?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push a4aeee26091b4888c4962ca8601c7905b237df2d
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
bloba701f36bfb241c0169c6e2f6c963655aed533907
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 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13 *
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2.1 of the License, or (at your option) any later version.
18 *
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
23 *
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
27 */
28
29 #include "config.h"
30
31 #include <math.h>
32 #include <stdio.h>
33
34 #include "wined3d_private.h"
35
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
39
40 #define GLINFO_LOCATION (*gl_info)
41
42 /********************************************************
43 * ARB_[vertex/fragment]_program helper functions follow
44 ********************************************************/
45
46 /**
47 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
48 * When constant_list == NULL, it will load all the constants.
49 *
50 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
51 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
52 */
53 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
54 unsigned int max_constants, float* constants, char *dirty_consts) {
55 local_constant* lconst;
56 DWORD i, j;
57 unsigned int ret;
58
59 if (TRACE_ON(d3d_shader)) {
60 for(i = 0; i < max_constants; i++) {
61 if(!dirty_consts[i]) continue;
62 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
63 constants[i * 4 + 0], constants[i * 4 + 1],
64 constants[i * 4 + 2], constants[i * 4 + 3]);
65 }
66 }
67 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
68 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
69 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
70 float lcl_const[4];
71 for(i = 0; i < max_constants; i++) {
72 if(!dirty_consts[i]) continue;
73 dirty_consts[i] = 0;
74
75 j = 4 * i;
76 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
77 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
78 else lcl_const[0] = constants[j + 0];
79
80 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
81 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
82 else lcl_const[1] = constants[j + 1];
83
84 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
85 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
86 else lcl_const[2] = constants[j + 2];
87
88 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
89 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
90 else lcl_const[3] = constants[j + 3];
91
92 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
93 }
94 } else {
95 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
96 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
97 * or just reloading *all* constants at once
98 *
99 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
100 */
101 for(i = 0; i < max_constants; i++) {
102 if(!dirty_consts[i]) continue;
103
104 /* Find the next block of dirty constants */
105 dirty_consts[i] = 0;
106 j = i;
107 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
108 dirty_consts[i] = 0;
109 }
110
111 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
112 }
113 } else {
114 for(i = 0; i < max_constants; i++) {
115 if(dirty_consts[i]) {
116 dirty_consts[i] = 0;
117 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
118 }
119 }
120 }
121 }
122 checkGLcall("glProgramEnvParameter4fvARB()");
123
124 /* Load immediate constants */
125 if(This->baseShader.load_local_constsF) {
126 if (TRACE_ON(d3d_shader)) {
127 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
128 GLfloat* values = (GLfloat*)lconst->value;
129 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
130 values[0], values[1], values[2], values[3]);
131 }
132 }
133 /* Immediate constants are clamped for 1.X shaders at loading times */
134 ret = 0;
135 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
136 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
137 ret = max(ret, lconst->idx + 1);
138 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
139 }
140 checkGLcall("glProgramEnvParameter4fvARB()");
141 return ret; /* The loaded immediate constants need reloading for the next shader */
142 } else {
143 return 0; /* No constants are dirty now */
144 }
145 }
146
147 /**
148 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
149 *
150 * We only support float constants in ARB at the moment, so don't
151 * worry about the Integers or Booleans
152 */
153 void shader_arb_load_constants(
154 IWineD3DDevice* device,
155 char usePixelShader,
156 char useVertexShader) {
157
158 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
159 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
160 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
161 unsigned char i;
162
163 if (useVertexShader) {
164 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
165
166 /* Load DirectX 9 float constants for vertex shader */
167 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
168 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
169 deviceImpl->highest_dirty_vs_const,
170 stateBlock->vertexShaderConstantF,
171 deviceImpl->activeContext->vshader_const_dirty);
172
173 /* Upload the position fixup */
174 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
175 }
176
177 if (usePixelShader) {
178
179 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
180 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
181
182 /* Load DirectX 9 float constants for pixel shader */
183 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
184 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
185 deviceImpl->highest_dirty_ps_const,
186 stateBlock->pixelShaderConstantF,
187 deviceImpl->activeContext->pshader_const_dirty);
188
189 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
190 /* The state manager takes care that this function is always called if the bump env matrix changes
191 */
192 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
193 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
194 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
195
196 if(psi->luminanceconst[i].const_num != -1) {
197 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
198 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
199 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
200 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
201 */
202 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
203 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
204 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
205 }
206 }
207
208 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
209 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
210 float comparison[4];
211 float mul_low[4];
212
213 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
214 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
215 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
216
217 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
218 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
219 } else {
220 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
221 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
222
223 mul_low[0] = 1.0; mul_low[1] = 1.0;
224 mul_low[2] = 1.0; mul_low[3] = 1.0;
225 }
226 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
227 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
228 checkGLcall("Load sRGB correction constants\n");
229 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
230 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
231
232 }
233 }
234 }
235
236 /* Generate the variable & register declarations for the ARB_vertex_program output target */
237 void shader_generate_arb_declarations(
238 IWineD3DBaseShader *iface,
239 shader_reg_maps* reg_maps,
240 SHADER_BUFFER* buffer,
241 WineD3D_GL_Info* gl_info) {
242
243 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
244 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
245 DWORD i, cur;
246 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
247 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
248 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
249 UINT extra_constants_needed = 0;
250 local_constant* lconst;
251
252 /* Temporary Output register */
253 shader_addline(buffer, "TEMP TMP_OUT;\n");
254
255 for(i = 0; i < This->baseShader.limits.temporary; i++) {
256 if (reg_maps->temporary[i])
257 shader_addline(buffer, "TEMP R%u;\n", i);
258 }
259
260 for (i = 0; i < This->baseShader.limits.address; i++) {
261 if (reg_maps->address[i])
262 shader_addline(buffer, "ADDRESS A%d;\n", i);
263 }
264
265 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
266 if (reg_maps->texcoord[i])
267 shader_addline(buffer,"TEMP T%u;\n", i);
268 }
269
270 /* Texture coordinate registers must be pre-loaded */
271 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
272 if (reg_maps->texcoord[i])
273 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
274 }
275
276 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
277 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
278 if(!reg_maps->bumpmat[i]) continue;
279
280 cur = ps->numbumpenvmatconsts;
281 ps->bumpenvmatconst[cur].const_num = -1;
282 ps->bumpenvmatconst[cur].texunit = i;
283 ps->luminanceconst[cur].const_num = -1;
284 ps->luminanceconst[cur].texunit = i;
285
286 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
287 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
288 * bump mapping.
289 */
290 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
291 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
292 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
293 i, ps->bumpenvmatconst[cur].const_num);
294 extra_constants_needed++;
295
296 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
297 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
298 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
299 i, ps->luminanceconst[cur].const_num);
300 extra_constants_needed++;
301 } else if(reg_maps->luminanceparams) {
302 FIXME("No free constant to load the luminance parameters\n");
303 }
304 } else {
305 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
306 }
307
308 ps->numbumpenvmatconsts = cur + 1;
309 }
310
311 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
312 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
313 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
314 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
315 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
316 * off again
317 */
318 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
319 /* The idea is that if srgb is enabled, then disabled, the constant loading code
320 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
321 * and comparison constants. If it disables it that way, the shader won't be recompiled
322 * and the code will stay in, so sRGB writing can be turned on again by setting the
323 * constants from the spec
324 */
325 ps_impl->srgb_mode_hardcoded = 0;
326 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
327 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
328 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
329 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
330 } else {
331 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
332 srgb_mul_low, srgb_mul_low, srgb_mul_low);
333 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
334 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
335 ps_impl->srgb_mode_hardcoded = 1;
336 }
337 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
338 * path if the comparison value is set to INF
339 */
340 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
341 srgb_pow, srgb_pow, srgb_pow);
342 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
343 srgb_mul_high, srgb_mul_high, srgb_mul_high);
344 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
345 srgb_sub_high, srgb_sub_high, srgb_sub_high);
346 ps_impl->srgb_enabled = 1;
347 } else if(pshader) {
348 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
349
350 /* Do not write any srgb fixup into the shader to save shader size and processing time.
