?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push 88671f85dcf7a7cd89c63d6e9f34ad6b6ad2ae64
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
blob84516c30d0f41789e50f80348c64e770f30878f9
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 != -1) {
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)->luminanceconst != -1) {
173 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
174 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
175 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
176 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
177 */
178 float *scale = (float *) &stateBlock->textureState[(int) psi->needsbumpmat][WINED3DTSS_BUMPENVLSCALE];
179 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst, scale));
180 }
181 }
182 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
183 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
184 float comparison[4];
185 float mul_low[4];
186
187 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
188 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
189 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
190
191 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
192 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
193 } else {
194 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
195 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
196
197 mul_low[0] = 1.0; mul_low[1] = 1.0;
198 mul_low[2] = 1.0; mul_low[3] = 1.0;
199 }
200 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
201 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
202 checkGLcall("Load sRGB correction constants\n");
203 }
204 }
205 }
206
207 /* Generate the variable & register declarations for the ARB_vertex_program output target */
208 void shader_generate_arb_declarations(
209 IWineD3DBaseShader *iface,
210 shader_reg_maps* reg_maps,
211 SHADER_BUFFER* buffer,
212 WineD3D_GL_Info* gl_info) {
213
214 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
215 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
216 DWORD i;
217 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
218 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
219 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
220 UINT extra_constants_needed = 0;
221
222 /* Temporary Output register */
223 shader_addline(buffer, "TEMP TMP_OUT;\n");
224
225 for(i = 0; i < This->baseShader.limits.temporary; i++) {
226 if (reg_maps->temporary[i])
227 shader_addline(buffer, "TEMP R%u;\n", i);
228 }
229
230 for (i = 0; i < This->baseShader.limits.address; i++) {
231 if (reg_maps->address[i])
232 shader_addline(buffer, "ADDRESS A%d;\n", i);
233 }
234
235 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
236 if (reg_maps->texcoord[i])
237 shader_addline(buffer,"TEMP T%u;\n", i);
238 }
239
240 /* Texture coordinate registers must be pre-loaded */
241 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
242 if (reg_maps->texcoord[i])
243 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
244 }
245
246 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = -1;
247 ((IWineD3DPixelShaderImpl *)This)->luminanceconst = -1;
248 if(reg_maps->bumpmat != -1 /* Only a pshader can use texbem */) {
249 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
250 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
251 * bump mapping.
252 */
253 if(max_constantsF < GL_LIMITS(pshader_constantsF)) {
254 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = max_constantsF;
255 shader_addline(buffer, "PARAM bumpenvmat = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst);
256
257 if(reg_maps->luminanceparams != -1 && max_constantsF +1 < GL_LIMITS(pshader_constantsF)) {
258 extra_constants_needed += 1;
259 ((IWineD3DPixelShaderImpl *)This)->luminanceconst = max_constantsF + 1;
260 shader_addline(buffer, "PARAM luminance = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->luminanceconst);
261 } else if(reg_maps->luminanceparams != -1) {
262 FIXME("No free constant to load the luminance parameters\n");
263 }
264 } else {
265 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
266 }
267 extra_constants_needed += 1;
268 }
269 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
270 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
271 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
272 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
273 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
274 * off again
275 */
276 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
277 /* The idea is that if srgb is enabled, then disabled, the constant loading code
278 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
279 * and comparison constants. If it disables it that way, the shader won't be recompiled
280 * and the code will stay in, so sRGB writing can be turned on again by setting the
281 * constants from the spec
282 */
283 ps_impl->srgb_mode_hardcoded = 0;
284 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
285 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
286 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
287 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
288 } else {
289 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
290 srgb_mul_low, srgb_mul_low, srgb_mul_low);
291 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
292 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
293 ps_impl->srgb_mode_hardcoded = 1;
294 }
295 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
296 * path if the comparison value is set to INF
297 */
298 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
299 srgb_pow, srgb_pow, srgb_pow);
300 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
301 srgb_mul_high, srgb_mul_high, srgb_mul_high);
302 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
303 srgb_sub_high, srgb_sub_high, srgb_sub_high);
304 ps_impl->srgb_enabled = 1;
305 } else if(pshader) {
306 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
307
308 /* Do not write any srgb fixup into the shader to save shader size and processing time.
