469 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
470 int i;
471 unsigned int extra_constants_needed = 0;
472 local_constant* lconst;
473
474 /* There are some minor differences between pixel and vertex shaders */
475 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
476 char prefix = pshader ? 'P' : 'V';
477
478 /* Prototype the subroutines */
479 for (i = 0; i < This->baseShader.limits.label; i++) {
480 if (reg_maps->labels[i])
481 shader_addline(buffer, "void subroutine%u();\n", i);
482 }
483
484 /* Declare the constants (aka uniforms) */
485 if (This->baseShader.limits.constant_float > 0) {
486 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
487 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
488 shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
489 }
490
491 if (This->baseShader.limits.constant_int > 0)
492 shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
493
494 if (This->baseShader.limits.constant_bool > 0)
495 shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
496
497 if(!pshader) {
498 shader_addline(buffer, "uniform vec4 posFixup;\n");
499 /* Predeclaration; This function is added at link time based on the pixel shader.
500 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
501 * that. We know the input to the reorder function at vertex shader compile time, so
502 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
503 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
504 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
505 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
506 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
507 * inout.
508 */
509 if(This->baseShader.hex_version >= WINED3DVS_VERSION(3, 0)) {
510 shader_addline(buffer, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT);
511 } else {
512 shader_addline(buffer, "void order_ps_input();\n");
513 }
514 } else {
515 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
516
517 if(reg_maps->bumpmat != -1) {
518 shader_addline(buffer, "uniform mat2 bumpenvmat;\n");
519 if(reg_maps->luminanceparams) {
520 shader_addline(buffer, "uniform float luminancescale;\n");
521 shader_addline(buffer, "uniform float luminanceoffset;\n");
522 extra_constants_needed++;
523 }
524 extra_constants_needed++;
525 }
526
527 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
528 ps_impl->srgb_enabled = 1;
529 if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
530 shader_addline(buffer, "uniform vec4 srgb_mul_low;\n");
531 shader_addline(buffer, "uniform vec4 srgb_comparison;\n");
532 ps_impl->srgb_mode_hardcoded = 0;
533 extra_constants_needed++;
534 } else {
535 ps_impl->srgb_mode_hardcoded = 1;
536 shader_addline(buffer, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
537 srgb_mul_low, srgb_mul_low, srgb_mul_low, srgb_mul_low);
538 shader_addline(buffer, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
539 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
540 }
541 } else {
542 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
543
544 /* Do not write any srgb fixup into the shader to save shader size and processing time.
545 * As a consequence, we can't toggle srgb write on without recompilation
546 */
547 ps_impl->srgb_enabled = 0;
548 ps_impl->srgb_mode_hardcoded = 1;
549 }
550 if(reg_maps->vpos || reg_maps->usesdsy) {
551 if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
552 shader_addline(buffer, "uniform vec4 ycorrection;\n");
553 ((IWineD3DPixelShaderImpl *) This)->vpos_uniform = 1;
554 extra_constants_needed++;
555 } else {
556 /* This happens because we do not have proper tracking of the constant registers that are
557 * actually used, only the max limit of the shader version
558 */
559 FIXME("Cannot find a free uniform for vpos correction params\n");
560 shader_addline(buffer, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
561 device->render_offscreen ? 0.0 : ((IWineD3DSurfaceImpl *) device->render_targets[0])->currentDesc.Height,
562 device->render_offscreen ? 1.0 : -1.0);
563 }
564 shader_addline(buffer, "vec4 vpos;\n");
565 }
566 }
567
568 /* Declare texture samplers */
569 for (i = 0; i < This->baseShader.limits.sampler; i++) {
570 if (reg_maps->samplers[i]) {
571
572 DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
573 switch (stype) {
574
575 case WINED3DSTT_1D:
576 shader_addline(buffer, "uniform sampler1D %csampler%u;\n", prefix, i);
577 break;
578 case WINED3DSTT_2D:
579 if(device->stateBlock->textures[i] &&
580 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[i]) == GL_TEXTURE_RECTANGLE_ARB) {
581 shader_addline(buffer, "uniform sampler2DRect %csampler%u;\n", prefix, i);
582 } else {
583 shader_addline(buffer, "uniform sampler2D %csampler%u;\n", prefix, i);
584 }
585 break;
586 case WINED3DSTT_CUBE:
587 shader_addline(buffer, "uniform samplerCube %csampler%u;\n", prefix, i);
588 break;
589 case WINED3DSTT_VOLUME:
590 shader_addline(buffer, "uniform sampler3D %csampler%u;\n", prefix, i);
591 break;
592 default:
593 shader_addline(buffer, "uniform unsupported_sampler %csampler%u;\n", prefix, i);
594 FIXME("Unrecognized sampler type: %#x\n", stype);
595 break;
596 }
597 }
598 }
599
600 /* Declare address variables */
601 for (i = 0; i < This->baseShader.limits.address; i++) {
602 if (reg_maps->address[i])
603 shader_addline(buffer, "ivec4 A%d;\n", i);
604 }
605
606 /* Declare texture coordinate temporaries and initialize them */
607 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
608 if (reg_maps->texcoord[i])
609 shader_addline(buffer, "vec4 T%u = gl_TexCoord[%u];\n", i, i);
610 }
611
612 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
613 * helper function shader that is linked in at link time
614 */
615 if(pshader && This->baseShader.hex_version >= WINED3DVS_VERSION(3, 0)) {
616 if(use_vs(device)) {
617 shader_addline(buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
618 } else {
619 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
620 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
621 * pixel shader that reads the fixed function color into the packed input registers.
622 */
623 shader_addline(buffer, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
624 }
625 }
626
627 /* Declare output register temporaries */
628 if(This->baseShader.limits.packed_output) {
629 shader_addline(buffer, "vec4 OUT[%u];\n", This->baseShader.limits.packed_output);
630 }
631
632 /* Declare temporary variables */
633 for(i = 0; i < This->baseShader.limits.temporary; i++) {
634 if (reg_maps->temporary[i])
635 shader_addline(buffer, "vec4 R%u;\n", i);
636 }
637
638 /* Declare attributes */
639 for (i = 0; i < This->baseShader.limits.attributes; i++) {
640 if (reg_maps->attributes[i])
641 shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
642 }
643
644 /* Declare loop registers aLx */
645 for (i = 0; i < reg_maps->loop_depth; i++) {
646 shader_addline(buffer, "int aL%u;\n", i);
647 shader_addline(buffer, "int tmpInt%u;\n", i);
648 }
649
650 /* Temporary variables for matrix operations */
651 shader_addline(buffer, "vec4 tmp0;\n");
652 shader_addline(buffer, "vec4 tmp1;\n");
653
654 /* Hardcodeable local constants */
655 if(!This->baseShader.load_local_constsF) {
656 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
657 float *value = (float *) lconst->value;
658 shader_addline(buffer, "const vec4 LC%u = vec4(%f, %f, %f, %f);\n", lconst->idx,
659 value[0], value[1], value[2], value[3]);
660 }
661 }
662
663 /* Start the main program */
664 shader_addline(buffer, "void main() {\n");
665 if(pshader && reg_maps->vpos) {
666 shader_addline(buffer, "vpos = vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1) - 0.5;\n");
667 }
668 }
669
670 /*****************************************************************************
671 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
672 *
673 * For more information, see http://wiki.winehq.org/DirectX-Shaders
674 ****************************************************************************/
675
676 /* Prototypes */
677 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
678 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
679
680 /** Used for opcode modifiers - They multiply the result by the specified amount */
681 static const char * const shift_glsl_tab[] = {
682 "", /* 0 (none) */
683 "2.0 * ", /* 1 (x2) */
684 "4.0 * ", /* 2 (x4) */
685 "8.0 * ", /* 3 (x8) */
686 "16.0 * ", /* 4 (x16) */
687 "32.0 * ", /* 5 (x32) */
688 "", /* 6 (x64) */
689 "", /* 7 (x128) */
690 "", /* 8 (d256) */
691 "", /* 9 (d128) */
692 "", /* 10 (d64) */
693 "", /* 11 (d32) */
694 "0.0625 * ", /* 12 (d16) */
695 "0.125 * ", /* 13 (d8) */
696 "0.25 * ", /* 14 (d4) */
697 "0.5 * " /* 15 (d2) */
698 };
699
700 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
701 static void shader_glsl_gen_modifier (
702 const DWORD instr,
703 const char *in_reg,
704 const char *in_regswizzle,
705 char *out_str) {
706
707 out_str[0] = 0;
708
709 if (instr == WINED3DSIO_TEXKILL)
710 return;
711
712 switch (instr & WINED3DSP_SRCMOD_MASK) {
713 case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
714 case WINED3DSPSM_DW:
715 case WINED3DSPSM_NONE:
716 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
717 break;
718 case WINED3DSPSM_NEG:
719 sprintf(out_str, "-%s%s", in_reg, in_regswizzle);
720 break;
721 case WINED3DSPSM_NOT:
722 sprintf(out_str, "!%s%s", in_reg, in_regswizzle);
723 break;
724 case WINED3DSPSM_BIAS:
725 sprintf(out_str, "(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
726 break;
727 case WINED3DSPSM_BIASNEG:
728 sprintf(out_str, "-(%s%s - vec4(0.5)%s)", in_reg, in_regswizzle, in_regswizzle);
729 break;
730 case WINED3DSPSM_SIGN:
731 sprintf(out_str, "(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
732 break;
733 case WINED3DSPSM_SIGNNEG:
734 sprintf(out_str, "-(2.0 * (%s%s - 0.5))", in_reg, in_regswizzle);
735 break;
736 case WINED3DSPSM_COMP:
737 sprintf(out_str, "(1.0 - %s%s)", in_reg, in_regswizzle);
738 break;
739 case WINED3DSPSM_X2:
740 sprintf(out_str, "(2.0 * %s%s)", in_reg, in_regswizzle);
741 break;
742 case WINED3DSPSM_X2NEG:
743 sprintf(out_str, "-(2.0 * %s%s)", in_reg, in_regswizzle);
744 break;
745 case WINED3DSPSM_ABS:
746 sprintf(out_str, "abs(%s%s)", in_reg, in_regswizzle);
747 break;
748 case WINED3DSPSM_ABSNEG:
749 sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
750 break;
751 default:
752 FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
753 sprintf(out_str, "%s%s", in_reg, in_regswizzle);
754 }
755 }
756
757 static BOOL constant_is_local(IWineD3DBaseShaderImpl* This, DWORD reg) {
758 local_constant* lconst;
759
760 if(This->baseShader.load_local_constsF) return FALSE;
761 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
762 if(lconst->idx == reg) return TRUE;
763 }
764 return FALSE;
765
766 }
767
768 /** Writes the GLSL variable name that corresponds to the register that the
769 * DX opcode parameter is trying to access */
770 static void shader_glsl_get_register_name(
771 const DWORD param,
772 const DWORD addr_token,
773 char* regstr,
774 BOOL* is_color,
775 SHADER_OPCODE_ARG* arg) {
776
777 /* oPos, oFog and oPts in D3D */
778 static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
779
780 DWORD reg = param & WINED3DSP_REGNUM_MASK;
781 DWORD regtype = shader_get_regtype(param);
782 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
783 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
784 WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
785
786 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
787 char tmpStr[50];
788
789 *is_color = FALSE;
790
791 switch (regtype) {
792 case WINED3DSPR_TEMP:
793 sprintf(tmpStr, "R%u", reg);
794 break;
795 case WINED3DSPR_INPUT:
796 if (pshader) {
797 /* Pixel shaders >= 3.0 */
798 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3) {
799 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
800 glsl_src_param_t rel_param;
801 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
802
803 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
804 * operation there
805 */
806 if(((IWineD3DPixelShaderImpl *) This)->input_reg_map[reg]) {
807 sprintf(tmpStr, "IN[%s + %u]", rel_param.param_str,
808 ((IWineD3DPixelShaderImpl *) This)->input_reg_map[reg]);
809 } else {
810 sprintf(tmpStr, "IN[%s]", rel_param.param_str);
811 }
812 } else {
813 sprintf(tmpStr, "IN[%u]",
814 ((IWineD3DPixelShaderImpl *) This)->input_reg_map[reg]);
815 }
816 } else {
817 if (reg==0)
818 strcpy(tmpStr, "gl_Color");
819 else
820 strcpy(tmpStr, "gl_SecondaryColor");
821 }
822 } else {
823 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
824 *is_color = TRUE;
825 sprintf(tmpStr, "attrib%u", reg);
826 }
827 break;
828 case WINED3DSPR_CONST:
829 {
830 const char* prefix = pshader? "PC":"VC";
831
832 /* Relative addressing */
833 if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
834
835 /* Relative addressing on shaders 2.0+ have a relative address token,
836 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
