?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push 9eb9af089d68d39110a91889d3a673043db63c4b
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
blob493c5d2df9a6012e98ea8b3913436140db0d2a2c
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 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
43 * so upload them above that
44 */
45 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
46 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
47
48 /* ARB_program_shader private data */
49 struct shader_arb_priv {
50 GLuint current_vprogram_id;
51 GLuint current_fprogram_id;
52 GLuint depth_blt_vprogram_id;
53 GLuint depth_blt_fprogram_id[tex_type_count];
54 BOOL use_arbfp_fixed_func;
55 struct hash_table_t *fragment_shaders;
56 };
57
58 /********************************************************
59 * ARB_[vertex/fragment]_program helper functions follow
60 ********************************************************/
61
62 /**
63 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
64 * When constant_list == NULL, it will load all the constants.
65 *
66 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
67 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
68 */
69 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
70 unsigned int max_constants, float* constants, char *dirty_consts) {
71 local_constant* lconst;
72 DWORD i, j;
73 unsigned int ret;
74
75 if (TRACE_ON(d3d_shader)) {
76 for(i = 0; i < max_constants; i++) {
77 if(!dirty_consts[i]) continue;
78 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
79 constants[i * 4 + 0], constants[i * 4 + 1],
80 constants[i * 4 + 2], constants[i * 4 + 3]);
81 }
82 }
83 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
84 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
85 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
86 float lcl_const[4];
87 for(i = 0; i < max_constants; i++) {
88 if(!dirty_consts[i]) continue;
89 dirty_consts[i] = 0;
90
91 j = 4 * i;
92 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
93 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
94 else lcl_const[0] = constants[j + 0];
95
96 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
97 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
98 else lcl_const[1] = constants[j + 1];
99
100 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
101 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
102 else lcl_const[2] = constants[j + 2];
103
104 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
105 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
106 else lcl_const[3] = constants[j + 3];
107
108 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
109 }
110 } else {
111 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
112 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
113 * or just reloading *all* constants at once
114 *
115 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
116 */
117 for(i = 0; i < max_constants; i++) {
118 if(!dirty_consts[i]) continue;
119
120 /* Find the next block of dirty constants */
121 dirty_consts[i] = 0;
122 j = i;
123 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
124 dirty_consts[i] = 0;
125 }
126
127 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
128 }
129 } else {
130 for(i = 0; i < max_constants; i++) {
131 if(dirty_consts[i]) {
132 dirty_consts[i] = 0;
133 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
134 }
135 }
136 }
137 }
138 checkGLcall("glProgramEnvParameter4fvARB()");
139
140 /* Load immediate constants */
141 if(This->baseShader.load_local_constsF) {
142 if (TRACE_ON(d3d_shader)) {
143 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
144 GLfloat* values = (GLfloat*)lconst->value;
145 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
146 values[0], values[1], values[2], values[3]);
147 }
148 }
149 /* Immediate constants are clamped for 1.X shaders at loading times */
150 ret = 0;
151 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
152 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
153 ret = max(ret, lconst->idx + 1);
154 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
155 }
156 checkGLcall("glProgramEnvParameter4fvARB()");
157 return ret; /* The loaded immediate constants need reloading for the next shader */
158 } else {
159 return 0; /* No constants are dirty now */
160 }
161 }
162
163 /**
164 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
165 *
166 * We only support float constants in ARB at the moment, so don't
167 * worry about the Integers or Booleans
168 */
169 static void shader_arb_load_constants(
170 IWineD3DDevice* device,
171 char usePixelShader,
172 char useVertexShader) {
173
174 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
175 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
176 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
177 unsigned char i;
178
179 if (useVertexShader) {
180 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
181
182 /* Load DirectX 9 float constants for vertex shader */
183 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
184 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
185 deviceImpl->highest_dirty_vs_const,
186 stateBlock->vertexShaderConstantF,
187 deviceImpl->activeContext->vshader_const_dirty);
188
189 /* Upload the position fixup */
190 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
191 }
192
193 if (usePixelShader) {
194
195 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
196 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
197
198 /* Load DirectX 9 float constants for pixel shader */
199 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
200 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
201 deviceImpl->highest_dirty_ps_const,
202 stateBlock->pixelShaderConstantF,
203 deviceImpl->activeContext->pshader_const_dirty);
204
205 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
206 /* The state manager takes care that this function is always called if the bump env matrix changes
207 */
208 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
209 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
210 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
211
212 if(psi->luminanceconst[i].const_num != -1) {
213 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
214 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
215 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
216 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
217 */
218 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
219 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
220 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
221 }
222 }
223
224 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
225 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
226 float comparison[4];
227 float mul_low[4];
228
229 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
230 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
231 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
232
233 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
234 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
235 } else {
236 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
237 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
238
239 mul_low[0] = 1.0; mul_low[1] = 1.0;
240 mul_low[2] = 1.0; mul_low[3] = 1.0;
241 }
242 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
243 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
244 checkGLcall("Load sRGB correction constants\n");
245 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
246 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
247
248 }
249 }
250 }
251
252 /* Generate the variable & register declarations for the ARB_vertex_program output target */
253 static void shader_generate_arb_declarations(
254 IWineD3DBaseShader *iface,
255 shader_reg_maps* reg_maps,
256 SHADER_BUFFER* buffer,
257 WineD3D_GL_Info* gl_info) {
258
259 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
260 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
261 DWORD i, cur;
262 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
263 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
264 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
265 UINT extra_constants_needed = 0;
266 local_constant* lconst;
267
268 /* Temporary Output register */
269 shader_addline(buffer, "TEMP TMP_OUT;\n");
270
271 for(i = 0; i < This->baseShader.limits.temporary; i++) {
272 if (reg_maps->temporary[i])
273 shader_addline(buffer, "TEMP R%u;\n", i);
274 }
275
276 for (i = 0; i < This->baseShader.limits.address; i++) {
277 if (reg_maps->address[i])
278 shader_addline(buffer, "ADDRESS A%d;\n", i);
279 }
280
281 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
282 if (reg_maps->texcoord[i])
283 shader_addline(buffer,"TEMP T%u;\n", i);
284 }
285
286 /* Texture coordinate registers must be pre-loaded */
287 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
288 if (reg_maps->texcoord[i])
289 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
290 }
291
292 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
293 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
294 if(!reg_maps->bumpmat[i]) continue;
295
296 cur = ps->numbumpenvmatconsts;
297 ps->bumpenvmatconst[cur].const_num = -1;
298 ps->bumpenvmatconst[cur].texunit = i;
299 ps->luminanceconst[cur].const_num = -1;
300 ps->luminanceconst[cur].texunit = i;
301
302 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
303 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
304 * bump mapping.
305 */
306 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
307 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
308 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
309 i, ps->bumpenvmatconst[cur].const_num);
310 extra_constants_needed++;
311
312 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
313 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
314 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
315 i, ps->luminanceconst[cur].const_num);
316 extra_constants_needed++;
317 } else if(reg_maps->luminanceparams) {
318 FIXME("No free constant to load the luminance parameters\n");
319 }
320 } else {
321 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
322 }
323
324 ps->numbumpenvmatconsts = cur + 1;
325 }
326
327 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
328 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
329 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
330 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
331 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
332 * off again
333 */
334 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
335 /* The idea is that if srgb is enabled, then disabled, the constant loading code
336 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
337 * and comparison constants. If it disables it that way, the shader won't be recompiled
338 * and the code will stay in, so sRGB writing can be turned on again by setting the
339 * constants from the spec
340 */
341 ps_impl->srgb_mode_hardcoded = 0;
342 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
343 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
344 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
345 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
346 } else {
347 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
348 srgb_mul_low, srgb_mul_low, srgb_mul_low);
349 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
350 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
351 ps_impl->srgb_mode_hardcoded = 1;
352 }
353 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
354 * path if the comparison value is set to INF
355 */
356 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
357 srgb_pow, srgb_pow, srgb_pow);
358 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
359 srgb_mul_high, srgb_mul_high, srgb_mul_high);
360 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
361 srgb_sub_high, srgb_sub_high, srgb_sub_high);
362 ps_impl->srgb_enabled = 1;
363 } else if(pshader) {
364 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
365
366 /* Do not write any srgb fixup into the shader to save shader size and processing time.
367 * As a consequence, we can't toggle srgb write on without recompilation
368 */
369 ps_impl->srgb_enabled = 0;
370 ps_impl->srgb_mode_hardcoded = 1;
371 }
372