351 * As a consequence, we can't toggle srgb write on without recompilation
352 */
353 ps_impl->srgb_enabled = 0;
354 ps_impl->srgb_mode_hardcoded = 1;
355 }
356
357 /* Load local constants using the program-local space,
358 * this avoids reloading them each time the shader is used
359 */
360 if(!This->baseShader.load_local_constsF) {
361 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
362 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
363 lconst->idx);
364 }
365 }
366
367 /* we use the array-based constants array if the local constants are marked for loading,
368 * because then we use indirect addressing, or when the local constant list is empty,
369 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
370 * local constants do not declare the loaded constants as an array because ARB compilers usually
371 * do not optimize unused constants away
372 */
373 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
374 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
375 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
376 max_constantsF, max_constantsF - 1);
377 } else {
378 for(i = 0; i < max_constantsF; i++) {
379 if(!shader_constant_is_local(This, i)) {
380 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
381 }
382 }
383 }
384 }
385
386 static const char * const shift_tab[] = {
387 "dummy", /* 0 (none) */
388 "coefmul.x", /* 1 (x2) */
389 "coefmul.y", /* 2 (x4) */
390 "coefmul.z", /* 3 (x8) */
391 "coefmul.w", /* 4 (x16) */
392 "dummy", /* 5 (x32) */
393 "dummy", /* 6 (x64) */
394 "dummy", /* 7 (x128) */
395 "dummy", /* 8 (d256) */
396 "dummy", /* 9 (d128) */
397 "dummy", /* 10 (d64) */
398 "dummy", /* 11 (d32) */
399 "coefdiv.w", /* 12 (d16) */
400 "coefdiv.z", /* 13 (d8) */
401 "coefdiv.y", /* 14 (d4) */
402 "coefdiv.x" /* 15 (d2) */
403 };
404
405 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
406 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
407 char *ptr = write_mask;
408 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
409
410 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
411 *ptr++ = '.';
412 *ptr++ = 'x';
413 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
414 *ptr++ = '.';
415 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
416 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
417 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
418 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
419 }
420
421 *ptr = '\0';
422 }
423
424 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
425 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
426 * but addressed as "rgba". To fix this we need to swap the register's x
427 * and z components. */
428 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
429 char *ptr = swizzle_str;
430
431 /* swizzle bits fields: wwzzyyxx */
432 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
433 DWORD swizzle_x = swizzle & 0x03;
434 DWORD swizzle_y = (swizzle >> 2) & 0x03;
435 DWORD swizzle_z = (swizzle >> 4) & 0x03;
436 DWORD swizzle_w = (swizzle >> 6) & 0x03;
437
438 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
439 * generate a swizzle string. Unless we need to our own swizzling. */
440 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
441 *ptr++ = '.';
442 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
443 *ptr++ = swizzle_chars[swizzle_x];
444 } else {
445 *ptr++ = swizzle_chars[swizzle_x];
446 *ptr++ = swizzle_chars[swizzle_y];
447 *ptr++ = swizzle_chars[swizzle_z];
448 *ptr++ = swizzle_chars[swizzle_w];
449 }
450 }
451
452 *ptr = '\0';
453 }
454
455 static void pshader_get_register_name(IWineD3DBaseShader* iface,
456 const DWORD param, char* regstr) {
457
458 DWORD reg = param & WINED3DSP_REGNUM_MASK;
459 DWORD regtype = shader_get_regtype(param);
460 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
461
462 switch (regtype) {
463 case WINED3DSPR_TEMP:
464 sprintf(regstr, "R%u", reg);
465 break;
466 case WINED3DSPR_INPUT:
467 if (reg==0) {
468 strcpy(regstr, "fragment.color.primary");
469 } else {
470 strcpy(regstr, "fragment.color.secondary");
471 }
472 break;
473 case WINED3DSPR_CONST:
474 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
475 sprintf(regstr, "C[%u]", reg);
476 } else {
477 sprintf(regstr, "C%u", reg);
478 }
479 break;
480 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
481 sprintf(regstr,"T%u", reg);
482 break;
483 case WINED3DSPR_COLOROUT:
484 if (reg == 0)
485 sprintf(regstr, "TMP_COLOR");
486 else {
487 /* TODO: See GL_ARB_draw_buffers */
488 FIXME("Unsupported write to render target %u\n", reg);
489 sprintf(regstr, "unsupported_register");
490 }
491 break;
492 case WINED3DSPR_DEPTHOUT:
493 sprintf(regstr, "result.depth");
494 break;
495 case WINED3DSPR_ATTROUT:
496 sprintf(regstr, "oD[%u]", reg);
497 break;
498 case WINED3DSPR_TEXCRDOUT:
499 sprintf(regstr, "oT[%u]", reg);
500 break;
501 default:
502 FIXME("Unhandled register name Type(%d)\n", regtype);
503 sprintf(regstr, "unrecognized_register");
504 break;
505 }
506 }
507
508 /* TODO: merge with pixel shader */
509 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
510
511 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
512
513 /* oPos, oFog and oPts in D3D */
514 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
515
516 DWORD reg = param & WINED3DSP_REGNUM_MASK;
517 DWORD regtype = shader_get_regtype(param);
518 char tmpReg[255];
519 BOOL is_color = FALSE;
520
521 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
522 strcat(hwLine, " -");
523 } else {
524 strcat(hwLine, " ");
525 }
526
527 switch (regtype) {
528 case WINED3DSPR_TEMP:
529 sprintf(tmpReg, "R%u", reg);
530 strcat(hwLine, tmpReg);
531 break;
532 case WINED3DSPR_INPUT:
533
534 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
535 is_color = TRUE;
536
537 sprintf(tmpReg, "vertex.attrib[%u]", reg);
538 strcat(hwLine, tmpReg);
539 break;
540 case WINED3DSPR_CONST:
541 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
542 if(reg >= This->rel_offset) {
543 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
544 } else {
545 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
546 }
547 } else {
548 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
549 sprintf(tmpReg, "C[%u]", reg);
550 } else {
551 sprintf(tmpReg, "C%u", reg);
552 }
553 }
554 strcat(hwLine, tmpReg);
555 break;
556 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
557 sprintf(tmpReg, "A%u", reg);
558 strcat(hwLine, tmpReg);
559 break;
560 case WINED3DSPR_RASTOUT:
561 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
562 strcat(hwLine, tmpReg);
563 break;
564 case WINED3DSPR_ATTROUT:
565 if (reg==0) {
566 strcat(hwLine, "result.color.primary");
567 } else {
568 strcat(hwLine, "result.color.secondary");
569 }
570 break;
571 case WINED3DSPR_TEXCRDOUT:
572 sprintf(tmpReg, "result.texcoord[%u]", reg);
573 strcat(hwLine, tmpReg);
574 break;
575 default:
576 FIXME("Unknown reg type %d %d\n", regtype, reg);
577 strcat(hwLine, "unrecognized_register");
578 break;
579 }
580
581 if (!is_input) {
582 char write_mask[6];
583 shader_arb_get_write_mask(arg, param, write_mask);
584 strcat(hwLine, write_mask);
585 } else {
586 char swizzle[6];
587 shader_arb_get_swizzle(param, is_color, swizzle);
588 strcat(hwLine, swizzle);
589 }
590 }
591
592 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
593 SHADER_BUFFER* buffer = arg->buffer;
594 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
595 const char *tex_type;
596
597 switch(sampler_type) {
598 case WINED3DSTT_1D:
599 tex_type = "1D";
600 break;
601
602 case WINED3DSTT_2D:
603 {
604 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
605 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
606 if(device->stateBlock->textures[sampler_idx] &&
607 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
608 tex_type = "RECT";
609 } else {
610 tex_type = "2D";
611 }
612 break;
613 }
614
615 case WINED3DSTT_VOLUME:
616 tex_type = "3D";
617 break;
618
619 case WINED3DSTT_CUBE:
620 tex_type = "CUBE";
621 break;
622
623 default:
624 ERR("Unexpected texture type %d\n", sampler_type);
625 tex_type = "";
626 }
627
628 if (bias) {
629 /* Shouldn't be possible, but let's check for it */
630 if(projected) FIXME("Biased and Projected texture sampling\n");
631 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
632 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
633 } else if (projected) {
634 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
635 } else {
636 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
637 }
638 }
639
640 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, const char *writemask,
641 const char *one, const char *two, WINED3DFORMAT fmt,
642 WineD3D_GL_Info *gl_info) {
643 switch(fmt) {
644 case WINED3DFMT_V8U8:
645 case WINED3DFMT_V16U16:
646 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
647 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
648 #if 0
649 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
650 * disabled until an application that needs it is found because it causes unneeded
651 * shader recompilation in some game
652 */
653 if(strlen(writemask) >= 4) {
654 shader_addline(buffer, "MOV %s.%c, %s;\n", reg, one);
655 }
656 #endif
657 } else {
658 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
659 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
660 * all registers, do so, this saves an instruction.
661 */
662 if(strlen(writemask) >= 5) {
663 shader_addline(buffer, "MAD %s, %s, %s, -%s;\n", reg, reg, two, one);
664 } else if(strlen(writemask) >= 3) {
665 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
666 reg, writemask[1],
667 reg, writemask[1],
668 two, one);
669 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
670 reg, writemask[2],
671 reg, writemask[2],
672 two, one);
673 } else if(strlen(writemask) == 2) {
674 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n", reg, writemask[1],
675 reg, writemask[1], two, one);
676 }
677 }
678 break;
679
680 case WINED3DFMT_X8L8V8U8:
681 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
682 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
683 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
684 */
685 if(strlen(writemask) >= 3) {
686 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
687 reg, writemask[1],
688 reg, writemask[1],
689 two, one);
690 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
691 reg, writemask[2],
692 reg, writemask[2],
693 two, one);
694 } else if(strlen(writemask) == 2) {
695 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
696 reg, writemask[1],
697 reg, writemask[1],
698 two, one);
699 }
700 }
701 break;
702
703 case WINED3DFMT_L6V5U5:
704 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
705 if(strlen(writemask) >= 4) {
706 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
707 shader_addline(buffer, "MOV TMP.g, %s.%c;\n",
708 reg, writemask[2]);
709 shader_addline(buffer, "MAD %s.%c%c, %s.%c%c, %s, -%s;\n",
710 reg, writemask[1], writemask[1],
711 reg, writemask[1], writemask[3],
712 two, one);
713 shader_addline(buffer, "MOV %s.%c, TMP.g;\n", reg,
714 writemask[3]);
715 } else if(strlen(writemask) == 3) {
716 /* This is bad: We have VL, but we need VU */
717 FIXME("2 components sampled from a converted L6V5U5 texture\n");
718 } else {
719 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
720 reg, writemask[1],
721 reg, writemask[1],
722 two, one);
723 }
724 }
725 break;
726
727 case WINED3DFMT_Q8W8V8U8:
728 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
729 /* Correct the sign in all channels */
730 switch(strlen(writemask)) {
731 case 4:
732 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
733 reg, writemask[3],
734 reg, writemask[3]);
735 /* drop through */
736 case 3:
737 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
738 reg, writemask[2],
739 reg, writemask[2]);
740 /* drop through */
741 case 2:
742 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
743 reg, writemask[1],
744 reg, writemask[1]);
745 break;
746
747 /* Should not occur, since it's at minimum '.' and a letter */
748 case 1:
749 ERR("Unexpected writemask: \"%s\"\n", writemask);
750 break;
751
752 case 5:
753 default:
754 shader_addline(buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
755 }
756 }
757 break;
758
759 case WINED3DFMT_ATI2N:
760 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
761 * which means the first one is replicated across .rgb, and the 2nd one is in
762 * .a. We need the 2nd in .g
763 *
764 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
765 * are swapped compared to d3d. So swap red and green.
