?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push ab3cc77eb114d2bd400734a9dccad24563f936a6
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
blob62b5c74e5c88c3e8e1cc95ac5985ce834bd1c28c
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, BOOL bias) {
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 (bias) {
517 /* Shouldn't be possible, but let's check for it */
518 if(projected) FIXME("Biased and Projected texture sampling\n");
519 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
520 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
521 } else if (projected) {
522 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
523 } else {
524 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
525 }
526 }
527
528 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
529 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
530 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
531 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
532 WINED3DFORMAT fmt;
533 WINED3DFORMAT conversion_group;
534 IWineD3DBaseTextureImpl *texture;
535 UINT i;
536 BOOL recorded = FALSE;
537 DWORD sampler_idx;
538 DWORD hex_version = shader->baseShader.hex_version;
539 char reg[256];
540 char writemask[6];
541
542 switch(arg->opcode->opcode) {
543 case WINED3DSIO_TEX:
544 if (hex_version < WINED3DPS_VERSION(2,0)) {
545 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
546 } else {
547 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
548 }
549 break;
550
551 case WINED3DSIO_TEXLDL:
552 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
553 return;
554
555 case WINED3DSIO_TEXDP3TEX:
556 case WINED3DSIO_TEXM3x3TEX:
557 case WINED3DSIO_TEXM3x3SPEC:
558 case WINED3DSIO_TEXM3x3VSPEC:
559 case WINED3DSIO_TEXBEM:
560 case WINED3DSIO_TEXREG2AR:
561 case WINED3DSIO_TEXREG2GB:
562 case WINED3DSIO_TEXREG2RGB:
563 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
564 break;
565
566 default:
567 /* Not a texture sampling instruction, nothing to do */
568 return;
569 };
570
571 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
572 if(texture) {
573 fmt = texture->resource.format;
574 conversion_group = texture->baseTexture.shader_conversion_group;
575 } else {
576 fmt = WINED3DFMT_UNKNOWN;
577 conversion_group = WINED3DFMT_UNKNOWN;
578 }
579
580 /* before doing anything, record the sampler with the format in the format conversion list,
581 * but check if it's not there already
582 */
583 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
584 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
585 recorded = TRUE;
586 }
587 }
588 if(!recorded) {
589 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
590 shader->baseShader.num_sampled_samplers++;
591 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
592 }
593
594 pshader_get_register_name(arg->dst, reg);
595 shader_arb_get_write_mask(arg, arg->dst, writemask);
596 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
597
598 switch(fmt) {
599 case WINED3DFMT_V8U8:
600 case WINED3DFMT_V16U16:
601 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
602 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
603 #if 0
604 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
605 * disabled until an application that needs it is found because it causes unneeded
606 * shader recompilation in some game
607 */
608 if(strlen(writemask) >= 4) {
609 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
610 }
611 #endif
612 } else {
613 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
614 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
615 * all registers, do so, this saves an instruction.