837 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
838 glsl_src_param_t rel_param;
839 shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
840 if(reg) {
841 sprintf(tmpStr, "%s[%s + %u]", prefix, rel_param.param_str, reg);
842 } else {
843 sprintf(tmpStr, "%s[%s]", prefix, rel_param.param_str);
844 }
845 } else {
846 if(reg) {
847 sprintf(tmpStr, "%s[A0.x + %u]", prefix, reg);
848 } else {
849 sprintf(tmpStr, "%s[A0.x]", prefix);
850 }
851 }
852
853 } else {
854 if(constant_is_local(This, reg)) {
855 sprintf(tmpStr, "LC%u", reg);
856 } else {
857 sprintf(tmpStr, "%s[%u]", prefix, reg);
858 }
859 }
860
861 break;
862 }
863 case WINED3DSPR_CONSTINT:
864 if (pshader)
865 sprintf(tmpStr, "PI[%u]", reg);
866 else
867 sprintf(tmpStr, "VI[%u]", reg);
868 break;
869 case WINED3DSPR_CONSTBOOL:
870 if (pshader)
871 sprintf(tmpStr, "PB[%u]", reg);
872 else
873 sprintf(tmpStr, "VB[%u]", reg);
874 break;
875 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
876 if (pshader) {
877 sprintf(tmpStr, "T%u", reg);
878 } else {
879 sprintf(tmpStr, "A%u", reg);
880 }
881 break;
882 case WINED3DSPR_LOOP:
883 sprintf(tmpStr, "aL%u", This->baseShader.cur_loop_regno - 1);
884 break;
885 case WINED3DSPR_SAMPLER:
886 if (pshader)
887 sprintf(tmpStr, "Psampler%u", reg);
888 else
889 sprintf(tmpStr, "Vsampler%u", reg);
890 break;
891 case WINED3DSPR_COLOROUT:
892 if (reg >= GL_LIMITS(buffers)) {
893 WARN("Write to render target %u, only %d supported\n", reg, 4);
894 }
895 if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
896 sprintf(tmpStr, "gl_FragData[%u]", reg);
897 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
898 sprintf(tmpStr, "gl_FragColor");
899 }
900 break;
901 case WINED3DSPR_RASTOUT:
902 sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
903 break;
904 case WINED3DSPR_DEPTHOUT:
905 sprintf(tmpStr, "gl_FragDepth");
906 break;
907 case WINED3DSPR_ATTROUT:
908 if (reg == 0) {
909 sprintf(tmpStr, "gl_FrontColor");
910 } else {
911 sprintf(tmpStr, "gl_FrontSecondaryColor");
912 }
913 break;
914 case WINED3DSPR_TEXCRDOUT:
915 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
916 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
917 sprintf(tmpStr, "OUT[%u]", reg);
918 else
919 sprintf(tmpStr, "gl_TexCoord[%u]", reg);
920 break;
921 case WINED3DSPR_MISCTYPE:
922 if (reg == 0) {
923 /* vPos */
924 sprintf(tmpStr, "vpos");
925 } else if (reg == 1){
926 /* Note that gl_FrontFacing is a bool, while vFace is
927 * a float for which the sign determines front/back
928 */
929 sprintf(tmpStr, "(gl_FrontFacing ? 1.0 : -1.0)");
930 } else {
931 FIXME("Unhandled misctype register %d\n", reg);
932 sprintf(tmpStr, "unrecognized_register");
933 }
934 break;
935 default:
936 FIXME("Unhandled register name Type(%d)\n", regtype);
937 sprintf(tmpStr, "unrecognized_register");
938 break;
939 }
940
941 strcat(regstr, tmpStr);
942 }
943
944 /* Get the GLSL write mask for the destination register */
945 static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
946 char *ptr = write_mask;
947 DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
948
949 if (shader_is_scalar(param)) {
950 mask = WINED3DSP_WRITEMASK_0;
951 } else {
952 *ptr++ = '.';
953 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
954 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
955 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
956 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
957 }
958
959 *ptr = '\0';
960
961 return mask;
962 }
963
964 static size_t shader_glsl_get_write_mask_size(DWORD write_mask) {
965 size_t size = 0;
966
967 if (write_mask & WINED3DSP_WRITEMASK_0) ++size;
968 if (write_mask & WINED3DSP_WRITEMASK_1) ++size;
969 if (write_mask & WINED3DSP_WRITEMASK_2) ++size;
970 if (write_mask & WINED3DSP_WRITEMASK_3) ++size;
971
972 return size;
973 }
974
975 static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
976 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
977 * but addressed as "rgba". To fix this we need to swap the register's x
978 * and z components. */
979 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
980 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
981 char *ptr = swizzle_str;
982
983 if (!shader_is_scalar(param)) {
984 *ptr++ = '.';
985 /* swizzle bits fields: wwzzyyxx */
986 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
987 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
988 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
989 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
990 }
991
992 *ptr = '\0';
993 }
994
995 /* From a given parameter token, generate the corresponding GLSL string.
996 * Also, return the actual register name and swizzle in case the
997 * caller needs this information as well. */
998 static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
999 const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
1000 BOOL is_color = FALSE;
1001 char swizzle_str[6];
1002
1003 src_param->reg_name[0] = '\0';
1004 src_param->param_str[0] = '\0';
1005 swizzle_str[0] = '\0';
1006
1007 shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
1008
1009 shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
1010 shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
1011 }
1012
1013 /* From a given parameter token, generate the corresponding GLSL string.
1014 * Also, return the actual register name and swizzle in case the
1015 * caller needs this information as well. */
1016 static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
1017 const DWORD addr_token, glsl_dst_param_t *dst_param) {
1018 BOOL is_color = FALSE;
1019
1020 dst_param->mask_str[0] = '\0';
1021 dst_param->reg_name[0] = '\0';
1022
1023 shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
1024 return shader_glsl_get_write_mask(param, dst_param->mask_str);
1025 }
1026
1027 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1028 static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
1029 glsl_dst_param_t dst_param;
1030 DWORD mask;
1031 int shift;
1032
1033 mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
1034
1035 if(mask) {
1036 shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1037 shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
1038 }
1039
1040 return mask;
1041 }
1042
1043 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1044 static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
1045 return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
1046 }
1047
1048 /** Process GLSL instruction modifiers */
1049 void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
1050
1051 DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
1052
1053 if (arg->opcode->dst_token && mask != 0) {
1054 glsl_dst_param_t dst_param;
1055
1056 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
1057
1058 if (mask & WINED3DSPDM_SATURATE) {
1059 /* _SAT means to clamp the value of the register to between 0 and 1 */
1060 shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
1061 dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
1062 }
1063 if (mask & WINED3DSPDM_MSAMPCENTROID) {
1064 FIXME("_centroid modifier not handled\n");
1065 }
1066 if (mask & WINED3DSPDM_PARTIALPRECISION) {
1067 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1068 }
1069 }
1070 }
1071
1072 static inline const char* shader_get_comp_op(
1073 const DWORD opcode) {
1074
1075 DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
1076 switch (op) {
1077 case COMPARISON_GT: return ">";
1078 case COMPARISON_EQ: return "==";
1079 case COMPARISON_GE: return ">=";
1080 case COMPARISON_LT: return "<";
1081 case COMPARISON_NE: return "!=";
1082 case COMPARISON_LE: return "<=";
1083 default:
1084 FIXME("Unrecognized comparison value: %u\n", op);
1085 return "(\?\?)";
1086 }
1087 }
1088
1089 static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, BOOL texrect, glsl_sample_function_t *sample_function) {
1090 /* Note that there's no such thing as a projected cube texture. */
1091 switch(sampler_type) {
1092 case WINED3DSTT_1D:
1093 sample_function->name = projected ? "texture1DProj" : "texture1D";
1094 sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
1095 break;
1096 case WINED3DSTT_2D:
1097 if(texrect) {
1098 sample_function->name = projected ? "texture2DRectProj" : "texture2DRect";
1099 } else {
1100 sample_function->name = projected ? "texture2DProj" : "texture2D";
1101 }
1102 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
1103 break;
1104 case WINED3DSTT_CUBE:
1105 sample_function->name = "textureCube";
1106 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1107 break;
1108 case WINED3DSTT_VOLUME:
1109 sample_function->name = projected ? "texture3DProj" : "texture3D";
1110 sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1111 break;
1112 default:
1113 sample_function->name = "";
1114 FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
1115 break;
1116 }
1117 }
1118
1119 static void shader_glsl_color_correction(SHADER_OPCODE_ARG* arg) {
1120 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1121 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
1122 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
1123 glsl_dst_param_t dst_param;
1124 glsl_dst_param_t dst_param2;
1125 WINED3DFORMAT fmt;
1126 WINED3DFORMAT conversion_group;
1127 IWineD3DBaseTextureImpl *texture;
1128 DWORD mask, mask_size;
1129 UINT i;
1130 BOOL recorded = FALSE;
1131 DWORD sampler_idx;
1132 DWORD hex_version = shader->baseShader.hex_version;
1133
1134 switch(arg->opcode->opcode) {
1135 case WINED3DSIO_TEX:
1136 if (hex_version < WINED3DPS_VERSION(2,0)) {
1137 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1138 } else {
1139 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
1140 }
1141 break;
1142
1143 case WINED3DSIO_TEXLDL:
1144 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
1145 return;
1146
1147 case WINED3DSIO_TEXDP3TEX:
1148 case WINED3DSIO_TEXM3x3TEX:
1149 case WINED3DSIO_TEXM3x3SPEC:
1150 case WINED3DSIO_TEXM3x3VSPEC:
1151 case WINED3DSIO_TEXBEM:
1152 case WINED3DSIO_TEXREG2AR:
1153 case WINED3DSIO_TEXREG2GB:
1154 case WINED3DSIO_TEXREG2RGB:
1155 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1156 break;
1157
1158 default:
1159 /* Not a texture sampling instruction, nothing to do */
1160 return;
1161 };
1162
1163 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
1164 if(texture) {
1165 fmt = texture->resource.format;
1166 conversion_group = texture->baseTexture.shader_conversion_group;
1167 } else {
1168 fmt = WINED3DFMT_UNKNOWN;
1169 conversion_group = WINED3DFMT_UNKNOWN;
1170 }
1171
1172 /* before doing anything, record the sampler with the format in the format conversion list,
1173 * but check if it's not there already
1174 */
1175 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
1176 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
1177 recorded = TRUE;
1178 break;
1179 }
1180 }
1181 if(!recorded) {
1182 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
1183 shader->baseShader.num_sampled_samplers++;
1184 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
1185 }
1186
1187 switch(fmt) {
1188 case WINED3DFMT_V8U8:
1189 case WINED3DFMT_V16U16:
1190 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
1191 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
1192 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
1193 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_2, &dst_param);
1194 mask_size = shader_glsl_get_write_mask_size(mask);
1195 if(mask_size >= 3) {
1196 shader_addline(arg->buffer, "%s.%c = 1.0;\n", dst_param.reg_name, dst_param.mask_str[3]);
1197 }
1198 } else {
1199 /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
1200 mask = shader_glsl_add_dst_param(arg, arg->dst,
1201 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1,
1202 &dst_param);
1203 mask_size = shader_glsl_get_write_mask_size(mask);
1204 if(mask_size >= 2) {
1205 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1206 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1207 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1208 } else if(mask_size == 1) {
1209 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str[1],
1210 dst_param.reg_name, dst_param.mask_str[1]);
1211 }
1212 }
1213 break;
1214
1215 case WINED3DFMT_X8L8V8U8:
1216 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1217 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
1218 * and a(X) is always 1.0
1219 */
1220 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1221 mask_size = shader_glsl_get_write_mask_size(mask);
1222 if(mask_size >= 2) {
1223 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1224 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
1225 dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
1226 } else if(mask_size == 1) {
1227 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1228 dst_param.reg_name, dst_param.mask_str[1],
1229 dst_param.reg_name, dst_param.mask_str[1]);
1230 }
1231 }
1232 break;
1233