766 */
767 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC)) {
768 shader_addline(buffer, "SWZ %s, %s, %c, %c, 1, 0;\n",
769 reg, reg, writemask[2], writemask[1]);
770 } else {
771 if(strlen(writemask) == 5) {
772 shader_addline(buffer, "MOV %s.%c, %s.%c;\n",
773 reg, writemask[2], reg, writemask[4]);
774 } else if(strlen(writemask) == 2) {
775 /* Nothing to do */
776 } else {
777 /* This is bad: We have VL, but we need VU */
778 FIXME("2 or 3 components sampled from a converted ATI2N texture\n");
779 }
780 }
781 break;
782
783 /* stupid compiler */
784 default:
785 break;
786 }
787 }
788
789 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
790 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
791 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
792 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
793 WINED3DFORMAT fmt;
794 WINED3DFORMAT conversion_group;
795 IWineD3DBaseTextureImpl *texture;
796 UINT i;
797 BOOL recorded = FALSE;
798 DWORD sampler_idx;
799 DWORD hex_version = shader->baseShader.hex_version;
800 char reg[256];
801 char writemask[6];
802
803 switch(arg->opcode->opcode) {
804 case WINED3DSIO_TEX:
805 if (hex_version < WINED3DPS_VERSION(2,0)) {
806 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
807 } else {
808 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
809 }
810 break;
811
812 case WINED3DSIO_TEXLDL:
813 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
814 return;
815
816 case WINED3DSIO_TEXDP3TEX:
817 case WINED3DSIO_TEXM3x3TEX:
818 case WINED3DSIO_TEXM3x3SPEC:
819 case WINED3DSIO_TEXM3x3VSPEC:
820 case WINED3DSIO_TEXBEM:
821 case WINED3DSIO_TEXREG2AR:
822 case WINED3DSIO_TEXREG2GB:
823 case WINED3DSIO_TEXREG2RGB:
824 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
825 break;
826
827 default:
828 /* Not a texture sampling instruction, nothing to do */
829 return;
830 };
831
832 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
833 if(texture) {
834 fmt = texture->resource.format;
835 conversion_group = texture->baseTexture.shader_conversion_group;
836 } else {
837 fmt = WINED3DFMT_UNKNOWN;
838 conversion_group = WINED3DFMT_UNKNOWN;
839 }
840
841 /* before doing anything, record the sampler with the format in the format conversion list,
842 * but check if it's not there already
843 */
844 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
845 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
846 recorded = TRUE;
847 }
848 }
849 if(!recorded) {
850 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
851 shader->baseShader.num_sampled_samplers++;
852 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
853 }
854
855 pshader_get_register_name(arg->shader, arg->dst, reg);
856 shader_arb_get_write_mask(arg, arg->dst, writemask);
857 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
858
859 gen_color_correction(arg->buffer, reg, writemask, "one", "coefmul.x", fmt, gl_info);
860
861 }
862
863
864 static void pshader_gen_input_modifier_line (
865 IWineD3DBaseShader *iface,
866 SHADER_BUFFER* buffer,
867 const DWORD instr,
868 int tmpreg,
869 char *outregstr) {
870
871 /* Generate a line that does the input modifier computation and return the input register to use */
872 char regstr[256];
873 char swzstr[20];
874 int insert_line;
875
876 /* Assume a new line will be added */
877 insert_line = 1;
878
879 /* Get register name */
880 pshader_get_register_name(iface, instr, regstr);
881 shader_arb_get_swizzle(instr, FALSE, swzstr);
882
883 switch (instr & WINED3DSP_SRCMOD_MASK) {
884 case WINED3DSPSM_NONE:
885 sprintf(outregstr, "%s%s", regstr, swzstr);
886 insert_line = 0;
887 break;
888 case WINED3DSPSM_NEG:
889 sprintf(outregstr, "-%s%s", regstr, swzstr);
890 insert_line = 0;
891 break;
892 case WINED3DSPSM_BIAS:
893 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
894 break;
895 case WINED3DSPSM_BIASNEG:
896 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
897 break;
898 case WINED3DSPSM_SIGN:
899 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
900 break;
901 case WINED3DSPSM_SIGNNEG:
902 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
903 break;
904 case WINED3DSPSM_COMP:
905 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
906 break;
907 case WINED3DSPSM_X2:
908 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
909 break;
910 case WINED3DSPSM_X2NEG:
911 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
912 break;
913 case WINED3DSPSM_DZ:
914 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
915 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
916 break;
917 case WINED3DSPSM_DW:
918 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
919 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
920 break;
921 default:
922 sprintf(outregstr, "%s%s", regstr, swzstr);
923 insert_line = 0;
924 }
925
926 /* Return modified or original register, with swizzle */
927 if (insert_line)
928 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
929 }
930
931 static inline void pshader_gen_output_modifier_line(
932 SHADER_BUFFER* buffer,
933 int saturate,
934 char *write_mask,
935 int shift,
936 char *regstr) {
937
938 /* Generate a line that does the output modifier computation */
939 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
940 regstr, write_mask, regstr, shift_tab[shift]);
941 }
942
943 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
944 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
945
946 SHADER_BUFFER* buffer = arg->buffer;
947 char dst_name[50];
948 char src_name[2][50];
949 char dst_wmask[20];
950 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
951 BOOL has_bumpmat = FALSE;
952 int i;
953
954 for(i = 0; i < This->numbumpenvmatconsts; i++) {
955 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
956 has_bumpmat = TRUE;
957 break;
958 }
959 }
960
961 pshader_get_register_name(arg->shader, arg->dst, dst_name);
962 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
963 strcat(dst_name, dst_wmask);
964
965 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
966 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
967
968 if(has_bumpmat) {
969 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
970 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
971 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
972 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
973 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
974
975 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
976 } else {
977 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
978 }
979 }
980
981 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
982
983 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
984 SHADER_BUFFER* buffer = arg->buffer;
985 char dst_wmask[20];
986 char dst_name[50];
987 char src_name[3][50];
988 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
989 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
990
991 /* FIXME: support output modifiers */
992
993 /* Handle output register */
994 pshader_get_register_name(arg->shader, arg->dst, dst_name);
995 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
996
997 /* Generate input register names (with modifiers) */
998 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
999 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1000 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1001
1002 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1003 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
1004 arg->opcode_token & WINED3DSI_COISSUE) {
1005 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
1006 } else {
1007 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
1008 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
1009 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
1010 }
1011 if (shift != 0)
1012 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1013 }
1014
1015 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
1016
1017 SHADER_BUFFER* buffer = arg->buffer;
1018 char dst_wmask[20];
1019 char dst_name[50];
1020 char src_name[3][50];
1021 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1022 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1023
1024 /* FIXME: support output modifiers */
1025
1026 /* Handle output register */
1027 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1028 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1029
1030 /* Generate input register names (with modifiers) */
1031 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1032 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1033 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1034
1035 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1036 src_name[0], src_name[2], src_name[1]);
1037
1038 if (shift != 0)
1039 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1040 }
1041
1042 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1043 * dst = dot2(src0, src1) + src2 */
1044 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1045 SHADER_BUFFER* buffer = arg->buffer;
1046 char dst_wmask[20];
1047 char dst_name[50];
1048 char src_name[3][50];
1049 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1050 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1051
1052 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1053 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1054
1055 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1056 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1057 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1058
1059 /* Emulate a DP2 with a DP3 and 0.0 */
1060 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1061 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1062 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1063 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1064
1065 if (shift != 0)
1066 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1067 }
1068
1069 /* Map the opcode 1-to-1 to the GL code */
1070 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1071
1072 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1073 SHADER_BUFFER* buffer = arg->buffer;
1074 DWORD dst = arg->dst;
1075 DWORD* src = arg->src;
1076
1077 unsigned int i;
1078 char tmpLine[256];
1079
1080 /* Output token related */
1081 char output_rname[256];
1082 char output_wmask[20];
1083 BOOL saturate = FALSE;
1084 BOOL centroid = FALSE;
1085 BOOL partialprecision = FALSE;
1086 DWORD shift;
1087
1088 strcpy(tmpLine, curOpcode->glname);
1089
1090 /* Process modifiers */
1091 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
1092 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1093
1094 saturate = mask & WINED3DSPDM_SATURATE;
1095 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1096 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1097 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1098 if (mask)
1099 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1100
1101 if (centroid)
1102 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1103 }
1104 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1105
1106 /* Generate input and output registers */
1107 if (curOpcode->num_params > 0) {
1108 char operands[4][100];
1109
1110 /* Generate input register names (with modifiers) */
1111 for (i = 1; i < curOpcode->num_params; ++i)
1112 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1113
1114 /* Handle output register */
1115 pshader_get_register_name(arg->shader, dst, output_rname);
1116 strcpy(operands[0], output_rname);
1117 shader_arb_get_write_mask(arg, dst, output_wmask);
1118 strcat(operands[0], output_wmask);
1119
1120 if (saturate && (shift == 0))
1121 strcat(tmpLine, "_SAT");
1122 strcat(tmpLine, " ");
1123 strcat(tmpLine, operands[0]);
1124 for (i = 1; i < curOpcode->num_params; i++) {
1125 strcat(tmpLine, ", ");
1126 strcat(tmpLine, operands[i]);
1127 }
1128 strcat(tmpLine,";\n");
1129 shader_addline(buffer, tmpLine);
1130
1131 /* A shift requires another line. */
1132 if (shift != 0)
1133 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1134 }
1135 }
1136
1137 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1138 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1139 DWORD hex_version = This->baseShader.hex_version;
1140 SHADER_BUFFER* buffer = arg->buffer;
1141 char reg_dest[40];
1142
1143 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1144 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1145 */
1146 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1147
1148 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1149 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1150 shader_addline(buffer, "KIL %s;\n", reg_dest);
1151 } else {
1152 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1153 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1154 */
1155 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1156 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1157 shader_addline(buffer, "KIL TMP;\n");
1158 }
1159 }
1160
1161 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1162 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1163 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1164
1165 DWORD dst = arg->dst;
1166 DWORD* src = arg->src;
1167 SHADER_BUFFER* buffer = arg->buffer;
1168 DWORD hex_version = This->baseShader.hex_version;
1169 BOOL projected = FALSE, bias = FALSE;
1170
1171 char reg_dest[40];
1172 char reg_coord[40];
1173 DWORD reg_dest_code;
1174 DWORD reg_sampler_code;
1175
1176 /* All versions have a destination register */
1177 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1178 pshader_get_register_name(arg->shader, dst, reg_dest);
1179
1180 /* 1.0-1.3: Use destination register as coordinate source.
1181 1.4+: Use provided coordinate source register. */
1182 if (hex_version < WINED3DPS_VERSION(1,4))
1183 strcpy(reg_coord, reg_dest);
1184 else
1185 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1186
1187 /* 1.0-1.4: Use destination register number as texture code.
