?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push eb25bf65c4616aa55a810ed5c29198b1a080b208
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
blob8d632946e261e0fae63c49ed3e2b9f15d2a0c77f
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 WINE_DECLARE_DEBUG_CHANNEL(d3d);
40
41 #define GLINFO_LOCATION (*gl_info)
42
43 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
44 * so upload them above that
45 */
46 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
47 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
48
49 /* ARB_program_shader private data */
50 struct shader_arb_priv {
51 GLuint current_vprogram_id;
52 GLuint current_fprogram_id;
53 GLuint depth_blt_vprogram_id;
54 GLuint depth_blt_fprogram_id[tex_type_count];
55 BOOL use_arbfp_fixed_func;
56 struct hash_table_t *fragment_shaders;
57 };
58
59 /********************************************************
60 * ARB_[vertex/fragment]_program helper functions follow
61 ********************************************************/
62
63 /**
64 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
65 * When constant_list == NULL, it will load all the constants.
66 *
67 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
68 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
69 */
70 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, const WineD3D_GL_Info *gl_info,
71 GLuint target_type, unsigned int max_constants, const float *constants, char *dirty_consts)
72 {
73 DWORD shader_version = This->baseShader.reg_maps.shader_version;
74 local_constant* lconst;
75 DWORD i, j;
76 unsigned int ret;
77
78 if (TRACE_ON(d3d_shader)) {
79 for(i = 0; i < max_constants; i++) {
80 if(!dirty_consts[i]) continue;
81 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
82 constants[i * 4 + 0], constants[i * 4 + 1],
83 constants[i * 4 + 2], constants[i * 4 + 3]);
84 }
85 }
86 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
87 if (target_type == GL_FRAGMENT_PROGRAM_ARB && WINED3DSHADER_VERSION_MAJOR(shader_version) == 1)
88 {
89 float lcl_const[4];
90 for(i = 0; i < max_constants; i++) {
91 if(!dirty_consts[i]) continue;
92 dirty_consts[i] = 0;
93
94 j = 4 * i;
95 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
96 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
97 else lcl_const[0] = constants[j + 0];
98
99 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
100 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
101 else lcl_const[1] = constants[j + 1];
102
103 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
104 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
105 else lcl_const[2] = constants[j + 2];
106
107 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
108 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
109 else lcl_const[3] = constants[j + 3];
110
111 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
112 }
113 } else {
114 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
115 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
116 * or just reloading *all* constants at once
117 *
118 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
119 */
120 for(i = 0; i < max_constants; i++) {
121 if(!dirty_consts[i]) continue;
122
123 /* Find the next block of dirty constants */
124 dirty_consts[i] = 0;
125 j = i;
126 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
127 dirty_consts[i] = 0;
128 }
129
130 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
131 }
132 } else {
133 for(i = 0; i < max_constants; i++) {
134 if(dirty_consts[i]) {
135 dirty_consts[i] = 0;
136 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
137 }
138 }
139 }
140 }
141 checkGLcall("glProgramEnvParameter4fvARB()");
142
143 /* Load immediate constants */
144 if(This->baseShader.load_local_constsF) {
145 if (TRACE_ON(d3d_shader)) {
146 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
147 GLfloat* values = (GLfloat*)lconst->value;
148 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
149 values[0], values[1], values[2], values[3]);
150 }
151 }
152 /* Immediate constants are clamped for 1.X shaders at loading times */
153 ret = 0;
154 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
155 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
156 ret = max(ret, lconst->idx + 1);
157 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
158 }
159 checkGLcall("glProgramEnvParameter4fvARB()");
160 return ret; /* The loaded immediate constants need reloading for the next shader */
161 } else {
162 return 0; /* No constants are dirty now */
163 }
164 }
165
166 /**
167 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
168 *
169 * We only support float constants in ARB at the moment, so don't
170 * worry about the Integers or Booleans
171 */
172 static void shader_arb_load_constants(
173 IWineD3DDevice* device,
174 char usePixelShader,
175 char useVertexShader) {
176
177 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
178 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
179 const WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
180 unsigned char i;
181
182 if (useVertexShader) {
183 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
184
185 /* Load DirectX 9 float constants for vertex shader */
186 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
187 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
188 deviceImpl->highest_dirty_vs_const,
189 stateBlock->vertexShaderConstantF,
190 deviceImpl->activeContext->vshader_const_dirty);
191
192 /* Upload the position fixup */
193 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
194 }
195
196 if (usePixelShader) {
197
198 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
199 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
200
201 /* Load DirectX 9 float constants for pixel shader */
202 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
203 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
204 deviceImpl->highest_dirty_ps_const,
205 stateBlock->pixelShaderConstantF,
206 deviceImpl->activeContext->pshader_const_dirty);
207
208 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
209 /* The state manager takes care that this function is always called if the bump env matrix changes
210 */
211 const float *data = (const float *)&stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
212 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
213 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
214
215 if (psi->luminanceconst[i].const_num != WINED3D_CONST_NUM_UNUSED)
216 {
217 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
218 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
219 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
220 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
221 */
222 const float *scale = (const float *)&stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
223 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
224 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
225 }
226 }
227 }
228 }
229
230 static void shader_arb_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count)
231 {
232 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
233
234 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
235 * context. On a context switch the old context will be fully dirtified */
236 memset(This->activeContext->vshader_const_dirty + start, 1,
237 sizeof(*This->activeContext->vshader_const_dirty) * count);
238 This->highest_dirty_vs_const = max(This->highest_dirty_vs_const, start + count + 1);
239 }
240
241 static void shader_arb_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count)
242 {
243 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
244
245 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
246 * context. On a context switch the old context will be fully dirtified */
247 memset(This->activeContext->pshader_const_dirty + start, 1,
248 sizeof(*This->activeContext->pshader_const_dirty) * count);
249 This->highest_dirty_ps_const = max(This->highest_dirty_ps_const, start + count + 1);
250 }
251
252 /* Generate the variable & register declarations for the ARB_vertex_program output target */
253 static void shader_generate_arb_declarations(IWineD3DBaseShader *iface, const shader_reg_maps *reg_maps,
254 SHADER_BUFFER *buffer, const WineD3D_GL_Info *gl_info)
255 {
256 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
257 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
258 DWORD i, cur;
259 char pshader = shader_is_pshader_version(reg_maps->shader_version);
260 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
261 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
262 UINT extra_constants_needed = 0;
263 const local_constant *lconst;
264
265 /* Temporary Output register */
266 shader_addline(buffer, "TEMP TMP_OUT;\n");
267
268 for(i = 0; i < This->baseShader.limits.temporary; i++) {
269 if (reg_maps->temporary[i])
270 shader_addline(buffer, "TEMP R%u;\n", i);
271 }
272
273 for (i = 0; i < This->baseShader.limits.address; i++) {
274 if (reg_maps->address[i])
275 shader_addline(buffer, "ADDRESS A%d;\n", i);
276 }
277
278 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
279 if (reg_maps->texcoord[i])
280 shader_addline(buffer,"TEMP T%u;\n", i);
281 }
282
283 /* Texture coordinate registers must be pre-loaded */
284 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
285 if (reg_maps->texcoord[i])
286 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
287 }
288
289 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
290 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
291 if(!reg_maps->bumpmat[i]) continue;
292
293 cur = ps->numbumpenvmatconsts;
294 ps->bumpenvmatconst[cur].const_num = -1;
295 ps->bumpenvmatconst[cur].texunit = i;
296 ps->luminanceconst[cur].const_num = -1;
297 ps->luminanceconst[cur].texunit = i;
298
299 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
300 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
301 * bump mapping.