1234 case WINED3DFMT_L6V5U5:
1235 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1236 mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
1237 mask_size = shader_glsl_get_write_mask_size(mask);
1238 shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_2, &dst_param2);
1239 if(mask_size >= 3) {
1240 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1241 shader_addline(arg->buffer, "tmp0.g = %s.%c;\n",
1242 dst_param.reg_name, dst_param.mask_str[2]);
1243 shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1244 dst_param.reg_name, dst_param.mask_str[2], dst_param.mask_str[1],
1245 dst_param2.reg_name, dst_param.mask_str[1], dst_param.mask_str[3]);
1246 shader_addline(arg->buffer, "%s.%c = tmp0.g;\n", dst_param.reg_name,
1247 dst_param.mask_str[3]);
1248 } else if(mask_size == 2) {
1249 /* This is bad: We have VL, but we need VU */
1250 FIXME("2 components sampled from a converted L6V5U5 texture\n");
1251 } else {
1252 shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1253 dst_param.reg_name, dst_param.mask_str[1],
1254 dst_param2.reg_name, dst_param.mask_str[1]);
1255 }
1256 }
1257 break;
1258
1259 case WINED3DFMT_Q8W8V8U8:
1260 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
1261 /* Correct the sign in all channels. The writemask just applies as-is, no
1262 * need for checking the mask size
1263 */
1264 shader_glsl_add_dst_param(arg, arg->dst,
1265 WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 |
1266 WINED3DSP_WRITEMASK_2 | WINED3DSP_WRITEMASK_3,
1267 &dst_param);
1268 shader_addline(arg->buffer, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str,
1269 dst_param.reg_name, dst_param.mask_str);
1270 }
1271 break;
1272
1273 /* stupid compiler */
1274 default:
1275 break;
1276 }
1277 }
1278
1279 /*****************************************************************************
1280 *
1281 * Begin processing individual instruction opcodes
1282 *
1283 ****************************************************************************/
1284
1285 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1286 void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
1287 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1288 SHADER_BUFFER* buffer = arg->buffer;
1289 glsl_src_param_t src0_param;
1290 glsl_src_param_t src1_param;
1291 DWORD write_mask;
1292 char op;
1293
1294 /* Determine the GLSL operator to use based on the opcode */
1295 switch (curOpcode->opcode) {
1296 case WINED3DSIO_MUL: op = '*'; break;
1297 case WINED3DSIO_ADD: op = '+'; break;
1298 case WINED3DSIO_SUB: op = '-'; break;
1299 default:
1300 op = ' ';
1301 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1302 break;
1303 }
1304
1305 write_mask = shader_glsl_append_dst(buffer, arg);
1306 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1307 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1308 shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
1309 }
1310
1311 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1312 void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
1313 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1314 SHADER_BUFFER* buffer = arg->buffer;
1315 glsl_src_param_t src0_param;
1316 DWORD write_mask;
1317
1318 write_mask = shader_glsl_append_dst(buffer, arg);
1319 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1320
1321 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1322 * shader versions WINED3DSIO_MOVA is used for this. */
1323 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
1324 !shader_is_pshader_version(shader->baseShader.hex_version) &&
1325 shader_get_regtype(arg->dst) == WINED3DSPR_ADDR) ||
1326 arg->opcode->opcode == WINED3DSIO_MOVA) {
1327 /* We need to *round* to the nearest int here. */
1328 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
1329 if (mask_size > 1) {
1330 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);
1331 } else {
1332 shader_addline(buffer, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param.param_str, src0_param.param_str);
1333 }
1334 } else {
1335 shader_addline(buffer, "%s);\n", src0_param.param_str);
1336 }
1337 }
1338
1339 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1340 void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
1341 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1342 SHADER_BUFFER* buffer = arg->buffer;
1343 glsl_src_param_t src0_param;
1344 glsl_src_param_t src1_param;
1345 DWORD dst_write_mask, src_write_mask;
1346 size_t dst_size = 0;
1347
1348 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1349 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1350
1351 /* dp3 works on vec3, dp4 on vec4 */
1352 if (curOpcode->opcode == WINED3DSIO_DP4) {
1353 src_write_mask = WINED3DSP_WRITEMASK_ALL;
1354 } else {
1355 src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1356 }
1357
1358 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
1359 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
1360
1361 if (dst_size > 1) {
1362 shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1363 } else {
1364 shader_addline(buffer, "dot(%s, %s));\n", src0_param.param_str, src1_param.param_str);
1365 }
1366 }
1367
1368 /* Note that this instruction has some restrictions. The destination write mask
1369 * can't contain the w component, and the source swizzles have to be .xyzw */
1370 void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
1371 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
1372 glsl_src_param_t src0_param;
1373 glsl_src_param_t src1_param;
1374 char dst_mask[6];
1375
1376 shader_glsl_get_write_mask(arg->dst, dst_mask);
1377 shader_glsl_append_dst(arg->buffer, arg);
1378 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
1379 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
1380 shader_addline(arg->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
1381 }
1382
1383 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1384 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1385 * GLSL uses the value as-is. */
1386 void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
1387 SHADER_BUFFER *buffer = arg->buffer;
1388 glsl_src_param_t src0_param;
1389 glsl_src_param_t src1_param;
1390 DWORD dst_write_mask;
1391 size_t dst_size;
1392
1393 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1394 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1395
1396 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1397 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1398
1399 if (dst_size > 1) {
1400 shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
1401 } else {
1402 shader_addline(buffer, "pow(abs(%s), %s));\n", src0_param.param_str, src1_param.param_str);
1403 }
1404 }
1405
1406 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1407 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1408 * GLSL uses the value as-is. */
1409 void shader_glsl_log(SHADER_OPCODE_ARG *arg) {
1410 SHADER_BUFFER *buffer = arg->buffer;
1411 glsl_src_param_t src0_param;
1412 DWORD dst_write_mask;
1413 size_t dst_size;
1414
1415 dst_write_mask = shader_glsl_append_dst(buffer, arg);
1416 dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
1417
1418 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1419
1420 if (dst_size > 1) {
1421 shader_addline(buffer, "vec%d(log2(abs(%s))));\n", dst_size, src0_param.param_str);
1422 } else {
1423 shader_addline(buffer, "log2(abs(%s)));\n", src0_param.param_str);
1424 }
1425 }
1426
1427 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1428 void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
1429 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1430 SHADER_BUFFER* buffer = arg->buffer;
1431 glsl_src_param_t src_param;
1432 const char *instruction;
1433 char arguments[256];
1434 DWORD write_mask;
1435 unsigned i;
1436
1437 /* Determine the GLSL function to use based on the opcode */
1438 /* TODO: Possibly make this a table for faster lookups */
1439 switch (curOpcode->opcode) {
1440 case WINED3DSIO_MIN: instruction = "min"; break;
1441 case WINED3DSIO_MAX: instruction = "max"; break;
1442 case WINED3DSIO_ABS: instruction = "abs"; break;
1443 case WINED3DSIO_FRC: instruction = "fract"; break;
1444 case WINED3DSIO_NRM: instruction = "normalize"; break;
1445 case WINED3DSIO_LOGP:
1446 case WINED3DSIO_LOG: instruction = "log2"; break;
1447 case WINED3DSIO_EXP: instruction = "exp2"; break;
1448 case WINED3DSIO_SGN: instruction = "sign"; break;
1449 case WINED3DSIO_DSX: instruction = "dFdx"; break;
1450 case WINED3DSIO_DSY: instruction = "ycorrection.y * dFdy"; break;
1451 default: instruction = "";
1452 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
1453 break;
1454 }
1455
1456 write_mask = shader_glsl_append_dst(buffer, arg);
1457
1458 arguments[0] = '\0';
1459 if (curOpcode->num_params > 0) {
1460 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
1461 strcat(arguments, src_param.param_str);
1462 for (i = 2; i < curOpcode->num_params; ++i) {
1463 strcat(arguments, ", ");
1464 shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
1465 strcat(arguments, src_param.param_str);
1466 }
1467 }
1468
1469 shader_addline(buffer, "%s(%s));\n", instruction, arguments);
1470 }
1471
1472 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1473 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1474 * dst.x = 2^(floor(src))
1475 * dst.y = src - floor(src)
1476 * dst.z = 2^src (partial precision is allowed, but optional)
1477 * dst.w = 1.0;
1478 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1479 * dst = 2^src; (partial precision is allowed, but optional)
1480 */
1481 void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
1482 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1483 glsl_src_param_t src_param;
1484
1485 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
1486
1487 if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1488 char dst_mask[6];
1489
1490 shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
1491 shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
1492 shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
1493 shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
1494
1495 shader_glsl_append_dst(arg->buffer, arg);
1496 shader_glsl_get_write_mask(arg->dst, dst_mask);
1497 shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
1498 } else {
1499 DWORD write_mask;
1500 size_t mask_size;
1501
1502 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1503 mask_size = shader_glsl_get_write_mask_size(write_mask);
1504
1505 if (mask_size > 1) {
1506 shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
1507 } else {
1508 shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
1509 }
1510 }
1511 }
1512
1513 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1514 void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
1515 glsl_src_param_t src_param;
1516 DWORD write_mask;
1517 size_t mask_size;
1518
1519 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1520 mask_size = shader_glsl_get_write_mask_size(write_mask);
1521 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1522
1523 if (mask_size > 1) {
1524 shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
1525 } else {
1526 shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
1527 }
1528 }
1529
1530 void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
1531 SHADER_BUFFER* buffer = arg->buffer;
1532 glsl_src_param_t src_param;
1533 DWORD write_mask;
1534 size_t mask_size;
1535
1536 write_mask = shader_glsl_append_dst(buffer, arg);
1537 mask_size = shader_glsl_get_write_mask_size(write_mask);
1538
1539 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
1540
1541 if (mask_size > 1) {
1542 shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
1543 } else {
1544 shader_addline(buffer, "inversesqrt(%s));\n", src_param.param_str);
1545 }
1546 }
1547
1548 /** Process signed comparison opcodes in GLSL. */
1549 void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
1550 glsl_src_param_t src0_param;
1551 glsl_src_param_t src1_param;
1552 DWORD write_mask;
1553 size_t mask_size;
1554
1555 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1556 mask_size = shader_glsl_get_write_mask_size(write_mask);
1557 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1558 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1559
1560 if (mask_size > 1) {
1561 const char *compare;
1562
1563 switch(arg->opcode->opcode) {
1564 case WINED3DSIO_SLT: compare = "lessThan"; break;
1565 case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
1566 default: compare = "";
1567 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1568 }
1569
1570 shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
1571 src0_param.param_str, src1_param.param_str);
1572 } else {
1573 switch(arg->opcode->opcode) {
1574 case WINED3DSIO_SLT:
1575 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1576 * to return 0.0 but step returns 1.0 because step is not < x
1577 * An alternative is a bvec compare padded with an unused secound component.