1188 2.0+: Use provided sampler number as texure code. */
1189 if (hex_version < WINED3DPS_VERSION(2,0))
1190 reg_sampler_code = reg_dest_code;
1191 else
1192 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1193
1194 /* projection flag:
1195 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1196 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1197 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1198 */
1199 if(hex_version < WINED3DPS_VERSION(1,4)) {
1200 DWORD flags = 0;
1201 if(reg_sampler_code < MAX_TEXTURES) {
1202 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1203 }
1204 if (flags & WINED3DTTFF_PROJECTED) {
1205 projected = TRUE;
1206 }
1207 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1208 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1209 if (src_mod == WINED3DSPSM_DZ) {
1210 projected = TRUE;
1211 } else if(src_mod == WINED3DSPSM_DW) {
1212 projected = TRUE;
1213 }
1214 } else {
1215 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1216 projected = TRUE;
1217 }
1218 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1219 bias = TRUE;
1220 }
1221 }
1222 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1223 }
1224
1225 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1226
1227 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1228 DWORD dst = arg->dst;
1229 SHADER_BUFFER* buffer = arg->buffer;
1230 DWORD hex_version = This->baseShader.hex_version;
1231
1232 char tmp[20];
1233 shader_arb_get_write_mask(arg, dst, tmp);
1234 if (hex_version != WINED3DPS_VERSION(1,4)) {
1235 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1236 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1237 } else {
1238 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1239 char reg_src[40];
1240
1241 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1242 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1243 }
1244 }
1245
1246 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1247
1248 SHADER_BUFFER* buffer = arg->buffer;
1249 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1250 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1251 DWORD flags;
1252
1253 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1254 char dst_str[8];
1255 char src_str[50];
1256
1257 sprintf(dst_str, "T%u", reg1);
1258 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1259 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1260 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1261 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1262 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1263 }
1264
1265 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1266
1267 SHADER_BUFFER* buffer = arg->buffer;
1268
1269 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1270 char dst_str[8];
1271 char src_str[50];
1272
1273 sprintf(dst_str, "T%u", reg1);
1274 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1275 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1276 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1277 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1278 }
1279
1280 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1281
1282 SHADER_BUFFER* buffer = arg->buffer;
1283 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1284 char dst_str[8];
1285 char src_str[50];
1286
1287 sprintf(dst_str, "T%u", reg1);
1288 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1289 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1290 }
1291
1292 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1293 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1294 BOOL has_bumpmat = FALSE;
1295 BOOL has_luminance = FALSE;
1296 int i;
1297
1298 DWORD dst = arg->dst;
1299 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1300 SHADER_BUFFER* buffer = arg->buffer;
1301
1302 char reg_coord[40];
1303 DWORD reg_dest_code;
1304
1305 /* All versions have a destination register */
1306 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1307 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1308 pshader_get_register_name(arg->shader, dst, reg_coord);
1309
1310 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1311 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1312 has_bumpmat = TRUE;
1313 break;
1314 }
1315 }
1316 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1317 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1318 has_luminance = TRUE;
1319 break;
1320 }
1321 }
1322
1323 if(has_bumpmat) {
1324 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1325
1326 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1327 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1328 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1329 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1330
1331 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1332 * so we can't let the GL handle this.
1333 */
1334 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1335 & WINED3DTTFF_PROJECTED) {
1336 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1337 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1338 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1339 } else {
1340 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1341 }
1342
1343 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1344
1345 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1346 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1347 src, reg_dest_code, reg_dest_code);
1348 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1349 }
1350
1351 } else {
1352 DWORD tf;
1353 if(reg_dest_code < MAX_TEXTURES) {
1354 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1355 } else {
1356 tf = 0;
1357 }
1358 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1359 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1360 }
1361 }
1362
1363 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1364
1365 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1366 SHADER_BUFFER* buffer = arg->buffer;
1367 char src0_name[50];
1368
1369 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1370 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1371 }
1372
1373 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1374
1375 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1376 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1377 DWORD flags;
1378 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1379 SHADER_BUFFER* buffer = arg->buffer;
1380 char dst_str[8];
1381 char src0_name[50];
1382
1383 sprintf(dst_str, "T%u", reg);
1384 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1385 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1386 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1387 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1388 }
1389
1390 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1391
1392 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1393 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1394 SHADER_BUFFER* buffer = arg->buffer;
1395 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1396 char src0_name[50];
1397
1398 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1399 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1400 current_state->texcoord_w[current_state->current_row++] = reg;
1401 }
1402
1403 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1404
1405 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1406 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1407 DWORD flags;
1408 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1409 SHADER_BUFFER* buffer = arg->buffer;
1410 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1411 char dst_str[8];
1412 char src0_name[50];
1413
1414 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1415 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1416
1417 /* Sample the texture using the calculated coordinates */
1418 sprintf(dst_str, "T%u", reg);
1419 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1420 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1421 current_state->current_row = 0;
1422 }
1423
1424 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1425
1426 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1427 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1428 DWORD flags;
1429 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1430 SHADER_BUFFER* buffer = arg->buffer;
1431 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1432 char dst_str[8];
1433 char src0_name[50];
1434
1435 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1436 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1437
1438 /* Construct the eye-ray vector from w coordinates */
1439 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1440 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1441 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1442
1443 /* Calculate reflection vector
1444 */
1445 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1446 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1447 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1448 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1449 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1450 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1451 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1452
1453 /* Sample the texture using the calculated coordinates */
1454 sprintf(dst_str, "T%u", reg);
1455 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1456 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1457 current_state->current_row = 0;
1458 }
1459
1460 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1461
1462 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1463 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1464 DWORD flags;
1465 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1466 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1467 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1468 SHADER_BUFFER* buffer = arg->buffer;
1469 char dst_str[8];
1470 char src0_name[50];
1471
1472 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1473 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1474
1475 /* Calculate reflection vector.
1476 *
1477 * dot(N, E)
1478 * TMP.xyz = 2 * --------- * N - E
1479 * dot(N, N)
1480 *
1481 * Which normalizes the normal vector
1482 */
1483 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1484 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1485 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1486 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1487 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1488 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1489
1490 /* Sample the texture using the calculated coordinates */
1491 sprintf(dst_str, "T%u", reg);
1492 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1493 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1494 current_state->current_row = 0;
1495 }
1496
1497 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1498 SHADER_BUFFER* buffer = arg->buffer;
1499 char dst_name[50];
1500
1501 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1502 * which is essentially an input, is the destination register because it is the first
1503 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1504 * here
1505 */
1506 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1507
1508 /* According to the msdn, the source register(must be r5) is unusable after
1509 * the texdepth instruction, so we're free to modify it
1510 */
1511 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1512
1513 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1514 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1515 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1516 */
1517 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1518 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1519 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1520 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1521 }
1522
1523 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1524 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1525 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1526 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1527 SHADER_BUFFER* buffer = arg->buffer;
1528 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1529 char src0[50];
1530 char dst_str[8];
1531
1532 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1533 shader_addline(buffer, "MOV TMP, 0.0;\n");
1534 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1535
1536 sprintf(dst_str, "T%u", sampler_idx);
1537 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1538 }
1539
1540 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1541 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1542 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1543 char src0[50];
1544 char dst_str[50];
1545 char dst_mask[6];
1546 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1547 SHADER_BUFFER* buffer = arg->buffer;
1548
1549 /* Handle output register */
1550 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1551 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1552
1553 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1554 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1555
1556 /* TODO: Handle output modifiers */
1557 }
1558
1559 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1560 * Perform the 3rd row of a 3x3 matrix multiply */
1561 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1562 SHADER_BUFFER* buffer = arg->buffer;
1563 char dst_str[50];
1564 char dst_mask[6];
1565 char src0[50];
1566 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1567
1568 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1569 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1570
1571 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1572 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1573 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1574
1575 /* TODO: Handle output modifiers */
1576 }
1577
1578 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1579 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1580 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1581 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1582 */
1583 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1584 SHADER_BUFFER* buffer = arg->buffer;
1585 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1586 char src0[50];
1587
1588 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1589 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1590
1591 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1592 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1593 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1594 */
1595 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1596 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1597 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1598 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1599 }
1600
1601 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1602 Vertex/Pixel shaders to ARB_vertex_program codes */
1603 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1604
1605 int i;
1606 int nComponents = 0;
1607 SHADER_OPCODE_ARG tmpArg;
1608
1609 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1610
1611 /* Set constants for the temporary argument */
1612 tmpArg.shader = arg->shader;
1613 tmpArg.buffer = arg->buffer;
1614 tmpArg.src[0] = arg->src[0];
1615 tmpArg.src_addr[0] = arg->src_addr[0];
1616 tmpArg.src_addr[1] = arg->src_addr[1];
1617 tmpArg.reg_maps = arg->reg_maps;
1618
1619 switch(arg->opcode->opcode) {
1620 case WINED3DSIO_M4x4:
1621 nComponents = 4;
1622 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1623 break;
1624 case WINED3DSIO_M4x3:
1625 nComponents = 3;
1626 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1627 break;
1628 case WINED3DSIO_M3x4:
1629 nComponents = 4;
1630 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1631 break;
1632 case WINED3DSIO_M3x3:
1633 nComponents = 3;
1634 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1635 break;
1636 case WINED3DSIO_M3x2:
1637 nComponents = 2;
1638 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1639 break;
1640 default:
1641 break;
1642 }
1643
1644 for (i = 0; i < nComponents; i++) {
1645 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1646 tmpArg.src[1] = arg->src[1]+i;
1647 vshader_hw_map2gl(&tmpArg);
1648 }
1649 }
1650
1651 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1652 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1653 SHADER_BUFFER* buffer = arg->buffer;
1654 DWORD dst = arg->dst;
1655 DWORD src = arg->src[0];
1656 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1657
1658 char tmpLine[256];
1659
1660 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1661 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1662 strcat(tmpLine, ",");
1663 vshader_program_add_param(arg, src, TRUE, tmpLine);
1664 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1665 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1666 * .w is used