302 */
303 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
304 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
305 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
306 i, ps->bumpenvmatconst[cur].const_num);
307 extra_constants_needed++;
308
309 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
310 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
311 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
312 i, ps->luminanceconst[cur].const_num);
313 extra_constants_needed++;
314 } else if(reg_maps->luminanceparams) {
315 FIXME("No free constant to load the luminance parameters\n");
316 }
317 } else {
318 FIXME("No free constant found to load environment bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
319 }
320
321 ps->numbumpenvmatconsts = cur + 1;
322 }
323
324 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
325 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
326 srgb_mul_low, srgb_mul_low, srgb_mul_low);
327 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
328 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
329 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
330 srgb_pow, srgb_pow, srgb_pow);
331 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
332 srgb_mul_high, srgb_mul_high, srgb_mul_high);
333 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
334 srgb_sub_high, srgb_sub_high, srgb_sub_high);
335 }
336
337 /* Load local constants using the program-local space,
338 * this avoids reloading them each time the shader is used
339 */
340 if(!This->baseShader.load_local_constsF) {
341 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
342 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
343 lconst->idx);
344 }
345 }
346
347 /* we use the array-based constants array if the local constants are marked for loading,
348 * because then we use indirect addressing, or when the local constant list is empty,
349 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
350 * local constants do not declare the loaded constants as an array because ARB compilers usually
351 * do not optimize unused constants away
352 */
353 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
354 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
355 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
356 max_constantsF, max_constantsF - 1);
357 } else {
358 for(i = 0; i < max_constantsF; i++) {
359 if(!shader_constant_is_local(This, i)) {
360 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
361 }
362 }
363 }
364 }
365
366 static const char * const shift_tab[] = {
367 "dummy", /* 0 (none) */
368 "coefmul.x", /* 1 (x2) */
369 "coefmul.y", /* 2 (x4) */
370 "coefmul.z", /* 3 (x8) */
371 "coefmul.w", /* 4 (x16) */
372 "dummy", /* 5 (x32) */
373 "dummy", /* 6 (x64) */
374 "dummy", /* 7 (x128) */
375 "dummy", /* 8 (d256) */
376 "dummy", /* 9 (d128) */
377 "dummy", /* 10 (d64) */
378 "dummy", /* 11 (d32) */
379 "coefdiv.w", /* 12 (d16) */
380 "coefdiv.z", /* 13 (d8) */
381 "coefdiv.y", /* 14 (d4) */
382 "coefdiv.x" /* 15 (d2) */
383 };
384
385 static void shader_arb_get_write_mask(const SHADER_OPCODE_ARG *arg, const DWORD param, char *write_mask)
386 {
387 char *ptr = write_mask;
388 char vshader = shader_is_vshader_version(arg->reg_maps->shader_version);
389
390 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
391 *ptr++ = '.';
392 *ptr++ = 'x';
393 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
394 *ptr++ = '.';
395 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
396 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
397 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
398 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
399 }
400
401 *ptr = '\0';
402 }
403
404 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
405 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
406 * but addressed as "rgba". To fix this we need to swap the register's x
407 * and z components. */
408 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
409 char *ptr = swizzle_str;
410
411 /* swizzle bits fields: wwzzyyxx */
412 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
413 DWORD swizzle_x = swizzle & 0x03;
414 DWORD swizzle_y = (swizzle >> 2) & 0x03;
415 DWORD swizzle_z = (swizzle >> 4) & 0x03;
416 DWORD swizzle_w = (swizzle >> 6) & 0x03;
417
418 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
419 * generate a swizzle string. Unless we need to our own swizzling. */
420 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
421 *ptr++ = '.';
422 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
423 *ptr++ = swizzle_chars[swizzle_x];
424 } else {
425 *ptr++ = swizzle_chars[swizzle_x];
426 *ptr++ = swizzle_chars[swizzle_y];
427 *ptr++ = swizzle_chars[swizzle_z];
428 *ptr++ = swizzle_chars[swizzle_w];
429 }
430 }
431
432 *ptr = '\0';
433 }
434
435 static void pshader_get_register_name(IWineD3DBaseShader* iface,
436 const DWORD param, char* regstr) {
437
438 DWORD reg = param & WINED3DSP_REGNUM_MASK;
439 DWORD regtype = shader_get_regtype(param);
440 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
441
442 switch (regtype) {
443 case WINED3DSPR_TEMP:
444 sprintf(regstr, "R%u", reg);
445 break;
446 case WINED3DSPR_INPUT:
447 if (reg==0) {
448 strcpy(regstr, "fragment.color.primary");
449 } else {
450 strcpy(regstr, "fragment.color.secondary");
451 }
452 break;
453 case WINED3DSPR_CONST:
454 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
455 sprintf(regstr, "C[%u]", reg);
456 } else {
457 sprintf(regstr, "C%u", reg);
458 }
459 break;
460 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
461 sprintf(regstr,"T%u", reg);
462 break;
463 case WINED3DSPR_COLOROUT:
464 if (reg == 0)
465 sprintf(regstr, "TMP_COLOR");
466 else {
467 /* TODO: See GL_ARB_draw_buffers */
468 FIXME("Unsupported write to render target %u\n", reg);
469 sprintf(regstr, "unsupported_register");
470 }
471 break;
472 case WINED3DSPR_DEPTHOUT:
473 sprintf(regstr, "result.depth");
474 break;
475 case WINED3DSPR_ATTROUT:
476 sprintf(regstr, "oD[%u]", reg);
477 break;
478 case WINED3DSPR_TEXCRDOUT:
479 sprintf(regstr, "oT[%u]", reg);
480 break;
481 default:
482 FIXME("Unhandled register name Type(%d)\n", regtype);
483 sprintf(regstr, "unrecognized_register");
484 break;
485 }
486 }
487
488 /* TODO: merge with pixel shader */
489 static void vshader_program_add_param(const SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine)
490 {
491 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
492
493 /* oPos, oFog and oPts in D3D */
494 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
495
496 DWORD reg = param & WINED3DSP_REGNUM_MASK;
497 DWORD regtype = shader_get_regtype(param);
498 char tmpReg[255];
499 BOOL is_color = FALSE;
500
501 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
502 strcat(hwLine, " -");
503 } else {
504 strcat(hwLine, " ");
505 }
506
507 switch (regtype) {
508 case WINED3DSPR_TEMP:
509 sprintf(tmpReg, "R%u", reg);
510 strcat(hwLine, tmpReg);
511 break;
512 case WINED3DSPR_INPUT:
513
514 if (This->cur_args->swizzle_map & (1 << reg)) is_color = TRUE;
515
516 sprintf(tmpReg, "vertex.attrib[%u]", reg);
517 strcat(hwLine, tmpReg);
518 break;
519 case WINED3DSPR_CONST:
520 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
521 if(reg >= This->rel_offset) {
522 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
523 } else {
524 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
525 }
526 } else {
527 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
528 sprintf(tmpReg, "C[%u]", reg);
529 } else {
530 sprintf(tmpReg, "C%u", reg);
531 }
532 }
533 strcat(hwLine, tmpReg);
534 break;
535 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
536 sprintf(tmpReg, "A%u", reg);
537 strcat(hwLine, tmpReg);
538 break;
539 case WINED3DSPR_RASTOUT:
540 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
541 strcat(hwLine, tmpReg);
542 break;
543 case WINED3DSPR_ATTROUT:
544 if (reg==0) {
545 strcat(hwLine, "result.color.primary");
546 } else {
547 strcat(hwLine, "result.color.secondary");
548 }
549 break;
550 case WINED3DSPR_TEXCRDOUT:
551 sprintf(tmpReg, "result.texcoord[%u]", reg);
552 strcat(hwLine, tmpReg);
553 break;
554 default:
555 FIXME("Unknown reg type %d %d\n", regtype, reg);
556 strcat(hwLine, "unrecognized_register");
557 break;
558 }
559
560 if (!is_input) {
561 char write_mask[6];
562 shader_arb_get_write_mask(arg, param, write_mask);
563 strcat(hwLine, write_mask);
564 } else {
565 char swizzle[6];
566 shader_arb_get_swizzle(param, is_color, swizzle);
567 strcat(hwLine, swizzle);
568 }
569 }
570
571 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source)
572 {
573 switch(channel_source)
574 {
575 case CHANNEL_SOURCE_ZERO: return "0";
576 case CHANNEL_SOURCE_ONE: return "1";
577 case CHANNEL_SOURCE_X: return "x";
578 case CHANNEL_SOURCE_Y: return "y";
579 case CHANNEL_SOURCE_Z: return "z";
580 case CHANNEL_SOURCE_W: return "w";
581 default:
582 FIXME("Unhandled channel source %#x\n", channel_source);
583 return "undefined";
584 }
585 }
586
587 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, DWORD dst_mask,
588 const char *one, const char *two, struct color_fixup_desc fixup)
589 {
590 DWORD mask;
591
592 if (is_yuv_fixup(fixup))
593 {
594 enum yuv_fixup yuv_fixup = get_yuv_fixup(fixup);
595 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup);
596 return;
597 }
598
599 mask = 0;
600 if (fixup.x_source != CHANNEL_SOURCE_X) mask |= WINED3DSP_WRITEMASK_0;
601 if (fixup.y_source != CHANNEL_SOURCE_Y) mask |= WINED3DSP_WRITEMASK_1;
602 if (fixup.z_source != CHANNEL_SOURCE_Z) mask |= WINED3DSP_WRITEMASK_2;
603 if (fixup.w_source != CHANNEL_SOURCE_W) mask |= WINED3DSP_WRITEMASK_3;
604 mask &= dst_mask;
605
606 if (mask)
607 {
608 shader_addline(buffer, "SWZ %s, %s, %s, %s, %s, %s;\n", reg, reg,
609 shader_arb_get_fixup_swizzle(fixup.x_source), shader_arb_get_fixup_swizzle(fixup.y_source),
610 shader_arb_get_fixup_swizzle(fixup.z_source), shader_arb_get_fixup_swizzle(fixup.w_source));
611 }
612
613 mask = 0;
614 if (fixup.x_sign_fixup) mask |= WINED3DSP_WRITEMASK_0;
615 if (fixup.y_sign_fixup) mask |= WINED3DSP_WRITEMASK_1;
616 if (fixup.z_sign_fixup) mask |= WINED3DSP_WRITEMASK_2;
617 if (fixup.w_sign_fixup) mask |= WINED3DSP_WRITEMASK_3;
618 mask &= dst_mask;
619
620 if (mask)
621 {
622 char reg_mask[6];
623 char *ptr = reg_mask;
624
625 if (mask != WINED3DSP_WRITEMASK_ALL)
626 {
627 *ptr++ = '.';
628 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
629 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
630 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
631 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
632 }
633 *ptr = '\0';
634
635 shader_addline(buffer, "MAD %s%s, %s, %s, -%s;\n", reg, reg_mask, reg, two, one);
