?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Add support for some unimplemented instructions to arb shaders.
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
blobfc543d6dd93fadeb7b65217ff817e47efc06bdac
1 /*
2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
4 *
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 *
13 * This library is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
17 *
18 * This library is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
22 *
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with this library; if not, write to the Free Software
25 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
26 */
27
28 #include "config.h"
29
30 #include <math.h>
31 #include <stdio.h>
32
33 #include "wined3d_private.h"
34
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
37
38 #define GLINFO_LOCATION (*gl_info)
39
40 /********************************************************
41 * ARB_[vertex/fragment]_program helper functions follow
42 ********************************************************/
43
44 /**
45 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
46 * When constant_list == NULL, it will load all the constants.
47 *
48 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
49 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
50 */
51 static void shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
52 unsigned int max_constants, float* constants, struct list *constant_list) {
53 constants_entry *constant;
54 local_constant* lconst;
55 DWORD i, j, k;
56 DWORD *idx;
57
58 if (TRACE_ON(d3d_shader)) {
59 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
60 idx = constant->idx;
61 j = constant->count;
62 while (j--) {
63 i = *idx++;
64 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
65 constants[i * 4 + 0], constants[i * 4 + 1],
66 constants[i * 4 + 2], constants[i * 4 + 3]);
67 }
68 }
69 }
70 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
71 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
72 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
73 float lcl_const[4];
74 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
75 idx = constant->idx;
76 j = constant->count;
77 while (j--) {
78 i = *idx++;
79 k = i * 4;
80 if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
81 else if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
82 else lcl_const[0] = constants[k + 0];
83
84 if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
85 else if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
86 else lcl_const[1] = constants[k + 1];
87
88 if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
89 else if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
90 else lcl_const[2] = constants[k + 2];
91
92 if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
93 else if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
94 else lcl_const[3] = constants[k + 3];
95
96 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
97 }
98 }
99 } else {
100 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
101 idx = constant->idx;
102 j = constant->count;
103 while (j--) {
104 i = *idx++;
105 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
106 }
107 }
108 }
109 checkGLcall("glProgramEnvParameter4fvARB()");
110
111 /* Load immediate constants */
112 if (TRACE_ON(d3d_shader)) {
113 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
114 GLfloat* values = (GLfloat*)lconst->value;
115 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
116 values[0], values[1], values[2], values[3]);
117 }
118 }
119 /* Immediate constants are clamped for 1.X shaders at loading times */
120 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
121 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
122 }
123 checkGLcall("glProgramEnvParameter4fvARB()");
124 }
125
126 /**
127 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
128 *
129 * We only support float constants in ARB at the moment, so don't
130 * worry about the Integers or Booleans
131 */
132 void shader_arb_load_constants(
133 IWineD3DDevice* device,
134 char usePixelShader,
135 char useVertexShader) {
136
137 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
138 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
139 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
140
141 if (useVertexShader) {
142 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
143
144 /* Load DirectX 9 float constants for vertex shader */
145 shader_arb_load_constantsF(vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
146 GL_LIMITS(vshader_constantsF),
147 stateBlock->vertexShaderConstantF,
148 &stateBlock->set_vconstantsF);
149
150 /* Upload the position fixup */
151 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
152 }
153
154 if (usePixelShader) {
155
156 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
157 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
158
159 /* Load DirectX 9 float constants for pixel shader */
160 shader_arb_load_constantsF(pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
161 GL_LIMITS(pshader_constantsF),
162 stateBlock->pixelShaderConstantF,
163 &stateBlock->set_pconstantsF);
164 if(((IWineD3DPixelShaderImpl *) pshader)->bumpenvmatconst) {
165 /* needsbumpmat stores the stage number from where to load the matrix. bumpenvmatconst stores the
166 * number of the constant to load the matrix into.
167 * The state manager takes care that this function is always called if the bump env matrix changes
168 */
169 float *data = (float *) &stateBlock->textureState[(int) psi->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
170 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst, data));
171 }
172 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
173 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
174 float comparison[4];
175 float mul_low[4];
176
177 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
178 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
179 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
180
181 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
182 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
183 } else {
184 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
185 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
186
187 mul_low[0] = 1.0; mul_low[1] = 1.0;
188 mul_low[2] = 1.0; mul_low[3] = 1.0;
189 }
190 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
191 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
192 checkGLcall("Load sRGB correction constants\n");
193 }
194 }
195 }
196
197 /* Generate the variable & register declarations for the ARB_vertex_program output target */
198 void shader_generate_arb_declarations(
199 IWineD3DBaseShader *iface,
200 shader_reg_maps* reg_maps,
201 SHADER_BUFFER* buffer,
202 WineD3D_GL_Info* gl_info) {
203
204 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
205 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
206 DWORD i;
207 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
208 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
209 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
210 UINT extra_constants_needed = 0;
211
212 /* Temporary Output register */
213 shader_addline(buffer, "TEMP TMP_OUT;\n");
214
215 for(i = 0; i < This->baseShader.limits.temporary; i++) {
216 if (reg_maps->temporary[i])
217 shader_addline(buffer, "TEMP R%u;\n", i);
218 }
219
220 for (i = 0; i < This->baseShader.limits.address; i++) {
221 if (reg_maps->address[i])
222 shader_addline(buffer, "ADDRESS A%d;\n", i);
223 }
224
225 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
226 if (reg_maps->texcoord[i])
227 shader_addline(buffer,"TEMP T%u;\n", i);
228 }
229
230 /* Texture coordinate registers must be pre-loaded */
231 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
232 if (reg_maps->texcoord[i])
233 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
234 }
235
236 if(reg_maps->bumpmat != -1 /* Only a pshader can use texbem */) {
237 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
238 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
239 * bump mapping.