373 /* Load local constants using the program-local space,
374 * this avoids reloading them each time the shader is used
375 */
376 if(!This->baseShader.load_local_constsF) {
377 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
378 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
379 lconst->idx);
380 }
381 }
382
383 /* we use the array-based constants array if the local constants are marked for loading,
384 * because then we use indirect addressing, or when the local constant list is empty,
385 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
386 * local constants do not declare the loaded constants as an array because ARB compilers usually
387 * do not optimize unused constants away
388 */
389 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
390 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
391 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
392 max_constantsF, max_constantsF - 1);
393 } else {
394 for(i = 0; i < max_constantsF; i++) {
395 if(!shader_constant_is_local(This, i)) {
396 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
397 }
398 }
399 }
400 }
401
402 static const char * const shift_tab[] = {
403 "dummy", /* 0 (none) */
404 "coefmul.x", /* 1 (x2) */
405 "coefmul.y", /* 2 (x4) */
406 "coefmul.z", /* 3 (x8) */
407 "coefmul.w", /* 4 (x16) */
408 "dummy", /* 5 (x32) */
409 "dummy", /* 6 (x64) */
410 "dummy", /* 7 (x128) */
411 "dummy", /* 8 (d256) */
412 "dummy", /* 9 (d128) */
413 "dummy", /* 10 (d64) */
414 "dummy", /* 11 (d32) */
415 "coefdiv.w", /* 12 (d16) */
416 "coefdiv.z", /* 13 (d8) */
417 "coefdiv.y", /* 14 (d4) */
418 "coefdiv.x" /* 15 (d2) */
419 };
420
421 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
422 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
423 char *ptr = write_mask;
424 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
425
426 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
427 *ptr++ = '.';
428 *ptr++ = 'x';
429 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
430 *ptr++ = '.';
431 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
432 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
433 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
434 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
435 }
436
437 *ptr = '\0';
438 }
439
440 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
441 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
442 * but addressed as "rgba". To fix this we need to swap the register's x
443 * and z components. */
444 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
445 char *ptr = swizzle_str;
446
447 /* swizzle bits fields: wwzzyyxx */
448 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
449 DWORD swizzle_x = swizzle & 0x03;
450 DWORD swizzle_y = (swizzle >> 2) & 0x03;
451 DWORD swizzle_z = (swizzle >> 4) & 0x03;
452 DWORD swizzle_w = (swizzle >> 6) & 0x03;
453
454 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
455 * generate a swizzle string. Unless we need to our own swizzling. */
456 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
457 *ptr++ = '.';
458 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
459 *ptr++ = swizzle_chars[swizzle_x];
460 } else {
461 *ptr++ = swizzle_chars[swizzle_x];
462 *ptr++ = swizzle_chars[swizzle_y];
463 *ptr++ = swizzle_chars[swizzle_z];
464 *ptr++ = swizzle_chars[swizzle_w];
465 }
466 }
467
468 *ptr = '\0';
469 }
470
471 static void pshader_get_register_name(IWineD3DBaseShader* iface,
472 const DWORD param, char* regstr) {
473
474 DWORD reg = param & WINED3DSP_REGNUM_MASK;
475 DWORD regtype = shader_get_regtype(param);
476 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
477
478 switch (regtype) {
479 case WINED3DSPR_TEMP:
480 sprintf(regstr, "R%u", reg);
481 break;
482 case WINED3DSPR_INPUT:
483 if (reg==0) {
484 strcpy(regstr, "fragment.color.primary");
485 } else {
486 strcpy(regstr, "fragment.color.secondary");
487 }
488 break;
489 case WINED3DSPR_CONST:
490 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
491 sprintf(regstr, "C[%u]", reg);
492 } else {
493 sprintf(regstr, "C%u", reg);
494 }
495 break;
496 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
497 sprintf(regstr,"T%u", reg);
498 break;
499 case WINED3DSPR_COLOROUT:
500 if (reg == 0)
501 sprintf(regstr, "TMP_COLOR");
502 else {
503 /* TODO: See GL_ARB_draw_buffers */
504 FIXME("Unsupported write to render target %u\n", reg);
505 sprintf(regstr, "unsupported_register");
506 }
507 break;
508 case WINED3DSPR_DEPTHOUT:
509 sprintf(regstr, "result.depth");
510 break;
511 case WINED3DSPR_ATTROUT:
512 sprintf(regstr, "oD[%u]", reg);
513 break;
514 case WINED3DSPR_TEXCRDOUT:
515 sprintf(regstr, "oT[%u]", reg);
516 break;
517 default:
518 FIXME("Unhandled register name Type(%d)\n", regtype);
519 sprintf(regstr, "unrecognized_register");
520 break;
521 }
522 }
523
524 /* TODO: merge with pixel shader */
525 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
526
527 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
528
529 /* oPos, oFog and oPts in D3D */
530 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
531
532 DWORD reg = param & WINED3DSP_REGNUM_MASK;
533 DWORD regtype = shader_get_regtype(param);
534 char tmpReg[255];
535 BOOL is_color = FALSE;
536
537 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
538 strcat(hwLine, " -");
539 } else {
540 strcat(hwLine, " ");
541 }
542
543 switch (regtype) {
544 case WINED3DSPR_TEMP:
545 sprintf(tmpReg, "R%u", reg);
546 strcat(hwLine, tmpReg);
547 break;
548 case WINED3DSPR_INPUT:
549
550 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
551 is_color = TRUE;
552
553 sprintf(tmpReg, "vertex.attrib[%u]", reg);
554 strcat(hwLine, tmpReg);
555 break;
556 case WINED3DSPR_CONST:
557 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
558 if(reg >= This->rel_offset) {
559 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
560 } else {
561 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
562 }
563 } else {
564 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
565 sprintf(tmpReg, "C[%u]", reg);
566 } else {
567 sprintf(tmpReg, "C%u", reg);
568 }
569 }
570 strcat(hwLine, tmpReg);
571 break;
572 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
573 sprintf(tmpReg, "A%u", reg);
574 strcat(hwLine, tmpReg);
575 break;
576 case WINED3DSPR_RASTOUT:
577 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
578 strcat(hwLine, tmpReg);
579 break;
580 case WINED3DSPR_ATTROUT:
581 if (reg==0) {
582 strcat(hwLine, "result.color.primary");
583 } else {
584 strcat(hwLine, "result.color.secondary");
585 }
586 break;
587 case WINED3DSPR_TEXCRDOUT:
588 sprintf(tmpReg, "result.texcoord[%u]", reg);
589 strcat(hwLine, tmpReg);
590 break;
591 default:
592 FIXME("Unknown reg type %d %d\n", regtype, reg);
593 strcat(hwLine, "unrecognized_register");
594 break;
595 }
596
597 if (!is_input) {
598 char write_mask[6];
599 shader_arb_get_write_mask(arg, param, write_mask);
600 strcat(hwLine, write_mask);
601 } else {
602 char swizzle[6];
603 shader_arb_get_swizzle(param, is_color, swizzle);
604 strcat(hwLine, swizzle);
605 }
606 }
607
608 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
609 SHADER_BUFFER* buffer = arg->buffer;
610 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
611 const char *tex_type;
612
613 switch(sampler_type) {
614 case WINED3DSTT_1D:
615 tex_type = "1D";
616 break;
617
618 case WINED3DSTT_2D:
619 {
620 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
621 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
622 if(device->stateBlock->textures[sampler_idx] &&
623 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
624 tex_type = "RECT";
625 } else {
626 tex_type = "2D";
627 }
628 break;
629 }
630
631 case WINED3DSTT_VOLUME:
632 tex_type = "3D";
633 break;
634
635 case WINED3DSTT_CUBE:
636 tex_type = "CUBE";
637 break;
638
639 default:
640 ERR("Unexpected texture type %d\n", sampler_type);
641 tex_type = "";
642 }
643
644 if (bias) {
645 /* Shouldn't be possible, but let's check for it */
646 if(projected) FIXME("Biased and Projected texture sampling\n");
647 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
648 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
649 } else if (projected) {
650 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
651 } else {
652 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
653 }
654 }
655
656 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, const char *writemask,
657 const char *one, const char *two, WINED3DFORMAT fmt,
658 WineD3D_GL_Info *gl_info) {
659 switch(fmt) {
660 case WINED3DFMT_V8U8:
661 case WINED3DFMT_V16U16:
662 if(GL_SUPPORT(NV_TEXTURE_SHADER) && fmt == WINED3DFMT_V8U8) {
663 if(0) {
664 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
665 * disabled until an application that needs it is found because it causes unneeded
666 * shader recompilation in some game
667 */
668 if(strlen(writemask) >= 4) {
669 shader_addline(buffer, "MOV %s.%c, %s;\n", reg, writemask[3], one);
670 }
671 }
672 } else {
673 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
674 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
675 * all registers, do so, this saves an instruction.
676 */
677 if(strlen(writemask) >= 5) {
678 shader_addline(buffer, "MAD %s, %s, %s, -%s;\n", reg, reg, two, one);
679 } else if(strlen(writemask) >= 3) {
680 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
681 reg, writemask[1],
682 reg, writemask[1],
683 two, one);
684 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
685 reg, writemask[2],
686 reg, writemask[2],
687 two, one);
688 } else if(strlen(writemask) == 2) {
689 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n", reg, writemask[1],
690 reg, writemask[1], two, one);
691 }
692 }
693 break;
694
695 case WINED3DFMT_X8L8V8U8:
696 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
697 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
698 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
699 */
700 if(strlen(writemask) >= 3) {
701 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
702 reg, writemask[1],
703 reg, writemask[1],
704 two, one);
705 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
706 reg, writemask[2],
707 reg, writemask[2],
708 two, one);
709 } else if(strlen(writemask) == 2) {
710 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
711 reg, writemask[1],
712 reg, writemask[1],
713 two, one);
714 }
715 }
716 break;
717
718 case WINED3DFMT_L6V5U5:
719 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
720 if(strlen(writemask) >= 4) {
721 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
722 shader_addline(buffer, "MOV TMP.g, %s.%c;\n",
723 reg, writemask[2]);
724 shader_addline(buffer, "MAD %s.%c%c, %s.%c%c%c%c, %s, -%s;\n",
725 reg, writemask[1], writemask[2],
726 reg, writemask[3], writemask[1], writemask[3], writemask[1],
727 two, one);
728 shader_addline(buffer, "MOV %s.%c, TMP.g;\n", reg,
729 writemask[3]);
730 } else if(strlen(writemask) == 3) {
731 /* This is bad: We have VL, but we need VU */
732 FIXME("2 components sampled from a converted L6V5U5 texture\n");
733 } else {
734 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
735 reg, writemask[1],
736 reg, writemask[1],
737 two, one);
738 }
739 }
740 break;
741
742 case WINED3DFMT_Q8W8V8U8:
743 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
744 /* Correct the sign in all channels */
745 switch(strlen(writemask)) {
746 case 4:
747 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
748 reg, writemask[3],
749 reg, writemask[3]);
750 /* drop through */
751 case 3:
752 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
753 reg, writemask[2],
754 reg, writemask[2]);
755 /* drop through */
756 case 2:
757 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
758 reg, writemask[1],
759 reg, writemask[1]);
760 break;
761
762 /* Should not occur, since it's at minimum '.' and a letter */
763 case 1:
764 ERR("Unexpected writemask: \"%s\"\n", writemask);
765 break;
766
767 case 5:
768 default:
769 shader_addline(buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
770 }
771 }
772 break;
773
774 case WINED3DFMT_ATI2N:
775 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
776 * which means the first one is replicated across .rgb, and the 2nd one is in
777 * .a. We need the 2nd in .g
778 *
779 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
780 * are swapped compared to d3d. So swap red and green.