1667 */
1668 strcat(tmpLine, ".w");
1669 }
1670
1671 shader_addline(buffer, "%s;\n", tmpLine);
1672 }
1673
1674 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1675 SHADER_BUFFER* buffer = arg->buffer;
1676 char dst_name[50];
1677 char src_name[50];
1678 char dst_wmask[20];
1679 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1680 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1681
1682 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1683 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1684
1685 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1686 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1687 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1688 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1689 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1690 src_name);
1691
1692 if (shift != 0)
1693 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1694 }
1695
1696 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1697 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1698 * must contain fixed constants. So we need a separate function to filter those constants and
1699 * can't use map2gl
1700 */
1701 SHADER_BUFFER* buffer = arg->buffer;
1702 char dst_name[50];
1703 char src_name[50];
1704 char dst_wmask[20];
1705 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1706 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1707
1708 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1709 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1710
1711 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1712 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1713 src_name);
1714
1715 if (shift != 0)
1716 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1717
1718 }
1719
1720 /* TODO: merge with pixel shader */
1721 /* Map the opcode 1-to-1 to the GL code */
1722 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1723
1724 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1725 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1726 SHADER_BUFFER* buffer = arg->buffer;
1727 DWORD dst = arg->dst;
1728 DWORD* src = arg->src;
1729
1730 DWORD dst_regtype = shader_get_regtype(dst);
1731 char tmpLine[256];
1732 unsigned int i;
1733
1734 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1735 memset(tmpLine, 0, sizeof(tmpLine));
1736 if(shader->rel_offset) {
1737 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1738 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1739 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1740 } else {
1741 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1742 * with more than one component. Thus replicate the first source argument over all
1743 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1744 */
1745 DWORD parm = src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1746 if((src[0] & WINED3DVS_X_W) == WINED3DVS_X_W) {
1747 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1748 } else if((src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z) {
1749 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1750 } else if((src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y) {
1751 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1752 } else if((src[0] & WINED3DVS_X_X) == WINED3DVS_X_X) {
1753 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1754 }
1755 vshader_program_add_param(arg, parm, TRUE, tmpLine);
1756 shader_addline(buffer, "ARL A0.x, %s;\n", tmpLine);
1757 }
1758 return;
1759 } else
1760 strcpy(tmpLine, curOpcode->glname);
1761
1762 if (curOpcode->num_params > 0) {
1763 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1764 for (i = 1; i < curOpcode->num_params; ++i) {
1765 strcat(tmpLine, ",");
1766 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1767 }
1768 }
1769 shader_addline(buffer, "%s;\n", tmpLine);
1770 }
1771
1772 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1773 GLuint program_id = 0;
1774 const char *blt_vprogram =
1775 "!!ARBvp1.0\n"
1776 "PARAM c[1] = { { 1, 0.5 } };\n"
1777 "MOV result.position, vertex.position;\n"
1778 "MOV result.color, c[0].x;\n"
1779 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1780 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1781 "END\n";
1782
1783 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1784 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1785 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1786
1787 if (glGetError() == GL_INVALID_OPERATION) {
1788 GLint pos;
1789 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1790 FIXME("Vertex program error at position %d: %s\n", pos,
1791 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1792 }
1793
1794 return program_id;
1795 }
1796
1797 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1798 GLuint program_id = 0;
1799 const char *blt_fprogram =
1800 "!!ARBfp1.0\n"
1801 "TEMP R0;\n"
1802 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1803 "MOV result.depth.z, R0.x;\n"
1804 "END\n";
1805
1806 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1807 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1808 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1809
1810 if (glGetError() == GL_INVALID_OPERATION) {
1811 GLint pos;
1812 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1813 FIXME("Fragment program error at position %d: %s\n", pos,
1814 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1815 }
1816
1817 return program_id;
1818 }
1819
1820 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1821 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1822 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1823 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1824
1825 if (useVS) {
1826 TRACE("Using vertex shader\n");
1827
1828 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId;
1829
1830 /* Bind the vertex program */
1831 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1832 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1833
1834 /* Enable OpenGL vertex programs */
1835 glEnable(GL_VERTEX_PROGRAM_ARB);
1836 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1837 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1838 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1839 priv->current_vprogram_id = 0;
1840 glDisable(GL_VERTEX_PROGRAM_ARB);
1841 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1842 }
1843
1844 if (usePS) {
1845 TRACE("Using pixel shader\n");
1846
1847 priv->current_fprogram_id = ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId;
1848
1849 /* Bind the fragment program */
1850 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1851 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1852
1853 /* Enable OpenGL fragment programs */
1854 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1855 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1856 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1857 } else {
1858 priv->current_fprogram_id = 0;
1859
1860 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1861 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1862 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1863 * replacement shader
1864 */
1865 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1866 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1867 }
1868 }
1869 }
1870
1871 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1872 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1873 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1874 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1875
1876 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1877 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1878 glEnable(GL_VERTEX_PROGRAM_ARB);
1879
1880 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1881 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1882 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1883 }
1884
1885 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1886 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1887 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1888 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1889
1890 if (priv->current_vprogram_id) {
1891 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1892 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1893
1894 glEnable(GL_VERTEX_PROGRAM_ARB);
1895 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1896
1897 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1898 } else {
1899 glDisable(GL_VERTEX_PROGRAM_ARB);
1900 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1901 }
1902
1903 if (priv->current_fprogram_id) {
1904 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1905 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1906
1907 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1908 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1909
1910 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1911 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1912 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1913 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1914 }
1915 }
1916
1917 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1918 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1919 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1920 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1921 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1922 }
1923
1924 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1925 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1926 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1927
1928 ENTER_GL();
1929 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1930 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1931 LEAVE_GL();
1932 This->baseShader.prgId = 0;
1933 This->baseShader.is_compiled = FALSE;
1934 }
1935
1936 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1937 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1938 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1939 return WINED3D_OK;
1940 }
1941
1942 static void shader_arb_free(IWineD3DDevice *iface) {
1943 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1944 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1945 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1946
1947 if(priv->depth_blt_vprogram_id) {
1948 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1949 }
1950 if(priv->depth_blt_fprogram_id) {
1951 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1952 }
1953
1954 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1955 }
1956
1957 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1958 return TRUE;
1959 }
1960
1961 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1962 const char *tmp2, const char *tmp3, const char *tmp4) {
1963 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1964
1965 /* Calculate the > 0.0031308 case */
1966 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1967 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1968 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1969 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1970 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1971 /* Calculate the < case */
1972 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1973 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1974 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1975 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1976 /* Store the components > 0.0031308 in the destination */
1977 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1978 /* Add the components that are < 0.0031308 */
1979 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1980 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1981 }
1982
1983 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
1984 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1985 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1986 CONST DWORD *function = This->baseShader.function;
1987 const char *fragcolor;
1988 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1989 local_constant* lconst;
1990
1991 /* Create the hw ARB shader */
1992 shader_addline(buffer, "!!ARBfp1.0\n");
1993
1994 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1995 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1996 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1997 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1998 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1999 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2000 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2001 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2002
2003 /* Base Declarations */
2004 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2005
2006 /* We need two variables for fog blending */
2007 shader_addline(buffer, "TEMP TMP_FOG;\n");
2008 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
2009 shader_addline(buffer, "TEMP TMP_COLOR;\n");
2010 }
2011
2012 /* Base Shader Body */
2013 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2014
2015 /* calculate fog and blend it
2016 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
2017 * -1/(e-s) and e/(e-s) respectively.
2018 */
2019 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
2020
2021 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
2022 fragcolor = "R0";
2023 } else {
2024 fragcolor = "TMP_COLOR";
2025 }
2026 if(This->srgb_enabled) {
2027 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
2028 }
2029 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
2030 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
2031 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
2032 }
2033
2034 shader_addline(buffer, "END\n");
2035
2036 /* TODO: change to resource.glObjectHandle or something like that */
2037 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2038
2039 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
2040 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
2041
2042 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
2043 /* Create the program and check for errors */
2044 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2045 buffer->bsize, buffer->buffer));
2046
2047 if (glGetError() == GL_INVALID_OPERATION) {
2048 GLint errPos;
2049 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2050 FIXME("HW PixelShader Error at position %d: %s\n",
2051 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2052 This->baseShader.prgId = -1;
2053 }
2054
2055 /* Load immediate constants */
2056 if(!This->baseShader.load_local_constsF) {
2057 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2058 float *value = (float *) lconst->value;
2059 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2060 checkGLcall("glProgramLocalParameter4fvARB");
2061 }
2062 }
2063 }
2064
2065 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
2066 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2067 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2068 CONST DWORD *function = This->baseShader.function;
2069 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
2070 local_constant* lconst;
2071
2072 /* Create the hw ARB shader */
2073 shader_addline(buffer, "!!ARBvp1.0\n");
2074 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2075
2076 /* Mesa supports only 95 constants */
2077 if (GL_VEND(MESA) || GL_VEND(WINE))
2078 This->baseShader.limits.constant_float =
2079 min(95, This->baseShader.limits.constant_float);
2080
2081 shader_addline(buffer, "TEMP TMP;\n");
2082
2083 /* Base Declarations */
2084 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2085
2086 /* We need a constant to fixup the final position */
2087 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2088
2089 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2090 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2091 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2092 * a replacement shader depend on the texcoord.w being set properly.
2093 *
2094 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2095 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2096 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2097 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2098 * this can eat a number of instructions, so skip it unless this cap is set as well
2099 */
2100 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2101 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2102
2103 if((GLINFO_LOCATION).set_texcoord_w) {
2104 int i;
2105 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2106 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2107 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2108 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2109 }
2110 }
2111 }
2112 }
2113
2114 /* Base Shader Body */
2115 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2116
2117 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2118 if (!reg_maps->fog)
2119 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2120
2121 /* Write the final position.