240 */
241 if(max_constantsF < GL_LIMITS(pshader_constantsF)) {
242 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = max_constantsF;
243 shader_addline(buffer, "PARAM bumpenvmat = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst);
244 } else {
245 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
246 }
247 extra_constants_needed += 1;
248 }
249 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
250 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
251 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
252 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
253 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
254 * off again
255 */
256 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
257 /* The idea is that if srgb is enabled, then disabled, the constant loading code
258 * can effectively disabling sRGB correction by passing 1.0 and INF as the multiplication
259 * and comparison constants. If it disables it that way, the shader won't be recompiled
260 * and the code will stay in, so sRGB writing can be turned on again by setting the
261 * constants from the spec
262 */
263 ps_impl->srgb_mode_hardcoded = 0;
264 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
265 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
266 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
267 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
268 } else {
269 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
270 srgb_mul_low, srgb_mul_low, srgb_mul_low);
271 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
272 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
273 ps_impl->srgb_mode_hardcoded = 1;
274 }
275 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
276 * path if the comparison value is set to INF
277 */
278 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
279 srgb_pow, srgb_pow, srgb_pow);
280 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
281 srgb_mul_high, srgb_mul_high, srgb_mul_high);
282 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
283 srgb_sub_high, srgb_sub_high, srgb_sub_high);
284 ps_impl->srgb_enabled = 1;
285 } else if(pshader) {
286 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
287
288 /* Do not write any srgb fixup into the shader to save shader size and processing time.
289 * As a consequence, we can't toggle srgb write on without recompilation
290 */
291 ps_impl->srgb_enabled = 0;
292 ps_impl->srgb_mode_hardcoded = 1;
293 }
294
295 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
296 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
297 max_constantsF, max_constantsF - 1);
298 }
299
300 static const char * const shift_tab[] = {
301 "dummy", /* 0 (none) */
302 "coefmul.x", /* 1 (x2) */
303 "coefmul.y", /* 2 (x4) */
304 "coefmul.z", /* 3 (x8) */
305 "coefmul.w", /* 4 (x16) */
306 "dummy", /* 5 (x32) */
307 "dummy", /* 6 (x64) */
308 "dummy", /* 7 (x128) */
309 "dummy", /* 8 (d256) */
310 "dummy", /* 9 (d128) */
311 "dummy", /* 10 (d64) */
312 "dummy", /* 11 (d32) */
313 "coefdiv.w", /* 12 (d16) */
314 "coefdiv.z", /* 13 (d8) */
315 "coefdiv.y", /* 14 (d4) */
316 "coefdiv.x" /* 15 (d2) */
317 };
318
319 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
320 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
321 char *ptr = write_mask;
322 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
323
324 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
325 *ptr++ = '.';
326 *ptr++ = 'x';
327 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
328 *ptr++ = '.';
329 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
330 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
331 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
332 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
333 }
334
335 *ptr = '\0';
336 }
337
338 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
339 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
340 * but addressed as "rgba". To fix this we need to swap the register's x
341 * and z components. */
342 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
343 char *ptr = swizzle_str;
344
345 /* swizzle bits fields: wwzzyyxx */
346 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
347 DWORD swizzle_x = swizzle & 0x03;
348 DWORD swizzle_y = (swizzle >> 2) & 0x03;
349 DWORD swizzle_z = (swizzle >> 4) & 0x03;
350 DWORD swizzle_w = (swizzle >> 6) & 0x03;
351
352 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
353 * generate a swizzle string. Unless we need to our own swizzling. */
354 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
355 *ptr++ = '.';
356 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
357 *ptr++ = swizzle_chars[swizzle_x];
358 } else {
359 *ptr++ = swizzle_chars[swizzle_x];
360 *ptr++ = swizzle_chars[swizzle_y];
361 *ptr++ = swizzle_chars[swizzle_z];
362 *ptr++ = swizzle_chars[swizzle_w];
363 }
364 }
365
366 *ptr = '\0';
367 }
368
369 static void pshader_get_register_name(
370 const DWORD param, char* regstr) {
371
372 DWORD reg = param & WINED3DSP_REGNUM_MASK;
373 DWORD regtype = shader_get_regtype(param);
374
375 switch (regtype) {
376 case WINED3DSPR_TEMP:
377 sprintf(regstr, "R%u", reg);
378 break;
379 case WINED3DSPR_INPUT:
380 if (reg==0) {
381 strcpy(regstr, "fragment.color.primary");
382 } else {
383 strcpy(regstr, "fragment.color.secondary");
384 }
385 break;
386 case WINED3DSPR_CONST:
387 sprintf(regstr, "C[%u]", reg);
388 break;
389 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
390 sprintf(regstr,"T%u", reg);
391 break;
392 case WINED3DSPR_COLOROUT:
393 if (reg == 0)
394 sprintf(regstr, "TMP_COLOR");
395 else {
396 /* TODO: See GL_ARB_draw_buffers */
397 FIXME("Unsupported write to render target %u\n", reg);
398 sprintf(regstr, "unsupported_register");
399 }
400 break;
401 case WINED3DSPR_DEPTHOUT:
402 sprintf(regstr, "result.depth");