781 */
782 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC)) {
783 shader_addline(buffer, "SWZ %s, %s, %c, %c, 1, 0;\n",
784 reg, reg, writemask[2], writemask[1]);
785 } else {
786 if(strlen(writemask) == 5) {
787 shader_addline(buffer, "MOV %s.%c, %s.%c;\n",
788 reg, writemask[2], reg, writemask[4]);
789 } else if(strlen(writemask) == 2) {
790 /* Nothing to do */
791 } else {
792 /* This is bad: We have VL, but we need VU */
793 FIXME("2 or 3 components sampled from a converted ATI2N texture\n");
794 }
795 }
796 break;
797
798 /* stupid compiler */
799 default:
800 break;
801 }
802 }
803
804 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
805 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
806 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
807 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
808 WINED3DFORMAT fmt;
809 WINED3DFORMAT conversion_group;
810 IWineD3DBaseTextureImpl *texture;
811 UINT i;
812 BOOL recorded = FALSE;
813 DWORD sampler_idx;
814 DWORD hex_version = shader->baseShader.hex_version;
815 char reg[256];
816 char writemask[6];
817
818 switch(arg->opcode->opcode) {
819 case WINED3DSIO_TEX:
820 if (hex_version < WINED3DPS_VERSION(2,0)) {
821 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
822 } else {
823 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
824 }
825 break;
826
827 case WINED3DSIO_TEXLDL:
828 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
829 return;
830
831 case WINED3DSIO_TEXDP3TEX:
832 case WINED3DSIO_TEXM3x3TEX:
833 case WINED3DSIO_TEXM3x3SPEC:
834 case WINED3DSIO_TEXM3x3VSPEC:
835 case WINED3DSIO_TEXBEM:
836 case WINED3DSIO_TEXREG2AR:
837 case WINED3DSIO_TEXREG2GB:
838 case WINED3DSIO_TEXREG2RGB:
839 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
840 break;
841
842 default:
843 /* Not a texture sampling instruction, nothing to do */
844 return;
845 };
846
847 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
848 if(texture) {
849 fmt = texture->resource.format;
850 conversion_group = texture->baseTexture.shader_conversion_group;
851 } else {
852 fmt = WINED3DFMT_UNKNOWN;
853 conversion_group = WINED3DFMT_UNKNOWN;
854 }
855
856 /* before doing anything, record the sampler with the format in the format conversion list,
857 * but check if it's not there already
858 */
859 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
860 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
861 recorded = TRUE;
862 }
863 }
864 if(!recorded) {
865 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
866 shader->baseShader.num_sampled_samplers++;
867 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
868 }
869
870 pshader_get_register_name(arg->shader, arg->dst, reg);
871 shader_arb_get_write_mask(arg, arg->dst, writemask);
872 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
873
874 gen_color_correction(arg->buffer, reg, writemask, "one", "coefmul.x", fmt, gl_info);
875
876 }
877
878
879 static void pshader_gen_input_modifier_line (
880 IWineD3DBaseShader *iface,
881 SHADER_BUFFER* buffer,
882 const DWORD instr,
883 int tmpreg,
884 char *outregstr) {
885
886 /* Generate a line that does the input modifier computation and return the input register to use */
887 char regstr[256];
888 char swzstr[20];
889 int insert_line;
890
891 /* Assume a new line will be added */
892 insert_line = 1;
893
894 /* Get register name */
895 pshader_get_register_name(iface, instr, regstr);
896 shader_arb_get_swizzle(instr, FALSE, swzstr);
897
898 switch (instr & WINED3DSP_SRCMOD_MASK) {
899 case WINED3DSPSM_NONE:
900 sprintf(outregstr, "%s%s", regstr, swzstr);
901 insert_line = 0;
902 break;
903 case WINED3DSPSM_NEG:
904 sprintf(outregstr, "-%s%s", regstr, swzstr);
905 insert_line = 0;
906 break;
907 case WINED3DSPSM_BIAS:
908 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
909 break;
910 case WINED3DSPSM_BIASNEG:
911 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
912 break;
913 case WINED3DSPSM_SIGN:
914 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
915 break;
916 case WINED3DSPSM_SIGNNEG:
917 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
918 break;
919 case WINED3DSPSM_COMP:
920 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
921 break;
922 case WINED3DSPSM_X2:
923 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
924 break;
925 case WINED3DSPSM_X2NEG:
926 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
927 break;
928 case WINED3DSPSM_DZ:
929 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
930 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
931 break;
932 case WINED3DSPSM_DW:
933 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
934 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
935 break;
936 default:
937 sprintf(outregstr, "%s%s", regstr, swzstr);
938 insert_line = 0;
939 }
940
941 /* Return modified or original register, with swizzle */
942 if (insert_line)
943 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
944 }
945
946 static inline void pshader_gen_output_modifier_line(
947 SHADER_BUFFER* buffer,
948 int saturate,
949 char *write_mask,
950 int shift,
951 char *regstr) {
952
953 /* Generate a line that does the output modifier computation */
954 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
955 regstr, write_mask, regstr, shift_tab[shift]);
956 }
957
958 static void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
959 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
960
961 SHADER_BUFFER* buffer = arg->buffer;
962 char dst_name[50];
963 char src_name[2][50];
964 char dst_wmask[20];
965 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
966 BOOL has_bumpmat = FALSE;
967 int i;
968
969 for(i = 0; i < This->numbumpenvmatconsts; i++) {
970 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
971 has_bumpmat = TRUE;
972 break;
973 }
974 }
975
976 pshader_get_register_name(arg->shader, arg->dst, dst_name);
977 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
978 strcat(dst_name, dst_wmask);
979
980 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
981 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
982
983 if(has_bumpmat) {
984 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
985 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
986 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
987 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
988 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
989
990 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
991 } else {
992 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
993 }
994 }
995
996 static void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
997
998 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
999 SHADER_BUFFER* buffer = arg->buffer;
1000 char dst_wmask[20];
1001 char dst_name[50];
1002 char src_name[3][50];
1003 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1004 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1005
1006 /* FIXME: support output modifiers */
1007
1008 /* Handle output register */
1009 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1010 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1011
1012 /* Generate input register names (with modifiers) */
1013 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1014 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1015 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1016
1017 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1018 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
1019 arg->opcode_token & WINED3DSI_COISSUE) {
1020 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
1021 } else {
1022 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
1023 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
1024 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
1025 }
1026 if (shift != 0)
1027 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1028 }
1029
1030 static void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
1031
1032 SHADER_BUFFER* buffer = arg->buffer;
1033 char dst_wmask[20];
1034 char dst_name[50];
1035 char src_name[3][50];
1036 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1037 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1038
1039 /* FIXME: support output modifiers */
1040
1041 /* Handle output register */
1042 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1043 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1044
1045 /* Generate input register names (with modifiers) */
1046 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1047 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1048 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1049
1050 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1051 src_name[0], src_name[2], src_name[1]);
1052
1053 if (shift != 0)
1054 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1055 }
1056
1057 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1058 * dst = dot2(src0, src1) + src2 */
1059 static void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1060 SHADER_BUFFER* buffer = arg->buffer;
1061 char dst_wmask[20];
1062 char dst_name[50];
1063 char src_name[3][50];
1064 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1065 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1066
1067 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1068 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1069
1070 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1071 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1072 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1073
1074 /* Emulate a DP2 with a DP3 and 0.0 */
1075 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1076 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1077 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1078 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1079
1080 if (shift != 0)
1081 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1082 }
1083
1084 /* Map the opcode 1-to-1 to the GL code */
1085 static void shader_hw_map2gl(SHADER_OPCODE_ARG* arg)
1086 {
1087 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1088 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1089 SHADER_BUFFER* buffer = arg->buffer;
1090 DWORD dst = arg->dst;
1091 DWORD* src = arg->src;
1092
1093 char tmpLine[256];
1094 unsigned int i;
1095
1096 if (shader_is_pshader_version(shader->baseShader.hex_version))
1097 {
1098 BOOL saturate = FALSE;
1099 BOOL centroid = FALSE;
1100 BOOL partialprecision = FALSE;
1101 DWORD shift;
1102
1103 strcpy(tmpLine, curOpcode->glname);
1104
1105 /* Process modifiers */
1106 if (dst & WINED3DSP_DSTMOD_MASK)
1107 {
1108 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1109
1110 saturate = mask & WINED3DSPDM_SATURATE;
1111 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1112 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1113 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1114 if (mask)
1115 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1116
1117 if (centroid)
1118 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1119 }
1120 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1121
1122 /* Generate input and output registers */
1123 if (curOpcode->num_params > 0)
1124 {
1125 /* Output token related */
1126 char output_rname[256];
1127 char output_wmask[20];
1128 char operands[4][100];
1129
1130 /* Generate input register names (with modifiers) */
1131 for (i = 1; i < curOpcode->num_params; ++i)
1132 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1133
1134 /* Handle output register */
1135 pshader_get_register_name(arg->shader, dst, output_rname);
1136 strcpy(operands[0], output_rname);
1137 shader_arb_get_write_mask(arg, dst, output_wmask);
1138 strcat(operands[0], output_wmask);
1139
1140 if (saturate && (shift == 0))
1141 strcat(tmpLine, "_SAT");
1142 strcat(tmpLine, " ");
1143 strcat(tmpLine, operands[0]);
1144 for (i = 1; i < curOpcode->num_params; i++)
1145 {
1146 strcat(tmpLine, ", ");
1147 strcat(tmpLine, operands[i]);
1148 }
1149 strcat(tmpLine,";\n");
1150 shader_addline(buffer, tmpLine);
1151
1152 /* A shift requires another line. */
1153 if (shift != 0)
1154 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1155 }
1156 } else {
1157 if ((curOpcode->opcode == WINED3DSIO_MOV && shader_get_regtype(dst) == WINED3DSPR_ADDR)
1158 || curOpcode->opcode == WINED3DSIO_MOVA) {
1159 memset(tmpLine, 0, sizeof(tmpLine));
1160 if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
1161 {
1162 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1163 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1164 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1165 } else {
1166 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1167 * with more than one component. Thus replicate the first source argument over all
1168 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1169 */
1170 DWORD parm = src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1171 if((src[0] & WINED3DVS_X_W) == WINED3DVS_X_W) {
1172 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1173 } else if((src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z) {
1174 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1175 } else if((src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y) {
1176 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1177 } else if((src[0] & WINED3DVS_X_X) == WINED3DVS_X_X) {
1178 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1179 }
1180 vshader_program_add_param(arg, parm, TRUE, tmpLine);
1181 shader_addline(buffer, "ARL A0.x, %s;\n", tmpLine);
1182 }
1183 return;
1184 } else
1185 strcpy(tmpLine, curOpcode->glname);
1186
1187 if (curOpcode->num_params > 0)
1188 {
1189 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1190 for (i = 1; i < curOpcode->num_params; ++i)
1191 {
1192 strcat(tmpLine, ",");
1193 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1194 }
1195 }
1196 shader_addline(buffer, "%s;\n", tmpLine);
1197 }
1198 }
1199
1200 static void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1201 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1202 DWORD hex_version = This->baseShader.hex_version;
1203 SHADER_BUFFER* buffer = arg->buffer;
1204 char reg_dest[40];
1205
1206 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1207 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1208 */
1209 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1210
1211 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1212 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1213 shader_addline(buffer, "KIL %s;\n", reg_dest);
1214 } else {
1215 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1216 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1217 */
1218 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1219 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1220 shader_addline(buffer, "KIL TMP;\n");
1221 }
1222 }
1223
1224 static void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1225 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1226 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1227
1228 DWORD dst = arg->dst;
1229 DWORD* src = arg->src;
1230 SHADER_BUFFER* buffer = arg->buffer;
1231 DWORD hex_version = This->baseShader.hex_version;
1232 BOOL projected = FALSE, bias = FALSE;
1233
1234 char reg_dest[40];
1235 char reg_coord[40];
1236 DWORD reg_dest_code;
1237 DWORD reg_sampler_code;
1238
1239 /* All versions have a destination register */
1240 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1241 pshader_get_register_name(arg->shader, dst, reg_dest);
1242
1243 /* 1.0-1.3: Use destination register as coordinate source.
1244 1.4+: Use provided coordinate source register. */
1245 if (hex_version < WINED3DPS_VERSION(1,4))
1246 strcpy(reg_coord, reg_dest);
1247 else
1248 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1249
1250 /* 1.0-1.4: Use destination register number as texture code.
1251 2.0+: Use provided sampler number as texure code. */
1252 if (hex_version < WINED3DPS_VERSION(2,0))
1253 reg_sampler_code = reg_dest_code;
1254 else
1255 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1256
1257 /* projection flag:
1258 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1259 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1260 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1261 */
1262 if(hex_version < WINED3DPS_VERSION(1,4)) {
1263 DWORD flags = 0;
1264 if(reg_sampler_code < MAX_TEXTURES) {
1265 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1266 }
1267 if (flags & WINED3DTTFF_PROJECTED) {
1268 projected = TRUE;
1269 }
1270 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1271 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1272 if (src_mod == WINED3DSPSM_DZ) {
1273 projected = TRUE;
1274 } else if(src_mod == WINED3DSPSM_DW) {
1275 projected = TRUE;
1276 }
1277 } else {
1278 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1279 projected = TRUE;
1280 }
1281 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1282 bias = TRUE;
1283 }
1284 }
1285 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1286 }
1287
1288 static void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1289
1290 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1291 DWORD dst = arg->dst;
1292 SHADER_BUFFER* buffer = arg->buffer;
1293 DWORD hex_version = This->baseShader.hex_version;
1294
1295 char tmp[20];
1296 shader_arb_get_write_mask(arg, dst, tmp);
1297 if (hex_version != WINED3DPS_VERSION(1,4)) {
1298 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1299 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1300 } else {
1301 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1302 char reg_src[40];
1303
1304 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1305 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1306 }
1307 }
1308
1309 static void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1310
1311 SHADER_BUFFER* buffer = arg->buffer;
1312 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1313 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1314 DWORD flags;
1315
1316 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1317 char dst_str[8];
1318 char src_str[50];
1319
1320 sprintf(dst_str, "T%u", reg1);
1321 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1322 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1323 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1324 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1325 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1326 }
1327
1328 static void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1329
1330 SHADER_BUFFER* buffer = arg->buffer;
1331
1332 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1333 char dst_str[8];
1334 char src_str[50];
1335
1336 sprintf(dst_str, "T%u", reg1);
1337 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1338 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1339 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1340 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1341 }
1342
1343 static void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1344
1345 SHADER_BUFFER* buffer = arg->buffer;
1346 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1347 char dst_str[8];
1348 char src_str[50];
1349
1350 sprintf(dst_str, "T%u", reg1);
1351 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1352 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1353 }
1354
1355 static void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1356 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1357 BOOL has_bumpmat = FALSE;
1358 BOOL has_luminance = FALSE;
1359 int i;
1360
1361 DWORD dst = arg->dst;
1362 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1363 SHADER_BUFFER* buffer = arg->buffer;
1364
1365 char reg_coord[40];
1366 DWORD reg_dest_code;
1367
1368 /* All versions have a destination register */
1369 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1370 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1371 pshader_get_register_name(arg->shader, dst, reg_coord);
1372
1373 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1374 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1375 has_bumpmat = TRUE;
1376 break;
1377 }
1378 }
1379 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1380 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1381 has_luminance = TRUE;
1382 break;
1383 }
1384 }
1385
1386 if(has_bumpmat) {
1387 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1388
1389 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1390 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1391 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1392 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1393
1394 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1395 * so we can't let the GL handle this.