1578 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1579 * issue. Playing with not() is not possible either because not() does not accept
1580 * a scalar.
1581 */
1582 shader_addline(arg->buffer, "(%s < %s) ? 1.0 : 0.0);\n", src0_param.param_str, src1_param.param_str);
1583 break;
1584 case WINED3DSIO_SGE:
1585 /* Here we can use the step() function and safe a conditional */
1586 shader_addline(arg->buffer, "step(%s, %s));\n", src1_param.param_str, src0_param.param_str);
1587 break;
1588 default:
1589 FIXME("Can't handle opcode %s\n", arg->opcode->name);
1590 }
1591
1592 }
1593 }
1594
1595 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1596 void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
1597 glsl_src_param_t src0_param;
1598 glsl_src_param_t src1_param;
1599 glsl_src_param_t src2_param;
1600 DWORD write_mask, cmp_channel = 0;
1601 unsigned int i, j;
1602 char mask_char[6];
1603 BOOL temp_destination = FALSE;
1604
1605 DWORD src0reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
1606 DWORD src1reg = arg->src[1] & WINED3DSP_REGNUM_MASK;
1607 DWORD src2reg = arg->src[2] & WINED3DSP_REGNUM_MASK;
1608 DWORD src0regtype = shader_get_regtype(arg->src[0]);
1609 DWORD src1regtype = shader_get_regtype(arg->src[1]);
1610 DWORD src2regtype = shader_get_regtype(arg->src[2]);
1611 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1612 DWORD dstregtype = shader_get_regtype(arg->dst);
1613
1614 /* Cycle through all source0 channels */
1615 for (i=0; i<4; i++) {
1616 write_mask = 0;
1617 /* Find the destination channels which use the current source0 channel */
1618 for (j=0; j<4; j++) {
1619 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1620 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1621 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1622 }
1623 }
1624
1625 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1626 * The first lines may overwrite source parameters of the following lines.
1627 * Deal with that by using a temporary destination register if needed
1628 */
1629 if((src0reg == dstreg && src0regtype == dstregtype) ||
1630 (src1reg == dstreg && src1regtype == dstregtype) ||
1631 (src2reg == dstreg && src2regtype == dstregtype)) {
1632
1633 write_mask = shader_glsl_get_write_mask(arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask), mask_char);
1634 if (!write_mask) continue;
1635 shader_addline(arg->buffer, "tmp0%s = (", mask_char);
1636 temp_destination = TRUE;
1637 } else {
1638 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1639 if (!write_mask) continue;
1640 }
1641
1642 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1643 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1644 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1645
1646 shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
1647 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1648 }
1649
1650 if(temp_destination) {
1651 shader_glsl_get_write_mask(arg->dst, mask_char);
1652 shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst);
1653 shader_addline(arg->buffer, "tmp0%s);\n", mask_char);
1654 }
1655
1656 }
1657
1658 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1659 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1660 * the compare is done per component of src0. */
1661 void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
1662 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1663 glsl_src_param_t src0_param;
1664 glsl_src_param_t src1_param;
1665 glsl_src_param_t src2_param;
1666 DWORD write_mask, cmp_channel = 0;
1667 unsigned int i, j;
1668
1669 if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
1670 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1671 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1672 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1673 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1674
1675 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1676 if(arg->opcode_token & WINED3DSI_COISSUE) {
1677 shader_addline(arg->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
1678 } else {
1679 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1680 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1681 }
1682 return;
1683 }
1684 /* Cycle through all source0 channels */
1685 for (i=0; i<4; i++) {
1686 write_mask = 0;
1687 /* Find the destination channels which use the current source0 channel */
1688 for (j=0; j<4; j++) {
1689 if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
1690 write_mask |= WINED3DSP_WRITEMASK_0 << j;
1691 cmp_channel = WINED3DSP_WRITEMASK_0 << j;
1692 }
1693 }
1694 write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
1695 if (!write_mask) continue;
1696
1697 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
1698 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1699 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1700
1701 shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
1702 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1703 }
1704 }
1705
1706 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1707 void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
1708 glsl_src_param_t src0_param;
1709 glsl_src_param_t src1_param;
1710 glsl_src_param_t src2_param;
1711 DWORD write_mask;
1712
1713 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1714 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1715 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1716 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1717 shader_addline(arg->buffer, "(%s * %s) + %s);\n",
1718 src0_param.param_str, src1_param.param_str, src2_param.param_str);
1719 }
1720
1721 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1722 Vertex shaders to GLSL codes */
1723 void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
1724 int i;
1725 int nComponents = 0;
1726 SHADER_OPCODE_ARG tmpArg;
1727
1728 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1729
1730 /* Set constants for the temporary argument */
1731 tmpArg.shader = arg->shader;
1732 tmpArg.buffer = arg->buffer;
1733 tmpArg.src[0] = arg->src[0];
1734 tmpArg.src_addr[0] = arg->src_addr[0];
1735 tmpArg.src_addr[1] = arg->src_addr[1];
1736 tmpArg.reg_maps = arg->reg_maps;
1737
1738 switch(arg->opcode->opcode) {
1739 case WINED3DSIO_M4x4:
1740 nComponents = 4;
1741 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1742 break;
1743 case WINED3DSIO_M4x3:
1744 nComponents = 3;
1745 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1746 break;
1747 case WINED3DSIO_M3x4:
1748 nComponents = 4;
1749 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1750 break;
1751 case WINED3DSIO_M3x3:
1752 nComponents = 3;
1753 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1754 break;
1755 case WINED3DSIO_M3x2:
1756 nComponents = 2;
1757 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1758 break;
1759 default:
1760 break;
1761 }
1762
1763 for (i = 0; i < nComponents; i++) {
1764 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1765 tmpArg.src[1] = arg->src[1]+i;
1766 shader_glsl_dot(&tmpArg);
1767 }
1768 }
1769
1770 /**
1771 The LRP instruction performs a component-wise linear interpolation
1772 between the second and third operands using the first operand as the
1773 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1774 This is equivalent to mix(src2, src1, src0);
1775 */
1776 void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
1777 glsl_src_param_t src0_param;
1778 glsl_src_param_t src1_param;
1779 glsl_src_param_t src2_param;
1780 DWORD write_mask;
1781
1782 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1783
1784 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
1785 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
1786 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
1787
1788 shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
1789 src2_param.param_str, src1_param.param_str, src0_param.param_str);
1790 }
1791
1792 /** Process the WINED3DSIO_LIT instruction in GLSL:
1793 * dst.x = dst.w = 1.0
1794 * dst.y = (src0.x > 0) ? src0.x
1795 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1796 * where src.w is clamped at +- 128
1797 */
1798 void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
1799 glsl_src_param_t src0_param;
1800 glsl_src_param_t src1_param;
1801 glsl_src_param_t src3_param;
1802 char dst_mask[6];
1803
1804 shader_glsl_append_dst(arg->buffer, arg);
1805 shader_glsl_get_write_mask(arg->dst, dst_mask);
1806
1807 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1808 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
1809 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
1810
1811 /* The sdk specifies the instruction like this
1812 * dst.x = 1.0;
1813 * if(src.x > 0.0) dst.y = src.x
1814 * else dst.y = 0.0.
1815 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
1816 * else dst.z = 0.0;
1817 * dst.w = 1.0;
1818 *
1819 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
1820 * dst.x = 1.0 ... No further explanation needed
1821 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
1822 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
1823 * dst.w = 1.0. ... Nothing fancy.
1824 *
1825 * So we still have one conditional in there. So do this:
1826 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
1827 *
1828 * step(0.0, x) will return 1 if src.x > 0.0, and 0 otherwise. So if y is 0 we get pow(0.0 * 1.0, power),
1829 * which sets dst.z to 0. If y > 0, but x = 0.0, we get pow(y * 0.0, power), which results in 0 too.
1830 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
1831 */
1832 shader_addline(arg->buffer, "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
1833 src0_param.param_str, src1_param.param_str, src0_param.param_str, src3_param.param_str, dst_mask);
1834 }
1835
1836 /** Process the WINED3DSIO_DST instruction in GLSL:
1837 * dst.x = 1.0
1838 * dst.y = src0.x * src0.y
1839 * dst.z = src0.z
1840 * dst.w = src1.w
1841 */
1842 void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
1843 glsl_src_param_t src0y_param;
1844 glsl_src_param_t src0z_param;
1845 glsl_src_param_t src1y_param;
1846 glsl_src_param_t src1w_param;
1847 char dst_mask[6];
1848
1849 shader_glsl_append_dst(arg->buffer, arg);
1850 shader_glsl_get_write_mask(arg->dst, dst_mask);
1851
1852 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
1853 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
1854 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
1855 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
1856
1857 shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1858 src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
1859 }
1860
1861 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1862 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1863 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1864 *
1865 * dst.x = cos(src0.?)
1866 * dst.y = sin(src0.?)