403 break;
404 case WINED3DSPR_ATTROUT:
405 sprintf(regstr, "oD[%u]", reg);
406 break;
407 case WINED3DSPR_TEXCRDOUT:
408 sprintf(regstr, "oT[%u]", reg);
409 break;
410 default:
411 FIXME("Unhandled register name Type(%d)\n", regtype);
412 sprintf(regstr, "unrecognized_register");
413 break;
414 }
415 }
416
417 /* TODO: merge with pixel shader */
418 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
419
420 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
421
422 /* oPos, oFog and oPts in D3D */
423 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
424
425 DWORD reg = param & WINED3DSP_REGNUM_MASK;
426 DWORD regtype = shader_get_regtype(param);
427 char tmpReg[255];
428 BOOL is_color = FALSE;
429
430 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
431 strcat(hwLine, " -");
432 } else {
433 strcat(hwLine, " ");
434 }
435
436 switch (regtype) {
437 case WINED3DSPR_TEMP:
438 sprintf(tmpReg, "R%u", reg);
439 strcat(hwLine, tmpReg);
440 break;
441 case WINED3DSPR_INPUT:
442
443 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
444 is_color = TRUE;
445
446 sprintf(tmpReg, "vertex.attrib[%u]", reg);
447 strcat(hwLine, tmpReg);
448 break;
449 case WINED3DSPR_CONST:
450 sprintf(tmpReg, "C[%s%u]", (param & WINED3DSHADER_ADDRMODE_RELATIVE) ? "A0.x + " : "", reg);
451 strcat(hwLine, tmpReg);
452 break;
453 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
454 sprintf(tmpReg, "A%u", reg);
455 strcat(hwLine, tmpReg);
456 break;
457 case WINED3DSPR_RASTOUT:
458 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
459 strcat(hwLine, tmpReg);
460 break;
461 case WINED3DSPR_ATTROUT:
462 if (reg==0) {
463 strcat(hwLine, "result.color.primary");
464 } else {
465 strcat(hwLine, "result.color.secondary");
466 }
467 break;
468 case WINED3DSPR_TEXCRDOUT:
469 sprintf(tmpReg, "result.texcoord[%u]", reg);
470 strcat(hwLine, tmpReg);
471 break;
472 default:
473 FIXME("Unknown reg type %d %d\n", regtype, reg);
474 strcat(hwLine, "unrecognized_register");
475 break;
476 }
477
478 if (!is_input) {
479 char write_mask[6];
480 shader_arb_get_write_mask(arg, param, write_mask);
481 strcat(hwLine, write_mask);
482 } else {
483 char swizzle[6];
484 shader_arb_get_swizzle(param, is_color, swizzle);
485 strcat(hwLine, swizzle);
486 }
487 }
488
489 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected) {
490 SHADER_BUFFER* buffer = arg->buffer;
491 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
492 const char *tex_type;
493
494 switch(sampler_type) {
495 case WINED3DSTT_1D:
496 tex_type = "1D";
497 break;
498
499 case WINED3DSTT_2D:
500 tex_type = "2D";
501 break;
502
503 case WINED3DSTT_VOLUME:
504 tex_type = "3D";
505 break;
506
507 case WINED3DSTT_CUBE:
508 tex_type = "CUBE";
509 break;
510
511 default:
512 ERR("Unexpected texture type %d\n", sampler_type);
513 tex_type = "";
514 }
515
516 if (projected) {
517 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
518 } else {
519 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
520 }
521 }
522
523 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
524 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
525 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
526 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
527 WINED3DFORMAT fmt;
528 WINED3DFORMAT conversion_group;
529 IWineD3DBaseTextureImpl *texture;
530 UINT i;
531 BOOL recorded = FALSE;
532 DWORD sampler_idx;
533 DWORD hex_version = shader->baseShader.hex_version;
534 char reg[256];
535 char writemask[6];
536
537 switch(arg->opcode->opcode) {
538 case WINED3DSIO_TEX:
539 if (hex_version < WINED3DPS_VERSION(2,0)) {
540 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
541 } else {
542 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
543 }
544 break;
545
546 case WINED3DSIO_TEXLDL:
547 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
548 return;
549
550 case WINED3DSIO_TEXDP3TEX:
551 case WINED3DSIO_TEXM3x3TEX:
552 case WINED3DSIO_TEXM3x3SPEC:
553 case WINED3DSIO_TEXM3x3VSPEC:
554 case WINED3DSIO_TEXBEM:
555 case WINED3DSIO_TEXREG2AR:
556 case WINED3DSIO_TEXREG2GB:
557 case WINED3DSIO_TEXREG2RGB:
558 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
559 break;
560
561 default:
562 /* Not a texture sampling instruction, nothing to do */
563 return;
564 };
565
566 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
567 if(texture) {
568 fmt = texture->resource.format;
569 conversion_group = texture->baseTexture.shader_conversion_group;
570 } else {
571 fmt = WINED3DFMT_UNKNOWN;
572 conversion_group = WINED3DFMT_UNKNOWN;
573 }
574
575 /* before doing anything, record the sampler with the format in the format conversion list,
576 * but check if it's not there already
577 */
578 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
579 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
580 recorded = TRUE;
581 }
582 }
583 if(!recorded) {
584 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
585 shader->baseShader.num_sampled_samplers++;
586 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
587 }
588
589 pshader_get_register_name(arg->dst, reg);
590 shader_arb_get_write_mask(arg, arg->dst, writemask);
591 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
592
593 switch(fmt) {
594 case WINED3DFMT_V8U8:
595 case WINED3DFMT_V16U16:
596 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
597 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
598 #if 0
599 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
600 * disabled until an application that needs it is found because it causes unneeded
601 * shader recompilation in some game
602 */
603 if(strlen(writemask) >= 4) {
604 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
605 }
606 #endif
607 } else {
608 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
609 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