1396 */
1397 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1398 & WINED3DTTFF_PROJECTED) {
1399 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1400 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1401 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1402 } else {
1403 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1404 }
1405
1406 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1407
1408 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1409 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1410 src, reg_dest_code, reg_dest_code);
1411 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1412 }
1413
1414 } else {
1415 DWORD tf;
1416 if(reg_dest_code < MAX_TEXTURES) {
1417 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1418 } else {
1419 tf = 0;
1420 }
1421 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1422 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1423 }
1424 }
1425
1426 static void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1427
1428 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1429 SHADER_BUFFER* buffer = arg->buffer;
1430 char src0_name[50];
1431
1432 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1433 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1434 }
1435
1436 static void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1437
1438 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1439 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1440 DWORD flags;
1441 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1442 SHADER_BUFFER* buffer = arg->buffer;
1443 char dst_str[8];
1444 char src0_name[50];
1445
1446 sprintf(dst_str, "T%u", reg);
1447 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1448 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1449 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1450 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1451 }
1452
1453 static void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1454
1455 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1456 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1457 SHADER_BUFFER* buffer = arg->buffer;
1458 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1459 char src0_name[50];
1460
1461 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1462 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1463 current_state->texcoord_w[current_state->current_row++] = reg;
1464 }
1465
1466 static void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1467
1468 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1469 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1470 DWORD flags;
1471 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1472 SHADER_BUFFER* buffer = arg->buffer;
1473 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1474 char dst_str[8];
1475 char src0_name[50];
1476
1477 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1478 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1479
1480 /* Sample the texture using the calculated coordinates */
1481 sprintf(dst_str, "T%u", reg);
1482 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1483 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1484 current_state->current_row = 0;
1485 }
1486
1487 static void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1488
1489 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1490 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1491 DWORD flags;
1492 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1493 SHADER_BUFFER* buffer = arg->buffer;
1494 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1495 char dst_str[8];
1496 char src0_name[50];
1497
1498 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1499 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1500
1501 /* Construct the eye-ray vector from w coordinates */
1502 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1503 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1504 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1505
1506 /* Calculate reflection vector
1507 */
1508 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1509 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1510 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1511 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1512 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1513 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1514 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1515
1516 /* Sample the texture using the calculated coordinates */
1517 sprintf(dst_str, "T%u", reg);
1518 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1519 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1520 current_state->current_row = 0;
1521 }
1522
1523 static void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1524
1525 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1526 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1527 DWORD flags;
1528 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1529 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1530 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1531 SHADER_BUFFER* buffer = arg->buffer;
1532 char dst_str[8];
1533 char src0_name[50];
1534
1535 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1536 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1537
1538 /* Calculate reflection vector.
1539 *
1540 * dot(N, E)
1541 * TMP.xyz = 2 * --------- * N - E
1542 * dot(N, N)
1543 *
1544 * Which normalizes the normal vector
1545 */
1546 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1547 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1548 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1549 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1550 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1551 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1552
1553 /* Sample the texture using the calculated coordinates */
1554 sprintf(dst_str, "T%u", reg);
1555 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1556 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1557 current_state->current_row = 0;
1558 }
1559
1560 static void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1561 SHADER_BUFFER* buffer = arg->buffer;
1562 char dst_name[50];
1563
1564 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1565 * which is essentially an input, is the destination register because it is the first
1566 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1567 * here
1568 */
1569 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1570
1571 /* According to the msdn, the source register(must be r5) is unusable after
1572 * the texdepth instruction, so we're free to modify it
1573 */
1574 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1575
1576 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1577 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1578 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1579 */
1580 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1581 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1582 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1583 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1584 }
1585
1586 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1587 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1588 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1589 static void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1590 SHADER_BUFFER* buffer = arg->buffer;
1591 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1592 char src0[50];
1593 char dst_str[8];
1594
1595 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1596 shader_addline(buffer, "MOV TMP, 0.0;\n");
1597 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1598
1599 sprintf(dst_str, "T%u", sampler_idx);
1600 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1601 }
1602
1603 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1604 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1605 static void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1606 char src0[50];
1607 char dst_str[50];
1608 char dst_mask[6];
1609 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1610 SHADER_BUFFER* buffer = arg->buffer;
1611
1612 /* Handle output register */
1613 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1614 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1615
1616 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1617 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1618
1619 /* TODO: Handle output modifiers */
1620 }
1621
1622 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1623 * Perform the 3rd row of a 3x3 matrix multiply */
1624 static void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1625 SHADER_BUFFER* buffer = arg->buffer;
1626 char dst_str[50];
1627 char dst_mask[6];
1628 char src0[50];
1629 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1630
1631 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1632 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1633
1634 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1635 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1636 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1637
1638 /* TODO: Handle output modifiers */
1639 }
1640
1641 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1642 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1643 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1644 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1645 */
1646 static void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1647 SHADER_BUFFER* buffer = arg->buffer;
1648 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1649 char src0[50];
1650
1651 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1652 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1653
1654 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1655 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1656 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1657 */
1658 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1659 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1660 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1661 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1662 }
1663
1664 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1665 Vertex/Pixel shaders to ARB_vertex_program codes */
1666 static void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1667
1668 int i;
1669 int nComponents = 0;
1670 SHADER_OPCODE_ARG tmpArg;
1671
1672 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1673
1674 /* Set constants for the temporary argument */
1675 tmpArg.shader = arg->shader;
1676 tmpArg.buffer = arg->buffer;
1677 tmpArg.src[0] = arg->src[0];
1678 tmpArg.src_addr[0] = arg->src_addr[0];
1679 tmpArg.src_addr[1] = arg->src_addr[1];
1680 tmpArg.reg_maps = arg->reg_maps;
1681
1682 switch(arg->opcode->opcode) {
1683 case WINED3DSIO_M4x4:
1684 nComponents = 4;
1685 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1686 break;
1687 case WINED3DSIO_M4x3:
1688 nComponents = 3;
1689 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1690 break;
1691 case WINED3DSIO_M3x4:
1692 nComponents = 4;
1693 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1694 break;
1695 case WINED3DSIO_M3x3:
1696 nComponents = 3;
1697 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1698 break;
1699 case WINED3DSIO_M3x2:
1700 nComponents = 2;
1701 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1702 break;
1703 default:
1704 break;
1705 }
1706
1707 for (i = 0; i < nComponents; i++) {
1708 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1709 tmpArg.src[1] = arg->src[1]+i;
1710 shader_hw_map2gl(&tmpArg);
1711 }
1712 }
1713
1714 static void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1715 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1716 SHADER_BUFFER* buffer = arg->buffer;
1717 DWORD dst = arg->dst;
1718 DWORD src = arg->src[0];
1719 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1720
1721 char tmpLine[256];
1722
1723 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1724 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1725 strcat(tmpLine, ",");
1726 vshader_program_add_param(arg, src, TRUE, tmpLine);
1727 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1728 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1729 * .w is used
1730 */
1731 strcat(tmpLine, ".w");
1732 }
1733
1734 shader_addline(buffer, "%s;\n", tmpLine);
1735 }
1736
1737 static void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1738 SHADER_BUFFER* buffer = arg->buffer;
1739 char dst_name[50];
1740 char src_name[50];
1741 char dst_wmask[20];
1742 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1743 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1744
1745 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1746 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1747
1748 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1749 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1750 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1751 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1752 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1753 src_name);
1754
1755 if (shift != 0)
1756 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1757 }
1758
1759 static void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1760 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1761 * must contain fixed constants. So we need a separate function to filter those constants and
1762 * can't use map2gl
1763 */
1764 SHADER_BUFFER* buffer = arg->buffer;
1765 char dst_name[50];
1766 char src_name[50];
1767 char dst_wmask[20];
1768 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1769 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1770
1771 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1772 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1773
1774 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1775 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1776 src_name);
1777
1778 if (shift != 0)
1779 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1780
1781 }
1782
1783 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1784 GLuint program_id = 0;
1785 const char *blt_vprogram =
1786 "!!ARBvp1.0\n"
1787 "PARAM c[1] = { { 1, 0.5 } };\n"
1788 "MOV result.position, vertex.position;\n"
1789 "MOV result.color, c[0].x;\n"
1790 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1791 "END\n";
1792
1793 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1794 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1795 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1796
1797 if (glGetError() == GL_INVALID_OPERATION) {
1798 GLint pos;
1799 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1800 FIXME("Vertex program error at position %d: %s\n", pos,
1801 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1802 }
1803
1804 return program_id;
1805 }
1806
1807 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1808 {
1809 GLuint program_id = 0;
1810 const char *blt_fprograms[tex_type_count] =
1811 {
1812 /* tex_1d */
1813 NULL,
1814 /* tex_2d */
1815 "!!ARBfp1.0\n"
1816 "TEMP R0;\n"
1817 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1818 "MOV result.depth.z, R0.x;\n"
1819 "END\n",
1820 /* tex_3d */
1821 NULL,
1822 /* tex_cube */
1823 "!!ARBfp1.0\n"
1824 "TEMP R0;\n"
1825 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1826 "MOV result.depth.z, R0.x;\n"
1827 "END\n",
1828 /* tex_rect */
1829 "!!ARBfp1.0\n"
1830 "TEMP R0;\n"
1831 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1832 "MOV result.depth.z, R0.x;\n"
1833 "END\n",
1834 };
1835
1836 if (!blt_fprograms[tex_type])
1837 {
1838 FIXME("tex_type %#x not supported\n", tex_type);
1839 tex_type = tex_2d;
1840 }
1841
1842 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1843 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1844 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1845
1846 if (glGetError() == GL_INVALID_OPERATION) {
1847 GLint pos;
1848 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1849 FIXME("Fragment program error at position %d: %s\n", pos,
1850 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1851 }
1852
1853 return program_id;
1854 }
1855
1856 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1857 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1858 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1859 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1860
1861 if (useVS) {
1862 TRACE("Using vertex shader\n");
1863
1864 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId;
1865
1866 /* Bind the vertex program */
1867 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1868 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1869
1870 /* Enable OpenGL vertex programs */
1871 glEnable(GL_VERTEX_PROGRAM_ARB);
1872 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1873 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1874 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1875 priv->current_vprogram_id = 0;
1876 glDisable(GL_VERTEX_PROGRAM_ARB);
1877 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1878 }
1879
1880 if (usePS) {
1881 TRACE("Using pixel shader\n");
1882
1883 priv->current_fprogram_id = ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId;
1884
1885 /* Bind the fragment program */
1886 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1887 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1888
1889 if(!priv->use_arbfp_fixed_func) {
1890 /* Enable OpenGL fragment programs */
1891 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1892 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1893 }
1894 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1895 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1896 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1897 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1898 * replacement shader
1899 */
1900 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1901 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1902 priv->current_fprogram_id = 0;
1903 }
1904 }
1905
1906 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1907 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1908 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1909 GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1910 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1911
1912 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1913 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1914 glEnable(GL_VERTEX_PROGRAM_ARB);
1915
1916 if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1917 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1918 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1919 }
1920
1921 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1922 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1923 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1924 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1925
1926 if (priv->current_vprogram_id) {
1927 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1928 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1929
1930 glEnable(GL_VERTEX_PROGRAM_ARB);
1931 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1932
1933 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1934 } else {
1935 glDisable(GL_VERTEX_PROGRAM_ARB);
1936 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1937 }
1938
1939 if (priv->current_fprogram_id) {
1940 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1941 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1942
1943 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1944 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1945
1946 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1947 } else {
1948 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1949 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1950 }
1951 }
1952
1953 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1954 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1955 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1956 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1957 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1958 }
1959
1960 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1961 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1962 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1963
1964 ENTER_GL();
1965 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1966 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1967 LEAVE_GL();
1968 This->baseShader.prgId = 0;
1969 This->baseShader.is_compiled = FALSE;
1970 }
1971
1972 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1973 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1974 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1975 return WINED3D_OK;
1976 }
1977
1978 static void shader_arb_free(IWineD3DDevice *iface) {
1979 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1980 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1981 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1982 int i;
1983
1984 if(priv->depth_blt_vprogram_id) {
1985 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1986 }
1987 for (i = 0; i < tex_type_count; ++i) {
1988 if (priv->depth_blt_fprogram_id[i]) {
1989 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
1990 }
1991 }
1992
1993 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1994 }
1995
1996 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1997 return TRUE;
1998 }
1999
2000 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
2001 const char *tmp2, const char *tmp3, const char *tmp4) {
2002 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
2003
2004 /* Calculate the > 0.0031308 case */
2005 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
2006 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
2007 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
2008 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
2009 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
2010 /* Calculate the < case */
2011 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
2012 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
2013 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
2014 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
2015 /* Store the components > 0.0031308 in the destination */
2016 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
2017 /* Add the components that are < 0.0031308 */
2018 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
2019 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
2020 }
2021
2022 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
2023 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
2024 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2025 CONST DWORD *function = This->baseShader.function;
2026 const char *fragcolor;
2027 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
2028 local_constant* lconst;
2029
2030 /* Create the hw ARB shader */
2031 shader_addline(buffer, "!!ARBfp1.0\n");
2032
2033 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
2034 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
2035 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
2036 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
2037 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
2038 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2039 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2040 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2041
2042 /* Base Declarations */
2043 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2044
2045 /* We need two variables for fog blending */
2046 shader_addline(buffer, "TEMP TMP_FOG;\n");
2047 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
2048 shader_addline(buffer, "TEMP TMP_COLOR;\n");
2049 }
2050
2051 /* Base Shader Body */
2052 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2053
2054 /* calculate fog and blend it
2055 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
2056 * -1/(e-s) and e/(e-s) respectively.