309 * As a consequence, we can't toggle srgb write on without recompilation
310 */
311 ps_impl->srgb_enabled = 0;
312 ps_impl->srgb_mode_hardcoded = 1;
313 }
314
315 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
316 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
317 max_constantsF, max_constantsF - 1);
318 }
319
320 static const char * const shift_tab[] = {
321 "dummy", /* 0 (none) */
322 "coefmul.x", /* 1 (x2) */
323 "coefmul.y", /* 2 (x4) */
324 "coefmul.z", /* 3 (x8) */
325 "coefmul.w", /* 4 (x16) */
326 "dummy", /* 5 (x32) */
327 "dummy", /* 6 (x64) */
328 "dummy", /* 7 (x128) */
329 "dummy", /* 8 (d256) */
330 "dummy", /* 9 (d128) */
331 "dummy", /* 10 (d64) */
332 "dummy", /* 11 (d32) */
333 "coefdiv.w", /* 12 (d16) */
334 "coefdiv.z", /* 13 (d8) */
335 "coefdiv.y", /* 14 (d4) */
336 "coefdiv.x" /* 15 (d2) */
337 };
338
339 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
340 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
341 char *ptr = write_mask;
342 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
343
344 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
345 *ptr++ = '.';
346 *ptr++ = 'x';
347 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
348 *ptr++ = '.';
349 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
350 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
351 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
352 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
353 }
354
355 *ptr = '\0';
356 }
357
358 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
359 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
360 * but addressed as "rgba". To fix this we need to swap the register's x
361 * and z components. */
362 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
363 char *ptr = swizzle_str;
364
365 /* swizzle bits fields: wwzzyyxx */
366 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
367 DWORD swizzle_x = swizzle & 0x03;
368 DWORD swizzle_y = (swizzle >> 2) & 0x03;
369 DWORD swizzle_z = (swizzle >> 4) & 0x03;
370 DWORD swizzle_w = (swizzle >> 6) & 0x03;
371
372 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
373 * generate a swizzle string. Unless we need to our own swizzling. */
374 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
375 *ptr++ = '.';
376 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
377 *ptr++ = swizzle_chars[swizzle_x];
378 } else {
379 *ptr++ = swizzle_chars[swizzle_x];
380 *ptr++ = swizzle_chars[swizzle_y];
381 *ptr++ = swizzle_chars[swizzle_z];
382 *ptr++ = swizzle_chars[swizzle_w];
383 }
384 }
385
386 *ptr = '\0';
387 }
388
389 static void pshader_get_register_name(
390 const DWORD param, char* regstr) {
391
392 DWORD reg = param & WINED3DSP_REGNUM_MASK;
393 DWORD regtype = shader_get_regtype(param);
394
395 switch (regtype) {
396 case WINED3DSPR_TEMP:
397 sprintf(regstr, "R%u", reg);
398 break;
399 case WINED3DSPR_INPUT:
400 if (reg==0) {
401 strcpy(regstr, "fragment.color.primary");
402 } else {
403 strcpy(regstr, "fragment.color.secondary");
404 }
405 break;
406 case WINED3DSPR_CONST:
407 sprintf(regstr, "C[%u]", reg);
408 break;
409 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
410 sprintf(regstr,"T%u", reg);
411 break;
412 case WINED3DSPR_COLOROUT:
413 if (reg == 0)
414 sprintf(regstr, "TMP_COLOR");
415 else {
416 /* TODO: See GL_ARB_draw_buffers */
417 FIXME("Unsupported write to render target %u\n", reg);
418 sprintf(regstr, "unsupported_register");
419 }
420 break;
421 case WINED3DSPR_DEPTHOUT:
422 sprintf(regstr, "result.depth");
423 break;
424 case WINED3DSPR_ATTROUT:
425 sprintf(regstr, "oD[%u]", reg);
426 break;
427 case WINED3DSPR_TEXCRDOUT:
428 sprintf(regstr, "oT[%u]", reg);
429 break;
430 default:
431 FIXME("Unhandled register name Type(%d)\n", regtype);
432 sprintf(regstr, "unrecognized_register");
433 break;
434 }
435 }
436
437 /* TODO: merge with pixel shader */
438 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
439
440 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
441
442 /* oPos, oFog and oPts in D3D */
443 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
444
445 DWORD reg = param & WINED3DSP_REGNUM_MASK;
446 DWORD regtype = shader_get_regtype(param);
447 char tmpReg[255];
448 BOOL is_color = FALSE;
449
450 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
451 strcat(hwLine, " -");
452 } else {
453 strcat(hwLine, " ");
454 }
455
456 switch (regtype) {
457 case WINED3DSPR_TEMP:
458 sprintf(tmpReg, "R%u", reg);
459 strcat(hwLine, tmpReg);
460 break;
461 case WINED3DSPR_INPUT:
462
463 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
464 is_color = TRUE;
465
466 sprintf(tmpReg, "vertex.attrib[%u]", reg);
467 strcat(hwLine, tmpReg);
468 break;
469 case WINED3DSPR_CONST:
470 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
471 if(reg >= This->rel_offset) {
472 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
473 } else {
474 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
475 }
476 } else {
477 sprintf(tmpReg, "C[%u]", reg);
478 }
479 strcat(hwLine, tmpReg);
480 break;