610 * all registers, do so, this saves an instruction.
611 */
612 if(strlen(writemask) >= 5) {
613 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
614 } else if(strlen(writemask) >= 3) {
615 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
616 reg, writemask[1],
617 reg, writemask[1]);
618 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
619 reg, writemask[2],
620 reg, writemask[2]);
621 } else if(strlen(writemask) == 2) {
622 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
623 reg, writemask[1]);
624 }
625 }
626 break;
627
628 case WINED3DFMT_X8L8V8U8:
629 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
630 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
631 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
632 */
633 if(strlen(writemask) >= 3) {
634 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
635 reg, writemask[1],
636 reg, writemask[1]);
637 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
638 reg, writemask[2],
639 reg, writemask[2]);
640 } else if(strlen(writemask) == 2) {
641 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
642 reg, writemask[1],
643 reg, writemask[1]);
644 }
645 }
646 break;
647
648 case WINED3DFMT_L6V5U5:
649 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
650 if(strlen(writemask) >= 4) {
651 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
652 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
653 reg, writemask[2]);
654 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
655 reg, writemask[1], writemask[1],
656 reg, writemask[1], writemask[3]);
657 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
658 writemask[3]);
659 } else if(strlen(writemask) == 3) {
660 /* This is bad: We have VL, but we need VU */
661 FIXME("2 components sampled from a converted L6V5U5 texture\n");
662 } else {
663 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
664 reg, writemask[1],
665 reg, writemask[1]);
666 }
667 }
668 break;
669
670 case WINED3DFMT_Q8W8V8U8:
671 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
672 /* Correct the sign in all channels */
673 switch(strlen(writemask)) {
674 case 4:
675 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
676 reg, writemask[3],
677 reg, writemask[3]);
678 /* drop through */
679 case 3:
680 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
681 reg, writemask[2],
682 reg, writemask[2]);
683 /* drop through */
684 case 2:
685 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
686 reg, writemask[1],
687 reg, writemask[1]);
688 break;
689
690 /* Should not occur, since it's at minimum '.' and a letter */
691 case 1:
692 ERR("Unexpected writemask: \"%s\"\n", writemask);
693 break;
694
695 case 5:
696 default:
697 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
698 }
699 }
700 break;
701
702 /* stupid compiler */
703 default:
704 break;
705 }
706 }
707
708
709 static void pshader_gen_input_modifier_line (
710 SHADER_BUFFER* buffer,
711 const DWORD instr,
712 int tmpreg,
713 char *outregstr) {
714
715 /* Generate a line that does the input modifier computation and return the input register to use */
716 char regstr[256];
717 char swzstr[20];
718 int insert_line;
719
720 /* Assume a new line will be added */
721 insert_line = 1;
722
723 /* Get register name */
724 pshader_get_register_name(instr, regstr);
725 shader_arb_get_swizzle(instr, FALSE, swzstr);
726
727 switch (instr & WINED3DSP_SRCMOD_MASK) {
728 case WINED3DSPSM_NONE:
729 sprintf(outregstr, "%s%s", regstr, swzstr);
730 insert_line = 0;
731 break;
732 case WINED3DSPSM_NEG:
733 sprintf(outregstr, "-%s%s", regstr, swzstr);
734 insert_line = 0;
735 break;
736 case WINED3DSPSM_BIAS:
737 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
738 break;
739 case WINED3DSPSM_BIASNEG:
740 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
741 break;
742 case WINED3DSPSM_SIGN:
743 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
744 break;
745 case WINED3DSPSM_SIGNNEG:
746 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
747 break;
748 case WINED3DSPSM_COMP:
749 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
750 break;
751 case WINED3DSPSM_X2:
752 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
753 break;
754 case WINED3DSPSM_X2NEG:
755 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
756 break;
757 case WINED3DSPSM_DZ:
758 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
759 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
760 break;
761 case WINED3DSPSM_DW:
762 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
763 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
764 break;
765 default:
766 sprintf(outregstr, "%s%s", regstr, swzstr);
767 insert_line = 0;
768 }
769
770 /* Return modified or original register, with swizzle */
771 if (insert_line)
772 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
773 }
774
775 static inline void pshader_gen_output_modifier_line(
776 SHADER_BUFFER* buffer,
777 int saturate,
778 char *write_mask,
779 int shift,
780 char *regstr) {
781
782 /* Generate a line that does the output modifier computation */
783 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
784 regstr, write_mask, regstr, shift_tab[shift]);
785 }
786
787 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
788 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
789
790 SHADER_BUFFER* buffer = arg->buffer;
791 char dst_name[50];
792 char src_name[2][50];
793 char dst_wmask[20];
794
795 pshader_get_register_name(arg->dst, dst_name);
796 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
797 strcat(dst_name, dst_wmask);
798
799 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
800 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
801
802 if(This->bumpenvmatconst != -1) {
803 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
804 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