2057 */
2058 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
2059
2060 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
2061 fragcolor = "R0";
2062 } else {
2063 fragcolor = "TMP_COLOR";
2064 }
2065 if(This->srgb_enabled) {
2066 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
2067 }
2068 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
2069 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
2070 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
2071 }
2072
2073 shader_addline(buffer, "END\n");
2074
2075 /* TODO: change to resource.glObjectHandle or something like that */
2076 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2077
2078 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
2079 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
2080
2081 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
2082 /* Create the program and check for errors */
2083 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2084 buffer->bsize, buffer->buffer));
2085
2086 if (glGetError() == GL_INVALID_OPERATION) {
2087 GLint errPos;
2088 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2089 FIXME("HW PixelShader Error at position %d: %s\n",
2090 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2091 This->baseShader.prgId = -1;
2092 }
2093
2094 /* Load immediate constants */
2095 if(!This->baseShader.load_local_constsF) {
2096 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2097 float *value = (float *) lconst->value;
2098 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2099 checkGLcall("glProgramLocalParameter4fvARB");
2100 }
2101 }
2102 }
2103
2104 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
2105 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2106 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2107 CONST DWORD *function = This->baseShader.function;
2108 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2109 WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2110 local_constant* lconst;
2111
2112 /* Create the hw ARB shader */
2113 shader_addline(buffer, "!!ARBvp1.0\n");
2114 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2115
2116 /* Mesa supports only 95 constants */
2117 if (GL_VEND(MESA) || GL_VEND(WINE))
2118 This->baseShader.limits.constant_float =
2119 min(95, This->baseShader.limits.constant_float);
2120
2121 shader_addline(buffer, "TEMP TMP;\n");
2122
2123 /* Base Declarations */
2124 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2125
2126 /* We need a constant to fixup the final position */
2127 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2128
2129 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2130 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2131 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2132 * a replacement shader depend on the texcoord.w being set properly.
2133 *
2134 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2135 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2136 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2137 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2138 * this can eat a number of instructions, so skip it unless this cap is set as well
2139 */
2140 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2141 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2142
2143 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2144 int i;
2145 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2146 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2147 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2148 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2149 }
2150 }
2151 }
2152 }
2153
2154 /* Base Shader Body */
2155 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2156
2157 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2158 if (!reg_maps->fog)
2159 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2160
2161 /* Write the final position.
2162 *
2163 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2164 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2165 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2166 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2167 */
2168 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2169 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2170 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2171
2172 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2173 * and the glsl equivalent
2174 */
2175 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2176
2177 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2178
2179 shader_addline(buffer, "END\n");
2180
2181 /* TODO: change to resource.glObjectHandle or something like that */
2182 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2183
2184 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2185 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2186
2187 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2188 /* Create the program and check for errors */
2189 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2190 buffer->bsize, buffer->buffer));
2191
2192 if (glGetError() == GL_INVALID_OPERATION) {
2193 GLint errPos;
2194 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2195 FIXME("HW VertexShader Error at position %d: %s\n",
2196 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2197 This->baseShader.prgId = -1;
2198 }
2199
2200 /* Load immediate constants */
2201 if(!This->baseShader.load_local_constsF) {
2202 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2203 float *value = (float *) lconst->value;
2204 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2205 }
2206 }
2207 }
2208
2209 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2210 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2211 * then overwrite the shader specific ones
2212 */
2213 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2214
2215 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2216 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2217 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2218 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2219 }
2220
2221 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2222 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2223 pCaps->PixelShader1xMaxValue = 8.0;
2224 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2225 }
2226 }
2227
2228 static BOOL shader_arb_conv_supported(WINED3DFORMAT fmt) {
2229 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt));
2230 switch(fmt) {
2231 case WINED3DFMT_V8U8:
2232 case WINED3DFMT_V16U16:
2233 case WINED3DFMT_X8L8V8U8:
2234 case WINED3DFMT_L6V5U5:
2235 case WINED3DFMT_Q8W8V8U8:
2236 case WINED3DFMT_ATI2N:
2237 TRACE("[OK]\n");
2238 return TRUE;
2239 default:
2240 TRACE("[FAILED\n");
2241 return FALSE;
2242 }
2243 }
2244
2245 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2246 {
2247 /* WINED3DSIH_ABS */ shader_hw_map2gl,
2248 /* WINED3DSIH_ADD */ shader_hw_map2gl,
2249 /* WINED3DSIH_BEM */ pshader_hw_bem,
2250 /* WINED3DSIH_BREAK */ NULL,
2251 /* WINED3DSIH_BREAKC */ NULL,
2252 /* WINED3DSIH_BREAKP */ NULL,
2253 /* WINED3DSIH_CALL */ NULL,
2254 /* WINED3DSIH_CALLNZ */ NULL,
2255 /* WINED3DSIH_CMP */ pshader_hw_cmp,
2256 /* WINED3DSIH_CND */ pshader_hw_cnd,
2257 /* WINED3DSIH_CRS */ shader_hw_map2gl,
2258 /* WINED3DSIH_DCL */ NULL,
2259 /* WINED3DSIH_DEF */ NULL,
2260 /* WINED3DSIH_DEFB */ NULL,
2261 /* WINED3DSIH_DEFI */ NULL,
2262 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add,
2263 /* WINED3DSIH_DP3 */ shader_hw_map2gl,
2264 /* WINED3DSIH_DP4 */ shader_hw_map2gl,
2265 /* WINED3DSIH_DST */ shader_hw_map2gl,
2266 /* WINED3DSIH_DSX */ NULL,
2267 /* WINED3DSIH_DSY */ NULL,
2268 /* WINED3DSIH_ELSE */ NULL,
2269 /* WINED3DSIH_ENDIF */ NULL,
2270 /* WINED3DSIH_ENDLOOP */ NULL,
2271 /* WINED3DSIH_ENDREP */ NULL,
2272 /* WINED3DSIH_EXP */ shader_hw_map2gl,
2273 /* WINED3DSIH_EXPP */ shader_hw_map2gl,
2274 /* WINED3DSIH_FRC */ shader_hw_map2gl,
2275 /* WINED3DSIH_IF */ NULL,
2276 /* WINED3DSIH_IFC */ NULL,
2277 /* WINED3DSIH_LABEL */ NULL,
2278 /* WINED3DSIH_LIT */ shader_hw_map2gl,
2279 /* WINED3DSIH_LOG */ shader_hw_map2gl,
2280 /* WINED3DSIH_LOGP */ shader_hw_map2gl,
2281 /* WINED3DSIH_LOOP */ NULL,
2282 /* WINED3DSIH_LRP */ shader_hw_map2gl,
2283 /* WINED3DSIH_M3x2 */ shader_hw_mnxn,
2284 /* WINED3DSIH_M3x3 */ shader_hw_mnxn,
2285 /* WINED3DSIH_M3x4 */ shader_hw_mnxn,
2286 /* WINED3DSIH_M4x3 */ shader_hw_mnxn,
2287 /* WINED3DSIH_M4x4 */ shader_hw_mnxn,
2288 /* WINED3DSIH_MAD */ shader_hw_map2gl,
2289 /* WINED3DSIH_MAX */ shader_hw_map2gl,
2290 /* WINED3DSIH_MIN */ shader_hw_map2gl,
2291 /* WINED3DSIH_MOV */ shader_hw_map2gl,
2292 /* WINED3DSIH_MOVA */ shader_hw_map2gl,
2293 /* WINED3DSIH_MUL */ shader_hw_map2gl,
2294 /* WINED3DSIH_NOP */ shader_hw_map2gl,
2295 /* WINED3DSIH_NRM */ shader_hw_nrm,
2296 /* WINED3DSIH_PHASE */ NULL,
2297 /* WINED3DSIH_POW */ shader_hw_map2gl,
2298 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp,
2299 /* WINED3DSIH_REP */ NULL,
2300 /* WINED3DSIH_RET */ NULL,
2301 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp,
2302 /* WINED3DSIH_SETP */ NULL,
2303 /* WINED3DSIH_SGE */ shader_hw_map2gl,
2304 /* WINED3DSIH_SGN */ NULL,
2305 /* WINED3DSIH_SINCOS */ shader_hw_sincos,
2306 /* WINED3DSIH_SLT */ shader_hw_map2gl,
2307 /* WINED3DSIH_SUB */ shader_hw_map2gl,
2308 /* WINED3DSIH_TEX */ pshader_hw_tex,
2309 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem,
2310 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem,
2311 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord,
2312 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth,
2313 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3,
2314 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex,
2315 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill,
2316 /* WINED3DSIH_TEXLDD */ NULL,
2317 /* WINED3DSIH_TEXLDL */ NULL,
2318 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth,
2319 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad,
2320 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex,
2321 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3,
2322 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
2323 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad,
2324 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec,
2325 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex,
2326 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec,
2327 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar,
2328 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb,
2329 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb,
2330 };
2331
2332 const shader_backend_t arb_program_shader_backend = {
2333 shader_arb_instruction_handler_table,
2334 shader_arb_select,
2335 shader_arb_select_depth_blt,
2336 shader_arb_deselect_depth_blt,
2337 shader_arb_load_constants,
2338 shader_arb_cleanup,
2339 shader_arb_color_correction,
2340 shader_arb_destroy,
2341 shader_arb_alloc,
2342 shader_arb_free,
2343 shader_arb_dirty_const,
2344 shader_arb_generate_pshader,
2345 shader_arb_generate_vshader,
2346 shader_arb_get_caps,
2347 shader_arb_conv_supported,
2348 };
2349
2350 /* ARB_fragment_program fixed function pipeline replacement definitions */
2351 #define ARB_FFP_CONST_TFACTOR 0
2352 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2353 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2354 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2355 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2356
2357 struct arbfp_ffp_desc
2358 {
2359 struct ffp_desc parent;
2360 GLuint shader;
2361 unsigned int num_textures_used;
2362 };
2363
2364 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2365 if(enable) {
2366 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2367 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2368 } else {
2369 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2370 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2371 }
2372 }
2373
2374 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2375 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2376 struct shader_arb_priv *priv;
2377 /* Share private data between the shader backend and the pipeline replacement, if both
2378 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2379 * if no pixel shader is bound or not
2380 */
2381 if(This->shader_backend == &arb_program_shader_backend) {
2382 This->fragment_priv = This->shader_priv;
2383 } else {
2384 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2385 if(!This->fragment_priv) return E_OUTOFMEMORY;
2386 }
2387 priv = (struct shader_arb_priv *) This->fragment_priv;
2388 priv->fragment_shaders = hash_table_create(ffp_program_key_hash, ffp_program_key_compare);
2389 priv->use_arbfp_fixed_func = TRUE;
2390 return WINED3D_OK;
2391 }
2392
2393 static void arbfp_free_ffpshader(void *value, void *gli) {
2394 WineD3D_GL_Info *gl_info = gli;
2395 struct arbfp_ffp_desc *entry_arb = value;
2396
2397 ENTER_GL();
2398 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2399 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2400 HeapFree(GetProcessHeap(), 0, entry_arb);
2401 LEAVE_GL();
2402 }
2403
2404 static void arbfp_free(IWineD3DDevice *iface) {
2405 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2406 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2407
2408 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2409 priv->use_arbfp_fixed_func = FALSE;
2410
2411 if(This->shader_backend != &arb_program_shader_backend) {
2412 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2413 }
2414 }
2415
2416 static void arbfp_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct fragment_caps *caps) {
2417 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2418 WINED3DTEXOPCAPS_SELECTARG1 |
2419 WINED3DTEXOPCAPS_SELECTARG2 |
2420 WINED3DTEXOPCAPS_MODULATE4X |
2421 WINED3DTEXOPCAPS_MODULATE2X |
2422 WINED3DTEXOPCAPS_MODULATE |
2423 WINED3DTEXOPCAPS_ADDSIGNED2X |
2424 WINED3DTEXOPCAPS_ADDSIGNED |
2425 WINED3DTEXOPCAPS_ADD |
2426 WINED3DTEXOPCAPS_SUBTRACT |
2427 WINED3DTEXOPCAPS_ADDSMOOTH |
2428 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2429 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2430 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2431 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2432 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2433 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2434 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2435 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2436 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2437 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2438 WINED3DTEXOPCAPS_MULTIPLYADD |
2439 WINED3DTEXOPCAPS_LERP |
2440 WINED3DTEXOPCAPS_BUMPENVMAP |
2441 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2442
2443 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2444
2445 caps->MaxTextureBlendStages = 8;
2446 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2447
2448 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2449 }
2450 #undef GLINFO_LOCATION
2451
2452 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2453 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2454 float col[4];
2455 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2456
2457 /* Do not overwrite pixel shader constants if a pshader is in use */
2458 if(use_ps(device)) return;
2459
2460 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2461 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2462 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2463
2464 if(device->shader_backend == &arb_program_shader_backend) {
2465 device = stateblock->wineD3DDevice;
2466 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2467 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2468 }
2469 }
2470
2471 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2472 float col[4];
2473 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2474
2475 /* Do not overwrite pixel shader constants if a pshader is in use */
2476 if(use_ps(device)) return;
2477
2478 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2479 /* The specular color has no alpha */
2480 col[0] = 1.0; col[1] = 1.0;
2481 col[2] = 1.0; col[3] = 0.0;
2482 } else {
2483 col[0] = 0.0; col[1] = 0.0;
2484 col[2] = 0.0; col[3] = 0.0;
2485 }
2486 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2487 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2488
2489 if(device->shader_backend == &arb_program_shader_backend) {
2490 device = stateblock->wineD3DDevice;
2491 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2492 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2493 }
2494 }
2495
2496 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2497 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2498 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2499 float mat[2][2];
2500
2501 if(use_ps(device)) {
2502 if(stage != 0 &&