616 */
617 if(strlen(writemask) >= 5) {
618 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
619 } else if(strlen(writemask) >= 3) {
620 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
621 reg, writemask[1],
622 reg, writemask[1]);
623 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
624 reg, writemask[2],
625 reg, writemask[2]);
626 } else if(strlen(writemask) == 2) {
627 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
628 reg, writemask[1]);
629 }
630 }
631 break;
632
633 case WINED3DFMT_X8L8V8U8:
634 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
635 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
636 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
637 */
638 if(strlen(writemask) >= 3) {
639 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
640 reg, writemask[1],
641 reg, writemask[1]);
642 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
643 reg, writemask[2],
644 reg, writemask[2]);
645 } else if(strlen(writemask) == 2) {
646 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
647 reg, writemask[1],
648 reg, writemask[1]);
649 }
650 }
651 break;
652
653 case WINED3DFMT_L6V5U5:
654 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
655 if(strlen(writemask) >= 4) {
656 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
657 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
658 reg, writemask[2]);
659 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
660 reg, writemask[1], writemask[1],
661 reg, writemask[1], writemask[3]);
662 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
663 writemask[3]);
664 } else if(strlen(writemask) == 3) {
665 /* This is bad: We have VL, but we need VU */
666 FIXME("2 components sampled from a converted L6V5U5 texture\n");
667 } else {
668 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
669 reg, writemask[1],
670 reg, writemask[1]);
671 }
672 }
673 break;
674
675 case WINED3DFMT_Q8W8V8U8:
676 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
677 /* Correct the sign in all channels */
678 switch(strlen(writemask)) {
679 case 4:
680 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
681 reg, writemask[3],
682 reg, writemask[3]);
683 /* drop through */
684 case 3:
685 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
686 reg, writemask[2],
687 reg, writemask[2]);
688 /* drop through */
689 case 2:
690 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
691 reg, writemask[1],
692 reg, writemask[1]);
693 break;
694
695 /* Should not occur, since it's at minimum '.' and a letter */
696 case 1:
697 ERR("Unexpected writemask: \"%s\"\n", writemask);
698 break;
699
700 case 5:
701 default:
702 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
703 }
704 }
705 break;
706
707 /* stupid compiler */
708 default:
709 break;
710 }
711 }
712
713
714 static void pshader_gen_input_modifier_line (
715 SHADER_BUFFER* buffer,
716 const DWORD instr,
717 int tmpreg,
718 char *outregstr) {
719
720 /* Generate a line that does the input modifier computation and return the input register to use */
721 char regstr[256];
722 char swzstr[20];
723 int insert_line;
724
725 /* Assume a new line will be added */
726 insert_line = 1;
727
728 /* Get register name */
729 pshader_get_register_name(instr, regstr);
730 shader_arb_get_swizzle(instr, FALSE, swzstr);
731
732 switch (instr & WINED3DSP_SRCMOD_MASK) {
733 case WINED3DSPSM_NONE:
734 sprintf(outregstr, "%s%s", regstr, swzstr);
735 insert_line = 0;
736 break;
737 case WINED3DSPSM_NEG:
738 sprintf(outregstr, "-%s%s", regstr, swzstr);
739 insert_line = 0;
740 break;
741 case WINED3DSPSM_BIAS:
742 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
743 break;
744 case WINED3DSPSM_BIASNEG:
745 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
746 break;
747 case WINED3DSPSM_SIGN:
748 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
749 break;
750 case WINED3DSPSM_SIGNNEG:
751 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
752 break;
753 case WINED3DSPSM_COMP:
754 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
755 break;
756 case WINED3DSPSM_X2:
757 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
758 break;
759 case WINED3DSPSM_X2NEG:
760 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
761 break;
762 case WINED3DSPSM_DZ:
763 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
764 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
765 break;
766 case WINED3DSPSM_DW:
767 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
768 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
769 break;
770 default:
771 sprintf(outregstr, "%s%s", regstr, swzstr);
772 insert_line = 0;
773 }
774
775 /* Return modified or original register, with swizzle */
776 if (insert_line)
777 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
778 }
779
780 static inline void pshader_gen_output_modifier_line(
781 SHADER_BUFFER* buffer,
782 int saturate,
783 char *write_mask,
784 int shift,
785 char *regstr) {
786
787 /* Generate a line that does the output modifier computation */
788 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
789 regstr, write_mask, regstr, shift_tab[shift]);
790 }
791
792 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
793 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
794
795 SHADER_BUFFER* buffer = arg->buffer;
796 char dst_name[50];
797 char src_name[2][50];
798 char dst_wmask[20];
799
800 pshader_get_register_name(arg->dst, dst_name);
801 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
802 strcat(dst_name, dst_wmask);
803
804 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
805 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
806
807 if(This->bumpenvmatconst != -1) {
808 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
809 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
810 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
811 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
812 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
813
814 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
815 } else {
816 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
817 }
818 }
819
820 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
821
822 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
823 SHADER_BUFFER* buffer = arg->buffer;
824 char dst_wmask[20];
825 char dst_name[50];
826 char src_name[3][50];
827 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
828 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
829
830 /* FIXME: support output modifiers */
831