481 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
482 sprintf(tmpReg, "A%u", reg);
483 strcat(hwLine, tmpReg);
484 break;
485 case WINED3DSPR_RASTOUT:
486 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
487 strcat(hwLine, tmpReg);
488 break;
489 case WINED3DSPR_ATTROUT:
490 if (reg==0) {
491 strcat(hwLine, "result.color.primary");
492 } else {
493 strcat(hwLine, "result.color.secondary");
494 }
495 break;
496 case WINED3DSPR_TEXCRDOUT:
497 sprintf(tmpReg, "result.texcoord[%u]", reg);
498 strcat(hwLine, tmpReg);
499 break;
500 default:
501 FIXME("Unknown reg type %d %d\n", regtype, reg);
502 strcat(hwLine, "unrecognized_register");
503 break;
504 }
505
506 if (!is_input) {
507 char write_mask[6];
508 shader_arb_get_write_mask(arg, param, write_mask);
509 strcat(hwLine, write_mask);
510 } else {
511 char swizzle[6];
512 shader_arb_get_swizzle(param, is_color, swizzle);
513 strcat(hwLine, swizzle);
514 }
515 }
516
517 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
518 SHADER_BUFFER* buffer = arg->buffer;
519 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
520 const char *tex_type;
521
522 switch(sampler_type) {
523 case WINED3DSTT_1D:
524 tex_type = "1D";
525 break;
526
527 case WINED3DSTT_2D:
528 {
529 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
530 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
531 if(device->stateBlock->textures[sampler_idx] &&
532 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
533 tex_type = "RECT";
534 } else {
535 tex_type = "2D";
536 }
537 break;
538 }
539
540 case WINED3DSTT_VOLUME:
541 tex_type = "3D";
542 break;
543
544 case WINED3DSTT_CUBE:
545 tex_type = "CUBE";
546 break;
547
548 default:
549 ERR("Unexpected texture type %d\n", sampler_type);
550 tex_type = "";
551 }
552
553 if (bias) {
554 /* Shouldn't be possible, but let's check for it */
555 if(projected) FIXME("Biased and Projected texture sampling\n");
556 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
557 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
558 } else if (projected) {
559 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
560 } else {
561 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
562 }
563 }
564
565 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
566 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
567 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
568 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
569 WINED3DFORMAT fmt;
570 WINED3DFORMAT conversion_group;
571 IWineD3DBaseTextureImpl *texture;
572 UINT i;
573 BOOL recorded = FALSE;
574 DWORD sampler_idx;
575 DWORD hex_version = shader->baseShader.hex_version;
576 char reg[256];
577 char writemask[6];
578
579 switch(arg->opcode->opcode) {
580 case WINED3DSIO_TEX:
581 if (hex_version < WINED3DPS_VERSION(2,0)) {
582 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
583 } else {
584 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
585 }
586 break;
587
588 case WINED3DSIO_TEXLDL:
589 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
590 return;
591
592 case WINED3DSIO_TEXDP3TEX:
593 case WINED3DSIO_TEXM3x3TEX:
594 case WINED3DSIO_TEXM3x3SPEC:
595 case WINED3DSIO_TEXM3x3VSPEC:
596 case WINED3DSIO_TEXBEM:
597 case WINED3DSIO_TEXREG2AR:
598 case WINED3DSIO_TEXREG2GB:
599 case WINED3DSIO_TEXREG2RGB:
600 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
601 break;
602
603 default:
604 /* Not a texture sampling instruction, nothing to do */
605 return;
606 };
607
608 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
609 if(texture) {
610 fmt = texture->resource.format;
611 conversion_group = texture->baseTexture.shader_conversion_group;
612 } else {
613 fmt = WINED3DFMT_UNKNOWN;
614 conversion_group = WINED3DFMT_UNKNOWN;
615 }
616
617 /* before doing anything, record the sampler with the format in the format conversion list,
618 * but check if it's not there already
619 */
620 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
621 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
622 recorded = TRUE;
623 }
624 }
625 if(!recorded) {
626 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
627 shader->baseShader.num_sampled_samplers++;
628 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
629 }
630
631 pshader_get_register_name(arg->dst, reg);
632 shader_arb_get_write_mask(arg, arg->dst, writemask);
633 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
634
635 switch(fmt) {
636 case WINED3DFMT_V8U8:
637 case WINED3DFMT_V16U16:
638 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
639 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
640 #if 0
641 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
642 * disabled until an application that needs it is found because it causes unneeded
643 * shader recompilation in some game
644 */
645 if(strlen(writemask) >= 4) {
646 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
647 }
648 #endif
649 } else {
650 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
651 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
652 * all registers, do so, this saves an instruction.