2503 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2504 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2505 * anyway
2506 */
2507 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2508 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2509 }
2510 }
2511 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2512 return;
2513 }
2514
2515 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2516 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2517 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2518 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2519
2520 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2521 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2522
2523 if(device->shader_backend == &arb_program_shader_backend) {
2524 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2525 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2526 }
2527 }
2528
2529 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2530 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2531 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2532 float param[4];
2533
2534 if(use_ps(device)) {
2535 if(stage != 0 &&
2536 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2537 /* The pixel shader has to know the luminance offset. Do a constants update if it
2538 * isn't scheduled anyway
2539 */
2540 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2541 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2542 }
2543 }
2544 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2545 return;
2546 }
2547
2548 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2549 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2550 param[2] = 0.0;
2551 param[3] = 0.0;
2552
2553 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2554 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2555
2556 if(device->shader_backend == &arb_program_shader_backend) {
2557 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2558 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2559 }
2560 }
2561
2562 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2563 const char *ret;
2564
2565 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2566
2567 switch(arg & WINED3DTA_SELECTMASK) {
2568 case WINED3DTA_DIFFUSE:
2569 ret = "fragment.color.primary"; break;
2570
2571 case WINED3DTA_CURRENT:
2572 if(stage == 0) ret = "fragment.color.primary";
2573 else ret = "ret";
2574 break;
2575
2576 case WINED3DTA_TEXTURE:
2577 switch(stage) {
2578 case 0: ret = "tex0"; break;
2579 case 1: ret = "tex1"; break;
2580 case 2: ret = "tex2"; break;
2581 case 3: ret = "tex3"; break;
2582 case 4: ret = "tex4"; break;
2583 case 5: ret = "tex5"; break;
2584 case 6: ret = "tex6"; break;
2585 case 7: ret = "tex7"; break;
2586 default: ret = "unknown texture";
2587 }
2588 break;
2589
2590 case WINED3DTA_TFACTOR:
2591 ret = "tfactor"; break;
2592
2593 case WINED3DTA_SPECULAR:
2594 ret = "fragment.color.secondary"; break;
2595
2596 case WINED3DTA_TEMP:
2597 ret = "tempreg"; break;
2598
2599 case WINED3DTA_CONSTANT:
2600 FIXME("Implement perstage constants\n");
2601 switch(stage) {
2602 case 0: ret = "const0"; break;
2603 case 1: ret = "const1"; break;
2604 case 2: ret = "const2"; break;
2605 case 3: ret = "const3"; break;
2606 case 4: ret = "const4"; break;
2607 case 5: ret = "const5"; break;
2608 case 6: ret = "const6"; break;
2609 case 7: ret = "const7"; break;
2610 default: ret = "unknown constant";
2611 }
2612 break;
2613
2614 default:
2615 return "unknown";
2616 }
2617
2618 if(arg & WINED3DTA_COMPLEMENT) {
2619 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2620 if(argnum == 0) ret = "arg0";
2621 if(argnum == 1) ret = "arg1";
2622 if(argnum == 2) ret = "arg2";
2623 }
2624 if(arg & WINED3DTA_ALPHAREPLICATE) {
2625 shader_addline(buffer, "MOV arg%u, %s.a;\n", argnum, ret);
2626 if(argnum == 0) ret = "arg0";
2627 if(argnum == 1) ret = "arg1";
2628 if(argnum == 2) ret = "arg2";
2629 }
2630 return ret;
2631 }
2632
2633 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2634 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2635 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2636 unsigned int mul = 1;
2637 BOOL mul_final_dest = FALSE;
2638
2639 if(color && alpha) dstmask = "";
2640 else if(color) dstmask = ".rgb";
2641 else dstmask = ".a";
2642
2643 if(dst == tempreg) dstreg = "tempreg";
2644 else dstreg = "ret";
2645
2646 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2647 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2648 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2649
2650 switch(op) {
2651 case WINED3DTOP_DISABLE:
2652 if(stage == 1) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2653 break;
2654
2655 case WINED3DTOP_SELECTARG2:
2656 arg1 = arg2;
2657 case WINED3DTOP_SELECTARG1:
2658 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2659 break;
2660
2661 case WINED3DTOP_MODULATE4X:
2662 mul = 2;
2663 case WINED3DTOP_MODULATE2X:
2664 mul *= 2;
2665 if(strcmp(dstreg, "result.color") == 0) {
2666 dstreg = "ret";
2667 mul_final_dest = TRUE;
2668 }
2669 case WINED3DTOP_MODULATE:
2670 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2671 break;
2672
2673 case WINED3DTOP_ADDSIGNED2X:
2674 mul = 2;
2675 if(strcmp(dstreg, "result.color") == 0) {
2676 dstreg = "ret";
2677 mul_final_dest = TRUE;
2678 }
2679 case WINED3DTOP_ADDSIGNED:
2680 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2681 arg2 = "arg2";
2682 case WINED3DTOP_ADD:
2683 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2684 break;
2685
2686 case WINED3DTOP_SUBTRACT:
2687 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2688 break;
2689
2690 case WINED3DTOP_ADDSMOOTH:
2691 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2692 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2693 break;
2694
2695 case WINED3DTOP_BLENDCURRENTALPHA:
2696 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2697 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2698 break;
2699 case WINED3DTOP_BLENDFACTORALPHA:
2700 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2701 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2702 break;
2703 case WINED3DTOP_BLENDTEXTUREALPHA:
2704 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2705 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2706 break;
2707 case WINED3DTOP_BLENDDIFFUSEALPHA:
2708 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2709 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2710 break;
2711
2712 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2713 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2714 shader_addline(buffer, "SUB arg0.a, const.x, %s.a;\n", arg0);
2715 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg, dstmask, arg2, arg1);
2716 break;
2717
2718 /* D3DTOP_PREMODULATE ???? */
2719
2720 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2721 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2722 shader_addline(buffer, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2723 break;
2724 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2725 shader_addline(buffer, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2726 break;
2727 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2728 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2729 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg, dstmask, arg2, arg1);
2730 break;
2731 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2732 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg, dstmask, arg1, arg2, arg1);
2733 break;
2734
2735 case WINED3DTOP_DOTPRODUCT3:
2736 mul = 4;
2737 if(strcmp(dstreg, "result.color") == 0) {
2738 dstreg = "ret";
2739 mul_final_dest = TRUE;
2740 }
2741 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2742 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2743 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2744 break;
2745
2746 case WINED3DTOP_MULTIPLYADD:
2747 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2748 break;
2749
2750 case WINED3DTOP_LERP:
2751 /* The msdn is not quite right here */
2752 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2753 break;
2754
2755 case WINED3DTOP_BUMPENVMAP:
2756 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2757 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2758 break;
2759
2760 default:
2761 FIXME("Unhandled texture op %08x\n", op);
2762 }
2763
2764 if(mul == 2) {
2765 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2766 } else if(mul == 4) {
2767 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2768 }
2769 }
2770
2771 /* The stateblock is passed for GLINFO_LOCATION */
2772 static GLuint gen_arbfp_ffp_shader(struct ffp_settings *settings, IWineD3DStateBlockImpl *stateblock) {
2773 unsigned int stage;
2774 SHADER_BUFFER buffer;
2775 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2776 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2777 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2778 const char *textype;
2779 const char *instr, *sat;
2780 char colorcor_dst[8];
2781 GLuint ret;
2782 DWORD arg0, arg1, arg2;
2783 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2784 BOOL op_equal;
2785 const char *final_combiner_src = "ret";
2786
2787 /* Find out which textures are read */
2788 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2789 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2790 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2791 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2792 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2793 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2794 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2795 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2796
2797 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2798 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2799 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2800 bump_used[stage] = TRUE;
2801 tex_read[stage] = TRUE;
2802 }
2803 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2804 bump_used[stage] = TRUE;
2805 tex_read[stage] = TRUE;
2806 luminance_used[stage] = TRUE;
2807 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2808 tfactor_used = TRUE;
2809 }
2810
2811 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2812 tfactor_used = TRUE;
2813 }
2814
2815 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2816 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2817 tempreg_used = TRUE;
2818 }
2819
2820 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2821 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2822 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2823 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2824 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2825 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2826 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2827
2828 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2829 tempreg_used = TRUE;
2830 }
2831 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2832 tfactor_used = TRUE;
2833 }
2834 }
2835
2836 /* Shader header */
2837 buffer.bsize = 0;
2838 buffer.lineNo = 0;
2839 buffer.newline = TRUE;
2840 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
2841
2842 shader_addline(&buffer, "!!ARBfp1.0\n");
2843
2844 switch(settings->fog) {
2845 case FOG_OFF: break;
2846 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2847 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2848 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2849 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2850 }
2851
2852 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2853 shader_addline(&buffer, "TEMP TMP;\n");
2854 shader_addline(&buffer, "TEMP ret;\n");
2855 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2856 shader_addline(&buffer, "TEMP arg0;\n");
2857 shader_addline(&buffer, "TEMP arg1;\n");
2858 shader_addline(&buffer, "TEMP arg2;\n");
2859 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2860 if(!tex_read[stage]) continue;
2861 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2862 if(!bump_used[stage]) continue;
2863 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2864 if(!luminance_used[stage]) continue;
2865 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2866 }
2867 if(tfactor_used) {
2868 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2869 }
2870 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2871
2872 if(settings->sRGB_write) {
2873 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2874 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2875 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2876 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2877 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2878 srgb_pow, srgb_pow, srgb_pow);
2879 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2880 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2881 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2882 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2883 }
2884
2885 /* Generate texture sampling instructions) */
2886 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2887 if(!tex_read[stage]) continue;
2888
2889 switch(settings->op[stage].tex_type) {
2890 case tex_1d: textype = "1D"; break;
2891 case tex_2d: textype = "2D"; break;
2892 case tex_3d: textype = "3D"; break;
2893 case tex_cube: textype = "CUBE"; break;
2894 case tex_rect: textype = "RECT"; break;
2895 default: textype = "unexpected_textype"; break;
2896 }
2897
2898 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2899 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2900 sat = "";
2901 } else {
2902 sat = "_SAT";
2903 }
2904
2905 if(settings->op[stage].projected == proj_none) {
2906 instr = "TEX";
2907 } else if(settings->op[stage].projected == proj_count4 ||
2908 settings->op[stage].projected == proj_count3) {
2909 instr = "TXP";
2910 } else {
2911 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2912 instr = "TXP";
2913 }
2914
2915 if(stage > 0 &&
2916 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2917 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2918 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2919 shader_addline(&buffer, "DP3 ret.r, arg1, tex%u;\n", stage - 1);
2920 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2921 shader_addline(&buffer, "DP3 ret.g, arg1, tex%u;\n", stage - 1);
2922
2923 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2924 * so multiply the displacement with the dividing parameter before passing it to TXP
2925 */
2926 if (settings->op[stage].projected != proj_none) {
2927 if(settings->op[stage].projected == proj_count4) {
2928 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].a;\n", stage);
2929 shader_addline(&buffer, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage, stage);
2930 } else {
2931 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].b;\n", stage);
2932 shader_addline(&buffer, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage, stage);
2933 }
2934 } else {
2935 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2936 }
2937
2938 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2939 instr, sat, stage, stage, textype);
2940 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2941 shader_addline(&buffer, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2942 stage - 1, stage - 1, stage - 1);
2943 shader_addline(&buffer, "MUL tex%u, tex%u, ret.r;\n", stage, stage);
2944 }
2945 } else if(settings->op[stage].projected == proj_count3) {
2946 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2947 shader_addline(&buffer, "MOV ret.a, ret.b;\n");
2948 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2949 instr, sat, stage, stage, textype);
2950 } else {
2951 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2952 instr, sat, stage, stage, stage, textype);
2953 }
2954
2955 sprintf(colorcor_dst, "tex%u", stage);
2956 gen_color_correction(&buffer, colorcor_dst, ".rgba", "const.x", "const.y",
2957 settings->op[stage].color_correction, &GLINFO_LOCATION);
2958 }
2959
2960 /* Generate the main shader */
2961 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2962 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2963 if(stage == 0) {
2964 final_combiner_src = "fragment.color.primary";
2965 }
2966 break;
2967 }
2968
2969 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2970 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2971 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2972 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2973 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2974 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2975 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2976 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2977 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2978 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2979 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2980 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2981 } else {