2122 *
2123 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2124 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2125 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2126 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2127 */
2128 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2129 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2130 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2131
2132 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2133 * and the glsl equivalent
2134 */
2135 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2136
2137 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2138
2139 shader_addline(buffer, "END\n");
2140
2141 /* TODO: change to resource.glObjectHandle or something like that */
2142 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2143
2144 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2145 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2146
2147 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2148 /* Create the program and check for errors */
2149 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2150 buffer->bsize, buffer->buffer));
2151
2152 if (glGetError() == GL_INVALID_OPERATION) {
2153 GLint errPos;
2154 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2155 FIXME("HW VertexShader Error at position %d: %s\n",
2156 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2157 This->baseShader.prgId = -1;
2158 }
2159
2160 /* Load immediate constants */
2161 if(!This->baseShader.load_local_constsF) {
2162 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2163 float *value = (float *) lconst->value;
2164 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2165 }
2166 }
2167 }
2168
2169 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2170 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2171 * then overwrite the shader specific ones
2172 */
2173 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2174
2175 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2176 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2177 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2178 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2179 }
2180
2181 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2182 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2183 pCaps->PixelShader1xMaxValue = 8.0;
2184 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2185 }
2186 }
2187
2188 static BOOL shader_arb_conv_supported(WINED3DFORMAT fmt) {
2189 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt));
2190 switch(fmt) {
2191 case WINED3DFMT_V8U8:
2192 case WINED3DFMT_V16U16:
2193 case WINED3DFMT_X8L8V8U8:
2194 case WINED3DFMT_L6V5U5:
2195 case WINED3DFMT_Q8W8V8U8:
2196 case WINED3DFMT_ATI2N:
2197 TRACE("[OK]\n");
2198 return TRUE;
2199 default:
2200 TRACE("[FAILED\n");
2201 return FALSE;
2202 }
2203 }
2204
2205 const shader_backend_t arb_program_shader_backend = {
2206 shader_arb_select,
2207 shader_arb_select_depth_blt,
2208 shader_arb_deselect_depth_blt,
2209 shader_arb_load_constants,
2210 shader_arb_cleanup,
2211 shader_arb_color_correction,
2212 shader_arb_destroy,
2213 shader_arb_alloc,
2214 shader_arb_free,
2215 shader_arb_dirty_const,
2216 shader_arb_generate_pshader,
2217 shader_arb_generate_vshader,
2218 shader_arb_get_caps,
2219 shader_arb_conv_supported,
2220 };
2221
2222 /* ARB_fragment_program fixed function pipeline replacement definitions */
2223 #define ARB_FFP_CONST_TFACTOR 0
2224 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_TFACTOR) + 1 + i)
2225 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2226
2227 struct arbfp_ffp_desc
2228 {
2229 struct ffp_desc parent;
2230 GLuint shader;
2231 unsigned int num_textures_used;
2232 };
2233
2234 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2235 if(enable) {
2236 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2237 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2238 } else {
2239 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2240 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2241 }
2242 }
2243
2244 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2245 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2246 struct shader_arb_priv *priv;
2247 /* Share private data between the shader backend and the pipeline replacement, if both
2248 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2249 * if no pixel shader is bound or not
2250 */
2251 if(This->shader_backend == &arb_program_shader_backend) {
2252 This->fragment_priv = This->shader_priv;
2253 } else {
2254 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2255 if(!This->fragment_priv) return E_OUTOFMEMORY;
2256 }
2257 priv = (struct shader_arb_priv *) This->fragment_priv;
2258 priv->fragment_shaders = hash_table_create(ffp_program_key_hash, ffp_program_key_compare);
2259 priv->use_arbfp_fixed_func = TRUE;
2260 return WINED3D_OK;
2261 }
2262
2263 static void arbfp_free_ffpshader(void *value, void *gli) {
2264 WineD3D_GL_Info *gl_info = gli;
2265 struct arbfp_ffp_desc *entry_arb = value;
2266
2267 ENTER_GL();
2268 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2269 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2270 HeapFree(GetProcessHeap(), 0, entry_arb);
2271 LEAVE_GL();
2272 }
2273
2274 static void arbfp_free(IWineD3DDevice *iface) {
2275 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2276 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2277
2278 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2279 priv->use_arbfp_fixed_func = FALSE;
2280
2281 if(This->shader_backend != &arb_program_shader_backend) {
2282 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2283 }
2284 }
2285
2286 static void arbfp_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct fragment_caps *caps) {
2287 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2288 WINED3DTEXOPCAPS_SELECTARG1 |
2289 WINED3DTEXOPCAPS_SELECTARG2 |
2290 WINED3DTEXOPCAPS_MODULATE4X |
2291 WINED3DTEXOPCAPS_MODULATE2X |
2292 WINED3DTEXOPCAPS_MODULATE |
2293 WINED3DTEXOPCAPS_ADDSIGNED2X |
2294 WINED3DTEXOPCAPS_ADDSIGNED |
2295 WINED3DTEXOPCAPS_ADD |
2296 WINED3DTEXOPCAPS_SUBTRACT |
2297 WINED3DTEXOPCAPS_ADDSMOOTH |
2298 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2299 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2300 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2301 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2302 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2303 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2304 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2305 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2306 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2307 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2308 WINED3DTEXOPCAPS_MULTIPLYADD |
2309 WINED3DTEXOPCAPS_LERP |
2310 WINED3DTEXOPCAPS_BUMPENVMAP;
2311
2312 /* TODO: Implement WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
2313 and WINED3DTEXOPCAPS_PREMODULATE */
2314
2315 caps->MaxTextureBlendStages = 8;
2316 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2317
2318 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2319 }
2320 #undef GLINFO_LOCATION
2321
2322 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2323 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2324 float col[4];
2325 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2326
2327 /* Do not overwrite pixel shader constants if a pshader is in use */
2328 if(use_ps(device)) return;
2329
2330 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2331 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2332 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2333
2334 if(device->shader_backend == &arb_program_shader_backend) {
2335 device = stateblock->wineD3DDevice;
2336 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2337 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2338 }
2339 }
2340
2341 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2342 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2343 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2344 float mat[2][2];
2345
2346 if(use_ps(device)) {
2347 if(stage != 0 &&
2348 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2349 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2350 * anyway
2351 */
2352 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2353 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2354 }
2355 }
2356 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2357 return;
2358 }
2359
2360 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2361 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2362 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2363 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2364
2365 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2366 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2367
2368 if(device->shader_backend == &arb_program_shader_backend) {
2369 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2370 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2371 }
2372 }
2373
2374 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2375 const char *ret;
2376
2377 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2378
2379 switch(arg & WINED3DTA_SELECTMASK) {
2380 case WINED3DTA_DIFFUSE:
2381 ret = "fragment.color.primary"; break;
2382
2383 case WINED3DTA_CURRENT:
2384 if(stage == 0) ret = "fragment.color.primary";
2385 else ret = "ret";
2386 break;
2387
2388 case WINED3DTA_TEXTURE:
2389 switch(stage) {
2390 case 0: ret = "tex0"; break;
2391 case 1: ret = "tex1"; break;
2392 case 2: ret = "tex2"; break;
2393 case 3: ret = "tex3"; break;
2394 case 4: ret = "tex4"; break;
2395 case 5: ret = "tex5"; break;
2396 case 6: ret = "tex6"; break;
2397 case 7: ret = "tex7"; break;
2398 default: ret = "unknown texture";
2399 }
2400 break;
2401
2402 case WINED3DTA_TFACTOR:
2403 ret = "tfactor"; break;
2404
2405 case WINED3DTA_SPECULAR:
2406 ret = "fragment.color.secondary"; break;
2407
2408 case WINED3DTA_TEMP:
2409 ret = "tempreg"; break;
2410
2411 case WINED3DTA_CONSTANT:
2412 FIXME("Implement perstage constants\n");
2413 switch(stage) {
2414 case 0: ret = "const0"; break;
2415 case 1: ret = "const1"; break;
2416 case 2: ret = "const2"; break;
2417 case 3: ret = "const3"; break;
2418 case 4: ret = "const4"; break;
2419 case 5: ret = "const5"; break;
2420 case 6: ret = "const6"; break;
2421 case 7: ret = "const7"; break;
2422 }
2423 default:
2424 return "unknown";
2425 }
2426
2427 if(arg & WINED3DTA_COMPLEMENT) {
2428 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2429 if(argnum == 0) ret = "arg0";
2430 if(argnum == 1) ret = "arg1";
2431 if(argnum == 2) ret = "arg2";
2432 }
2433 return ret;
2434 }
2435
2436 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha, BOOL last,
2437 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2438 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2439 unsigned int mul = 1;
2440 BOOL mul_final_dest = FALSE;
2441
2442 if(color && alpha) dstmask = "";
2443 else if(color) dstmask = ".rgb";
2444 else dstmask = ".a";
2445
2446 if(dst == tempreg && last) FIXME("Last texture stage writes to D3DTA_TEMP\n");
2447 if(dst == tempreg) dstreg = "tempreg";
2448 else if(last) dstreg = "result.color";
2449 else dstreg = "ret";
2450
2451 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2452 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2453 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2454
2455 switch(op) {
2456 case WINED3DTOP_DISABLE:
2457 if(stage == 1) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2458 break;
2459
2460 case WINED3DTOP_SELECTARG2:
2461 arg1 = arg2;
2462 case WINED3DTOP_SELECTARG1:
2463 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2464 break;
2465
2466 case WINED3DTOP_MODULATE4X:
2467 mul = 2;
2468 case WINED3DTOP_MODULATE2X:
2469 mul *= 2;
2470 if(strcmp(dstreg, "result.color") == 0) {
2471 dstreg = "ret";
2472 mul_final_dest = TRUE;
2473 }
2474 case WINED3DTOP_MODULATE:
2475 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2476 break;
2477
2478 case WINED3DTOP_ADDSIGNED2X:
2479 mul = 2;
2480 if(strcmp(dstreg, "result.color") == 0) {
2481 dstreg = "ret";
2482 mul_final_dest = TRUE;
2483 }
2484 case WINED3DTOP_ADDSIGNED:
2485 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2486 arg2 = "arg2";
2487 case WINED3DTOP_ADD:
2488 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2489 break;
2490
2491 case WINED3DTOP_SUBTRACT:
2492 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2493 break;
2494
2495 case WINED3DTOP_ADDSMOOTH:
2496 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2497 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2498 break;