653 */
654 if(strlen(writemask) >= 5) {
655 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
656 } else if(strlen(writemask) >= 3) {
657 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
658 reg, writemask[1],
659 reg, writemask[1]);
660 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
661 reg, writemask[2],
662 reg, writemask[2]);
663 } else if(strlen(writemask) == 2) {
664 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
665 reg, writemask[1]);
666 }
667 }
668 break;
669
670 case WINED3DFMT_X8L8V8U8:
671 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
672 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
673 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
674 */
675 if(strlen(writemask) >= 3) {
676 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
677 reg, writemask[1],
678 reg, writemask[1]);
679 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
680 reg, writemask[2],
681 reg, writemask[2]);
682 } else if(strlen(writemask) == 2) {
683 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
684 reg, writemask[1],
685 reg, writemask[1]);
686 }
687 }
688 break;
689
690 case WINED3DFMT_L6V5U5:
691 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
692 if(strlen(writemask) >= 4) {
693 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
694 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
695 reg, writemask[2]);
696 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
697 reg, writemask[1], writemask[1],
698 reg, writemask[1], writemask[3]);
699 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
700 writemask[3]);
701 } else if(strlen(writemask) == 3) {
702 /* This is bad: We have VL, but we need VU */
703 FIXME("2 components sampled from a converted L6V5U5 texture\n");
704 } else {
705 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
706 reg, writemask[1],
707 reg, writemask[1]);
708 }
709 }
710 break;
711
712 case WINED3DFMT_Q8W8V8U8:
713 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
714 /* Correct the sign in all channels */
715 switch(strlen(writemask)) {
716 case 4:
717 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
718 reg, writemask[3],
719 reg, writemask[3]);
720 /* drop through */
721 case 3:
722 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
723 reg, writemask[2],
724 reg, writemask[2]);
725 /* drop through */
726 case 2:
727 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
728 reg, writemask[1],
729 reg, writemask[1]);
730 break;
731
732 /* Should not occur, since it's at minimum '.' and a letter */
733 case 1:
734 ERR("Unexpected writemask: \"%s\"\n", writemask);
735 break;
736
737 case 5:
738 default:
739 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
740 }
741 }
742 break;
743
744 /* stupid compiler */
745 default:
746 break;
747 }
748 }
749
750
751 static void pshader_gen_input_modifier_line (
752 SHADER_BUFFER* buffer,
753 const DWORD instr,
754 int tmpreg,
755 char *outregstr) {
756
757 /* Generate a line that does the input modifier computation and return the input register to use */
758 char regstr[256];
759 char swzstr[20];
760 int insert_line;
761
762 /* Assume a new line will be added */
763 insert_line = 1;
764
765 /* Get register name */
766 pshader_get_register_name(instr, regstr);
767 shader_arb_get_swizzle(instr, FALSE, swzstr);
768
769 switch (instr & WINED3DSP_SRCMOD_MASK) {
770 case WINED3DSPSM_NONE:
771 sprintf(outregstr, "%s%s", regstr, swzstr);
772 insert_line = 0;
773 break;
774 case WINED3DSPSM_NEG:
775 sprintf(outregstr, "-%s%s", regstr, swzstr);
776 insert_line = 0;
777 break;
778 case WINED3DSPSM_BIAS:
779 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
780 break;
781 case WINED3DSPSM_BIASNEG:
782 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
783 break;
784 case WINED3DSPSM_SIGN:
785 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
786 break;
787 case WINED3DSPSM_SIGNNEG:
788 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
789 break;
790 case WINED3DSPSM_COMP:
791 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
792 break;
793 case WINED3DSPSM_X2:
794 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
795 break;
796 case WINED3DSPSM_X2NEG:
797 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
798 break;
799 case WINED3DSPSM_DZ:
800 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
801 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
802 break;
803 case WINED3DSPSM_DW:
804 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
805 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
806 break;
807 default:
808 sprintf(outregstr, "%s%s", regstr, swzstr);
809 insert_line = 0;
810 }
811
812 /* Return modified or original register, with swizzle */
813 if (insert_line)
814 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
815 }
816
817 static inline void pshader_gen_output_modifier_line(
818 SHADER_BUFFER* buffer,
819 int saturate,
820 char *write_mask,
821 int shift,
822 char *regstr) {
823
824 /* Generate a line that does the output modifier computation */
825 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
826 regstr, write_mask, regstr, shift_tab[shift]);
827 }
828
829 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
830 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
831
832 SHADER_BUFFER* buffer = arg->buffer;
833 char dst_name[50];
834 char src_name[2][50];
835 char dst_wmask[20];
836
837 pshader_get_register_name(arg->dst, dst_name);
838 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
839 strcat(dst_name, dst_wmask);
840
841 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
842 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
843
844 if(This->bumpenvmatconst != -1) {
845 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
846 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
847 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
848 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
849 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
850
851 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
852 } else {
853 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
854 }
855 }
856
857 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
858
859 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
860 SHADER_BUFFER* buffer = arg->buffer;
861 char dst_wmask[20];
862 char dst_name[50];
863 char src_name[3][50];
864 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
865 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
866
867 /* FIXME: support output modifiers */
868
869 /* Handle output register */
870 pshader_get_register_name(arg->dst, dst_name);
871 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
872