2982 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
2983 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2984 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2985 settings->op[stage].carg2 == settings->op[stage].aarg2;
2986 }
2987
2988 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2989 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2990 settings->op[stage].cop, settings->op[stage].carg0,
2991 settings->op[stage].carg1, settings->op[stage].carg2);
2992 if(stage == 0) {
2993 shader_addline(&buffer, "MOV ret.a, fragment.color.primary.a;\n");
2994 }
2995 } else if(op_equal) {
2996 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2997 settings->op[stage].cop, settings->op[stage].carg0,
2998 settings->op[stage].carg1, settings->op[stage].carg2);
2999 } else {
3000 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
3001 settings->op[stage].cop, settings->op[stage].carg0,
3002 settings->op[stage].carg1, settings->op[stage].carg2);
3003 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
3004 settings->op[stage].aop, settings->op[stage].aarg0,
3005 settings->op[stage].aarg1, settings->op[stage].aarg2);
3006 }
3007 }
3008
3009 if(settings->sRGB_write) {
3010 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
3011 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
3012 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
3013 } else {
3014 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
3015 }
3016
3017 /* Footer */
3018 shader_addline(&buffer, "END\n");
3019
3020 /* Generate the shader */
3021 GL_EXTCALL(glGenProgramsARB(1, &ret));
3022 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
3023 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3024
3025 if (glGetError() == GL_INVALID_OPERATION) {
3026 GLint pos;
3027 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3028 FIXME("Fragment program error at position %d: %s\n", pos,
3029 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3030 }
3031 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3032 return ret;
3033 }
3034
3035 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3036 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
3037 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
3038 BOOL use_pshader = use_ps(device);
3039 BOOL use_vshader = use_vs(device);
3040 struct ffp_settings settings;
3041 struct arbfp_ffp_desc *desc;
3042 unsigned int i;
3043
3044 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
3045 if(use_pshader) {
3046 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
3047 } else if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3048 /* Reload fixed function constants since they collide with the pixel shader constants */
3049 for(i = 0; i < MAX_TEXTURES; i++) {
3050 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3051 }
3052 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3053 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3054 }
3055 return;
3056 }
3057
3058 if(use_pshader) {
3059 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
3060 } else {
3061 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
3062 gen_ffp_op(stateblock, &settings, FALSE);
3063 desc = (struct arbfp_ffp_desc *) find_ffp_shader(priv->fragment_shaders, &settings);
3064 if(!desc) {
3065 desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*desc));
3066 if(!desc) {
3067 ERR("Out of memory\n");
3068 return;
3069 }
3070 desc->num_textures_used = 0;
3071 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
3072 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
3073 desc->num_textures_used = i;
3074 }
3075
3076 memcpy(&desc->parent.settings, &settings, sizeof(settings));
3077 desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
3078 add_ffp_shader(priv->fragment_shaders, &desc->parent);
3079 TRACE("Allocated fixed function replacement shader descriptor %p\n", desc);
3080 }
3081
3082 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3083 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3084 * deactivate it.
3085 */
3086 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
3087 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3088 priv->current_fprogram_id = desc->shader;
3089
3090 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3091 /* Reload fixed function constants since they collide with the pixel shader constants */
3092 for(i = 0; i < MAX_TEXTURES; i++) {
3093 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3094 }
3095 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3096 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3097 }
3098 }
3099
3100 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3101 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3102 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3103 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3104 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3105 *
3106 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3107 * to be compiled before activating them(needs some cleanups in the shader backend interface)
3108 */
3109 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
3110 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
3111
3112 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
3113 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
3114 }
3115 }
3116 if(use_pshader) {
3117 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
3118 }
3119 }
3120
3121 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3122 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3123 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3124 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3125 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3126 */
3127 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3128 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3129 fragment_prog_arbfp(state, stateblock, context);
3130 }
3131 }
3132
3133 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3134 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3135 fragment_prog_arbfp(state, stateblock, context);
3136 }
3137 }
3138
3139 #undef GLINFO_LOCATION
3140
3141 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3142 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3143 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3144 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3145 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3146 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3147 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3148 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3149 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3150 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3151 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3152 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3153 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3154 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3155 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3156 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3157 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3158 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3159 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3160 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3161 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3162 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3163 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3164 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3165 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3166 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3167 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3168 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3169 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3170 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3171 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3172 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3173 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3174 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3175 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3176 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3177 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3178 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3179 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3180 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3181 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3182 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3183 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3184 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3185 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3186 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3187 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3188 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3189 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3190 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3191 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3192 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3193 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3194 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3195 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3196 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3197 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3198 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3199 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3200 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3201 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3202 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3203 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3204 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3205 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3206 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3207 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3208 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3209 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3210 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3211 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3212 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3213 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3214 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3215 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3216 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3217 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3218 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3219 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3220 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3221 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3222 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3223 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3224 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3225 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3226 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3227 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3228 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3229 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3230 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3231 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3232 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3233 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3234 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3235 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3236 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3237 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3238 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3239 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3240 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3241 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3242 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3243 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3244 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3245 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3246 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3247 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3248 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3249 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3250 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3251 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3252 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3253 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3254 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3255 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3256 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3257 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3258 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3259 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3260 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3261 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3262 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3263 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3264 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3265 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3266 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3267 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3268 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3269 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3270 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3271 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3272 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3273 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3274 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3275 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3276 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3277 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3278 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3279 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3280 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3281 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3282 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3283 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3284 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3285 {0 /* Terminate */, { 0, 0 }, 0 },
3286 };
3287
3288 const struct fragment_pipeline arbfp_fragment_pipeline = {
3289 arbfp_enable,
3290 arbfp_get_caps,
3291 arbfp_alloc,
3292 arbfp_free,
3293 shader_arb_conv_supported,
3294 arbfp_fragmentstate_template,
3295 TRUE /* We can disable projected textures */
3296 };
3297
3298 #define GLINFO_LOCATION device->adapter->gl_info
3299
3300 struct arbfp_blit_priv {
3301 GLenum yuy2_rect_shader, yuy2_2d_shader;
3302 GLenum uyvy_rect_shader, uyvy_2d_shader;
3303 GLenum yv12_rect_shader, yv12_2d_shader;
3304 };
3305
3306 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3307 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3308 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3309 if(!device->blit_priv) {
3310 ERR("Out of memory\n");
3311 return E_OUTOFMEMORY;
3312 }
3313 return WINED3D_OK;
3314 }
3315 static void arbfp_blit_free(IWineD3DDevice *iface) {
3316 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3317 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3318
3319 ENTER_GL();
3320 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3321 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3322 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3323 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3324 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3325 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3326 checkGLcall("Delete yuv programs\n");
3327 LEAVE_GL();
3328 }
3329
3330 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3331 char chroma;
3332 const char *tex, *texinstr;
3333
3334 if(fmt == WINED3DFMT_UYVY) {
3335 chroma = 'r';
3336 *luminance = 'a';
3337 } else {
3338 chroma = 'a';
3339 *luminance = 'r';
3340 }
3341 switch(textype) {
3342 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3343 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3344 default:
3345 /* This is more tricky than just replacing the texture type - we have to navigate
3346 * properly in the texture to find the correct chroma values
3347 */
3348 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3349 return FALSE;
3350 }
3351
3352 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3353 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3354 * filtering when we sample the texture.