832 /* Handle output register */
833 pshader_get_register_name(arg->dst, dst_name);
834 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
835
836 /* Generate input register names (with modifiers) */
837 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
838 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
839 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
840
841 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
842 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
843 arg->opcode_token & WINED3DSI_COISSUE) {
844 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
845 } else {
846 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
847 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
848 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
849 }
850 if (shift != 0)
851 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
852 }
853
854 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
855
856 SHADER_BUFFER* buffer = arg->buffer;
857 char dst_wmask[20];
858 char dst_name[50];
859 char src_name[3][50];
860 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
861 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
862
863 /* FIXME: support output modifiers */
864
865 /* Handle output register */
866 pshader_get_register_name(arg->dst, dst_name);
867 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
868
869 /* Generate input register names (with modifiers) */
870 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
871 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
872 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
873
874 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
875 src_name[0], src_name[2], src_name[1]);
876
877 if (shift != 0)
878 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
879 }
880
881 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
882 * dst = dot2(src0, src1) + src2 */
883 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
884 SHADER_BUFFER* buffer = arg->buffer;
885 char dst_wmask[20];
886 char dst_name[50];
887 char src_name[3][50];
888 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
889 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
890
891 pshader_get_register_name(arg->dst, dst_name);
892 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
893
894 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
895 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
896 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
897
898 /* Emulate a DP2 with a DP3 and 0.0 */
899 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
900 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
901 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
902 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
903
904 if (shift != 0)
905 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
906 }
907
908 /* Map the opcode 1-to-1 to the GL code */
909 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
910
911 CONST SHADER_OPCODE* curOpcode = arg->opcode;
912 SHADER_BUFFER* buffer = arg->buffer;
913 DWORD dst = arg->dst;
914 DWORD* src = arg->src;
915
916 unsigned int i;
917 char tmpLine[256];
918
919 /* Output token related */
920 char output_rname[256];
921 char output_wmask[20];
922 BOOL saturate = FALSE;
923 BOOL centroid = FALSE;
924 BOOL partialprecision = FALSE;
925 DWORD shift;
926
927 strcpy(tmpLine, curOpcode->glname);
928
929 /* Process modifiers */
930 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
931 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
932
933 saturate = mask & WINED3DSPDM_SATURATE;
934 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
935 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
936 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
937 if (mask)
938 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
939
940 if (centroid)
941 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
942 }
943 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
944
945 /* Generate input and output registers */
946 if (curOpcode->num_params > 0) {
947 char operands[4][100];
948
949 /* Generate input register names (with modifiers) */
950 for (i = 1; i < curOpcode->num_params; ++i)
951 pshader_gen_input_modifier_line(buffer, src[i-1], i-1, operands[i]);
952
953 /* Handle output register */
954 pshader_get_register_name(dst, output_rname);
955 strcpy(operands[0], output_rname);
956 shader_arb_get_write_mask(arg, dst, output_wmask);
957 strcat(operands[0], output_wmask);
958
959 if (saturate && (shift == 0))
960 strcat(tmpLine, "_SAT");
961 strcat(tmpLine, " ");
962 strcat(tmpLine, operands[0]);
963 for (i = 1; i < curOpcode->num_params; i++) {
964 strcat(tmpLine, ", ");
965 strcat(tmpLine, operands[i]);
966 }
967 strcat(tmpLine,";\n");
968 shader_addline(buffer, tmpLine);
969
970 /* A shift requires another line. */
971 if (shift != 0)
972 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
973 }
974 }
975
976 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
977 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
978 DWORD hex_version = This->baseShader.hex_version;
979 SHADER_BUFFER* buffer = arg->buffer;
980 char reg_dest[40];
981
982 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
983 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
984 */
985 pshader_get_register_name(arg->dst, reg_dest);
986
987 if(hex_version >= WINED3DPS_VERSION(2,0)) {
988 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
989 shader_addline(buffer, "KIL %s;\n", reg_dest);
990 } else {
991 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
992 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
993 */
994 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
995 shader_addline(buffer, "MOV TMP.w, one.w;\n", reg_dest);
996 shader_addline(buffer, "KIL TMP;\n", reg_dest);
997 }
998 }
999
1000 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1001 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1002 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1003
1004 DWORD dst = arg->dst;
1005 DWORD* src = arg->src;
1006 SHADER_BUFFER* buffer = arg->buffer;
1007 DWORD hex_version = This->baseShader.hex_version;
1008 BOOL projected = FALSE, bias = FALSE;
1009
1010 char reg_dest[40];
1011 char reg_coord[40];
1012 DWORD reg_dest_code;
1013 DWORD reg_sampler_code;
1014
1015 /* All versions have a destination register */
1016 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1017 pshader_get_register_name(dst, reg_dest);
1018
1019 /* 1.0-1.3: Use destination register as coordinate source.