1867 * dst.z = dst.z
1868 * dst.w = dst.w
1869 */
1870 void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
1871 glsl_src_param_t src0_param;
1872 DWORD write_mask;
1873
1874 write_mask = shader_glsl_append_dst(arg->buffer, arg);
1875 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1876
1877 switch (write_mask) {
1878 case WINED3DSP_WRITEMASK_0:
1879 shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
1880 break;
1881
1882 case WINED3DSP_WRITEMASK_1:
1883 shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
1884 break;
1885
1886 case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
1887 shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
1888 break;
1889
1890 default:
1891 ERR("Write mask should be .x, .y or .xy\n");
1892 break;
1893 }
1894 }
1895
1896 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1897 * Start a for() loop where src1.y is the initial value of aL,
1898 * increment aL by src1.z for a total of src1.x iterations.
1899 * Need to use a temporary variable for this operation.
1900 */
1901 /* FIXME: I don't think nested loops will work correctly this way. */
1902 void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
1903 glsl_src_param_t src1_param;
1904 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1905 DWORD regtype = shader_get_regtype(arg->src[1]);
1906 DWORD reg = arg->src[1] & WINED3DSP_REGNUM_MASK;
1907 const DWORD *control_values = NULL;
1908 local_constant *constant;
1909
1910 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
1911
1912 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
1913 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
1914 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
1915 * addressing.
1916 */
1917 if(regtype == WINED3DSPR_CONSTINT) {
1918 LIST_FOR_EACH_ENTRY(constant, &shader->baseShader.constantsI, local_constant, entry) {
1919 if(constant->idx == reg) {
1920 control_values = constant->value;
1921 break;
1922 }
1923 }
1924 }
1925
1926 if(control_values) {
1927 if(control_values[2] > 0) {
1928 shader_addline(arg->buffer, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
1929 shader->baseShader.cur_loop_depth, control_values[1],
1930 shader->baseShader.cur_loop_depth, control_values[0], control_values[2], control_values[1],
1931 shader->baseShader.cur_loop_depth, control_values[2]);
1932 } else if(control_values[2] == 0) {
1933 shader_addline(arg->buffer, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
1934 shader->baseShader.cur_loop_depth, control_values[1], shader->baseShader.cur_loop_depth,
1935 shader->baseShader.cur_loop_depth, control_values[0],
1936 shader->baseShader.cur_loop_depth);
1937 } else {
1938 shader_addline(arg->buffer, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
1939 shader->baseShader.cur_loop_depth, control_values[1],
1940 shader->baseShader.cur_loop_depth, control_values[0], control_values[2], control_values[1],
1941 shader->baseShader.cur_loop_depth, control_values[2]);
1942 }
1943 } else {
1944 shader_addline(arg->buffer, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
1945 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno,
1946 src1_param.reg_name, shader->baseShader.cur_loop_depth, src1_param.reg_name,
1947 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno, src1_param.reg_name);
1948 }
1949
1950 shader->baseShader.cur_loop_depth++;
1951 shader->baseShader.cur_loop_regno++;
1952 }
1953
1954 void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
1955 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1956
1957 shader_addline(arg->buffer, "}\n");
1958
1959 if(arg->opcode->opcode == WINED3DSIO_ENDLOOP) {
1960 shader->baseShader.cur_loop_depth--;
1961 shader->baseShader.cur_loop_regno--;
1962 }
1963 if(arg->opcode->opcode == WINED3DSIO_ENDREP) {
1964 shader->baseShader.cur_loop_depth--;
1965 }
1966 }
1967
1968 void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
1969 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
1970 glsl_src_param_t src0_param;
1971
1972 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1973 shader_addline(arg->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
1974 shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_depth,
1975 src0_param.param_str, shader->baseShader.cur_loop_depth);
1976 shader->baseShader.cur_loop_depth++;
1977 }
1978
1979 void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
1980 glsl_src_param_t src0_param;
1981
1982 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1983 shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
1984 }
1985
1986 void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
1987 glsl_src_param_t src0_param;
1988 glsl_src_param_t src1_param;
1989
1990 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
1991 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
1992
1993 shader_addline(arg->buffer, "if (%s %s %s) {\n",
1994 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
1995 }
1996
1997 void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
1998 shader_addline(arg->buffer, "} else {\n");
1999 }
2000
2001 void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
2002 shader_addline(arg->buffer, "break;\n");
2003 }
2004
2005 /* FIXME: According to MSDN the compare is done per component. */
2006 void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
2007 glsl_src_param_t src0_param;
2008 glsl_src_param_t src1_param;
2009
2010 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
2011 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
2012
2013 shader_addline(arg->buffer, "if (%s %s %s) break;\n",
2014 src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
2015 }
2016
2017 void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
2018
2019 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
2020 shader_addline(arg->buffer, "}\n");
2021 shader_addline(arg->buffer, "void subroutine%u () {\n", snum);
2022 }
2023
2024 void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
2025 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
2026 shader_addline(arg->buffer, "subroutine%u();\n", snum);
2027 }
2028
2029 void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
2030 glsl_src_param_t src1_param;
2031
2032 DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
2033 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
2034 shader_addline(arg->buffer, "if (%s) subroutine%u();\n", src1_param.param_str, snum);
2035 }
2036
2037 /*********************************************
2038 * Pixel Shader Specific Code begins here
2039 ********************************************/
2040 void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
2041 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2042 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2043 DWORD hex_version = This->baseShader.hex_version;
2044 char dst_swizzle[6];
2045 glsl_sample_function_t sample_function;
2046 DWORD sampler_type;
2047 DWORD sampler_idx;
2048 BOOL projected, texrect = FALSE;
2049 DWORD mask = 0;
2050
2051 /* All versions have a destination register */
2052 shader_glsl_append_dst(arg->buffer, arg);
2053
2054 /* 1.0-1.4: Use destination register as sampler source.
2055 * 2.0+: Use provided sampler source. */
2056 if (hex_version < WINED3DPS_VERSION(1,4)) {
2057 DWORD flags;
2058
2059 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2060 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
2061
2062 if (flags & WINED3DTTFF_PROJECTED) {
2063 projected = TRUE;
2064 switch (flags & ~WINED3DTTFF_PROJECTED) {
2065 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2066 case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
2067 case WINED3DTTFF_COUNT3: mask = WINED3DSP_WRITEMASK_2; break;
2068 case WINED3DTTFF_COUNT4:
2069 case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
2070 }
2071 } else {
2072 projected = FALSE;
2073 }
2074 } else if (hex_version < WINED3DPS_VERSION(2,0)) {
2075 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
2076 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2077
2078 if (src_mod == WINED3DSPSM_DZ) {
2079 projected = TRUE;
2080 mask = WINED3DSP_WRITEMASK_2;
2081 } else if (src_mod == WINED3DSPSM_DW) {
2082 projected = TRUE;
2083 mask = WINED3DSP_WRITEMASK_3;
2084 } else {
2085 projected = FALSE;
2086 }
2087 } else {
2088 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
2089 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
2090 /* ps 2.0 texldp instruction always divides by the fourth component. */
2091 projected = TRUE;
2092 mask = WINED3DSP_WRITEMASK_3;
2093 } else {
2094 projected = FALSE;
2095 }
2096 }
2097
2098 if(deviceImpl->stateBlock->textures[sampler_idx] &&
2099 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
2100 texrect = TRUE;
2101 }
2102
2103 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2104 shader_glsl_get_sample_function(sampler_type, projected, texrect, &sample_function);
2105 mask |= sample_function.coord_mask;
2106
2107 if (hex_version < WINED3DPS_VERSION(2,0)) {
2108 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
2109 } else {
2110 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
2111 }
2112
2113 /* 1.0-1.3: Use destination register as coordinate source.
2114 1.4+: Use provided coordinate source register. */
2115 if (hex_version < WINED3DPS_VERSION(1,4)) {
2116 char coord_mask[6];
2117 shader_glsl_get_write_mask(mask, coord_mask);
2118 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
2119 sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
2120 } else {
2121 glsl_src_param_t coord_param;
2122 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
2123 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
2124 glsl_src_param_t bias;
2125 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &bias);
2126
2127 shader_addline(arg->buffer, "%s(Psampler%u, %s, %s)%s);\n",
2128 sample_function.name, sampler_idx, coord_param.param_str,
2129 bias.param_str, dst_swizzle);
2130 } else {
2131 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
2132 sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
2133 }
2134 }
2135 }
2136
2137 void shader_glsl_texldl(SHADER_OPCODE_ARG* arg) {
2138 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*)arg->shader;
2139 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2140 glsl_sample_function_t sample_function;
2141 glsl_src_param_t coord_param, lod_param;
2142 char dst_swizzle[6];
2143 DWORD sampler_type;
2144 DWORD sampler_idx;
2145 BOOL texrect = FALSE;
2146
2147 shader_glsl_append_dst(arg->buffer, arg);
2148 shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
2149
2150 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
2151 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2152 if(deviceImpl->stateBlock->textures[sampler_idx] &&
2153 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
2154 texrect = TRUE;
2155 }
2156 shader_glsl_get_sample_function(sampler_type, FALSE, texrect, &sample_function); shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &coord_param);
2157
2158 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &lod_param);
2159
2160 if (shader_is_pshader_version(This->baseShader.hex_version)) {
2161 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2162 * However, they seem to work just fine in fragment shaders as well. */
2163 WARN("Using %sLod in fragment shader.\n", sample_function.name);
2164 shader_addline(arg->buffer, "%sLod(Psampler%u, %s, %s)%s);\n",
2165 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
2166 } else {
2167 shader_addline(arg->buffer, "%sLod(Vsampler%u, %s, %s)%s);\n",
2168 sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
2169 }
2170 }
2171
2172 void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
2173
2174 /* FIXME: Make this work for more than just 2D textures */
2175
2176 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2177 SHADER_BUFFER* buffer = arg->buffer;
2178 DWORD hex_version = This->baseShader.hex_version;
2179 DWORD write_mask;
2180 char dst_mask[6];
2181
2182 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2183 shader_glsl_get_write_mask(write_mask, dst_mask);
2184
2185 if (hex_version != WINED3DPS_VERSION(1,4)) {
2186 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2187 shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
2188 } else {
2189 DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
2190 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
2191 char dst_swizzle[6];
2192
2193 shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
2194
2195 if (src_mod == WINED3DSPSM_DZ) {
2196 glsl_src_param_t div_param;
2197 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
2198 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
2199
2200 if (mask_size > 1) {
2201 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
2202 } else {
2203 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
2204 }
2205 } else if (src_mod == WINED3DSPSM_DW) {
2206 glsl_src_param_t div_param;
2207 size_t mask_size = shader_glsl_get_write_mask_size(write_mask);
2208 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
2209
2210 if (mask_size > 1) {
2211 shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
2212 } else {
2213 shader_addline(buffer, "gl_TexCoord[%u]%s / %s);\n", reg, dst_swizzle, div_param.param_str);
2214 }
2215 } else {
2216 shader_addline(buffer, "gl_TexCoord[%u]%s);\n", reg, dst_swizzle);
2217 }
2218 }
2219 }
2220
2221 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2222 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2223 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2224 void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
2225 glsl_src_param_t src0_param;
2226 char dst_mask[6];
2227 glsl_sample_function_t sample_function;
2228 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2229 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2230 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2231
2232 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2233
2234 shader_glsl_append_dst(arg->buffer, arg);
2235 shader_glsl_get_write_mask(arg->dst, dst_mask);
2236
2237 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2238 * scalar, and projected sampling would require 4.