805 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
806 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
807 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
808
809 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
810 } else {
811 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
812 }
813 }
814
815 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
816
817 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
818 SHADER_BUFFER* buffer = arg->buffer;
819 char dst_wmask[20];
820 char dst_name[50];
821 char src_name[3][50];
822
823 /* FIXME: support output modifiers */
824
825 /* Handle output register */
826 pshader_get_register_name(arg->dst, dst_name);
827 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
828 strcat(dst_name, dst_wmask);
829
830 /* Generate input register names (with modifiers) */
831 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
832 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
833 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
834
835 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
836 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
837 arg->opcode_token & WINED3DSI_COISSUE) {
838 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[1]);
839 } else {
840 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
841 shader_addline(buffer, "CMP %s, TMP, %s, %s;\n", dst_name, src_name[1], src_name[2]);
842 }
843 }
844
845 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
846
847 SHADER_BUFFER* buffer = arg->buffer;
848 char dst_wmask[20];
849 char dst_name[50];
850 char src_name[3][50];
851
852 /* FIXME: support output modifiers */
853
854 /* Handle output register */
855 pshader_get_register_name(arg->dst, dst_name);
856 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
857 strcat(dst_name, dst_wmask);
858
859 /* Generate input register names (with modifiers) */
860 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
861 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
862 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
863
864 shader_addline(buffer, "CMP %s, %s, %s, %s;\n", dst_name,
865 src_name[0], src_name[2], src_name[1]);
866 }
867
868 /* Map the opcode 1-to-1 to the GL code */
869 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
870
871 CONST SHADER_OPCODE* curOpcode = arg->opcode;
872 SHADER_BUFFER* buffer = arg->buffer;
873 DWORD dst = arg->dst;
874 DWORD* src = arg->src;
875
876 unsigned int i;
877 char tmpLine[256];
878
879 /* Output token related */
880 char output_rname[256];
881 char output_wmask[20];
882 BOOL saturate = FALSE;
883 BOOL centroid = FALSE;
884 BOOL partialprecision = FALSE;
885 DWORD shift;
886
887 strcpy(tmpLine, curOpcode->glname);
888
889 /* Process modifiers */
890 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
891 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
892
893 saturate = mask & WINED3DSPDM_SATURATE;
894 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
895 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
896 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
897 if (mask)
898 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
899
900 if (centroid)
901 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
902 }
903 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
904
905 /* Generate input and output registers */
906 if (curOpcode->num_params > 0) {
907 char operands[4][100];
908
909 /* Generate input register names (with modifiers) */
910 for (i = 1; i < curOpcode->num_params; ++i)
911 pshader_gen_input_modifier_line(buffer, src[i-1], i-1, operands[i]);
912
913 /* Handle output register */
914 pshader_get_register_name(dst, output_rname);
915 strcpy(operands[0], output_rname);
916 shader_arb_get_write_mask(arg, dst, output_wmask);
917 strcat(operands[0], output_wmask);
918
919 if (saturate && (shift == 0))
920 strcat(tmpLine, "_SAT");
921 strcat(tmpLine, " ");
922 strcat(tmpLine, operands[0]);
923 for (i = 1; i < curOpcode->num_params; i++) {
924 strcat(tmpLine, ", ");
925 strcat(tmpLine, operands[i]);
926 }
927 strcat(tmpLine,";\n");
928 shader_addline(buffer, tmpLine);
929
930 /* A shift requires another line. */
931 if (shift != 0)
932 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
933 }
934 }
935
936 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
937 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
938 DWORD hex_version = This->baseShader.hex_version;
939 SHADER_BUFFER* buffer = arg->buffer;
940 char reg_dest[40];
941
942 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
943 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
944 */
945 pshader_get_register_name(arg->dst, reg_dest);
946
947 if(hex_version >= WINED3DPS_VERSION(2,0)) {
948 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
949 shader_addline(buffer, "KIL %s;\n", reg_dest);
950 } else {
951 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
952 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
953 */
954 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
955 shader_addline(buffer, "MOV TMP.w, one.w;\n", reg_dest);
956 shader_addline(buffer, "KIL TMP;\n", reg_dest);
957 }
958 }
959
960 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
961 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
962 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
963
964 DWORD dst = arg->dst;
965 DWORD* src = arg->src;
966 SHADER_BUFFER* buffer = arg->buffer;
967 DWORD hex_version = This->baseShader.hex_version;
968 BOOL projected = FALSE;
969
970 char reg_dest[40];
971 char reg_coord[40];
972 DWORD reg_dest_code;
973 DWORD reg_sampler_code;
974
975 /* All versions have a destination register */
976 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
977 pshader_get_register_name(dst, reg_dest);
978
979 /* 1.0-1.3: Use destination register as coordinate source.