1020 1.4+: Use provided coordinate source register. */
1021 if (hex_version < WINED3DPS_VERSION(1,4))
1022 strcpy(reg_coord, reg_dest);
1023 else
1024 pshader_gen_input_modifier_line(buffer, src[0], 0, reg_coord);
1025
1026 /* 1.0-1.4: Use destination register number as texture code.
1027 2.0+: Use provided sampler number as texure code. */
1028 if (hex_version < WINED3DPS_VERSION(2,0))
1029 reg_sampler_code = reg_dest_code;
1030 else
1031 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1032
1033 /* projection flag:
1034 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1035 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1036 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1037 */
1038 if(hex_version < WINED3DPS_VERSION(1,4)) {
1039 DWORD flags = 0;
1040 if(reg_sampler_code < MAX_TEXTURES) {
1041 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1042 }
1043 if (flags & WINED3DTTFF_PROJECTED) {
1044 projected = TRUE;
1045 }
1046 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1047 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1048 if (src_mod == WINED3DSPSM_DZ) {
1049 projected = TRUE;
1050 } else if(src_mod == WINED3DSPSM_DW) {
1051 projected = TRUE;
1052 }
1053 } else {
1054 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1055 projected = TRUE;
1056 }
1057 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1058 bias = TRUE;
1059 }
1060 }
1061 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1062 }
1063
1064 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1065
1066 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1067 DWORD dst = arg->dst;
1068 SHADER_BUFFER* buffer = arg->buffer;
1069 DWORD hex_version = This->baseShader.hex_version;
1070
1071 char tmp[20];
1072 shader_arb_get_write_mask(arg, dst, tmp);
1073 if (hex_version != WINED3DPS_VERSION(1,4)) {
1074 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1075 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1076 } else {
1077 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1078 char reg_src[40];
1079
1080 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, reg_src);
1081 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1082 }
1083 }
1084
1085 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1086
1087 SHADER_BUFFER* buffer = arg->buffer;
1088 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1089 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1090 DWORD flags;
1091
1092 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1093 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1094 char dst_str[8];
1095
1096 sprintf(dst_str, "T%u", reg1);
1097 shader_addline(buffer, "MOV TMP.r, T%u.a;\n", reg2);
1098 shader_addline(buffer, "MOV TMP.g, T%u.r;\n", reg2);
1099 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1100 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1101 }
1102
1103 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1104
1105 SHADER_BUFFER* buffer = arg->buffer;
1106 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1107 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1108 DWORD flags;
1109
1110 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1111 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1112 char dst_str[8];
1113
1114 sprintf(dst_str, "T%u", reg1);
1115 shader_addline(buffer, "MOV TMP.r, T%u.g;\n", reg2);
1116 shader_addline(buffer, "MOV TMP.g, T%u.b;\n", reg2);
1117 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1118 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1119 }
1120
1121 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1122 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1123
1124 DWORD dst = arg->dst;
1125 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1126 SHADER_BUFFER* buffer = arg->buffer;
1127
1128 char reg_coord[40];
1129 DWORD reg_dest_code;
1130
1131 /* All versions have a destination register */
1132 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1133 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1134 pshader_get_register_name(dst, reg_coord);
1135
1136 if(This->bumpenvmatconst != -1) {
1137 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1138
1139 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
1140 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1141 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
1142 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1143
1144 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1145 * so we can't let the GL handle this.