2239 *
2240 * It is a dependent read - not valid with conditional NP2 textures
2241 */
2242 shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
2243
2244 switch(count_bits(sample_function.coord_mask)) {
2245 case 1:
2246 shader_addline(arg->buffer, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2247 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2248 break;
2249
2250 case 2:
2251 shader_addline(arg->buffer, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2252 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2253 break;
2254
2255 case 3:
2256 shader_addline(arg->buffer, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2257 sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
2258 break;
2259 default:
2260 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function.coord_mask));
2261 }
2262 }
2263
2264 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2265 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2266 void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
2267 glsl_src_param_t src0_param;
2268 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2269 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2270 DWORD dst_mask;
2271 size_t mask_size;
2272
2273 dst_mask = shader_glsl_append_dst(arg->buffer, arg);
2274 mask_size = shader_glsl_get_write_mask_size(dst_mask);
2275 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2276
2277 if (mask_size > 1) {
2278 shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
2279 } else {
2280 shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
2281 }
2282 }
2283
2284 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2285 * Calculate the depth as dst.x / dst.y */
2286 void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
2287 glsl_dst_param_t dst_param;
2288
2289 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2290
2291 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2292 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2293 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2294 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2295 * >= 1.0 or < 0.0
2296 */
2297 shader_addline(arg->buffer, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param.reg_name, dst_param.reg_name);
2298 }
2299
2300 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2301 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2302 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2303 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2304 */
2305 void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
2306 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2307 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
2308 glsl_src_param_t src0_param;
2309
2310 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2311
2312 shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
2313 shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2314 }
2315
2316 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2317 * Calculate the 1st of a 2-row matrix multiplication. */
2318 void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
2319 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2320 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2321 SHADER_BUFFER* buffer = arg->buffer;
2322 glsl_src_param_t src0_param;
2323
2324 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2325 shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2326 }
2327
2328 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2329 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2330 void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
2331
2332 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2333 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2334 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2335 SHADER_BUFFER* buffer = arg->buffer;
2336 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2337 glsl_src_param_t src0_param;
2338
2339 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2340 shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
2341 current_state->texcoord_w[current_state->current_row++] = reg;
2342 }
2343
2344 void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
2345 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2346 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2347 SHADER_BUFFER* buffer = arg->buffer;
2348 glsl_src_param_t src0_param;
2349 char dst_mask[6];
2350
2351 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2352 shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2353
2354 shader_glsl_append_dst(buffer, arg);
2355 shader_glsl_get_write_mask(arg->dst, dst_mask);
2356
2357 /* Sample the texture using the calculated coordinates */
2358 shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
2359 }
2360
2361 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2362 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2363 void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
2364 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2365 glsl_src_param_t src0_param;
2366 char dst_mask[6];
2367 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2368 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2369 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2370 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2371 glsl_sample_function_t sample_function;
2372
2373 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2374 shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2375
2376 shader_glsl_append_dst(arg->buffer, arg);
2377 shader_glsl_get_write_mask(arg->dst, dst_mask);
2378 /* Dependent read, not valid with conditional NP2 */
2379 shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
2380
2381 /* Sample the texture using the calculated coordinates */
2382 shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2383
2384 current_state->current_row = 0;
2385 }
2386
2387 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2388 * Perform the 3rd row of a 3x3 matrix multiply */
2389 void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
2390 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2391 glsl_src_param_t src0_param;
2392 char dst_mask[6];
2393 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2394 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2395 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
2396
2397 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2398
2399 shader_glsl_append_dst(arg->buffer, arg);
2400 shader_glsl_get_write_mask(arg->dst, dst_mask);
2401 shader_addline(arg->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
2402
2403 current_state->current_row = 0;
2404 }
2405
2406 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2407 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2408 void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
2409
2410 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2411 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2412 glsl_src_param_t src0_param;
2413 glsl_src_param_t src1_param;
2414 char dst_mask[6];
2415 SHADER_BUFFER* buffer = arg->buffer;
2416 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2417 DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2418 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2419 glsl_sample_function_t sample_function;
2420
2421 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2422 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
2423
2424 /* Perform the last matrix multiply operation */
2425 shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
2426 /* Reflection calculation */
2427 shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
2428
2429 shader_glsl_append_dst(buffer, arg);
2430 shader_glsl_get_write_mask(arg->dst, dst_mask);
2431 /* Dependent read, not valid with conditional NP2 */
2432 shader_glsl_get_sample_function(stype, FALSE, FALSE, &sample_function);
2433
2434 /* Sample the texture */
2435 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2436
2437 current_state->current_row = 0;
2438 }
2439
2440 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2441 * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2442 void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
2443
2444 IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
2445 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
2446 SHADER_BUFFER* buffer = arg->buffer;
2447 SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
2448 glsl_src_param_t src0_param;
2449 char dst_mask[6];
2450 DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
2451 DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
2452 glsl_sample_function_t sample_function;
2453
2454 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
2455
2456 /* Perform the last matrix multiply operation */
2457 shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
2458
2459 /* Construct the eye-ray vector from w coordinates */
2460 shader_addline(buffer, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2461 current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
2462 shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2463
2464 shader_glsl_append_dst(buffer, arg);
2465 shader_glsl_get_write_mask(arg->dst, dst_mask);
2466 /* Dependent read, not valid with conditional NP2 */
2467 shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
2468
2469 /* Sample the texture using the calculated coordinates */
2470 shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
2471
2472 current_state->current_row = 0;
2473 }
2474
2475 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2476 * Apply a fake bump map transform.
2477 * texbem is pshader <= 1.3 only, this saves a few version checks
2478 */
2479 void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
2480 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2481 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
2482 char dst_swizzle[6];
2483 glsl_sample_function_t sample_function;
2484 glsl_src_param_t coord_param;
2485 DWORD sampler_type;
2486 DWORD sampler_idx;
2487 DWORD mask;
2488 DWORD flags;
2489 char coord_mask[6];
2490
2491 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2492 flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
2493
2494 sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2495 /* Dependent read, not valid with conditional NP2 */
2496 shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
2497 mask = sample_function.coord_mask;
2498
2499 shader_glsl_get_write_mask(arg->dst, dst_swizzle);
2500
2501 shader_glsl_get_write_mask(mask, coord_mask);
2502
2503 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2504 * so we can't let the GL handle this.
2505 */
2506 if (flags & WINED3DTTFF_PROJECTED) {
2507 DWORD div_mask=0;
2508 char coord_div_mask[3];
2509 switch (flags & ~WINED3DTTFF_PROJECTED) {
2510 case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2511 case WINED3DTTFF_COUNT2: div_mask = WINED3DSP_WRITEMASK_1; break;
2512 case WINED3DTTFF_COUNT3: div_mask = WINED3DSP_WRITEMASK_2; break;
2513 case WINED3DTTFF_COUNT4:
2514 case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
2515 }
2516 shader_glsl_get_write_mask(div_mask, coord_div_mask);
2517 shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
2518 }
2519
2520 shader_glsl_append_dst(arg->buffer, arg);
2521 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
2522 if(arg->opcode->opcode == WINED3DSIO_TEXBEML) {
2523 glsl_src_param_t luminance_param;
2524 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &luminance_param);
2525 shader_addline(arg->buffer, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )*(%s * luminancescale + luminanceoffset))%s);\n",
2526 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask,
2527 luminance_param.param_str, dst_swizzle);
2528 } else {
2529 shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat * %s, 0.0, 0.0)%s )%s);\n",
2530 sample_function.name, sampler_idx, sampler_idx, coord_mask, coord_param.param_str, coord_mask, dst_swizzle);
2531 }
2532 }
2533
2534 void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
2535 glsl_src_param_t src0_param, src1_param;
2536
2537 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
2538 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
2539
2540 shader_glsl_append_dst(arg->buffer, arg);
2541 shader_addline(arg->buffer, "%s + bumpenvmat * %s);\n",
2542 src0_param.param_str, src1_param.param_str);
2543 }
2544
2545 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2546 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2547 void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
2548
2549 glsl_src_param_t src0_param;
2550 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2551 char dst_mask[6];
2552
2553 shader_glsl_append_dst(arg->buffer, arg);
2554 shader_glsl_get_write_mask(arg->dst, dst_mask);
2555 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2556
2557 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2558 }
2559
2560 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2561 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2562 void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
2563 glsl_src_param_t src0_param;
2564 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2565 char dst_mask[6];
2566
2567 shader_glsl_append_dst(arg->buffer, arg);
2568 shader_glsl_get_write_mask(arg->dst, dst_mask);
2569 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
2570
2571 shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
2572 }
2573
2574 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2575 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2576 void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
2577 glsl_src_param_t src0_param;
2578 char dst_mask[6];
2579 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
2580 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
2581 glsl_sample_function_t sample_function;
2582
2583 shader_glsl_append_dst(arg->buffer, arg);
2584 shader_glsl_get_write_mask(arg->dst, dst_mask);
2585 /* Dependent read, not valid with conditional NP2 */
2586 shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
2587 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
2588
2589 shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
2590 }
2591
2592 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2593 * If any of the first 3 components are < 0, discard this pixel */