980 1.4+: Use provided coordinate source register. */
981 if (hex_version < WINED3DPS_VERSION(1,4))
982 strcpy(reg_coord, reg_dest);
983 else
984 pshader_gen_input_modifier_line(buffer, src[0], 0, reg_coord);
985
986 /* 1.0-1.4: Use destination register number as texture code.
987 2.0+: Use provided sampler number as texure code. */
988 if (hex_version < WINED3DPS_VERSION(2,0))
989 reg_sampler_code = reg_dest_code;
990 else
991 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
992
993 /* projection flag:
994 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
995 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
996 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
997 */
998 if(hex_version < WINED3DPS_VERSION(1,4)) {
999 DWORD flags = 0;
1000 if(reg_sampler_code < MAX_TEXTURES) {
1001 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1002 }
1003 if (flags & WINED3DTTFF_PROJECTED) {
1004 projected = TRUE;
1005 }
1006 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1007 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1008 if (src_mod == WINED3DSPSM_DZ) {
1009 projected = TRUE;
1010 } else if(src_mod == WINED3DSPSM_DW) {
1011 projected = TRUE;
1012 }
1013 } else {
1014 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1015 projected = TRUE;
1016 }
1017 }
1018 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected);
1019 }
1020
1021 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1022
1023 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1024 DWORD dst = arg->dst;
1025 SHADER_BUFFER* buffer = arg->buffer;
1026 DWORD hex_version = This->baseShader.hex_version;
1027
1028 char tmp[20];
1029 shader_arb_get_write_mask(arg, dst, tmp);
1030 if (hex_version != WINED3DPS_VERSION(1,4)) {
1031 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1032 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1033 } else {
1034 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1035 char reg_src[40];
1036
1037 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, reg_src);
1038 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1039 }
1040 }
1041
1042 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1043
1044 SHADER_BUFFER* buffer = arg->buffer;
1045 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1046 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1047 DWORD flags;
1048
1049 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1050 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1051 char dst_str[8];
1052
1053 sprintf(dst_str, "T%u", reg1);
1054 shader_addline(buffer, "MOV TMP.r, T%u.a;\n", reg2);
1055 shader_addline(buffer, "MOV TMP.g, T%u.r;\n", reg2);
1056 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1057 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED);
1058 }
1059
1060 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1061
1062 SHADER_BUFFER* buffer = arg->buffer;
1063 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1064 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1065 DWORD flags;
1066
1067 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1068 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1069 char dst_str[8];
1070
1071 sprintf(dst_str, "T%u", reg1);
1072 shader_addline(buffer, "MOV TMP.r, T%u.g;\n", reg2);
1073 shader_addline(buffer, "MOV TMP.g, T%u.b;\n", reg2);
1074 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1075 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE);
1076 }
1077
1078 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1079 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1080
1081 DWORD dst = arg->dst;
1082 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1083 SHADER_BUFFER* buffer = arg->buffer;
1084
1085 char reg_coord[40];
1086 DWORD reg_dest_code;
1087
1088 /* All versions have a destination register */
1089 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1090 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1091 pshader_get_register_name(dst, reg_coord);
1092
1093 if(This->bumpenvmatconst != -1) {
1094 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1095
1096 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
1097 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1098 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
1099 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1100
1101 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1102 * so we can't let the GL handle this.