1146 */
1147 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1148 & WINED3DTTFF_PROJECTED) {
1149 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1150 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1151 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1152 } else {
1153 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1154 }
1155
1156 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1157 } else {
1158 DWORD tf;
1159 if(reg_dest_code < MAX_TEXTURES) {
1160 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1161 } else {
1162 tf = 0;
1163 }
1164 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1165 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1166 }
1167 }
1168
1169 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1170
1171 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1172 SHADER_BUFFER* buffer = arg->buffer;
1173 char src0_name[50];
1174
1175 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1176 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1177 }
1178
1179 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1180
1181 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1182 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1183 DWORD flags;
1184 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1185 SHADER_BUFFER* buffer = arg->buffer;
1186 char dst_str[8];
1187 char src0_name[50];
1188
1189 sprintf(dst_str, "T%u", reg);
1190 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1191 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1192 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1193 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1194 }
1195
1196 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1197
1198 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1199 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1200 SHADER_BUFFER* buffer = arg->buffer;
1201 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1202 char src0_name[50];
1203
1204 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1205 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1206 current_state->texcoord_w[current_state->current_row++] = reg;
1207 }
1208
1209 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1210
1211 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1212 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1213 DWORD flags;
1214 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1215 SHADER_BUFFER* buffer = arg->buffer;
1216 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1217 char dst_str[8];
1218 char src0_name[50];
1219
1220 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1221 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1222
1223 /* Sample the texture using the calculated coordinates */
1224 sprintf(dst_str, "T%u", reg);
1225 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1226 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1227 current_state->current_row = 0;
1228 }
1229
1230 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1231
1232 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1233 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1234 DWORD flags;
1235 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1236 SHADER_BUFFER* buffer = arg->buffer;
1237 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1238 char dst_str[8];
1239 char src0_name[50];
1240
1241 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1242 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1243
1244 /* Construct the eye-ray vector from w coordinates */
1245 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1246 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1247 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1248
1249 /* Calculate reflection vector
1250 */
1251 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1252 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1253 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1254 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1255 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1256 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1257 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1258
1259 /* Sample the texture using the calculated coordinates */
1260 sprintf(dst_str, "T%u", reg);
1261 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1262 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1263 current_state->current_row = 0;
1264 }
1265
1266 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1267
1268 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1269 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1270 DWORD flags;
1271 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1272 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1273 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1274 SHADER_BUFFER* buffer = arg->buffer;
1275 char dst_str[8];
1276 char src0_name[50];
1277
1278 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1279 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1280
1281 /* Calculate reflection vector.
1282 *
1283 * dot(N, E)
1284 * TMP.xyz = 2 * --------- * N - E
1285 * dot(N, N)
1286 *
1287 * Which normalizes the normal vector
1288 */
1289 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1290 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1291 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1292 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1293 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1294 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1295
1296 /* Sample the texture using the calculated coordinates */
1297 sprintf(dst_str, "T%u", reg);
1298 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1299 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1300 current_state->current_row = 0;
1301 }
1302
1303 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1304 SHADER_BUFFER* buffer = arg->buffer;
1305 char dst_name[50];
1306
1307 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1308 * which is essentially an input, is the destiantion register because it is the first
1309 * param. According to the msdn, this must be register r5, but let's keep it more flexible
1310 * here
1311 */
1312 pshader_get_register_name(arg->dst, dst_name);
1313
1314 /* According to the msdn, the source register(must be r5) is unusable after