2594 void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
2595 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
2596 DWORD hex_version = This->baseShader.hex_version;
2597 glsl_dst_param_t dst_param;
2598
2599 /* The argument is a destination parameter, and no writemasks are allowed */
2600 shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
2601 if((hex_version >= WINED3DPS_VERSION(2,0))) {
2602 /* 2.0 shaders compare all 4 components in texkill */
2603 shader_addline(arg->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
2604 } else {
2605 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2606 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2607 * 4 components are defined, only the first 3 are used
2608 */
2609 shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
2610 }
2611 }
2612
2613 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2614 * dst = dot2(src0, src1) + src2 */
2615 void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
2616 glsl_src_param_t src0_param;
2617 glsl_src_param_t src1_param;
2618 glsl_src_param_t src2_param;
2619 DWORD write_mask;
2620 size_t mask_size;
2621
2622 write_mask = shader_glsl_append_dst(arg->buffer, arg);
2623 mask_size = shader_glsl_get_write_mask_size(write_mask);
2624
2625 shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
2626 shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
2627 shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
2628
2629 shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
2630 }
2631
2632 void pshader_glsl_input_pack(
2633 SHADER_BUFFER* buffer,
2634 semantic* semantics_in,
2635 IWineD3DPixelShader *iface) {
2636
2637 unsigned int i;
2638 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
2639
2640 for (i = 0; i < MAX_REG_INPUT; i++) {
2641
2642 DWORD usage_token = semantics_in[i].usage;
2643 DWORD register_token = semantics_in[i].reg;
2644 DWORD usage, usage_idx;
2645 char reg_mask[6];
2646
2647 /* Uninitialized */
2648 if (!usage_token) continue;
2649 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2650 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2651 shader_glsl_get_write_mask(register_token, reg_mask);
2652
2653 switch(usage) {
2654
2655 case WINED3DDECLUSAGE_TEXCOORD:
2656 if(usage_idx < 8 && This->vertexprocessing == pretransformed) {
2657 shader_addline(buffer, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2658 This->input_reg_map[i], reg_mask, usage_idx, reg_mask);
2659 } else {
2660 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2661 This->input_reg_map[i], reg_mask, reg_mask);
2662 }
2663 break;
2664
2665 case WINED3DDECLUSAGE_COLOR:
2666 if (usage_idx == 0)
2667 shader_addline(buffer, "IN[%u]%s = vec4(gl_Color)%s;\n",
2668 This->input_reg_map[i], reg_mask, reg_mask);
2669 else if (usage_idx == 1)
2670 shader_addline(buffer, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2671 This->input_reg_map[i], reg_mask, reg_mask);
2672 else
2673 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2674 This->input_reg_map[i], reg_mask, reg_mask);
2675 break;
2676
2677 default:
2678 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2679 This->input_reg_map[i], reg_mask, reg_mask);
2680 }
2681 }
2682 }
2683
2684 /*********************************************
2685 * Vertex Shader Specific Code begins here
2686 ********************************************/
2687
2688 static void add_glsl_program_entry(IWineD3DDeviceImpl *device, struct glsl_shader_prog_link *entry) {
2689 glsl_program_key_t *key;
2690
2691 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2692 key->vshader = entry->vshader;
2693 key->pshader = entry->pshader;
2694
2695 hash_table_put(device->glsl_program_lookup, key, entry);
2696 }
2697
2698 static struct glsl_shader_prog_link *get_glsl_program_entry(IWineD3DDeviceImpl *device,
2699 GLhandleARB vshader, GLhandleARB pshader) {
2700 glsl_program_key_t key;
2701
2702 key.vshader = vshader;
2703 key.pshader = pshader;
2704
2705 return (struct glsl_shader_prog_link *)hash_table_get(device->glsl_program_lookup, &key);
2706 }
2707
2708 void delete_glsl_program_entry(IWineD3DDevice *iface, struct glsl_shader_prog_link *entry) {
2709 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2710 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2711 glsl_program_key_t *key;
2712
2713 key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
2714 key->vshader = entry->vshader;
2715 key->pshader = entry->pshader;
2716 hash_table_remove(This->glsl_program_lookup, key);
2717
2718 GL_EXTCALL(glDeleteObjectARB(entry->programId));
2719 if (entry->vshader) list_remove(&entry->vshader_entry);
2720 if (entry->pshader) list_remove(&entry->pshader_entry);
2721 HeapFree(GetProcessHeap(), 0, entry->vuniformF_locations);
2722 HeapFree(GetProcessHeap(), 0, entry->puniformF_locations);
2723 HeapFree(GetProcessHeap(), 0, entry);
2724 }
2725
2726 static void handle_ps3_input(SHADER_BUFFER *buffer, semantic *semantics_in, semantic *semantics_out, WineD3D_GL_Info *gl_info, DWORD *map) {
2727 unsigned int i, j;
2728 DWORD usage_token, usage_token_out;
2729 DWORD register_token, register_token_out;
2730 DWORD usage, usage_idx, usage_out, usage_idx_out;
2731 DWORD *set;
2732 char reg_mask[6], reg_mask_out[6];
2733
2734 set = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*set) * (GL_LIMITS(glsl_varyings) / 4));
2735
2736 for(i = 0; i < MAX_REG_INPUT; i++) {
2737 usage_token = semantics_in[i].usage;
2738 if (!usage_token) continue;
2739 if(map[i] >= (GL_LIMITS(glsl_varyings) / 4)) {
2740 FIXME("More input varyings declared than supported, expect issues\n");
2741 continue;
2742 } else if(map[i] == -1) {
2743 /* Declared, but not read register */
2744 continue;
2745 }
2746 register_token = semantics_in[i].reg;
2747
2748 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2749 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2750 set[map[i]] = shader_glsl_get_write_mask(register_token, reg_mask);
2751
2752 if(!semantics_out) {
2753 switch(usage) {
2754 case WINED3DDECLUSAGE_COLOR:
2755 if (usage_idx == 0)
2756 shader_addline(buffer, "IN[%u]%s = gl_FrontColor%s;\n",
2757 map[i], reg_mask, reg_mask);
2758 else if (usage_idx == 1)
2759 shader_addline(buffer, "IN[%u]%s = gl_FrontSecondaryColor%s;\n",
2760 map[i], reg_mask, reg_mask);
2761 else
2762 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2763 map[i], reg_mask, reg_mask);
2764 break;
2765
2766 case WINED3DDECLUSAGE_TEXCOORD:
2767 if (usage_idx < 8) {
2768 shader_addline(buffer, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2769 map[i], reg_mask, usage_idx, reg_mask);
2770 } else {
2771 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2772 map[i], reg_mask, reg_mask);
2773 }
2774 break;
2775
2776 case WINED3DDECLUSAGE_FOG:
2777 shader_addline(buffer, "IN[%u]%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
2778 map[i], reg_mask, reg_mask);
2779 break;
2780
2781 default:
2782 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2783 map[i], reg_mask, reg_mask);
2784 }
2785 } else {
2786 BOOL found = FALSE;
2787 for(j = 0; j < MAX_REG_OUTPUT; j++) {
2788 usage_token_out = semantics_out[j].usage;
2789 if (!usage_token_out) continue;
2790 register_token_out = semantics_out[j].reg;
2791
2792 usage_out = (usage_token_out & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2793 usage_idx_out = (usage_token_out & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2794 shader_glsl_get_write_mask(register_token_out, reg_mask_out);
2795
2796 if(usage == usage_out &&
2797 usage_idx == usage_idx_out) {
2798 shader_addline(buffer, "IN[%u]%s = OUT[%u]%s;\n",
2799 map[i], reg_mask, j, reg_mask);
2800 found = TRUE;
2801 }
2802 }
2803 if(!found) {
2804 shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2805 map[i], reg_mask, reg_mask);
2806 }
2807 }
2808 }
2809
2810 /* This is solely to make the compiler / linker happy and avoid warning about undefined
2811 * varyings. It shouldn't result in any real code executed on the GPU, since all read
2812 * input varyings are assigned above, if the optimizer works properly.
2813 */
2814 for(i = 0; i < GL_LIMITS(glsl_varyings) / 4; i++) {
2815 if(set[i] != WINED3DSP_WRITEMASK_ALL) {
2816 unsigned int size = 0;
2817 memset(reg_mask, 0, sizeof(reg_mask));
2818 if(!(set[i] & WINED3DSP_WRITEMASK_0)) {
2819 reg_mask[size] = 'x';
2820 size++;
2821 }
2822 if(!(set[i] & WINED3DSP_WRITEMASK_1)) {
2823 reg_mask[size] = 'y';
2824 size++;
2825 }
2826 if(!(set[i] & WINED3DSP_WRITEMASK_2)) {
2827 reg_mask[size] = 'z';
2828 size++;
2829 }
2830 if(!(set[i] & WINED3DSP_WRITEMASK_3)) {
2831 reg_mask[size] = 'w';
2832 size++;
2833 }
2834 switch(size) {
2835 case 1:
2836 shader_addline(buffer, "IN[%u].%s = 0.0;\n", i, reg_mask);
2837 break;
2838 case 2:
2839 shader_addline(buffer, "IN[%u].%s = vec2(0.0, 0.0);\n", i, reg_mask);
2840 break;
2841 case 3:
2842 shader_addline(buffer, "IN[%u].%s = vec3(0.0, 0.0, 0.0);\n", i, reg_mask);
2843 break;
2844 case 4:
2845 shader_addline(buffer, "IN[%u].%s = vec4(0.0, 0.0, 0.0, 0.0);\n", i, reg_mask);
2846 break;
2847 }
2848 }
2849 }
2850
2851 HeapFree(GetProcessHeap(), 0, set);
2852 }
2853
2854 static GLhandleARB generate_param_reorder_function(IWineD3DVertexShader *vertexshader,
2855 IWineD3DPixelShader *pixelshader,
2856 WineD3D_GL_Info *gl_info) {
2857 GLhandleARB ret = 0;
2858 IWineD3DVertexShaderImpl *vs = (IWineD3DVertexShaderImpl *) vertexshader;
2859 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) pixelshader;
2860 DWORD vs_major = vs ? WINED3DSHADER_VERSION_MAJOR(vs->baseShader.hex_version) : 0;
2861 DWORD ps_major = ps ? WINED3DSHADER_VERSION_MAJOR(ps->baseShader.hex_version) : 0;
2862 unsigned int i;
2863 SHADER_BUFFER buffer;
2864 DWORD usage_token;
2865 DWORD register_token;
2866 DWORD usage, usage_idx;
2867 char reg_mask[6];
2868 semantic *semantics_out, *semantics_in;
2869
2870 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
2871 buffer.bsize = 0;
2872 buffer.lineNo = 0;
2873 buffer.newline = TRUE;
2874
2875 if(vs_major < 3 && ps_major < 3) {
2876 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them */
2877 shader_addline(&buffer, "void order_ps_input() { /* do nothing */ }\n");
2878 } else if(ps_major < 3 && vs_major >= 3) {
2879 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
2880 semantics_out = vs->semantics_out;
2881
2882 shader_addline(&buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
2883 for(i = 0; i < MAX_REG_OUTPUT; i++) {
2884 usage_token = semantics_out[i].usage;
2885 if (!usage_token) continue;
2886 register_token = semantics_out[i].reg;
2887
2888 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2889 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2890 shader_glsl_get_write_mask(register_token, reg_mask);
2891
2892 switch(usage) {
2893 case WINED3DDECLUSAGE_COLOR:
2894 if (usage_idx == 0)
2895 shader_addline(&buffer, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
2896 else if (usage_idx == 1)
2897 shader_addline(&buffer, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
2898 break;
2899
2900 case WINED3DDECLUSAGE_POSITION:
2901 shader_addline(&buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
2902 break;
2903
2904 case WINED3DDECLUSAGE_TEXCOORD:
2905 if (usage_idx < 8) {
2906 shader_addline(&buffer, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
2907 usage_idx, reg_mask, i, reg_mask);
2908 }
2909 break;
2910
2911 case WINED3DDECLUSAGE_PSIZE:
2912 shader_addline(&buffer, "gl_PointSize = OUT[%u].x;\n", i);
2913 break;
2914
2915 case WINED3DDECLUSAGE_FOG:
2916 shader_addline(&buffer, "gl_FogFragCoord = OUT[%u].%c;\n", i, reg_mask[1]);
2917 break;
2918
2919 default:
2920 break;
2921 }
2922 }
2923 shader_addline(&buffer, "}\n");
2924
2925 } else if(ps_major >= 3 && vs_major >= 3) {
2926 semantics_out = vs->semantics_out;
2927 semantics_in = ps->semantics_in;
2928
2929 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
2930 shader_addline(&buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
2931 shader_addline(&buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
2932
2933 /* First, sort out position and point size. Those are not passed to the pixel shader */
2934 for(i = 0; i < MAX_REG_OUTPUT; i++) {
2935 usage_token = semantics_out[i].usage;
2936 if (!usage_token) continue;
2937 register_token = semantics_out[i].reg;
2938
2939 usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
2940 usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
2941 shader_glsl_get_write_mask(register_token, reg_mask);
2942
2943 switch(usage) {
2944 case WINED3DDECLUSAGE_POSITION:
2945 shader_addline(&buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
2946 break;
2947
2948 case WINED3DDECLUSAGE_PSIZE:
2949 shader_addline(&buffer, "gl_PointSize = OUT[%u].x;\n", i);
2950 break;
2951
2952 default:
2953 break;
2954 }
2955 }
2956
2957 /* Then, fix the pixel shader input */
2958 handle_ps3_input(&buffer, semantics_in, semantics_out, gl_info, ps->input_reg_map);
2959
2960 shader_addline(&buffer, "}\n");
2961 } else if(ps_major >= 3 && vs_major < 3) {
2962 semantics_in = ps->semantics_in;
2963
2964 shader_addline(&buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
2965 shader_addline(&buffer, "void order_ps_input() {\n");
2966 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
2967 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
2968 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
2969 */
2970 handle_ps3_input(&buffer, semantics_in, NULL, gl_info, ps->input_reg_map);
2971 shader_addline(&buffer, "}\n");
2972 } else {
2973 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major, ps_major);
2974 }
2975
2976 ret = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
2977 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
2978 GL_EXTCALL(glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL));
2979 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
2980 GL_EXTCALL(glCompileShaderARB(ret));
2981 checkGLcall("glCompileShaderARB(ret)");
2982
2983 HeapFree(GetProcessHeap(), 0, buffer.buffer);
2984 return ret;
2985 }
2986
2987 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
2988 * It sets the programId on the current StateBlock (because it should be called
2989 * inside of the DrawPrimitive() part of the render loop).