873 /* Generate input register names (with modifiers) */
874 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
875 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
876 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
877
878 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
879 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
880 arg->opcode_token & WINED3DSI_COISSUE) {
881 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
882 } else {
883 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
884 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
885 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
886 }
887 if (shift != 0)
888 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
889 }
890
891 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
892
893 SHADER_BUFFER* buffer = arg->buffer;
894 char dst_wmask[20];
895 char dst_name[50];
896 char src_name[3][50];
897 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
898 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
899
900 /* FIXME: support output modifiers */
901
902 /* Handle output register */
903 pshader_get_register_name(arg->dst, dst_name);
904 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
905
906 /* Generate input register names (with modifiers) */
907 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
908 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
909 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
910
911 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
912 src_name[0], src_name[2], src_name[1]);
913
914 if (shift != 0)
915 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
916 }
917
918 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
919 * dst = dot2(src0, src1) + src2 */
920 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
921 SHADER_BUFFER* buffer = arg->buffer;
922 char dst_wmask[20];
923 char dst_name[50];
924 char src_name[3][50];
925 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
926 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
927
928 pshader_get_register_name(arg->dst, dst_name);
929 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
930
931 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
932 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
933 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
934
935 /* Emulate a DP2 with a DP3 and 0.0 */
936 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
937 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
938 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
939 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
940
941 if (shift != 0)
942 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
943 }
944
945 /* Map the opcode 1-to-1 to the GL code */
946 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
947
948 CONST SHADER_OPCODE* curOpcode = arg->opcode;
949 SHADER_BUFFER* buffer = arg->buffer;
950 DWORD dst = arg->dst;
951 DWORD* src = arg->src;
952
953 unsigned int i;
954 char tmpLine[256];
955
956 /* Output token related */
957 char output_rname[256];
958 char output_wmask[20];
959 BOOL saturate = FALSE;
960 BOOL centroid = FALSE;
961 BOOL partialprecision = FALSE;
962 DWORD shift;
963
964 strcpy(tmpLine, curOpcode->glname);
965
966 /* Process modifiers */
967 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
968 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
969
970 saturate = mask & WINED3DSPDM_SATURATE;
971 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
972 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
973 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
974 if (mask)
975 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
976
977 if (centroid)
978 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
979 }
980 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
981
982 /* Generate input and output registers */
983 if (curOpcode->num_params > 0) {
984 char operands[4][100];
985
986 /* Generate input register names (with modifiers) */
987 for (i = 1; i < curOpcode->num_params; ++i)
988 pshader_gen_input_modifier_line(buffer, src[i-1], i-1, operands[i]);
989
990 /* Handle output register */
991 pshader_get_register_name(dst, output_rname);
992 strcpy(operands[0], output_rname);
993 shader_arb_get_write_mask(arg, dst, output_wmask);
994 strcat(operands[0], output_wmask);
995
996 if (saturate && (shift == 0))
997 strcat(tmpLine, "_SAT");
998 strcat(tmpLine, " ");
999 strcat(tmpLine, operands[0]);
1000 for (i = 1; i < curOpcode->num_params; i++) {
1001 strcat(tmpLine, ", ");
1002 strcat(tmpLine, operands[i]);
1003 }
1004 strcat(tmpLine,";\n");
1005 shader_addline(buffer, tmpLine);
1006
1007 /* A shift requires another line. */
1008 if (shift != 0)
1009 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1010 }
1011 }
1012
1013 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1014 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1015 DWORD hex_version = This->baseShader.hex_version;
1016 SHADER_BUFFER* buffer = arg->buffer;
1017 char reg_dest[40];
1018
1019 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1020 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1021 */
1022 pshader_get_register_name(arg->dst, reg_dest);
1023
1024 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1025 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1026 shader_addline(buffer, "KIL %s;\n", reg_dest);
1027 } else {
1028 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1029 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1030 */
1031 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1032 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1033 shader_addline(buffer, "KIL TMP;\n");
1034 }
1035 }
1036
1037 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1038 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1039 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1040
1041 DWORD dst = arg->dst;
1042 DWORD* src = arg->src;
1043 SHADER_BUFFER* buffer = arg->buffer;
1044 DWORD hex_version = This->baseShader.hex_version;
1045 BOOL projected = FALSE, bias = FALSE;
1046
1047 char reg_dest[40];
1048 char reg_coord[40];
1049 DWORD reg_dest_code;
1050 DWORD reg_sampler_code;
1051
1052 /* All versions have a destination register */
1053 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1054 pshader_get_register_name(dst, reg_dest);
1055
1056 /* 1.0-1.3: Use destination register as coordinate source.
1057 1.4+: Use provided coordinate source register. */
1058 if (hex_version < WINED3DPS_VERSION(1,4))
1059 strcpy(reg_coord, reg_dest);
1060 else
1061 pshader_gen_input_modifier_line(buffer, src[0], 0, reg_coord);
1062
1063 /* 1.0-1.4: Use destination register number as texture code.