1315 * the texdepth instruction, so we're free to modify it
1316 */
1317 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1318
1319 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1320 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1321 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1322 */
1323 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1324 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1325 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n", dst_name, dst_name);
1326 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n", dst_name, dst_name);
1327 }
1328
1329 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1330 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1331 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1332 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1333 SHADER_BUFFER* buffer = arg->buffer;
1334 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1335 char src0[50];
1336 char dst_str[8];
1337
1338 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1339 shader_addline(buffer, "MOV TMP, 0.0;\n");
1340 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1341
1342 sprintf(dst_str, "T%u", sampler_idx);
1343 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1344 }
1345
1346 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1347 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1348 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1349 char src0[50];
1350 char dst_str[50];
1351 char dst_mask[6];
1352 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1353 SHADER_BUFFER* buffer = arg->buffer;
1354
1355 /* Handle output register */
1356 pshader_get_register_name(arg->dst, dst_str);
1357 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1358
1359 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1360 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1361
1362 /* TODO: Handle output modifiers */
1363 }
1364
1365 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1366 * Perform the 3rd row of a 3x3 matrix multiply */
1367 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1368 SHADER_BUFFER* buffer = arg->buffer;
1369 char dst_str[50];
1370 char dst_mask[6];
1371 char src0[50];
1372 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1373
1374 pshader_get_register_name(arg->dst, dst_str);
1375 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1376
1377 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1378 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1379 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1380
1381 /* TODO: Handle output modifiers */
1382 }
1383
1384 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1385 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1386 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1387 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1388 */
1389 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1390 SHADER_BUFFER* buffer = arg->buffer;
1391 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1392 char src0[50];
1393
1394 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1395 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1396
1397 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1398 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1399 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1400 */
1401 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1402 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1403 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1404 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1405 }
1406
1407 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1408 Vertex shaders to ARB_vertex_program codes */
1409 void vshader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1410
1411 int i;
1412 int nComponents = 0;
1413 SHADER_OPCODE_ARG tmpArg;
1414
1415 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1416
1417 /* Set constants for the temporary argument */
1418 tmpArg.shader = arg->shader;
1419 tmpArg.buffer = arg->buffer;
1420 tmpArg.src[0] = arg->src[0];
1421 tmpArg.src_addr[0] = arg->src_addr[0];
1422 tmpArg.src_addr[1] = arg->src_addr[1];
1423 tmpArg.reg_maps = arg->reg_maps;
1424
1425 switch(arg->opcode->opcode) {
1426 case WINED3DSIO_M4x4:
1427 nComponents = 4;
1428 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1429 break;
1430 case WINED3DSIO_M4x3:
1431 nComponents = 3;
1432 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1433 break;
1434 case WINED3DSIO_M3x4:
1435 nComponents = 4;
1436 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1437 break;
1438 case WINED3DSIO_M3x3:
1439 nComponents = 3;
1440 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1441 break;
1442 case WINED3DSIO_M3x2:
1443 nComponents = 2;
1444 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1445 break;
1446 default:
1447 break;
1448 }
1449
1450 for (i = 0; i < nComponents; i++) {
1451 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1452 tmpArg.src[1] = arg->src[1]+i;
1453 vshader_hw_map2gl(&tmpArg);
1454 }
1455 }
1456
1457 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1458 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1459 SHADER_BUFFER* buffer = arg->buffer;
1460 DWORD dst = arg->dst;
1461 DWORD src = arg->src[0];
1462 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1463
1464 char tmpLine[256];
1465
1466 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1467 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1468 strcat(tmpLine, ",");
1469 vshader_program_add_param(arg, src, TRUE, tmpLine);
1470 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1471 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1472 * .w is used
1473 */
1474 strcat(tmpLine, ".w");
1475 }
1476
1477 shader_addline(buffer, "%s;\n", tmpLine);
1478 }
1479
1480 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1481 SHADER_BUFFER* buffer = arg->buffer;
1482 char dst_name[50];
1483 char src_name[50];
1484 char dst_wmask[20];
1485 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1486 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1487
1488 pshader_get_register_name(arg->dst, dst_name);
1489 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1490
1491 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1492 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1493 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1494 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1495 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1496 src_name);