1103 */
1104 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1105 & WINED3DTTFF_PROJECTED) {
1106 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1107 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1108 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1109 } else {
1110 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1111 }
1112
1113 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE);
1114 } else {
1115 DWORD tf;
1116 if(reg_dest_code < MAX_TEXTURES) {
1117 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1118 } else {
1119 tf = 0;
1120 }
1121 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1122 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED);
1123 }
1124 }
1125
1126 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1127
1128 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1129 SHADER_BUFFER* buffer = arg->buffer;
1130 char src0_name[50];
1131
1132 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1133 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1134 }
1135
1136 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1137
1138 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1139 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1140 DWORD flags;
1141 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1142 SHADER_BUFFER* buffer = arg->buffer;
1143 char dst_str[8];
1144 char src0_name[50];
1145
1146 sprintf(dst_str, "T%u", reg);
1147 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1148 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1149 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1150 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED);
1151 }
1152
1153 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1154
1155 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1156 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1157 SHADER_BUFFER* buffer = arg->buffer;
1158 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1159 char src0_name[50];
1160
1161 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1162 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1163 current_state->texcoord_w[current_state->current_row++] = reg;
1164 }
1165
1166 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1167
1168 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1169 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1170 DWORD flags;
1171 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1172 SHADER_BUFFER* buffer = arg->buffer;
1173 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1174 char dst_str[8];
1175 char src0_name[50];
1176
1177 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1178 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1179
1180 /* Sample the texture using the calculated coordinates */
1181 sprintf(dst_str, "T%u", reg);
1182 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1183 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED);
1184 current_state->current_row = 0;
1185 }
1186
1187 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1188
1189 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1190 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1191 DWORD flags;
1192 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1193 SHADER_BUFFER* buffer = arg->buffer;
1194 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1195 char dst_str[8];
1196 char src0_name[50];
1197
1198 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1199 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1200
1201 /* Construct the eye-ray vector from w coordinates */
1202 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1203 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1204 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1205
1206 /* Calculate reflection vector
1207 */
1208 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1209 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1210 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1211 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1212 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1213 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1214 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1215
1216 /* Sample the texture using the calculated coordinates */
1217 sprintf(dst_str, "T%u", reg);
1218 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1219 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED);
1220 current_state->current_row = 0;
1221 }
1222
1223 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1224
1225 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1226 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1227 DWORD flags;
1228 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1229 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1230 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1231 SHADER_BUFFER* buffer = arg->buffer;
1232 char dst_str[8];
1233 char src0_name[50];
1234
1235 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1236 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1237
1238 /* Calculate reflection vector.
1239 *
1240 * dot(N, E)
1241 * TMP.xyz = 2 * --------- * N - E
1242 * dot(N, N)
1243 *
1244 * Which normalizes the normal vector
1245 */
1246 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1247 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1248 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1249 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1250 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1251 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1252
1253 /* Sample the texture using the calculated coordinates */
1254 sprintf(dst_str, "T%u", reg);
1255 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1256 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED);
1257 current_state->current_row = 0;
1258 }
1259
1260 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1261 SHADER_BUFFER* buffer = arg->buffer;
1262 char dst_name[50];
1263
1264 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1265 * which is essentially an input, is the destiantion register because it is the first
1266 * param. According to the msdn, this must be register r5, but let's keep it more flexible
1267 * here
1268 */
1269 pshader_get_register_name(arg->dst, dst_name);
1270
1271 /* According to the msdn, the source register(must be r5) is unusable after
1272 * the texdepth instruction, so we're free to modify it
1273 */
1274 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1275
1276 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1277 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1278 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1279 */
1280 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1281 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1282 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n", dst_name, dst_name);
1283 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n", dst_name, dst_name);
1284 }
1285
1286 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1287 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1288 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1289 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1290 SHADER_BUFFER* buffer = arg->buffer;
1291 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1292 char src0[50];
1293 char dst_str[8];
1294
1295 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1296 shader_addline(buffer, "MOV TMP, 0.0;\n");
1297 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1298
1299 sprintf(dst_str, "T%u", sampler_idx);
1300 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */);
1301 }
1302
1303 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1304 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1305 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1306 char src0[50];
1307 char dst_str[50];
1308 char dst_mask[6];
1309 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1310 SHADER_BUFFER* buffer = arg->buffer;
1311
1312 /* Handle output register */
1313 pshader_get_register_name(arg->dst, dst_str);
1314 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1315
1316 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1317 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1318
1319 /* TODO: Handle output modifiers */
1320 }
1321
1322 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1323 * Perform the 3rd row of a 3x3 matrix multiply */
1324 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1325 SHADER_BUFFER* buffer = arg->buffer;
1326 char dst_str[50];
1327 char dst_mask[6];
1328 char src0[50];
1329 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1330
1331 pshader_get_register_name(arg->dst, dst_str);
1332 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1333