1064 2.0+: Use provided sampler number as texure code. */
1065 if (hex_version < WINED3DPS_VERSION(2,0))
1066 reg_sampler_code = reg_dest_code;
1067 else
1068 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1069
1070 /* projection flag:
1071 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1072 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1073 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1074 */
1075 if(hex_version < WINED3DPS_VERSION(1,4)) {
1076 DWORD flags = 0;
1077 if(reg_sampler_code < MAX_TEXTURES) {
1078 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1079 }
1080 if (flags & WINED3DTTFF_PROJECTED) {
1081 projected = TRUE;
1082 }
1083 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1084 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1085 if (src_mod == WINED3DSPSM_DZ) {
1086 projected = TRUE;
1087 } else if(src_mod == WINED3DSPSM_DW) {
1088 projected = TRUE;
1089 }
1090 } else {
1091 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1092 projected = TRUE;
1093 }
1094 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1095 bias = TRUE;
1096 }
1097 }
1098 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1099 }
1100
1101 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1102
1103 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1104 DWORD dst = arg->dst;
1105 SHADER_BUFFER* buffer = arg->buffer;
1106 DWORD hex_version = This->baseShader.hex_version;
1107
1108 char tmp[20];
1109 shader_arb_get_write_mask(arg, dst, tmp);
1110 if (hex_version != WINED3DPS_VERSION(1,4)) {
1111 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1112 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1113 } else {
1114 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1115 char reg_src[40];
1116
1117 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, reg_src);
1118 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1119 }
1120 }
1121
1122 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1123
1124 SHADER_BUFFER* buffer = arg->buffer;
1125 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1126 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1127 DWORD flags;
1128
1129 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1130 char dst_str[8];
1131 char src_str[50];
1132
1133 sprintf(dst_str, "T%u", reg1);
1134 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1135 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1136 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1137 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1138 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1139 }
1140
1141 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1142
1143 SHADER_BUFFER* buffer = arg->buffer;
1144 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1145 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1146 DWORD flags;
1147
1148 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1149 char dst_str[8];
1150 char src_str[50];
1151
1152 sprintf(dst_str, "T%u", reg1);
1153 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1154 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1155 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1156 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1157 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1158 }
1159
1160 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1161
1162 SHADER_BUFFER* buffer = arg->buffer;
1163 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1164 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1165 DWORD flags;
1166 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1167 char dst_str[8];
1168 char src_str[50];
1169
1170 sprintf(dst_str, "T%u", reg1);
1171 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1172 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1173 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1174 }
1175
1176 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1177 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1178
1179 DWORD dst = arg->dst;
1180 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1181 SHADER_BUFFER* buffer = arg->buffer;
1182
1183 char reg_coord[40];
1184 DWORD reg_dest_code;
1185
1186 /* All versions have a destination register */
1187 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1188 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1189 pshader_get_register_name(dst, reg_coord);
1190
1191 if(This->bumpenvmatconst != -1) {
1192 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1193
1194 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
1195 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1196 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
1197 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1198
1199 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1200 * so we can't let the GL handle this.
1201 */
1202 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1203 & WINED3DTTFF_PROJECTED) {
1204 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1205 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1206 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1207 } else {
1208 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1209 }
1210
1211 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1212
1213 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && This->luminanceconst != -1) {
1214 shader_addline(buffer, "MAD TMP, T%u.z, luminance.x, luminance.y;\n", src);
1215 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1216 }
1217
1218 } else {
1219 DWORD tf;
1220 if(reg_dest_code < MAX_TEXTURES) {
1221 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1222 } else {
1223 tf = 0;
1224 }
1225 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1226 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1227 }
1228 }
1229
1230 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1231
1232 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1233 SHADER_BUFFER* buffer = arg->buffer;
1234 char src0_name[50];
1235
1236 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1237 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1238 }
1239
1240 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1241
1242 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1243 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1244 DWORD flags;
1245 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1246 SHADER_BUFFER* buffer = arg->buffer;
1247 char dst_str[8];
1248 char src0_name[50];
1249
1250 sprintf(dst_str, "T%u", reg);
1251 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1252 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1253 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1254 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1255 }
1256
1257 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1258
1259 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1260 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1261 SHADER_BUFFER* buffer = arg->buffer;
1262 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1263 char src0_name[50];
1264
1265 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1266 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1267 current_state->texcoord_w[current_state->current_row++] = reg;
1268 }
1269
1270 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1271
1272 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1273 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1274 DWORD flags;
1275 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1276 SHADER_BUFFER* buffer = arg->buffer;
1277 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1278 char dst_str[8];
1279 char src0_name[50];
1280
1281 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1282 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1283
1284 /* Sample the texture using the calculated coordinates */
1285 sprintf(dst_str, "T%u", reg);
1286 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1287 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1288 current_state->current_row = 0;
1289 }
1290
1291 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1292
1293 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1294 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1295 DWORD flags;
1296 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1297 SHADER_BUFFER* buffer = arg->buffer;
1298 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1299 char dst_str[8];
1300 char src0_name[50];
1301
1302 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1303 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1304
1305 /* Construct the eye-ray vector from w coordinates */
1306 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1307 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1308 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1309
1310 /* Calculate reflection vector
1311 */
1312 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1313 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1314 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1315 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1316 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1317 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1318 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1319
1320 /* Sample the texture using the calculated coordinates */
1321 sprintf(dst_str, "T%u", reg);
1322 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1323 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1324 current_state->current_row = 0;
1325 }
1326
1327 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1328
1329 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1330 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1331 DWORD flags;
1332 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1333 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1334 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1335 SHADER_BUFFER* buffer = arg->buffer;
1336 char dst_str[8];
1337 char src0_name[50];
1338
1339 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1340 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1341
1342 /* Calculate reflection vector.