636 }
637 }
638
639 static void shader_hw_sample(const SHADER_OPCODE_ARG *arg, DWORD sampler_idx, const char *dst_str,
640 const char *coord_reg, BOOL projected, BOOL bias)
641 {
642 SHADER_BUFFER* buffer = arg->buffer;
643 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
644 const char *tex_type;
645 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
646 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
647
648 switch(sampler_type) {
649 case WINED3DSTT_1D:
650 tex_type = "1D";
651 break;
652
653 case WINED3DSTT_2D:
654 if(device->stateBlock->textures[sampler_idx] &&
655 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
656 tex_type = "RECT";
657 } else {
658 tex_type = "2D";
659 }
660 break;
661
662 case WINED3DSTT_VOLUME:
663 tex_type = "3D";
664 break;
665
666 case WINED3DSTT_CUBE:
667 tex_type = "CUBE";
668 break;
669
670 default:
671 ERR("Unexpected texture type %d\n", sampler_type);
672 tex_type = "";
673 }
674
675 if (bias) {
676 /* Shouldn't be possible, but let's check for it */
677 if(projected) FIXME("Biased and Projected texture sampling\n");
678 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
679 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
680 } else if (projected) {
681 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
682 } else {
683 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
684 }
685
686 if(shader_is_pshader_version(arg->reg_maps->shader_version)) {
687 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) arg->shader;
688 gen_color_correction(buffer, dst_str, arg->dst & WINED3DSP_WRITEMASK_ALL, "one", "coefmul.x",
689 ps->cur_args->color_fixup[sampler_idx]);
690 }
691 }
692
693 static void pshader_gen_input_modifier_line (
694 IWineD3DBaseShader *iface,
695 SHADER_BUFFER* buffer,
696 const DWORD instr,
697 int tmpreg,
698 char *outregstr) {
699
700 /* Generate a line that does the input modifier computation and return the input register to use */
701 char regstr[256];
702 char swzstr[20];
703 int insert_line;
704
705 /* Assume a new line will be added */
706 insert_line = 1;
707
708 /* Get register name */
709 pshader_get_register_name(iface, instr, regstr);
710 shader_arb_get_swizzle(instr, FALSE, swzstr);
711
712 switch (instr & WINED3DSP_SRCMOD_MASK) {
713 case WINED3DSPSM_NONE:
714 sprintf(outregstr, "%s%s", regstr, swzstr);
715 insert_line = 0;
716 break;
717 case WINED3DSPSM_NEG:
718 sprintf(outregstr, "-%s%s", regstr, swzstr);
719 insert_line = 0;
720 break;
721 case WINED3DSPSM_BIAS:
722 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
723 break;
724 case WINED3DSPSM_BIASNEG:
725 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
726 break;
727 case WINED3DSPSM_SIGN:
728 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
729 break;
730 case WINED3DSPSM_SIGNNEG:
731 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
732 break;
733 case WINED3DSPSM_COMP:
734 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
735 break;
736 case WINED3DSPSM_X2:
737 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
738 break;
739 case WINED3DSPSM_X2NEG:
740 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
741 break;
742 case WINED3DSPSM_DZ:
743 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
744 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
745 break;
746 case WINED3DSPSM_DW:
747 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
748 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
749 break;
750 default:
751 sprintf(outregstr, "%s%s", regstr, swzstr);
752 insert_line = 0;
753 }
754
755 /* Return modified or original register, with swizzle */
756 if (insert_line)
757 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
758 }
759
760 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER *buffer, int saturate, const char *write_mask,
761 int shift, const char *regstr)
762 {
763 /* Generate a line that does the output modifier computation */
764 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
765 regstr, write_mask, regstr, shift_tab[shift]);
766 }
767
768 static void pshader_hw_bem(const SHADER_OPCODE_ARG *arg)
769 {
770 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
771
772 SHADER_BUFFER* buffer = arg->buffer;
773 char dst_name[50];
774 char src_name[2][50];
775 char dst_wmask[20];
776 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
777 BOOL has_bumpmat = FALSE;
778 int i;
779
780 for(i = 0; i < This->numbumpenvmatconsts; i++) {
781 if (This->bumpenvmatconst[i].const_num != WINED3D_CONST_NUM_UNUSED
782 && This->bumpenvmatconst[i].texunit == sampler_code)
783 {
784 has_bumpmat = TRUE;
785 break;
786 }
787 }
788
789 pshader_get_register_name(arg->shader, arg->dst, dst_name);
790 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
791 strcat(dst_name, dst_wmask);
792
793 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
794 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
795
796 if(has_bumpmat) {
797 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
798 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
799 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
800 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
801 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
802
803 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
804 } else {
805 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
806 }
807 }
808
809 static void pshader_hw_cnd(const SHADER_OPCODE_ARG *arg)
810 {
811 SHADER_BUFFER* buffer = arg->buffer;
812 char dst_wmask[20];
813 char dst_name[50];
814 char src_name[3][50];
815 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
816 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
817
818 /* FIXME: support output modifiers */
819
820 /* Handle output register */
821 pshader_get_register_name(arg->shader, arg->dst, dst_name);
822 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
823
824 /* Generate input register names (with modifiers) */
825 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
826 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
827 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
828
829 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
830 if (arg->reg_maps->shader_version <= WINED3DPS_VERSION(1, 3) && arg->opcode_token & WINED3DSI_COISSUE)
831 {
832 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
833 } else {
834 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
835 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
836 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
837 }
838 if (shift != 0)
839 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
840 }
841
842 static void pshader_hw_cmp(const SHADER_OPCODE_ARG *arg)
843 {
844 SHADER_BUFFER* buffer = arg->buffer;
845 char dst_wmask[20];
846 char dst_name[50];
847 char src_name[3][50];
848 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
849 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
850
851 /* FIXME: support output modifiers */
852
853 /* Handle output register */
854 pshader_get_register_name(arg->shader, arg->dst, dst_name);
855 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
856
857 /* Generate input register names (with modifiers) */
858 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
859 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
860 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
861
862 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
863 src_name[0], src_name[2], src_name[1]);
864
865 if (shift != 0)
866 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
867 }
868
869 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
870 * dst = dot2(src0, src1) + src2 */
871 static void pshader_hw_dp2add(const SHADER_OPCODE_ARG *arg)
872 {
873 SHADER_BUFFER* buffer = arg->buffer;
874 char dst_wmask[20];
875 char dst_name[50];
876 char src_name[3][50];
877 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
878 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
879
880 pshader_get_register_name(arg->shader, arg->dst, dst_name);
881 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
882
883 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
884 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
885 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
886
887 /* Emulate a DP2 with a DP3 and 0.0 */
888 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
889 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
890 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
891 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
892
893 if (shift != 0)
894 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
895 }
896
897 /* Map the opcode 1-to-1 to the GL code */
898 static void shader_hw_map2gl(const SHADER_OPCODE_ARG *arg)
899 {
900 CONST SHADER_OPCODE* curOpcode = arg->opcode;
901 SHADER_BUFFER* buffer = arg->buffer;
902 DWORD dst = arg->dst;
903 const DWORD *src = arg->src;
904 const char *instruction;
905 char arguments[256];
906 unsigned int i;
907
908 switch (curOpcode->opcode)
909 {
910 case WINED3DSIO_ABS: instruction = "ABS"; break;
911 case WINED3DSIO_ADD: instruction = "ADD"; break;
912 case WINED3DSIO_CRS: instruction = "XPD"; break;
913 case WINED3DSIO_DP3: instruction = "DP3"; break;
914 case WINED3DSIO_DP4: instruction = "DP4"; break;
915 case WINED3DSIO_DST: instruction = "DST"; break;
916 case WINED3DSIO_EXP: instruction = "EX2"; break;
917 case WINED3DSIO_EXPP: instruction = "EXP"; break;
918 case WINED3DSIO_FRC: instruction = "FRC"; break;
919 case WINED3DSIO_LIT: instruction = "LIT"; break;
920 case WINED3DSIO_LOG: instruction = "LG2"; break;
921 case WINED3DSIO_LOGP: instruction = "LOG"; break;
922 case WINED3DSIO_LRP: instruction = "LRP"; break;
923 case WINED3DSIO_MAD: instruction = "MAD"; break;
924 case WINED3DSIO_MAX: instruction = "MAX"; break;
925 case WINED3DSIO_MIN: instruction = "MIN"; break;
926 case WINED3DSIO_MOV: instruction = "MOV"; break;
927 case WINED3DSIO_MUL: instruction = "MUL"; break;
928 case WINED3DSIO_NOP: instruction = "NOP"; break;
929 case WINED3DSIO_POW: instruction = "POW"; break;
930 case WINED3DSIO_SGE: instruction = "SGE"; break;
931 case WINED3DSIO_SLT: instruction = "SLT"; break;
932 case WINED3DSIO_SUB: instruction = "SUB"; break;
933 default: instruction = "";
934 FIXME("Unhandled opcode %s\n", curOpcode->name);
935 break;
936 }
937
938 if (shader_is_pshader_version(arg->reg_maps->shader_version))
939 {
940 /* Output token related */
941 char output_rname[256];
942 char output_wmask[20];
943 char operands[4][100];
944 BOOL saturate = FALSE;
945 BOOL centroid = FALSE;
946 BOOL partialprecision = FALSE;
947 const char *modifier;
948 DWORD shift;
949
950 if (!curOpcode->num_params)
951 {