1497
1498 if (shift != 0)
1499 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1500 }
1501
1502 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1503 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1504 * must contain fixed constants. So we need a separate functin to filter those constants and
1505 * can't use map2gl
1506 */
1507 SHADER_BUFFER* buffer = arg->buffer;
1508 char dst_name[50];
1509 char src_name[50];
1510 char dst_wmask[20];
1511 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1512 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1513
1514 pshader_get_register_name(arg->dst, dst_name);
1515 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1516
1517 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1518 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1519 src_name);
1520
1521 if (shift != 0)
1522 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1523
1524 }
1525
1526 /* TODO: merge with pixel shader */
1527 /* Map the opcode 1-to-1 to the GL code */
1528 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1529
1530 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1531 SHADER_BUFFER* buffer = arg->buffer;
1532 DWORD dst = arg->dst;
1533 DWORD* src = arg->src;
1534
1535 DWORD dst_regtype = shader_get_regtype(dst);
1536 char tmpLine[256];
1537 unsigned int i;
1538
1539 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA)
1540 strcpy(tmpLine, "ARL");
1541 else
1542 strcpy(tmpLine, curOpcode->glname);
1543
1544 if (curOpcode->num_params > 0) {
1545 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1546 for (i = 1; i < curOpcode->num_params; ++i) {
1547 strcat(tmpLine, ",");
1548 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1549 }
1550 }
1551 shader_addline(buffer, "%s;\n", tmpLine);
1552 }
1553
1554 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1555 GLuint program_id = 0;
1556 const char *blt_vprogram =
1557 "!!ARBvp1.0\n"
1558 "PARAM c[1] = { { 1, 0.5 } };\n"
1559 "MOV result.position, vertex.position;\n"
1560 "MOV result.color, c[0].x;\n"
1561 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1562 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1563 "END\n";
1564
1565 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1566 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1567 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1568
1569 if (glGetError() == GL_INVALID_OPERATION) {
1570 GLint pos;
1571 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1572 FIXME("Vertex program error at position %d: %s\n", pos,
1573 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1574 }
1575
1576 return program_id;
1577 }
1578
1579 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1580 GLuint program_id = 0;
1581 const char *blt_fprogram =
1582 "!!ARBfp1.0\n"
1583 "TEMP R0;\n"
1584 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1585 "MOV result.depth.z, R0.x;\n"
1586 "END\n";
1587
1588 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1589 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1590 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1591
1592 if (glGetError() == GL_INVALID_OPERATION) {
1593 GLint pos;
1594 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1595 FIXME("Fragment program error at position %d: %s\n", pos,
1596 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1597 }
1598
1599 return program_id;
1600 }
1601
1602 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1603 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1604 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1605
1606 if (useVS) {
1607 TRACE("Using vertex shader\n");
1608
1609 /* Bind the vertex program */
1610 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1611 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1612 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1613
1614 /* Enable OpenGL vertex programs */
1615 glEnable(GL_VERTEX_PROGRAM_ARB);
1616 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1617 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1618 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1619 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1620 glDisable(GL_VERTEX_PROGRAM_ARB);
1621 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1622 }
1623
1624 if (usePS) {
1625 TRACE("Using pixel shader\n");
1626
1627 /* Bind the fragment program */
1628 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1629 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1630 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1631
1632 /* Enable OpenGL fragment programs */
1633 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1634 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1635 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1636 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1637 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1638 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1639 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1640 }
1641 }
1642
1643 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1644 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1645 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1646 static GLuint vprogram_id = 0;
1647 static GLuint fprogram_id = 0;
1648
1649 if (!vprogram_id) vprogram_id = create_arb_blt_vertex_program(gl_info);
1650 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vprogram_id));
1651 glEnable(GL_VERTEX_PROGRAM_ARB);
1652
1653 if (!fprogram_id) fprogram_id = create_arb_blt_fragment_program(gl_info);
1654 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, fprogram_id));
1655 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1656 }
1657
1658 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1659 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1660 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1661 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1662 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1663 }
1664
1665 const shader_backend_t arb_program_shader_backend = {
1666 &shader_arb_select,
1667 &shader_arb_select_depth_blt,
1668 &shader_arb_load_constants,
1669 &shader_arb_cleanup,
1670 &shader_arb_color_correction
1671 };
useful hacks and other patches not (yet) suitable for mainline