1343 *
1344 * dot(N, E)
1345 * TMP.xyz = 2 * --------- * N - E
1346 * dot(N, N)
1347 *
1348 * Which normalizes the normal vector
1349 */
1350 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1351 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1352 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1353 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1354 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1355 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1356
1357 /* Sample the texture using the calculated coordinates */
1358 sprintf(dst_str, "T%u", reg);
1359 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1360 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1361 current_state->current_row = 0;
1362 }
1363
1364 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1365 SHADER_BUFFER* buffer = arg->buffer;
1366 char dst_name[50];
1367
1368 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1369 * which is essentially an input, is the destination register because it is the first
1370 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1371 * here
1372 */
1373 pshader_get_register_name(arg->dst, dst_name);
1374
1375 /* According to the msdn, the source register(must be r5) is unusable after
1376 * the texdepth instruction, so we're free to modify it
1377 */
1378 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1379
1380 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1381 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1382 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1383 */
1384 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1385 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1386 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1387 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1388 }
1389
1390 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1391 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1392 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1393 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1394 SHADER_BUFFER* buffer = arg->buffer;
1395 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1396 char src0[50];
1397 char dst_str[8];
1398
1399 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1400 shader_addline(buffer, "MOV TMP, 0.0;\n");
1401 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1402
1403 sprintf(dst_str, "T%u", sampler_idx);
1404 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1405 }
1406
1407 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1408 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1409 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1410 char src0[50];
1411 char dst_str[50];
1412 char dst_mask[6];
1413 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1414 SHADER_BUFFER* buffer = arg->buffer;
1415
1416 /* Handle output register */
1417 pshader_get_register_name(arg->dst, dst_str);
1418 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1419
1420 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1421 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1422
1423 /* TODO: Handle output modifiers */
1424 }
1425
1426 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1427 * Perform the 3rd row of a 3x3 matrix multiply */
1428 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1429 SHADER_BUFFER* buffer = arg->buffer;
1430 char dst_str[50];
1431 char dst_mask[6];
1432 char src0[50];
1433 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1434
1435 pshader_get_register_name(arg->dst, dst_str);
1436 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1437
1438 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1439 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1440 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1441
1442 /* TODO: Handle output modifiers */
1443 }
1444
1445 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1446 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1447 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1448 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1449 */
1450 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1451 SHADER_BUFFER* buffer = arg->buffer;
1452 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1453 char src0[50];
1454
1455 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1456 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1457
1458 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1459 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1460 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1461 */
1462 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1463 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1464 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1465 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1466 }
1467
1468 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1469 Vertex/Pixel shaders to ARB_vertex_program codes */
1470 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1471
1472 int i;
1473 int nComponents = 0;
1474 SHADER_OPCODE_ARG tmpArg;
1475
1476 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1477
1478 /* Set constants for the temporary argument */
1479 tmpArg.shader = arg->shader;
1480 tmpArg.buffer = arg->buffer;
1481 tmpArg.src[0] = arg->src[0];
1482 tmpArg.src_addr[0] = arg->src_addr[0];
1483 tmpArg.src_addr[1] = arg->src_addr[1];
1484 tmpArg.reg_maps = arg->reg_maps;
1485
1486 switch(arg->opcode->opcode) {
1487 case WINED3DSIO_M4x4:
1488 nComponents = 4;
1489 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1490 break;
1491 case WINED3DSIO_M4x3:
1492 nComponents = 3;
1493 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1494 break;
1495 case WINED3DSIO_M3x4:
1496 nComponents = 4;
1497 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1498 break;
1499 case WINED3DSIO_M3x3:
1500 nComponents = 3;
1501 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1502 break;
1503 case WINED3DSIO_M3x2:
1504 nComponents = 2;
1505 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1506 break;
1507 default:
1508 break;
1509 }
1510
1511 for (i = 0; i < nComponents; i++) {
1512 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1513 tmpArg.src[1] = arg->src[1]+i;
1514 vshader_hw_map2gl(&tmpArg);
1515 }
1516 }
1517
1518 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1519 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1520 SHADER_BUFFER* buffer = arg->buffer;
1521 DWORD dst = arg->dst;
1522 DWORD src = arg->src[0];
1523 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1524
1525 char tmpLine[256];
1526
1527 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1528 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1529 strcat(tmpLine, ",");
1530 vshader_program_add_param(arg, src, TRUE, tmpLine);
1531 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1532 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1533 * .w is used
1534 */
1535 strcat(tmpLine, ".w");
1536 }
1537
1538 shader_addline(buffer, "%s;\n", tmpLine);
1539 }
1540
1541 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1542 SHADER_BUFFER* buffer = arg->buffer;
1543 char dst_name[50];
1544 char src_name[50];
1545 char dst_wmask[20];
1546 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1547 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1548
1549 pshader_get_register_name(arg->dst, dst_name);
1550 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1551
1552 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1553 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1554 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1555 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1556 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1557 src_name);