2499
2500 case WINED3DTOP_BLENDCURRENTALPHA:
2501 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2502 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2503 break;
2504 case WINED3DTOP_BLENDFACTORALPHA:
2505 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2506 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2507 break;
2508 case WINED3DTOP_BLENDTEXTUREALPHA:
2509 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2510 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2511 break;
2512 case WINED3DTOP_BLENDDIFFUSEALPHA:
2513 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2514 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2515 break;
2516
2517 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2518 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2519 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg, dstmask, arg2, arg1);
2520 break;
2521
2522 /* D3DTOP_PREMODULATE ???? */
2523
2524 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2525 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2526 shader_addline(buffer, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2527 break;
2528 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2529 shader_addline(buffer, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2530 break;
2531 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2532 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2533 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg, dstmask, arg2, arg1);
2534 break;
2535 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2536 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg, dstmask, arg1, arg2, arg1);
2537 break;
2538
2539 case WINED3DTOP_DOTPRODUCT3:
2540 mul = 4;
2541 if(strcmp(dstreg, "result.color") == 0) {
2542 dstreg = "ret";
2543 mul_final_dest = TRUE;
2544 }
2545 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2546 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2547 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2548 break;
2549
2550 case WINED3DTOP_MULTIPLYADD:
2551 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2552 break;
2553
2554 case WINED3DTOP_LERP:
2555 /* The msdn is not quite right here */
2556 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2557 break;
2558
2559 case WINED3DTOP_BUMPENVMAP:
2560 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2561 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2562 break;
2563
2564 default:
2565 FIXME("Unhandled texture op %08x\n", op);
2566 }
2567
2568 if(mul == 2) {
2569 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2570 } else if(mul == 4) {
2571 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2572 }
2573 }
2574
2575 /* The stateblock is passed for GLINFO_LOCATION */
2576 static GLuint gen_arbfp_ffp_shader(struct ffp_settings *settings, IWineD3DStateBlockImpl *stateblock) {
2577 unsigned int stage;
2578 SHADER_BUFFER buffer;
2579 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2580 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2581 const char *textype;
2582 const char *instr, *sat;
2583 char colorcor_dst[8];
2584 GLuint ret;
2585 DWORD arg0, arg1, arg2;
2586 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2587 BOOL last = FALSE;
2588
2589 /* Find out which textures are read */
2590 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2591 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2592 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2593 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2594 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2595 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2596 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2597 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2598
2599 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2600 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2601 bump_used[stage] = TRUE;
2602 tex_read[stage] = TRUE;
2603 }
2604 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2605 bump_used[stage] = TRUE;
2606 tex_read[stage] = TRUE;
2607 }
2608
2609 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2610 tfactor_used = TRUE;
2611 }
2612
2613 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2614 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2615 tempreg_used = TRUE;
2616 }
2617
2618 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2619 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2620 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2621 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2622 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2623 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2624 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2625
2626 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2627 tempreg_used = TRUE;
2628 }
2629 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2630 tfactor_used = TRUE;
2631 }
2632 }
2633
2634 /* Shader header */
2635 buffer.bsize = 0;
2636 buffer.lineNo = 0;
2637 buffer.newline = TRUE;
2638 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
2639
2640 shader_addline(&buffer, "!!ARBfp1.0\n");
2641
2642 switch(settings->fog) {
2643 case FOG_OFF: break;
2644 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2645 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2646 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2647 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2648 }
2649
2650 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2651 shader_addline(&buffer, "TEMP ret;\n");
2652 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2653 shader_addline(&buffer, "TEMP arg0;\n");
2654 shader_addline(&buffer, "TEMP arg1;\n");
2655 shader_addline(&buffer, "TEMP arg2;\n");
2656 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2657 if(!tex_read[stage]) continue;
2658 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2659 if(!bump_used[stage]) continue;
2660 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2661 }
2662 if(tfactor_used) {
2663 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2664 }
2665
2666 if(settings->sRGB_write) {
2667 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2668 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2669 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2670 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2671 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2672 srgb_pow, srgb_pow, srgb_pow);
2673 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2674 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2675 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2676 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2677 }
2678
2679 /* Generate texture sampling instructions) */
2680 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2681 if(!tex_read[stage]) continue;
2682
2683 switch(settings->op[stage].tex_type) {
2684 case tex_1d: textype = "1D"; break;
2685 case tex_2d: textype = "2D"; break;
2686 case tex_3d: textype = "3D"; break;
2687 case tex_cube: textype = "CUBE"; break;
2688 case tex_rect: textype = "RECT"; break;
2689 default: textype = "unexpected_textype"; break;
2690 }
2691
2692 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2693 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2694 sat = "";
2695 } else {
2696 sat = "_SAT";
2697 }
2698
2699 if(settings->op[stage].projected == proj_none) {
2700 instr = "TEX";
2701 } else if(settings->op[stage].projected == proj_count4) {
2702 instr = "TXP";
2703 } else {
2704 instr = "TXP";
2705 ERR("Implement proj_count3\n");
2706 }
2707
2708 if(stage > 0 &&
2709 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2710 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2711 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2712 shader_addline(&buffer, "DP3 ret.r, arg1, tex%u;\n", stage - 1);
2713 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2714 shader_addline(&buffer, "DP3 ret.g, arg1, tex%u;\n", stage - 1);
2715 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2716 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2717 instr, sat, stage, stage, textype);
2718 } else {
2719 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2720 instr, sat, stage, stage, stage, textype);
2721 }
2722
2723 sprintf(colorcor_dst, "tex%u", stage);
2724 gen_color_correction(&buffer, colorcor_dst, ".rgba", "const.x", "const.y",
2725 settings->op[stage].color_correction, &GLINFO_LOCATION);
2726 }
2727
2728 /* Generate the main shader */
2729 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2730 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2731 if(stage == 0) {
2732 shader_addline(&buffer, "MOV result.color, fragment.color.primary;\n");
2733 }
2734 break;
2735 } else if(settings->sRGB_write) {
2736 last = FALSE;
2737 } else if(stage == (MAX_TEXTURES - 1)) {
2738 last = TRUE;
2739 } else if(settings->op[stage + 1].cop == WINED3DTOP_DISABLE) {
2740 last = TRUE;
2741 }
2742
2743 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2744 gen_ffp_instr(&buffer, stage, TRUE, FALSE, last, settings->op[stage].dst,
2745 settings->op[stage].cop, settings->op[stage].carg0,
2746 settings->op[stage].carg1, settings->op[stage].carg2);
2747 if(last && stage == 0) {
2748 shader_addline(&buffer, "MOV result.color.a, fragment.color.primary.a;\n");
2749 } else if(last) {
2750 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
2751 } else if(stage == 0) {
2752 shader_addline(&buffer, "MOV ret.a, fragment.color.primary.a;\n");
2753 }
2754 } else if(settings->op[stage].aop == settings->op[stage].cop &&
2755 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2756 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2757 settings->op[stage].carg2 == settings->op[stage].aarg2) {
2758 gen_ffp_instr(&buffer, stage, TRUE, TRUE, last, settings->op[stage].dst,
2759 settings->op[stage].cop, settings->op[stage].carg0,
2760 settings->op[stage].carg1, settings->op[stage].carg2);
2761 } else {
2762 gen_ffp_instr(&buffer, stage, TRUE, FALSE, last, settings->op[stage].dst,
2763 settings->op[stage].cop, settings->op[stage].carg0,
2764 settings->op[stage].carg1, settings->op[stage].carg2);
2765 gen_ffp_instr(&buffer, stage, FALSE, TRUE, last, settings->op[stage].dst,
2766 settings->op[stage].aop, settings->op[stage].aarg0,
2767 settings->op[stage].aarg1, settings->op[stage].aarg2);
2768 }
2769 }
2770
2771 if(settings->sRGB_write) {
2772 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2773 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
2774 }
2775
2776 /* Footer */
2777 shader_addline(&buffer, "END\n");
2778
2779 /* Generate the shader */
2780 GL_EXTCALL(glGenProgramsARB(1, &ret));
2781 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2782 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2783
2784 if (glGetError() == GL_INVALID_OPERATION) {
2785 GLint pos;
2786 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2787 FIXME("Vertex program error at position %d: %s\n", pos,
2788 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2789 }
2790 return ret;
2791 }
2792
2793 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2794 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2795 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
2796 BOOL use_pshader = use_ps(device);
2797 BOOL use_vshader = use_vs(device);
2798 struct ffp_settings settings;
2799 struct arbfp_ffp_desc *desc;
2800 unsigned int i;
2801
2802 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2803 if(use_pshader) {
2804 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
2805 }
2806 return;
2807 }
2808
2809 if(use_pshader) {
2810 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
2811 } else {
2812 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2813 gen_ffp_op(stateblock, &settings, FALSE);
2814 desc = (struct arbfp_ffp_desc *) find_ffp_shader(priv->fragment_shaders, &settings);
2815 if(!desc) {
2816 desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*desc));
2817 if(!desc) {
2818 ERR("Out of memory\n");
2819 return;
2820 }
2821 desc->num_textures_used = 0;
2822 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
2823 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
2824 desc->num_textures_used = i;
2825 }
2826
2827 memcpy(&desc->parent.settings, &settings, sizeof(settings));
2828 desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
2829 add_ffp_shader(priv->fragment_shaders, &desc->parent);
2830 TRACE("Allocated fixed function replacement shader descriptor %p\n", desc);
2831 }
2832
2833 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
2834 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
2835 * deactivate it.