952 ERR("Opcode \"%s\" has no parameters\n", curOpcode->name);
953 return;
954 }
955
956 /* Process modifiers */
957 if (dst & WINED3DSP_DSTMOD_MASK)
958 {
959 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
960
961 saturate = mask & WINED3DSPDM_SATURATE;
962 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
963 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
964 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
965 if (mask)
966 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
967
968 if (centroid)
969 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
970 }
971 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
972 modifier = (saturate && !shift) ? "_SAT" : "";
973
974 /* Generate input register names (with modifiers) */
975 for (i = 1; i < curOpcode->num_params; ++i)
976 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
977
978 /* Handle output register */
979 pshader_get_register_name(arg->shader, dst, output_rname);
980 strcpy(operands[0], output_rname);
981 shader_arb_get_write_mask(arg, dst, output_wmask);
982 strcat(operands[0], output_wmask);
983
984 arguments[0] = '\0';
985 strcat(arguments, operands[0]);
986 for (i = 1; i < curOpcode->num_params; i++)
987 {
988 strcat(arguments, ", ");
989 strcat(arguments, operands[i]);
990 }
991 shader_addline(buffer, "%s%s %s;\n", instruction, modifier, arguments);
992
993 /* A shift requires another line. */
994 if (shift) pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
995 } else {
996 /* Note that vshader_program_add_param() adds spaces. */
997
998 arguments[0] = '\0';
999 if (curOpcode->num_params > 0)
1000 {
1001 vshader_program_add_param(arg, dst, FALSE, arguments);
1002 for (i = 1; i < curOpcode->num_params; ++i)
1003 {
1004 strcat(arguments, ",");
1005 vshader_program_add_param(arg, src[i-1], TRUE, arguments);
1006 }
1007 }
1008 shader_addline(buffer, "%s%s;\n", instruction, arguments);
1009 }
1010 }
1011
1012 static void shader_hw_mov(const SHADER_OPCODE_ARG *arg)
1013 {
1014 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1015
1016 if ((WINED3DSHADER_VERSION_MAJOR(arg->reg_maps->shader_version) == 1
1017 && !shader_is_pshader_version(arg->reg_maps->shader_version)
1018 && shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)
1019 || arg->opcode->opcode == WINED3DSIO_MOVA)
1020 {
1021 SHADER_BUFFER *buffer = arg->buffer;
1022 char src0_param[256];
1023
1024 if (arg->opcode->opcode == WINED3DSIO_MOVA)
1025 FIXME("mova should round\n");
1026
1027 src0_param[0] = '\0';
1028 if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
1029 {
1030 vshader_program_add_param(arg, arg->src[0], TRUE, src0_param);
1031 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", src0_param);
1032 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1033 }
1034 else
1035 {
1036 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1037 * with more than one component. Thus replicate the first source argument over all
1038 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1039 */
1040 DWORD parm = arg->src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1041 if((arg->src[0] & WINED3DVS_X_W) == WINED3DVS_X_W)
1042 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1043 else if((arg->src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z)
1044 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1045 else if((arg->src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y)
1046 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1047 else if((arg->src[0] & WINED3DVS_X_X) == WINED3DVS_X_X)
1048 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1049 vshader_program_add_param(arg, parm, TRUE, src0_param);
1050 shader_addline(buffer, "ARL A0.x, %s;\n", src0_param);
1051 }
1052 }
1053 else
1054 {
1055 shader_hw_map2gl(arg);
1056 }
1057 }
1058
1059 static void pshader_hw_texkill(const SHADER_OPCODE_ARG *arg)
1060 {
1061 DWORD shader_version = arg->reg_maps->shader_version;
1062 SHADER_BUFFER* buffer = arg->buffer;
1063 char reg_dest[40];
1064
1065 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1066 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1067 */
1068 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1069
1070 if (shader_version >= WINED3DPS_VERSION(2,0))
1071 {
1072 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1073 shader_addline(buffer, "KIL %s;\n", reg_dest);
1074 } else {
1075 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1076 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1077 */
1078 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1079 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1080 shader_addline(buffer, "KIL TMP;\n");
1081 }
1082 }
1083
1084 static void pshader_hw_tex(const SHADER_OPCODE_ARG *arg)
1085 {
1086 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1087 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1088
1089 DWORD dst = arg->dst;
1090 const DWORD *src = arg->src;
1091 SHADER_BUFFER* buffer = arg->buffer;
1092 DWORD shader_version = arg->reg_maps->shader_version;
1093 BOOL projected = FALSE, bias = FALSE;
1094
1095 char reg_dest[40];
1096 char reg_coord[40];
1097 DWORD reg_dest_code;
1098 DWORD reg_sampler_code;
1099
1100 /* All versions have a destination register */
1101 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1102 pshader_get_register_name(arg->shader, dst, reg_dest);
1103
1104 /* 1.0-1.3: Use destination register as coordinate source.
1105 1.4+: Use provided coordinate source register. */
1106 if (shader_version < WINED3DPS_VERSION(1,4))
1107 strcpy(reg_coord, reg_dest);
1108 else
1109 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1110
1111 /* 1.0-1.4: Use destination register number as texture code.
1112 2.0+: Use provided sampler number as texure code. */
1113 if (shader_version < WINED3DPS_VERSION(2,0))
1114 reg_sampler_code = reg_dest_code;
1115 else
1116 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1117
1118 /* projection flag:
1119 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1120 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1121 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1122 */
1123 if (shader_version < WINED3DPS_VERSION(1,4))
1124 {
1125 DWORD flags = 0;
1126 if(reg_sampler_code < MAX_TEXTURES) {
1127 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1128 }
1129 if (flags & WINED3DTTFF_PROJECTED) {
1130 projected = TRUE;
1131 }
1132 }
1133 else if (shader_version < WINED3DPS_VERSION(2,0))
1134 {
1135 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1136 if (src_mod == WINED3DSPSM_DZ) {
1137 projected = TRUE;
1138 } else if(src_mod == WINED3DSPSM_DW) {
1139 projected = TRUE;
1140 }
1141 } else {
1142 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1143 projected = TRUE;
1144 }
1145 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1146 bias = TRUE;
1147 }
1148 }
1149 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1150 }
1151
1152 static void pshader_hw_texcoord(const SHADER_OPCODE_ARG *arg)
1153 {
1154 DWORD dst = arg->dst;
1155 SHADER_BUFFER* buffer = arg->buffer;
1156
1157 char tmp[20];
1158 shader_arb_get_write_mask(arg, dst, tmp);
1159 if (arg->reg_maps->shader_version != WINED3DPS_VERSION(1,4))
1160 {
1161 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1162 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1163 } else {
1164 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1165 char reg_src[40];
1166
1167 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1168 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1169 }
1170 }
1171
1172 static void pshader_hw_texreg2ar(const SHADER_OPCODE_ARG *arg)
1173 {
1174 SHADER_BUFFER* buffer = arg->buffer;
1175 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1176 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1177 DWORD flags;
1178
1179 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1180 char dst_str[8];
1181 char src_str[50];
1182
1183 sprintf(dst_str, "T%u", reg1);
1184 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1185 shader_addline(buffer, "MOV TMP.x, %s.w;\n", src_str);
1186 shader_addline(buffer, "MOV TMP.y, %s.x;\n", src_str);
1187 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1188 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1189 }
1190
1191 static void pshader_hw_texreg2gb(const SHADER_OPCODE_ARG *arg)
1192 {
1193 SHADER_BUFFER* buffer = arg->buffer;
1194
1195 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1196 char dst_str[8];
1197 char src_str[50];
1198
1199 sprintf(dst_str, "T%u", reg1);
1200 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1201 shader_addline(buffer, "MOV TMP.x, %s.y;\n", src_str);
1202 shader_addline(buffer, "MOV TMP.y, %s.z;\n", src_str);
1203 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1204 }
1205
1206 static void pshader_hw_texreg2rgb(const SHADER_OPCODE_ARG *arg)
1207 {
1208 SHADER_BUFFER* buffer = arg->buffer;
1209 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1210 char dst_str[8];
1211 char src_str[50];
1212
1213 sprintf(dst_str, "T%u", reg1);
1214 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1215 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1216 }
1217
1218 static void pshader_hw_texbem(const SHADER_OPCODE_ARG *arg)
1219 {
1220 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1221 BOOL has_bumpmat = FALSE;
1222 BOOL has_luminance = FALSE;
1223 int i;
1224
1225 DWORD dst = arg->dst;
1226 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1227 SHADER_BUFFER* buffer = arg->buffer;
1228
1229 char reg_coord[40];
1230 DWORD reg_dest_code;
1231
1232 /* All versions have a destination register */
1233 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1234 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1235 pshader_get_register_name(arg->shader, dst, reg_coord);
1236
1237 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1238 if (This->bumpenvmatconst[i].const_num != WINED3D_CONST_NUM_UNUSED
1239 && reg_dest_code == This->bumpenvmatconst[i].texunit)
1240 {
1241 has_bumpmat = TRUE;
1242 break;
1243 }
1244 }
1245 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1246 if (This->luminanceconst[i].const_num != WINED3D_CONST_NUM_UNUSED
1247 && reg_dest_code == This->luminanceconst[i].texunit)
1248 {
1249 has_luminance = TRUE;
1250 break;
1251 }
1252 }
1253
1254 if(has_bumpmat) {
1255 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1256
1257 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1258 shader_addline(buffer, "DP3 TMP.x, TMP2, T%u;\n", src);
1259 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1260 shader_addline(buffer, "DP3 TMP.y, TMP2, T%u;\n", src);
1261
1262 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1263 * so we can't let the GL handle this.