3355 *
3356 * These are the rules for reading the chroma:
3357 *
3358 * Even pixel: Cr
3359 * Even pixel: U
3360 * Odd pixel: V
3361 *
3362 * So we have to get the sampling x position in non-normalized coordinates in integers
3363 */
3364 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3365 shader_addline(buffer, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3366 shader_addline(buffer, "MOV texcrd.a, size.x;\n");
3367 } else {
3368 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3369 }
3370 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3371 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3372 * 0.5, so add 0.5.
3373 */
3374 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3375 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3376
3377 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3378 * even and odd pixels respectively
3379 */
3380 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3381 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3382
3383 /* Sample Pixel 1 */
3384 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3385
3386 /* Put the value into either of the chroma values */
3387 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3388 shader_addline(buffer, "MUL chroma.r, luminance.%c, temp.x;\n", chroma);
3389 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3390 shader_addline(buffer, "MUL chroma.g, luminance.%c, temp.x;\n", chroma);
3391
3392 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3393 * the pixel right to the current one. Otherwise, sample the left pixel.
3394 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3395 */
3396 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3397 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3398 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3399
3400 /* Put the value into the other chroma */
3401 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3402 shader_addline(buffer, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma);
3403 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3404 shader_addline(buffer, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma);
3405
3406 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3407 * the current one and lerp the two U and V values
3408 */
3409
3410 /* This gives the correctly filtered luminance value */
3411 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3412
3413 return TRUE;
3414 }
3415
3416 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3417 const char *tex;
3418
3419 switch(textype) {
3420 case GL_TEXTURE_2D: tex = "2D"; break;
3421 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3422 default:
3423 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3424 return FALSE;
3425 }
3426
3427 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3428 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3429 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3430 * pitch of the luminance plane, the packing into the gl texture is a bit
3431 * unfortunate. If the whole texture is interpreted as luminance data it looks
3432 * approximately like this:
3433 *
3434 * +----------------------------------+----
3435 * | |
3436 * | |
3437 * | |
3438 * | |
3439 * | | 2
3440 * | LUMINANCE | -
3441 * | | 3
3442 * | |
3443 * | |
3444 * | |
3445 * | |
3446 * +----------------+-----------------+----
3447 * | | |
3448 * | U even rows | U odd rows |
3449 * | | | 1
3450 * +----------------+------------------ -
3451 * | | | 3
3452 * | V even rows | V odd rows |
3453 * | | |
3454 * +----------------+-----------------+----
3455 * | | |
3456 * | 0.5 | 0.5 |
3457 *
3458 * So it appears as if there are 4 chroma images, but in fact the odd rows
3459 * in the chroma images are in the same row as the even ones. So its is
3460 * kinda tricky to read
3461 *
3462 * When reading from rectangle textures, keep in mind that the input y coordinates
3463 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3464 */
3465 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3466 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3467
3468 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3469 /* the chroma planes have only half the width */
3470 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3471
3472 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3473 * the coordinate. Also read the right side of the image when reading odd lines
3474 *
3475 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3476 * bleeding
3477 */
3478 if(textype == GL_TEXTURE_2D) {
3479
3480 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3481
3482 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3483
3484 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3485 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3486
3487 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3488 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3489 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3490 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3491 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3492
3493 /* clamp, keep the half pixel origin in mind */
3494 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3495 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3496 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3497 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3498 } else {
3499 /* Read from [size - size+size/4] */
3500 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3501 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3502
3503 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3504 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3505 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3506 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3507 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3508 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3509
3510 /* Make sure to read exactly from the pixel center */
3511 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3512 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3513
3514 /* Clamp */
3515 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3516 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3517 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3518 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3519 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3520 }
3521 /* Read the texture, put the result into the output register */
3522 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3523 shader_addline(buffer, "MOV chroma.r, temp.a;\n");
3524
3525 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3526 * No need to clamp because we're just reusing the already clamped value from above
3527 */
3528 if(textype == GL_TEXTURE_2D) {
3529 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3530 } else {
3531 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3532 }
3533 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3534 shader_addline(buffer, "MOV chroma.g, temp.a;\n");
3535
3536 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3537 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3538 * values due to filtering
3539 */
3540 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3541 if(textype == GL_TEXTURE_2D) {
3542 /* Multiply the y coordinate by 2/3 and clamp it */
3543 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3544 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3545 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3546 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3547 } else {
3548 /* Reading from texture_rectangles is pretty streightforward, just use the unmodified
3549 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3550 * is bigger
3551 */
3552 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3553 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3554 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3555 }
3556 *luminance = 'a';
3557
3558 return TRUE;
3559 }
3560
3561 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, WINED3DFORMAT fmt, GLenum textype) {
3562 GLenum shader;
3563 SHADER_BUFFER buffer;
3564 char luminance_component;
3565 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3566
3567 /* Shader header */
3568 buffer.bsize = 0;
3569 buffer.lineNo = 0;
3570 buffer.newline = TRUE;
3571 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
3572
3573 ENTER_GL();
3574 GL_EXTCALL(glGenProgramsARB(1, &shader));
3575 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3576 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3577 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3578 LEAVE_GL();
3579 if(!shader) {
3580 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3581 return 0;
3582 }
3583
3584 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3585 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3586 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3587 * each single pixel it contains, and one U and one V value shared between both
3588 * pixels.
3589 *
3590 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3591 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3592 * take the format into account when generating the read swizzles
3593 *
3594 * Reading the Y value is streightforward - just sample the texture. The hardware
3595 * takes care of filtering in the horizontal and vertical direction.
3596 *
3597 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3598 * because that would mix the U and V values of one pixel or two adjacent pixels.
3599 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3600 * regardless of the filtering setting. Vertical filtering works automatically
3601 * though - the U and V values of two rows are mixed nicely.
3602 *
3603 * Appart of avoiding filtering issues, the code has to know which value it just
3604 * read, and where it can find the other one. To determine this, it checks if
3605 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3606 *
3607 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3608 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3609 *
3610 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3611 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3612 * in an unfiltered situation. Finding the luminance on the other hand requires
3613 * finding out if it is an odd or even pixel. The real drawback of this approach
3614 * is filtering. This would have to be emulated completely in the shader, reading
3615 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3616 * vertically. Beyond that it would require adjustments to the texture handling
3617 * code to deal with the width scaling
3618 */
3619 shader_addline(&buffer, "!!ARBfp1.0\n");
3620 shader_addline(&buffer, "TEMP luminance;\n");
3621 shader_addline(&buffer, "TEMP temp;\n");
3622 shader_addline(&buffer, "TEMP chroma;\n");
3623 shader_addline(&buffer, "TEMP texcrd;\n");
3624 shader_addline(&buffer, "TEMP texcrd2;\n");
3625 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3626 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3627 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3628
3629 if(fmt == WINED3DFMT_UYVY || fmt ==WINED3DFMT_YUY2) {
3630 if(gen_planar_yuv_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3631 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3632 return 0;
3633 }
3634 } else {
3635 if(gen_yv12_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3636 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3637 return 0;
3638 }
3639 }
3640
3641 /* Calculate the final result. Formula is taken from
3642 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3643 * ranges from -0.5 to 0.5
3644 */
3645 shader_addline(&buffer, "SUB chroma.rg, chroma, coef.y;\n");
3646
3647 shader_addline(&buffer, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component);
3648 shader_addline(&buffer, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component);
3649 shader_addline(&buffer, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3650 shader_addline(&buffer, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component);
3651 shader_addline(&buffer, "END\n");
3652
3653 ENTER_GL();
3654 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3655
3656 if (glGetError() == GL_INVALID_OPERATION) {
3657 GLint pos;
3658 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3659 FIXME("Fragment program error at position %d: %s\n", pos,
3660 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3661 }
3662 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3663 LEAVE_GL();
3664
3665 if(fmt == WINED3DFMT_YUY2) {
3666 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3667 priv->yuy2_rect_shader = shader;
3668 } else {
3669 priv->yuy2_2d_shader = shader;
3670 }
3671 } else if(fmt == WINED3DFMT_UYVY) {
3672 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3673 priv->uyvy_rect_shader = shader;
3674 } else {
3675 priv->uyvy_2d_shader = shader;
3676 }
3677 } else {
3678 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3679 priv->yv12_rect_shader = shader;
3680 } else {
3681 priv->yv12_2d_shader = shader;
3682 }
3683 }
3684 return shader;
3685 }
3686
3687 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3688 GLenum shader;
3689 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3690 float size[4] = {width, height, 1, 1};
3691 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3692 const GlPixelFormatDesc *glDesc;
3693
3694 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3695
3696 if(glDesc->conversion_group != WINED3DFMT_YUY2 && glDesc->conversion_group != WINED3DFMT_UYVY &&
3697 glDesc->conversion_group != WINED3DFMT_YV12) {
3698 TRACE("Format: %s\n", debug_d3dformat(glDesc->conversion_group));
3699 /* Don't bother setting up a shader for unconverted formats */
3700 ENTER_GL();
3701 glEnable(textype);
3702 checkGLcall("glEnable(textype)");
3703 LEAVE_GL();
3704 return WINED3D_OK;
3705 }
3706
3707 if(glDesc->conversion_group == WINED3DFMT_YUY2) {
3708 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3709 shader = priv->yuy2_rect_shader;
3710 } else {
3711 shader = priv->yuy2_2d_shader;
3712 }
3713 } else if(glDesc->conversion_group == WINED3DFMT_UYVY) {
3714 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3715 shader = priv->uyvy_rect_shader;
3716 } else {
3717 shader = priv->uyvy_2d_shader;
3718 }
3719 } else {
3720 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3721 shader = priv->yv12_rect_shader;
3722 } else {
3723 shader = priv->yv12_2d_shader;
3724 }
3725 }
3726
3727 if(!shader) {
3728 shader = gen_yuv_shader(device, glDesc->conversion_group, textype);
3729 }
3730
3731 ENTER_GL();
3732 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3733 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3734 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3735 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3736 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3737 checkGLcall("glProgramLocalParameter4fvARB");
3738 LEAVE_GL();
3739
3740 return WINED3D_OK;
3741 }
3742
3743 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3744 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3745
3746 ENTER_GL();
3747 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3748 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3749 glDisable(GL_TEXTURE_2D);
3750 checkGLcall("glDisable(GL_TEXTURE_2D)");
3751 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3752 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3753 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3754 }
3755 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3756 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3757 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3758 }
3759 LEAVE_GL();
3760 }
3761
3762 static BOOL arbfp_blit_conv_supported(WINED3DFORMAT fmt) {
3763 TRACE("Checking blit format support for format %s:", debug_d3dformat(fmt));
3764 switch(fmt) {
3765 case WINED3DFMT_YUY2:
3766 case WINED3DFMT_UYVY:
3767 case WINED3DFMT_YV12:
3768 TRACE("[OK]\n");
3769 return TRUE;
3770 default:
3771 TRACE("[FAILED]\n");
3772 return FALSE;
3773 }
3774 }
3775
3776 const struct blit_shader arbfp_blit = {
3777 arbfp_blit_alloc,
3778 arbfp_blit_free,
3779 arbfp_blit_set,
3780 arbfp_blit_unset,
3781 arbfp_blit_conv_supported
3782 };
3783
3784 #undef GLINFO_LOCATION
useful hacks and other patches not (yet) suitable for mainline