2836 */
2837 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
2838 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
2839
2840 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2841 /* Reload fixed function constants since they collide with the pixel shader constants */
2842 for(i = 0; i < MAX_TEXTURES; i++) {
2843 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2844 }
2845 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2846 }
2847 }
2848
2849 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
2850 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
2851 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
2852 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
2853 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
2854 *
2855 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
2856 * to be compiled before activating them(needs some cleanups in the shader backend interface)
2857 */
2858 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2859 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2860
2861 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
2862 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
2863 }
2864 }
2865 if(use_pshader) {
2866 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2867 }
2868 }
2869
2870 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
2871 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
2872 * state table, so we need to handle that with a forwarding function. The other invisible side effect
2873 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
2874 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
2875 */
2876 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2877 if(!isStateDirty(context, STATE_PIXELSHADER)) {
2878 fragment_prog_arbfp(state, stateblock, context);
2879 }
2880 }
2881
2882 #undef GLINFO_LOCATION
2883
2884 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
2885 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
2886 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2887 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2888 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2889 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2890 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2891 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2892 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2893 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2894 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2895 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2896 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2897 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2898 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2899 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2900 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2901 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2902 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2903 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2904 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2905 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2906 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2907 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2908 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2909 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2910 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2911 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2912 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2913 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2914 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2915 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2916 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2917 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2918 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2919 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2920 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2921 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2922 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2923 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2924 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2925 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2926 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2927 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2928 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2929 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2930 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2931 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2932 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2933 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2934 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2935 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2936 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2937 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2938 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2939 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2940 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2941 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2942 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2943 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2944 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2945 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2946 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2947 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2948 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2949 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2950 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2951 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2952 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2953 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2954 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2955 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2956 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2957 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2958 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2959 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2960 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2961 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2962 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2963 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2964 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2965 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2966 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2967 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2968 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2969 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2970 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2971 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2972 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2973 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2974 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2975 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2976 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2977 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2978 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2979 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2980 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2981 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2982 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2983 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2984 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2985 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2986 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2987 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2988 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2989 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
2990 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
2991 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
2992 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
2993 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
2994 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
2995 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
2996 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
2997 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
2998 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
2999 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3000 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3001 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3002 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3003 {0 /* Terminate */, { 0, 0 }, 0 },
3004 };
3005
3006 const struct fragment_pipeline arbfp_fragment_pipeline = {
3007 arbfp_enable,
3008 arbfp_get_caps,
3009 arbfp_alloc,
3010 arbfp_free,
3011 shader_arb_conv_supported,
3012 arbfp_fragmentstate_template
3013 };
3014
3015 #define GLINFO_LOCATION device->adapter->gl_info
3016
3017 struct arbfp_blit_priv {
3018 GLenum yuy2_rect_shader, yuy2_2d_shader;
3019 GLenum uyvy_rect_shader, uyvy_2d_shader;
3020 };
3021
3022 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3023 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3024 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3025 if(!device->blit_priv) {
3026 ERR("Out of memory\n");
3027 return E_OUTOFMEMORY;
3028 }
3029 return WINED3D_OK;
3030 }
3031 static void arbfp_blit_free(IWineD3DDevice *iface) {
3032 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3033 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3034
3035 ENTER_GL();
3036 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3037 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3038 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3039 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3040 checkGLcall("Delete yuv programs\n");
3041 LEAVE_GL();
3042 }
3043
3044 GLenum gen_yuv_shader(IWineD3DDeviceImpl *device, WINED3DFORMAT fmt, GLenum textype) {
3045 GLenum shader;
3046 SHADER_BUFFER buffer;
3047 const char *tex, *texinstr;
3048 char chroma, luminance;
3049 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3050
3051 /* Shader header */
3052 buffer.bsize = 0;
3053 buffer.lineNo = 0;
3054 buffer.newline = TRUE;
3055 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
3056
3057 switch(textype) {
3058 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3059 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3060 default:
3061 /* This is more tricky than just replacing the texture type - we have to navigate
3062 * properly in the texture to find the correct chroma values
3063 */
3064 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3065 return 0;
3066 }
3067
3068 if(fmt == WINED3DFMT_UYVY) {
3069 chroma = 'r';
3070 luminance = 'a';
3071 } else {
3072 chroma = 'a';
3073 luminance = 'r';
3074 }
3075
3076 GL_EXTCALL(glGenProgramsARB(1, &shader));
3077 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3078 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3079 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3080 if(!shader) return 0;
3081
3082 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3083 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3084 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3085 * each single pixel it contains, and one U and one V value shared between both
3086 * pixels.
3087 *
3088 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3089 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3090 * take the format into account when generating the read swizzles
3091 *
3092 * Reading the Y value is streightforward - just sample the texture. The hardware
3093 * takes care of filtering in the horizontal and vertical direction.
3094 *
3095 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3096 * because that would mix the U and V values of one pixel or two adjacent pixels.
3097 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3098 * regardless of the filtering setting. Vertical filtering works automatically
3099 * though - the U and V values of two rows are mixed nicely.
3100 *
3101 * Appart of avoiding filtering issues, the code has to know which value it just
3102 * read, and where it can find the other one. To determine this, it checks if
3103 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3104 *
3105 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3106 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3107 *
3108 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3109 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3110 * in an unfiltered situation. Finding the luminance on the other hand requires
3111 * finding out if it is an odd or even pixel. The real drawback of this approach
3112 * is filtering. This would have to be emulated completely in the shader, reading
3113 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3114 * vertically. Beyond that it would require adjustments to the texture handling
3115 * code to deal with the width scaling
3116 */
3117 shader_addline(&buffer, "!!ARBfp1.0\n");
3118 shader_addline(&buffer, "TEMP luminance;\n");
3119 shader_addline(&buffer, "TEMP temp;\n");
3120 shader_addline(&buffer, "TEMP chroma;\n");
3121 shader_addline(&buffer, "TEMP texcrd;\n");
3122 shader_addline(&buffer, "TEMP texcrd2;\n");
3123 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.0};\n");
3124 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3125 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3126
3127 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3128 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3129 * filtering when we sample the texture.
3130 *
3131 * These are the rules for reading the chroma:
3132 *
3133 * Even pixel: Cr
3134 * Even pixel: U
3135 * Odd pixel: V
3136 *
3137 * So we have to get the sampling x position in non-normalized coordinates in integers
3138 */
3139 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3140 shader_addline(&buffer, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3141 shader_addline(&buffer, "MOV texcrd.a, size.x;\n");
3142 } else {
3143 shader_addline(&buffer, "MOV texcrd, fragment.texcoord[0];\n");
3144 }
3145 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3146 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3147 * 0.5, so add 0.5.
3148 */
3149 shader_addline(&buffer, "FLR texcrd.x, texcrd.x;\n");
3150 shader_addline(&buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3151
3152 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3153 * even and odd pixels respectively
3154 */
3155 shader_addline(&buffer, "MUL texcrd2, texcrd, coef.y;\n");
3156 shader_addline(&buffer, "FRC texcrd2, texcrd2;\n");
3157
3158 /* Sample Pixel 1 */
3159 shader_addline(&buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3160
3161 /* Put the value into either of the chroma values */
3162 shader_addline(&buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3163 shader_addline(&buffer, "MUL chroma.r, luminance.%c, temp.x;\n", chroma);
3164 shader_addline(&buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3165 shader_addline(&buffer, "MUL chroma.g, luminance.%c, temp.x;\n", chroma);
3166
3167 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3168 * the pixel right to the current one. Otherwise, sample the left pixel.
3169 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3170 */
3171 shader_addline(&buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3172 shader_addline(&buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3173 shader_addline(&buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3174
3175 /* Put the value into the other chroma */
3176 shader_addline(&buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3177 shader_addline(&buffer, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma);
3178 shader_addline(&buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3179 shader_addline(&buffer, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma);
3180
3181 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3182 * the current one and lerp the two U and V values
3183 */
3184
3185 /* This gives the correctly filtered luminance value */
3186 shader_addline(&buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3187
3188 /* Calculate the final result. Formula is taken from
3189 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3190 * ranges from -0.5 to 0.5
3191 */
3192 shader_addline(&buffer, "SUB chroma.rg, chroma, coef.y;\n");
3193
3194 shader_addline(&buffer, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance);
3195 shader_addline(&buffer, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance);
3196 shader_addline(&buffer, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3197 shader_addline(&buffer, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance);
3198 shader_addline(&buffer, "END\n");
3199
3200 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3201
3202 if (glGetError() == GL_INVALID_OPERATION) {
3203 GLint pos;
3204 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3205 FIXME("Fragment program error at position %d: %s\n", pos,
3206 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3207 }
3208
3209 if(fmt == WINED3DFMT_YUY2) {
3210 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3211 priv->yuy2_rect_shader = shader;
3212 } else {
3213 priv->yuy2_2d_shader = shader;
3214 }
3215 } else {
3216 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3217 priv->uyvy_rect_shader = shader;
3218 } else {
3219 priv->uyvy_2d_shader = shader;
3220 }
3221 }
3222 return shader;
3223 }
3224
3225 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3226 GLenum shader;
3227 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3228 float size[4] = {width, height, 1, 1};
3229 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3230 const GlPixelFormatDesc *glDesc;
3231
3232 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3233
3234 if(glDesc->conversion_group != WINED3DFMT_YUY2 && glDesc->conversion_group != WINED3DFMT_UYVY) {
3235 /* Don't bother setting up a shader for unconverted formats */
3236 glEnable(textype);
3237 checkGLcall("glEnable(textype)");
3238 return WINED3D_OK;
3239 }
3240
3241 if(glDesc->conversion_group == WINED3DFMT_YUY2) {
3242 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3243 shader = priv->yuy2_rect_shader;
3244 } else {
3245 shader = priv->yuy2_2d_shader;
3246 }
3247 } else {
3248 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3249 shader = priv->uyvy_rect_shader;
3250 } else {
3251 shader = priv->uyvy_2d_shader;
3252 }
3253 }
3254
3255 if(!shader) {
3256 shader = gen_yuv_shader(device, glDesc->conversion_group, textype);
3257 }
3258
3259 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3260 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3261 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3262 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3263 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3264 checkGLcall("glProgramLocalParameter4fvARB");
3265
3266 return WINED3D_OK;
3267 }
3268
3269 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3270 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3271 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3272 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3273 glDisable(GL_TEXTURE_2D);
3274 checkGLcall("glDisable(GL_TEXTURE_2D)");
3275 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3276 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3277 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3278 }
3279 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3280 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3281 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3282 }
3283 }
3284
3285 static BOOL arbfp_blit_conv_supported(WINED3DFORMAT fmt) {
3286 TRACE("Checking blit format support for format %s:", debug_d3dformat(fmt));
3287 switch(fmt) {
3288 case WINED3DFMT_YUY2:
3289 case WINED3DFMT_UYVY:
3290 TRACE("[OK]\n");
3291 return TRUE;
3292 default:
3293 TRACE("[FAILED]\n");
3294 return FALSE;
3295 }
3296 }
3297
3298 const struct blit_shader arbfp_blit = {
3299 arbfp_blit_alloc,
3300 arbfp_blit_free,
3301 arbfp_blit_set,
3302 arbfp_blit_unset,
3303 arbfp_blit_conv_supported
3304 };
3305
3306 #undef GLINFO_LOCATION
useful hacks and other patches not (yet) suitable for mainline