1264 */
1265 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1266 & WINED3DTTFF_PROJECTED) {
1267 shader_addline(buffer, "RCP TMP2.w, %s.w;\n", reg_coord);
1268 shader_addline(buffer, "MUL TMP2.xy, %s, TMP2.w;\n", reg_coord);
1269 shader_addline(buffer, "ADD TMP.xy, TMP, TMP2;\n");
1270 } else {
1271 shader_addline(buffer, "ADD TMP.xy, TMP, %s;\n", reg_coord);
1272 }
1273
1274 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1275
1276 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1277 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1278 src, reg_dest_code, reg_dest_code);
1279 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1280 }
1281
1282 } else {
1283 DWORD tf;
1284 if(reg_dest_code < MAX_TEXTURES) {
1285 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1286 } else {
1287 tf = 0;
1288 }
1289 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1290 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1291 }
1292 }
1293
1294 static void pshader_hw_texm3x2pad(const SHADER_OPCODE_ARG *arg)
1295 {
1296 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1297 SHADER_BUFFER* buffer = arg->buffer;
1298 char src0_name[50];
1299
1300 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1301 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1302 }
1303
1304 static void pshader_hw_texm3x2tex(const SHADER_OPCODE_ARG *arg)
1305 {
1306 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1307 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1308 DWORD flags;
1309 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1310 SHADER_BUFFER* buffer = arg->buffer;
1311 char dst_str[8];
1312 char src0_name[50];
1313
1314 sprintf(dst_str, "T%u", reg);
1315 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1316 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1317 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1318 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1319 }
1320
1321 static void pshader_hw_texm3x3pad(const SHADER_OPCODE_ARG *arg)
1322 {
1323 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1324 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1325 SHADER_BUFFER* buffer = arg->buffer;
1326 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1327 char src0_name[50];
1328
1329 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1330 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1331 current_state->texcoord_w[current_state->current_row++] = reg;
1332 }
1333
1334 static void pshader_hw_texm3x3tex(const SHADER_OPCODE_ARG *arg)
1335 {
1336 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1337 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1338 DWORD flags;
1339 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1340 SHADER_BUFFER* buffer = arg->buffer;
1341 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1342 char dst_str[8];
1343 char src0_name[50];
1344
1345 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1346 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1347
1348 /* Sample the texture using the calculated coordinates */
1349 sprintf(dst_str, "T%u", reg);
1350 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1351 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1352 current_state->current_row = 0;
1353 }
1354
1355 static void pshader_hw_texm3x3vspec(const SHADER_OPCODE_ARG *arg)
1356 {
1357 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1358 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1359 DWORD flags;
1360 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1361 SHADER_BUFFER* buffer = arg->buffer;
1362 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1363 char dst_str[8];
1364 char src0_name[50];
1365
1366 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1367 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1368
1369 /* Construct the eye-ray vector from w coordinates */
1370 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1371 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1372 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1373
1374 /* Calculate reflection vector
1375 */
1376 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1377 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1378 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1379 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1380 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1381 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1382 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1383
1384 /* Sample the texture using the calculated coordinates */
1385 sprintf(dst_str, "T%u", reg);
1386 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1387 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1388 current_state->current_row = 0;
1389 }
1390
1391 static void pshader_hw_texm3x3spec(const SHADER_OPCODE_ARG *arg)
1392 {
1393 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1394 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1395 DWORD flags;
1396 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1397 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1398 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1399 SHADER_BUFFER* buffer = arg->buffer;
1400 char dst_str[8];
1401 char src0_name[50];
1402
1403 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1404 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1405
1406 /* Calculate reflection vector.
1407 *
1408 * dot(N, E)
1409 * TMP.xyz = 2 * --------- * N - E
1410 * dot(N, N)
1411 *
1412 * Which normalizes the normal vector
1413 */
1414 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1415 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1416 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1417 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1418 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1419 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1420
1421 /* Sample the texture using the calculated coordinates */
1422 sprintf(dst_str, "T%u", reg);
1423 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1424 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1425 current_state->current_row = 0;
1426 }
1427
1428 static void pshader_hw_texdepth(const SHADER_OPCODE_ARG *arg)
1429 {
1430 SHADER_BUFFER* buffer = arg->buffer;
1431 char dst_name[50];
1432
1433 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1434 * which is essentially an input, is the destination register because it is the first
1435 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1436 * here
1437 */
1438 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1439
1440 /* According to the msdn, the source register(must be r5) is unusable after
1441 * the texdepth instruction, so we're free to modify it
1442 */
1443 shader_addline(buffer, "MIN %s.y, %s.y, one.y;\n", dst_name, dst_name);
1444
1445 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1446 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1447 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1448 */
1449 shader_addline(buffer, "RCP %s.y, %s.y;\n", dst_name, dst_name);
1450 shader_addline(buffer, "MUL TMP.x, %s.x, %s.y;\n", dst_name, dst_name);
1451 shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1452 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1453 }
1454
1455 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1456 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1457 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1458 static void pshader_hw_texdp3tex(const SHADER_OPCODE_ARG *arg)
1459 {
1460 SHADER_BUFFER* buffer = arg->buffer;
1461 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1462 char src0[50];
1463 char dst_str[8];
1464
1465 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1466 shader_addline(buffer, "MOV TMP, 0.0;\n");
1467 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1468
1469 sprintf(dst_str, "T%u", sampler_idx);
1470 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1471 }
1472
1473 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1474 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1475 static void pshader_hw_texdp3(const SHADER_OPCODE_ARG *arg)
1476 {
1477 char src0[50];
1478 char dst_str[50];
1479 char dst_mask[6];
1480 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1481 SHADER_BUFFER* buffer = arg->buffer;
1482
1483 /* Handle output register */
1484 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1485 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1486
1487 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1488 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1489
1490 /* TODO: Handle output modifiers */
1491 }
1492
1493 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1494 * Perform the 3rd row of a 3x3 matrix multiply */
1495 static void pshader_hw_texm3x3(const SHADER_OPCODE_ARG *arg)
1496 {
1497 SHADER_BUFFER* buffer = arg->buffer;
1498 char dst_str[50];
1499 char dst_mask[6];
1500 char src0[50];
1501 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1502
1503 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1504 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1505
1506 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1507 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1508 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1509
1510 /* TODO: Handle output modifiers */
1511 }
1512
1513 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1514 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1515 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1516 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1517 */
1518 static void pshader_hw_texm3x2depth(const SHADER_OPCODE_ARG *arg)
1519 {
1520 SHADER_BUFFER* buffer = arg->buffer;
1521 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1522 char src0[50];
1523
1524 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1525 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1526
1527 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1528 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1529 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1530 */
1531 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1532 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1533 shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1534 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1535 }
1536
1537 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1538 Vertex/Pixel shaders to ARB_vertex_program codes */
1539 static void shader_hw_mnxn(const SHADER_OPCODE_ARG *arg)
1540 {
1541 int i;
1542 int nComponents = 0;
1543 SHADER_OPCODE_ARG tmpArg;
1544 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1545 const SHADER_OPCODE *opcode_table = shader->baseShader.shader_ins;
1546 DWORD shader_version = arg->reg_maps->shader_version;
1547
1548 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1549
1550 /* Set constants for the temporary argument */
1551 tmpArg.shader = arg->shader;
1552 tmpArg.buffer = arg->buffer;
1553 tmpArg.src[0] = arg->src[0];
1554 tmpArg.src_addr[0] = arg->src_addr[0];
1555 tmpArg.src_addr[1] = arg->src_addr[1];
1556 tmpArg.reg_maps = arg->reg_maps;
1557
1558 switch(arg->opcode->opcode) {