1558
1559 if (shift != 0)
1560 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1561 }
1562
1563 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1564 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1565 * must contain fixed constants. So we need a separate function to filter those constants and
1566 * can't use map2gl
1567 */
1568 SHADER_BUFFER* buffer = arg->buffer;
1569 char dst_name[50];
1570 char src_name[50];
1571 char dst_wmask[20];
1572 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1573 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1574
1575 pshader_get_register_name(arg->dst, dst_name);
1576 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1577
1578 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1579 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1580 src_name);
1581
1582 if (shift != 0)
1583 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1584
1585 }
1586
1587 /* TODO: merge with pixel shader */
1588 /* Map the opcode 1-to-1 to the GL code */
1589 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1590
1591 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1592 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1593 SHADER_BUFFER* buffer = arg->buffer;
1594 DWORD dst = arg->dst;
1595 DWORD* src = arg->src;
1596
1597 DWORD dst_regtype = shader_get_regtype(dst);
1598 char tmpLine[256];
1599 unsigned int i;
1600
1601 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1602 if(shader->rel_offset) {
1603 memset(tmpLine, 0, sizeof(tmpLine));
1604 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1605 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1606 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1607 return;
1608 } else {
1609 strcpy(tmpLine, "ARL");
1610 }
1611 } else
1612 strcpy(tmpLine, curOpcode->glname);
1613
1614 if (curOpcode->num_params > 0) {
1615 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1616 for (i = 1; i < curOpcode->num_params; ++i) {
1617 strcat(tmpLine, ",");
1618 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1619 }
1620 }
1621 shader_addline(buffer, "%s;\n", tmpLine);
1622 }
1623
1624 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1625 GLuint program_id = 0;
1626 const char *blt_vprogram =
1627 "!!ARBvp1.0\n"
1628 "PARAM c[1] = { { 1, 0.5 } };\n"
1629 "MOV result.position, vertex.position;\n"
1630 "MOV result.color, c[0].x;\n"
1631 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1632 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1633 "END\n";
1634
1635 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1636 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1637 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1638
1639 if (glGetError() == GL_INVALID_OPERATION) {
1640 GLint pos;
1641 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1642 FIXME("Vertex program error at position %d: %s\n", pos,
1643 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1644 }
1645
1646 return program_id;
1647 }
1648
1649 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1650 GLuint program_id = 0;
1651 const char *blt_fprogram =
1652 "!!ARBfp1.0\n"
1653 "TEMP R0;\n"
1654 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1655 "MOV result.depth.z, R0.x;\n"
1656 "END\n";
1657
1658 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1659 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1660 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1661
1662 if (glGetError() == GL_INVALID_OPERATION) {
1663 GLint pos;
1664 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1665 FIXME("Fragment program error at position %d: %s\n", pos,
1666 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1667 }
1668
1669 return program_id;
1670 }
1671
1672 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1673 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1674 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1675
1676 if (useVS) {
1677 TRACE("Using vertex shader\n");
1678
1679 /* Bind the vertex program */
1680 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1681 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1682 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1683
1684 /* Enable OpenGL vertex programs */
1685 glEnable(GL_VERTEX_PROGRAM_ARB);
1686 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1687 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1688 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1689 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1690 glDisable(GL_VERTEX_PROGRAM_ARB);
1691 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1692 }
1693
1694 if (usePS) {
1695 TRACE("Using pixel shader\n");
1696
1697 /* Bind the fragment program */
1698 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1699 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1700 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1701
1702 /* Enable OpenGL fragment programs */
1703 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1704 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1705 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1706 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1707 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1708 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1709 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1710 }
1711 }
1712
1713 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1714 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1715 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1716
1717 if (!This->depth_blt_vprogram_id) This->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1718 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->depth_blt_vprogram_id));
1719 glEnable(GL_VERTEX_PROGRAM_ARB);
1720
1721 if (!This->depth_blt_fprogram_id) This->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1722 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->depth_blt_fprogram_id));
1723 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1724 }
1725
1726 static void shader_arb_destroy_depth_blt(IWineD3DDevice *iface) {
1727 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1728 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1729
1730 if(This->depth_blt_vprogram_id) {
1731 GL_EXTCALL(glDeleteProgramsARB(1, &This->depth_blt_vprogram_id));
1732 This->depth_blt_vprogram_id = 0;
1733 }
1734 if(This->depth_blt_fprogram_id) {
1735 GL_EXTCALL(glDeleteProgramsARB(1, &This->depth_blt_fprogram_id));
1736 This->depth_blt_fprogram_id = 0;
1737 }
1738 }
1739
1740 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1741 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1742 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1743 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1744 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1745 }
1746
1747 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1748 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1749 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1750
1751 ENTER_GL();
1752 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1753 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1754 LEAVE_GL();
1755 This->baseShader.prgId = 0;
1756 This->baseShader.is_compiled = FALSE;
1757 }
1758
1759 const shader_backend_t arb_program_shader_backend = {
1760 &shader_arb_select,
1761 &shader_arb_select_depth_blt,
1762 &shader_arb_destroy_depth_blt,
1763 &shader_arb_load_constants,
1764 &shader_arb_cleanup,
1765 &shader_arb_color_correction,
1766 &shader_arb_destroy
1767 };
useful hacks and other patches not (yet) suitable for mainline