1334 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1335 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1336 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1337
1338 /* TODO: Handle output modifiers */
1339 }
1340
1341 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1342 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1343 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1344 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1345 */
1346 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1347 SHADER_BUFFER* buffer = arg->buffer;
1348 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1349 char src0[50];
1350
1351 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1352 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1353
1354 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1355 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1356 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1357 */
1358 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1359 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1360 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1361 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1362 }
1363
1364 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1365 Vertex shaders to ARB_vertex_program codes */
1366 void vshader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1367
1368 int i;
1369 int nComponents = 0;
1370 SHADER_OPCODE_ARG tmpArg;
1371
1372 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1373
1374 /* Set constants for the temporary argument */
1375 tmpArg.shader = arg->shader;
1376 tmpArg.buffer = arg->buffer;
1377 tmpArg.src[0] = arg->src[0];
1378 tmpArg.src_addr[0] = arg->src_addr[0];
1379 tmpArg.src_addr[1] = arg->src_addr[1];
1380 tmpArg.reg_maps = arg->reg_maps;
1381
1382 switch(arg->opcode->opcode) {
1383 case WINED3DSIO_M4x4:
1384 nComponents = 4;
1385 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1386 break;
1387 case WINED3DSIO_M4x3:
1388 nComponents = 3;
1389 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1390 break;
1391 case WINED3DSIO_M3x4:
1392 nComponents = 4;
1393 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1394 break;
1395 case WINED3DSIO_M3x3:
1396 nComponents = 3;
1397 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1398 break;
1399 case WINED3DSIO_M3x2:
1400 nComponents = 2;
1401 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1402 break;
1403 default:
1404 break;
1405 }
1406
1407 for (i = 0; i < nComponents; i++) {
1408 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1409 tmpArg.src[1] = arg->src[1]+i;
1410 vshader_hw_map2gl(&tmpArg);
1411 }
1412 }
1413
1414 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1415 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1416 SHADER_BUFFER* buffer = arg->buffer;
1417 DWORD dst = arg->dst;
1418 DWORD src = arg->src[0];
1419 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1420
1421 char tmpLine[256];
1422
1423 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1424 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1425 strcat(tmpLine, ",");
1426 vshader_program_add_param(arg, src, TRUE, tmpLine);
1427 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1428 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1429 * .w is used
1430 */
1431 strcat(tmpLine, ".w");
1432 }
1433
1434 shader_addline(buffer, "%s;\n", tmpLine);
1435 }
1436
1437 /* TODO: merge with pixel shader */
1438 /* Map the opcode 1-to-1 to the GL code */
1439 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1440
1441 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1442 SHADER_BUFFER* buffer = arg->buffer;
1443 DWORD dst = arg->dst;
1444 DWORD* src = arg->src;
1445
1446 DWORD dst_regtype = shader_get_regtype(dst);
1447 char tmpLine[256];
1448 unsigned int i;
1449
1450 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA)
1451 strcpy(tmpLine, "ARL");
1452 else
1453 strcpy(tmpLine, curOpcode->glname);
1454
1455 if (curOpcode->num_params > 0) {
1456 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1457 for (i = 1; i < curOpcode->num_params; ++i) {
1458 strcat(tmpLine, ",");
1459 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1460 }
1461 }
1462 shader_addline(buffer, "%s;\n", tmpLine);
1463 }
1464
1465 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1466 GLuint program_id = 0;
1467 const char *blt_vprogram =
1468 "!!ARBvp1.0\n"
1469 "PARAM c[1] = { { 1, 0.5 } };\n"
1470 "MOV result.position, vertex.position;\n"
1471 "MOV result.color, c[0].x;\n"
1472 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1473 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1474 "END\n";
1475
1476 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1477 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1478 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1479
1480 if (glGetError() == GL_INVALID_OPERATION) {
1481 GLint pos;
1482 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1483 FIXME("Vertex program error at position %d: %s\n", pos,
1484 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1485 }
1486
1487 return program_id;
1488 }
1489
1490 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1491 GLuint program_id = 0;
1492 const char *blt_fprogram =
1493 "!!ARBfp1.0\n"
1494 "TEMP R0;\n"
1495 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1496 "MOV result.depth.z, R0.x;\n"
1497 "END\n";
1498
1499 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1500 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1501 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1502
1503 if (glGetError() == GL_INVALID_OPERATION) {
1504 GLint pos;
1505 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1506 FIXME("Fragment program error at position %d: %s\n", pos,
1507 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1508 }
1509
1510 return program_id;
1511 }
1512
1513 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1514 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1515 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1516
1517 if (useVS) {
1518 TRACE("Using vertex shader\n");
1519
1520 /* Bind the vertex program */
1521 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1522 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1523 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1524
1525 /* Enable OpenGL vertex programs */
1526 glEnable(GL_VERTEX_PROGRAM_ARB);
1527 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1528 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1529 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1530 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1531 glDisable(GL_VERTEX_PROGRAM_ARB);
1532 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1533 }
1534
1535 if (usePS) {
1536 TRACE("Using pixel shader\n");
1537
1538 /* Bind the fragment program */
1539 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1540 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1541 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1542
1543 /* Enable OpenGL fragment programs */
1544 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1545 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1546 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1547 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1548 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1549 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1550 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1551 }
1552 }
1553
1554 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1555 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1556 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1557 static GLuint vprogram_id = 0;
1558 static GLuint fprogram_id = 0;
1559
1560 if (!vprogram_id) vprogram_id = create_arb_blt_vertex_program(gl_info);
1561 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vprogram_id));
1562 glEnable(GL_VERTEX_PROGRAM_ARB);
1563
1564 if (!fprogram_id) fprogram_id = create_arb_blt_fragment_program(gl_info);
1565 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, fprogram_id));
1566 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1567 }
1568
1569 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1570 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1571 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1572 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1573 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1574 }
1575
1576 const shader_backend_t arb_program_shader_backend = {
1577 &shader_arb_select,
1578 &shader_arb_select_depth_blt,
1579 &shader_arb_load_constants,
1580 &shader_arb_cleanup,
1581 &shader_arb_color_correction
1582 };
useful hacks and other patches not (yet) suitable for mainline