1559 case WINED3DSIO_M4x4:
1560 nComponents = 4;
1561 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1562 break;
1563 case WINED3DSIO_M4x3:
1564 nComponents = 3;
1565 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1566 break;
1567 case WINED3DSIO_M3x4:
1568 nComponents = 4;
1569 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1570 break;
1571 case WINED3DSIO_M3x3:
1572 nComponents = 3;
1573 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1574 break;
1575 case WINED3DSIO_M3x2:
1576 nComponents = 2;
1577 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1578 break;
1579 default:
1580 break;
1581 }
1582
1583 for (i = 0; i < nComponents; i++) {
1584 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1585 tmpArg.src[1] = arg->src[1]+i;
1586 shader_hw_map2gl(&tmpArg);
1587 }
1588 }
1589
1590 static void vshader_hw_rsq_rcp(const SHADER_OPCODE_ARG *arg)
1591 {
1592 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1593 SHADER_BUFFER* buffer = arg->buffer;
1594 DWORD dst = arg->dst;
1595 DWORD src = arg->src[0];
1596 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1597 const char *instruction;
1598
1599 char tmpLine[256];
1600
1601 switch(curOpcode->opcode)
1602 {
1603 case WINED3DSIO_RSQ: instruction = "RSQ"; break;
1604 case WINED3DSIO_RCP: instruction = "RCP"; break;
1605 default: instruction = "";
1606 FIXME("Unhandled opcode %s\n", curOpcode->name);
1607 break;
1608 }
1609
1610 strcpy(tmpLine, instruction);
1611 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1612 strcat(tmpLine, ",");
1613 vshader_program_add_param(arg, src, TRUE, tmpLine);
1614 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1615 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1616 * .w is used
1617 */
1618 strcat(tmpLine, ".w");
1619 }
1620
1621 shader_addline(buffer, "%s;\n", tmpLine);
1622 }
1623
1624 static void shader_hw_nrm(const SHADER_OPCODE_ARG *arg)
1625 {
1626 SHADER_BUFFER* buffer = arg->buffer;
1627 char dst_name[50];
1628 char src_name[50];
1629 char dst_wmask[20];
1630 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1631 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1632
1633 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1634 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1635
1636 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1637 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1638 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1639 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1640 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1641 src_name);
1642
1643 if (shift != 0)
1644 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1645 }
1646
1647 static void shader_hw_sincos(const SHADER_OPCODE_ARG *arg)
1648 {
1649 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1650 * must contain fixed constants. So we need a separate function to filter those constants and
1651 * can't use map2gl
1652 */
1653 SHADER_BUFFER* buffer = arg->buffer;
1654 char dst_name[50];
1655 char src_name[50];
1656 char dst_wmask[20];
1657 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1658 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1659
1660 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1661 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1662
1663 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1664 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1665 src_name);
1666
1667 if (shift != 0)
1668 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1669
1670 }
1671
1672 static GLuint create_arb_blt_vertex_program(const WineD3D_GL_Info *gl_info)
1673 {
1674 GLuint program_id = 0;
1675 const char *blt_vprogram =
1676 "!!ARBvp1.0\n"
1677 "PARAM c[1] = { { 1, 0.5 } };\n"
1678 "MOV result.position, vertex.position;\n"
1679 "MOV result.color, c[0].x;\n"
1680 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1681 "END\n";
1682
1683 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1684 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1685 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1686
1687 if (glGetError() == GL_INVALID_OPERATION) {
1688 GLint pos;
1689 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1690 FIXME("Vertex program error at position %d: %s\n", pos,
1691 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1692 }
1693
1694 return program_id;
1695 }
1696
1697 static GLuint create_arb_blt_fragment_program(const WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1698 {
1699 GLuint program_id = 0;
1700 static const char * const blt_fprograms[tex_type_count] =
1701 {
1702 /* tex_1d */
1703 NULL,
1704 /* tex_2d */
1705 "!!ARBfp1.0\n"
1706 "TEMP R0;\n"
1707 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1708 "MOV result.depth.z, R0.x;\n"
1709 "END\n",
1710 /* tex_3d */
1711 NULL,
1712 /* tex_cube */
1713 "!!ARBfp1.0\n"
1714 "TEMP R0;\n"
1715 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1716 "MOV result.depth.z, R0.x;\n"
1717 "END\n",
1718 /* tex_rect */
1719 "!!ARBfp1.0\n"
1720 "TEMP R0;\n"
1721 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1722 "MOV result.depth.z, R0.x;\n"
1723 "END\n",
1724 };
1725
1726 if (!blt_fprograms[tex_type])
1727 {
1728 FIXME("tex_type %#x not supported\n", tex_type);
1729 tex_type = tex_2d;
1730 }
1731
1732 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1733 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1734 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1735
1736 if (glGetError() == GL_INVALID_OPERATION) {
1737 GLint pos;
1738 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1739 FIXME("Fragment program error at position %d: %s\n", pos,
1740 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1741 }
1742
1743 return program_id;
1744 }
1745
1746 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1747 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1748 struct shader_arb_priv *priv = This->shader_priv;
1749 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1750
1751 if (useVS) {
1752 struct vs_compile_args compile_args;
1753
1754 TRACE("Using vertex shader\n");
1755 find_vs_compile_args((IWineD3DVertexShaderImpl *) This->stateBlock->vertexShader, This->stateBlock, &compile_args);
1756 priv->current_vprogram_id = find_gl_vshader((IWineD3DVertexShaderImpl *) This->stateBlock->vertexShader, &compile_args);
1757
1758 /* Bind the vertex program */
1759 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1760 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
1761
1762 /* Enable OpenGL vertex programs */
1763 glEnable(GL_VERTEX_PROGRAM_ARB);
1764 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1765 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1766 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1767 priv->current_vprogram_id = 0;
1768 glDisable(GL_VERTEX_PROGRAM_ARB);
1769 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1770 }
1771
1772 if (usePS) {
1773 struct ps_compile_args compile_args;
1774 TRACE("Using pixel shader\n");
1775 find_ps_compile_args((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader, This->stateBlock, &compile_args);
1776 priv->current_fprogram_id = find_gl_pshader((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader,
1777 &compile_args);
1778
1779 /* Bind the fragment program */
1780 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1781 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
1782
1783 if(!priv->use_arbfp_fixed_func) {
1784 /* Enable OpenGL fragment programs */
1785 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1786 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1787 }
1788 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1789 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1790 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1791 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1792 * replacement shader
1793 */
1794 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1795 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1796 priv->current_fprogram_id = 0;
1797 }
1798 }
1799
1800 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1801 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1802 struct shader_arb_priv *priv = This->shader_priv;
1803 GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1804 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1805
1806 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1807 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1808 glEnable(GL_VERTEX_PROGRAM_ARB);
1809
1810 if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1811 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1812 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1813 }
1814
1815 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1816 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1817 struct shader_arb_priv *priv = This->shader_priv;
1818 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1819
1820 if (priv->current_vprogram_id) {
1821 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1822 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1823
1824 glEnable(GL_VERTEX_PROGRAM_ARB);
1825 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1826
1827 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1828 } else {
1829 glDisable(GL_VERTEX_PROGRAM_ARB);
1830 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1831 }
1832
1833 if (priv->current_fprogram_id) {
1834 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1835 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1836
1837 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1838 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1839
1840 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1841 } else {
1842 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1843 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1844 }
1845 }
1846
1847 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1848 IWineD3DBaseShaderImpl *baseShader = (IWineD3DBaseShaderImpl *) iface;
1849 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)baseShader->baseShader.device)->adapter->gl_info;
1850
1851 if (shader_is_pshader_version(baseShader->baseShader.reg_maps.shader_version))
1852 {
1853 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
1854 UINT i;
1855
1856 ENTER_GL();
1857 for(i = 0; i < This->num_gl_shaders; i++) {
1858 GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1859 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1860 }
1861 LEAVE_GL();
1862 HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1863 This->gl_shaders = NULL;
1864 This->num_gl_shaders = 0;
1865 This->shader_array_size = 0;
1866 } else {
1867 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *) iface;
1868 UINT i;
1869
1870 ENTER_GL();
1871 for(i = 0; i < This->num_gl_shaders; i++) {
1872 GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1873 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1874 }
1875 LEAVE_GL();
1876 HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1877 This->gl_shaders = NULL;
1878 This->num_gl_shaders = 0;
1879 This->shader_array_size = 0;
1880 }
1881 }
1882
1883 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1884 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1885 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1886 return WINED3D_OK;