2990 *
2991 * If a program for the given combination does not exist, create one, and store
2992 * the program in the hash table. If it creates a program, it will link the
2993 * given objects, too.
2994 */
2995 static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
2996 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
2997 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
2998 IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
2999 IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
3000 struct glsl_shader_prog_link *entry = NULL;
3001 GLhandleARB programId = 0;
3002 GLhandleARB reorder_shader_id = 0;
3003 int i;
3004 char glsl_name[8];
3005
3006 GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
3007 GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
3008 entry = get_glsl_program_entry(This, vshader_id, pshader_id);
3009 if (entry) {
3010 This->stateBlock->glsl_program = entry;
3011 return;
3012 }
3013
3014 /* If we get to this point, then no matching program exists, so we create one */
3015 programId = GL_EXTCALL(glCreateProgramObjectARB());
3016 TRACE("Created new GLSL shader program %u\n", programId);
3017
3018 /* Create the entry */
3019 entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
3020 entry->programId = programId;
3021 entry->vshader = vshader_id;
3022 entry->pshader = pshader_id;
3023 /* Add the hash table entry */
3024 add_glsl_program_entry(This, entry);
3025
3026 /* Set the current program */
3027 This->stateBlock->glsl_program = entry;
3028
3029 /* Attach GLSL vshader */
3030 if (vshader_id) {
3031 int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
3032 char tmp_name[10];
3033
3034 reorder_shader_id = generate_param_reorder_function(vshader, pshader, gl_info);
3035 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id, programId);
3036 GL_EXTCALL(glAttachObjectARB(programId, reorder_shader_id));
3037 checkGLcall("glAttachObjectARB");
3038 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3039 * is destroyed
3040 */
3041 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id));
3042
3043 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id, programId);
3044 GL_EXTCALL(glAttachObjectARB(programId, vshader_id));
3045 checkGLcall("glAttachObjectARB");
3046
3047 /* Bind vertex attributes to a corresponding index number to match
3048 * the same index numbers as ARB_vertex_programs (makes loading
3049 * vertex attributes simpler). With this method, we can use the
3050 * exact same code to load the attributes later for both ARB and
3051 * GLSL shaders.
3052 *
3053 * We have to do this here because we need to know the Program ID
3054 * in order to make the bindings work, and it has to be done prior
3055 * to linking the GLSL program. */
3056 for (i = 0; i < max_attribs; ++i) {
3057 if (((IWineD3DBaseShaderImpl*)vshader)->baseShader.reg_maps.attributes[i]) {
3058 snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
3059 GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
3060 }
3061 }
3062 checkGLcall("glBindAttribLocationARB");
3063
3064 list_add_head(&((IWineD3DBaseShaderImpl *)vshader)->baseShader.linked_programs, &entry->vshader_entry);
3065 }
3066
3067 /* Attach GLSL pshader */
3068 if (pshader_id) {
3069 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
3070 GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
3071 checkGLcall("glAttachObjectARB");
3072
3073 list_add_head(&((IWineD3DBaseShaderImpl *)pshader)->baseShader.linked_programs, &entry->pshader_entry);
3074 }
3075
3076 /* Link the program */
3077 TRACE("Linking GLSL shader program %u\n", programId);
3078 GL_EXTCALL(glLinkProgramARB(programId));
3079 print_glsl_info_log(&GLINFO_LOCATION, programId);
3080
3081 entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
3082 for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
3083 snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
3084 entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
3085 }
3086 for (i = 0; i < MAX_CONST_I; ++i) {
3087 snprintf(glsl_name, sizeof(glsl_name), "VI[%i]", i);
3088 entry->vuniformI_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
3089 }
3090 entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
3091 for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
3092 snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
3093 entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
3094 }
3095 for (i = 0; i < MAX_CONST_I; ++i) {
3096 snprintf(glsl_name, sizeof(glsl_name), "PI[%i]", i);
3097 entry->puniformI_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
3098 }
3099
3100 entry->posFixup_location = GL_EXTCALL(glGetUniformLocationARB(programId, "posFixup"));
3101 entry->bumpenvmat_location = GL_EXTCALL(glGetUniformLocationARB(programId, "bumpenvmat"));
3102 entry->luminancescale_location = GL_EXTCALL(glGetUniformLocationARB(programId, "luminancescale"));
3103 entry->luminanceoffset_location = GL_EXTCALL(glGetUniformLocationARB(programId, "luminanceoffset"));
3104 entry->srgb_comparison_location = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_comparison"));
3105 entry->srgb_mul_low_location = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_mul_low"));
3106 entry->ycorrection_location = GL_EXTCALL(glGetUniformLocationARB(programId, "ycorrection"));
3107 checkGLcall("Find glsl program uniform locations");
3108
3109 /* Set the shader to allow uniform loading on it */
3110 GL_EXTCALL(glUseProgramObjectARB(programId));
3111 checkGLcall("glUseProgramObjectARB(programId)");
3112
3113 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3114 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3115 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3116 * vertex shader with fixed function pixel processing is used we make sure that the card
3117 * supports enough samplers to allow the max number of vertex samplers with all possible
3118 * fixed function fragment processing setups. So once the program is linked these samplers
3119 * won't change.
3120 */
3121 if(vshader_id) {
3122 /* Load vertex shader samplers */
3123 shader_glsl_load_vsamplers(gl_info, (IWineD3DStateBlock*)This->stateBlock, programId);
3124 }
3125 if(pshader_id) {
3126 /* Load pixel shader samplers */
3127 shader_glsl_load_psamplers(gl_info, (IWineD3DStateBlock*)This->stateBlock, programId);
3128 }
3129 }
3130
3131 static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
3132 GLhandleARB program_id;
3133 GLhandleARB vshader_id, pshader_id;
3134 const char *blt_vshader[] = {
3135 "void main(void)\n"
3136 "{\n"
3137 " gl_Position = gl_Vertex;\n"
3138 " gl_FrontColor = vec4(1.0);\n"
3139 " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
3140 " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
3141 "}\n"
3142 };
3143
3144 const char *blt_pshader[] = {
3145 "uniform sampler2D sampler;\n"
3146 "void main(void)\n"
3147 "{\n"
3148 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3149 "}\n"
3150 };
3151
3152 vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
3153 GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
3154 GL_EXTCALL(glCompileShaderARB(vshader_id));
3155
3156 pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
3157 GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
3158 GL_EXTCALL(glCompileShaderARB(pshader_id));
3159
3160 program_id = GL_EXTCALL(glCreateProgramObjectARB());
3161 GL_EXTCALL(glAttachObjectARB(program_id, vshader_id));
3162 GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
3163 GL_EXTCALL(glLinkProgramARB(program_id));
3164
3165 print_glsl_info_log(&GLINFO_LOCATION, program_id);
3166
3167 return program_id;
3168 }
3169
3170 static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
3171 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
3172 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
3173 GLhandleARB program_id = 0;
3174
3175 if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
3176 else This->stateBlock->glsl_program = NULL;
3177
3178 program_id = This->stateBlock->glsl_program ? This->stateBlock->glsl_program->programId : 0;
3179 if (program_id) TRACE("Using GLSL program %u\n", program_id);
3180 GL_EXTCALL(glUseProgramObjectARB(program_id));
3181 checkGLcall("glUseProgramObjectARB");
3182 }
3183
3184 static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
3185 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
3186 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
3187 static GLhandleARB program_id = 0;
3188 static GLhandleARB loc = -1;
3189
3190 if (!program_id) {
3191 program_id = create_glsl_blt_shader(gl_info);
3192 loc = GL_EXTCALL(glGetUniformLocationARB(program_id, "sampler"));
3193 }
3194
3195 GL_EXTCALL(glUseProgramObjectARB(program_id));
3196 GL_EXTCALL(glUniform1iARB(loc, 0));
3197 }
3198
3199 static void shader_glsl_cleanup(IWineD3DDevice *iface) {
3200 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
3201 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
3202 GL_EXTCALL(glUseProgramObjectARB(0));
3203 }
3204
3205 static void shader_glsl_destroy(IWineD3DBaseShader *iface) {
3206 struct list *linked_programs;
3207 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
3208 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
3209
3210 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3211 * can be called from IWineD3DBaseShader::Release
3212 */
3213 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
3214
3215 if(This->baseShader.prgId == 0) return;
3216 linked_programs = &This->baseShader.linked_programs;
3217
3218 TRACE("Deleting linked programs\n");
3219 if (linked_programs->next) {
3220 struct glsl_shader_prog_link *entry, *entry2;
3221
3222 if(pshader) {
3223 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs, struct glsl_shader_prog_link, pshader_entry) {
3224 delete_glsl_program_entry(This->baseShader.device, entry);
3225 }
3226 } else {
3227 LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs, struct glsl_shader_prog_link, vshader_entry) {
3228 delete_glsl_program_entry(This->baseShader.device, entry);
3229 }
3230 }
3231 }
3232
3233 TRACE("Deleting shader object %u\n", This->baseShader.prgId);
3234 GL_EXTCALL(glDeleteObjectARB(This->baseShader.prgId));
3235 checkGLcall("glDeleteObjectARB");
3236 }
3237
3238 const shader_backend_t glsl_shader_backend = {
3239 &shader_glsl_select,
3240 &shader_glsl_select_depth_blt,
3241 &shader_glsl_load_constants,
3242 &shader_glsl_cleanup,
3243 &shader_glsl_color_correction,
3244 &shader_glsl_destroy
3245 };
DirectPlay implementation for GSoC