1887 }
1888
1889 static void shader_arb_free(IWineD3DDevice *iface) {
1890 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1891 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1892 struct shader_arb_priv *priv = This->shader_priv;
1893 int i;
1894
1895 if(priv->depth_blt_vprogram_id) {
1896 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1897 }
1898 for (i = 0; i < tex_type_count; ++i) {
1899 if (priv->depth_blt_fprogram_id[i]) {
1900 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
1901 }
1902 }
1903
1904 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1905 }
1906
1907 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1908 return TRUE;
1909 }
1910
1911 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1912 const char *tmp2, const char *tmp3, const char *tmp4) {
1913 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1914
1915 /* Calculate the > 0.0031308 case */
1916 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1917 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1918 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1919 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1920 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1921 /* Calculate the < case */
1922 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1923 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1924 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1925 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1926 /* Store the components > 0.0031308 in the destination */
1927 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1928 /* Add the components that are < 0.0031308 */
1929 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1930 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1931 }
1932
1933 static GLuint shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer, const struct ps_compile_args *args) {
1934 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1935 const shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1936 CONST DWORD *function = This->baseShader.function;
1937 DWORD shader_version = reg_maps->shader_version;
1938 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1939 const local_constant *lconst;
1940 GLuint retval;
1941 const char *fragcolor;
1942
1943 /* Create the hw ARB shader */
1944 shader_addline(buffer, "!!ARBfp1.0\n");
1945
1946 if (shader_version < WINED3DPS_VERSION(3,0)) {
1947 switch(args->fog) {
1948 case FOG_OFF:
1949 break;
1950 case FOG_LINEAR:
1951 shader_addline(buffer, "OPTION ARB_fog_linear;\n");
1952 break;
1953 case FOG_EXP:
1954 shader_addline(buffer, "OPTION ARB_fog_exp;\n");
1955 break;
1956 case FOG_EXP2:
1957 shader_addline(buffer, "OPTION ARB_fog_exp2;\n");
1958 break;
1959 }
1960 }
1961
1962 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1963 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1964 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1965 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1966 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1967 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1968 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1969 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1970
1971 if (shader_version < WINED3DPS_VERSION(2,0)) {
1972 fragcolor = "R0";
1973 } else {
1974 shader_addline(buffer, "TEMP TMP_COLOR;\n");
1975 fragcolor = "TMP_COLOR";
1976 }
1977
1978 /* Base Declarations */
1979 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1980
1981 /* Base Shader Body */
1982 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1983
1984 if(args->srgb_correction) {
1985 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
1986 }
1987 shader_addline(buffer, "MOV result.color, %s;\n", fragcolor);
1988 shader_addline(buffer, "END\n");
1989
1990 /* TODO: change to resource.glObjectHandle or something like that */
1991 GL_EXTCALL(glGenProgramsARB(1, &retval));
1992
1993 TRACE("Creating a hw pixel shader, prg=%d\n", retval);
1994 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, retval));
1995
1996 TRACE("Created hw pixel shader, prg=%d\n", retval);
1997 /* Create the program and check for errors */
1998 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1999 buffer->bsize, buffer->buffer));
2000
2001 if (glGetError() == GL_INVALID_OPERATION) {
2002 GLint errPos;
2003 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2004 FIXME("HW PixelShader Error at position %d: %s\n",
2005 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2006 retval = 0;
2007 }
2008
2009 /* Load immediate constants */
2010 if(!This->baseShader.load_local_constsF) {
2011 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2012 const float *value = (const float *)lconst->value;
2013 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2014 checkGLcall("glProgramLocalParameter4fvARB");
2015 }
2016 }
2017
2018 return retval;
2019 }
2020
2021 static GLuint shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer, const struct vs_compile_args *args) {
2022 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2023 const shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
2024 CONST DWORD *function = This->baseShader.function;
2025 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2026 const WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2027 const local_constant *lconst;
2028 GLuint ret;
2029
2030 /* Create the hw ARB shader */
2031 shader_addline(buffer, "!!ARBvp1.0\n");
2032 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2033
2034 /* Mesa supports only 95 constants */
2035 if (GL_VEND(MESA) || GL_VEND(WINE))
2036 This->baseShader.limits.constant_float =
2037 min(95, This->baseShader.limits.constant_float);
2038
2039 shader_addline(buffer, "TEMP TMP;\n");
2040
2041 /* Base Declarations */
2042 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2043
2044 /* We need a constant to fixup the final position */
2045 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2046
2047 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2048 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2049 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2050 * a replacement shader depend on the texcoord.w being set properly.
2051 *
2052 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2053 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2054 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2055 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2056 * this can eat a number of instructions, so skip it unless this cap is set as well
2057 */
2058 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2059 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2060
2061 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2062 int i;
2063 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2064 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2065 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2066 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2067 }
2068 }
2069 }
2070 }
2071
2072 /* Base Shader Body */
2073 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2074
2075 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
2076 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
2077 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
2078 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
2079 */
2080 if(args->fog_src == VS_FOG_Z) {
2081 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2082 } else if (!reg_maps->fog) {
2083 shader_addline(buffer, "MOV result.fogcoord, 0.0;\n");
2084 }
2085
2086 /* Write the final position.
2087 *
2088 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2089 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2090 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2091 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2092 */
2093 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2094 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2095 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2096
2097 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2098 * and the glsl equivalent
2099 */
2100 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2101
2102 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2103
2104 shader_addline(buffer, "END\n");
2105
2106 /* TODO: change to resource.glObjectHandle or something like that */
2107 GL_EXTCALL(glGenProgramsARB(1, &ret));
2108
2109 TRACE("Creating a hw vertex shader, prg=%d\n", ret);
2110 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, ret));
2111
2112 TRACE("Created hw vertex shader, prg=%d\n", ret);
2113 /* Create the program and check for errors */
2114 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2115 buffer->bsize, buffer->buffer));
2116
2117 if (glGetError() == GL_INVALID_OPERATION) {
2118 GLint errPos;
2119 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2120 FIXME("HW VertexShader Error at position %d: %s\n",
2121 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2122 ret = -1;
2123 } else {
2124 /* Load immediate constants */
2125 if(!This->baseShader.load_local_constsF) {
2126 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2127 const float *value = (const float *)lconst->value;
2128 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2129 }
2130 }
2131 }
2132 return ret;
2133 }
2134
2135 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
2136 {
2137 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2138 * then overwrite the shader specific ones
2139 */
2140 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2141
2142 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2143 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2144 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2145 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2146 }
2147
2148 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2149 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2150 pCaps->PixelShader1xMaxValue = 8.0;
2151 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2152 }
2153 }
2154
2155 static BOOL shader_arb_color_fixup_supported(struct color_fixup_desc fixup)
2156 {
2157 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
2158 {
2159 TRACE("Checking support for color_fixup:\n");
2160 dump_color_fixup_desc(fixup);
2161 }
2162
2163 /* We support everything except YUV conversions. */
2164 if (!is_yuv_fixup(fixup))
2165 {
2166 TRACE("[OK]\n");
2167 return TRUE;
2168 }
2169
2170 TRACE("[FAILED]\n");
2171 return FALSE;
2172 }
2173
2174 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2175 {
2176 /* WINED3DSIH_ABS */ shader_hw_map2gl,
2177 /* WINED3DSIH_ADD */ shader_hw_map2gl,
2178 /* WINED3DSIH_BEM */ pshader_hw_bem,
2179 /* WINED3DSIH_BREAK */ NULL,
2180 /* WINED3DSIH_BREAKC */ NULL,
2181 /* WINED3DSIH_BREAKP */ NULL,
2182 /* WINED3DSIH_CALL */ NULL,
2183 /* WINED3DSIH_CALLNZ */ NULL,
2184 /* WINED3DSIH_CMP */ pshader_hw_cmp,
2185 /* WINED3DSIH_CND */ pshader_hw_cnd,
2186 /* WINED3DSIH_CRS */ shader_hw_map2gl,
2187 /* WINED3DSIH_DCL */ NULL,
2188 /* WINED3DSIH_DEF */ NULL,
2189 /* WINED3DSIH_DEFB */ NULL,
2190 /* WINED3DSIH_DEFI */ NULL,
2191 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add,
2192 /* WINED3DSIH_DP3 */ shader_hw_map2gl,
2193 /* WINED3DSIH_DP4 */ shader_hw_map2gl,
2194 /* WINED3DSIH_DST */ shader_hw_map2gl,
2195 /* WINED3DSIH_DSX */ NULL,
2196 /* WINED3DSIH_DSY */ NULL,
2197 /* WINED3DSIH_ELSE */ NULL,
2198 /* WINED3DSIH_ENDIF */ NULL,
2199 /* WINED3DSIH_ENDLOOP */ NULL,
2200 /* WINED3DSIH_ENDREP */ NULL,
2201 /* WINED3DSIH_EXP */ shader_hw_map2gl,
2202 /* WINED3DSIH_EXPP */ shader_hw_map2gl,
2203 /* WINED3DSIH_FRC */ shader_hw_map2gl,
2204 /* WINED3DSIH_IF */ NULL,
2205 /* WINED3DSIH_IFC */ NULL,
2206 /* WINED3DSIH_LABEL */ NULL,
2207 /* WINED3DSIH_LIT */ shader_hw_map2gl,
2208 /* WINED3DSIH_LOG */ shader_hw_map2gl,
2209 /* WINED3DSIH_LOGP */ shader_hw_map2gl,
2210 /* WINED3DSIH_LOOP */ NULL,
2211 /* WINED3DSIH_LRP */ shader_hw_map2gl,
2212 /* WINED3DSIH_M3x2 */ shader_hw_mnxn,
2213 /* WINED3DSIH_M3x3 */ shader_hw_mnxn,
2214 /* WINED3DSIH_M3x4 */ shader_hw_mnxn,
2215 /* WINED3DSIH_M4x3 */ shader_hw_mnxn,
2216 /* WINED3DSIH_M4x4 */ shader_hw_mnxn,
2217 /* WINED3DSIH_MAD */ shader_hw_map2gl,
2218 /* WINED3DSIH_MAX */ shader_hw_map2gl,
2219 /* WINED3DSIH_MIN */ shader_hw_map2gl,
2220 /* WINED3DSIH_MOV */ shader_hw_mov,
2221 /* WINED3DSIH_MOVA */ shader_hw_mov,
2222 /* WINED3DSIH_MUL */ shader_hw_map2gl,
2223 /* WINED3DSIH_NOP */ shader_hw_map2gl,
2224 /* WINED3DSIH_NRM */ shader_hw_nrm,
2225 /* WINED3DSIH_PHASE */ NULL,
2226 /* WINED3DSIH_POW */ shader_hw_map2gl,
2227 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp,
2228 /* WINED3DSIH_REP */ NULL,
2229 /* WINED3DSIH_RET */ NULL,
2230 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp,
2231 /* WINED3DSIH_SETP */ NULL,
2232 /* WINED3DSIH_SGE */ shader_hw_map2gl,
2233 /* WINED3DSIH_SGN */ NULL,
2234 /* WINED3DSIH_SINCOS */ shader_hw_sincos,
2235 /* WINED3DSIH_SLT */ shader_hw_map2gl,
2236 /* WINED3DSIH_SUB */ shader_hw_map2gl,
2237 /* WINED3DSIH_TEX */ pshader_hw_tex,
2238 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem,
2239 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem,
2240 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord,
2241 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth,
2242 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3,
2243 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex,
2244 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill,
2245 /* WINED3DSIH_TEXLDD */ NULL,
2246 /* WINED3DSIH_TEXLDL */ NULL,
2247 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth,
2248 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad,
2249 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex,
2250 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3,
2251 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
2252 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad,
2253 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec,
2254 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex,
2255 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec,
2256 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar,
2257 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb,
2258 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb,
2259 };
2260
2261 const shader_backend_t arb_program_shader_backend = {
2262 shader_arb_instruction_handler_table,
2263 shader_arb_select,
2264 shader_arb_select_depth_blt,
2265 shader_arb_deselect_depth_blt,
2266 shader_arb_update_float_vertex_constants,
2267 shader_arb_update_float_pixel_constants,
2268 shader_arb_load_constants,
2269 shader_arb_destroy,
2270 shader_arb_alloc,
2271 shader_arb_free,
2272 shader_arb_dirty_const,
2273 shader_arb_generate_pshader,
2274 shader_arb_generate_vshader,
2275 shader_arb_get_caps,
2276 shader_arb_color_fixup_supported,
2277 };
2278
2279 /* ARB_fragment_program fixed function pipeline replacement definitions */
2280 #define ARB_FFP_CONST_TFACTOR 0
2281 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2282 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2283 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2284 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2285
2286 struct arbfp_ffp_desc
2287 {
2288 struct ffp_frag_desc parent;
2289 GLuint shader;
2290 unsigned int num_textures_used;
2291 };
2292
2293 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2294 if(enable) {
2295 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2296 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2297 } else {
2298 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2299 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2300 }
2301 }
2302
2303 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2304 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2305 struct shader_arb_priv *priv;
2306 /* Share private data between the shader backend and the pipeline replacement, if both
2307 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2308 * if no pixel shader is bound or not
2309 */
2310 if(This->shader_backend == &arb_program_shader_backend) {
2311 This->fragment_priv = This->shader_priv;
2312 } else {
2313 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2314 if(!This->fragment_priv) return E_OUTOFMEMORY;
2315 }
2316 priv = This->fragment_priv;
2317 priv->fragment_shaders = hash_table_create(ffp_frag_program_key_hash, ffp_frag_program_key_compare);
2318 priv->use_arbfp_fixed_func = TRUE;
2319 return WINED3D_OK;
2320 }
2321
2322 static void arbfp_free_ffpshader(void *value, void *gli) {
2323 const WineD3D_GL_Info *gl_info = gli;
2324 struct arbfp_ffp_desc *entry_arb = value;
2325
2326 ENTER_GL();
2327 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2328 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2329 HeapFree(GetProcessHeap(), 0, entry_arb);
2330 LEAVE_GL();
2331 }
2332
2333 static void arbfp_free(IWineD3DDevice *iface) {
2334 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2335 struct shader_arb_priv *priv = This->fragment_priv;
2336
2337 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2338 priv->use_arbfp_fixed_func = FALSE;
2339
2340 if(This->shader_backend != &arb_program_shader_backend) {
2341 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2342 }
2343 }
2344
2345 static void arbfp_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct fragment_caps *caps)
2346 {
2347 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2348 WINED3DTEXOPCAPS_SELECTARG1 |
2349 WINED3DTEXOPCAPS_SELECTARG2 |
2350 WINED3DTEXOPCAPS_MODULATE4X |
2351 WINED3DTEXOPCAPS_MODULATE2X |
2352 WINED3DTEXOPCAPS_MODULATE |
2353 WINED3DTEXOPCAPS_ADDSIGNED2X |
2354 WINED3DTEXOPCAPS_ADDSIGNED |
2355 WINED3DTEXOPCAPS_ADD |
2356 WINED3DTEXOPCAPS_SUBTRACT |
2357 WINED3DTEXOPCAPS_ADDSMOOTH |
2358 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2359 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2360 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2361 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2362 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2363 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2364 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2365 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2366 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2367 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2368 WINED3DTEXOPCAPS_MULTIPLYADD |
2369 WINED3DTEXOPCAPS_LERP |
2370 WINED3DTEXOPCAPS_BUMPENVMAP |
2371 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2372
2373 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2374
2375 caps->MaxTextureBlendStages = 8;
2376 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2377
2378 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2379 }
2380 #undef GLINFO_LOCATION
2381
2382 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2383 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2384 float col[4];
2385 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2386
2387 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2388 * application provided constants
2389 */
2390 if(device->shader_backend == &arb_program_shader_backend) {
2391 if (use_ps(stateblock)) return;
2392
2393 device = stateblock->wineD3DDevice;
2394 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2395 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2396 }
2397
2398 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2399 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2400 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2401
2402 }
2403
2404 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2405 float col[4];
2406 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2407
2408 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2409 * application provided constants
2410 */
2411 if(device->shader_backend == &arb_program_shader_backend) {
2412 if (use_ps(stateblock)) return;
2413
2414 device = stateblock->wineD3DDevice;
2415 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2416 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2417 }
2418
2419 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2420 /* The specular color has no alpha */
2421 col[0] = 1.0; col[1] = 1.0;
2422 col[2] = 1.0; col[3] = 0.0;
2423 } else {
2424 col[0] = 0.0; col[1] = 0.0;
2425 col[2] = 0.0; col[3] = 0.0;
2426 }
2427 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2428 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2429 }
2430
2431 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2432 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE + 1);
2433 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2434 float mat[2][2];
2435
2436 if (use_ps(stateblock))
2437 {
2438 if(stage != 0 &&
2439 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2440 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2441 * anyway
2442 */
2443 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2444 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2445 }
2446 }
2447
2448 if(device->shader_backend == &arb_program_shader_backend) {
2449 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2450 return;
2451 }
2452 } else if(device->shader_backend == &arb_program_shader_backend) {
2453 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2454 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2455 }
2456
2457 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2458 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2459 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2460 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2461
2462 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2463 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2464 }
2465
2466 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2467 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE + 1);
2468 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2469 float param[4];
2470
2471 if (use_ps(stateblock))
2472 {
2473 if(stage != 0 &&
2474 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2475 /* The pixel shader has to know the luminance offset. Do a constants update if it
2476 * isn't scheduled anyway
2477 */
2478 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2479 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2480 }
2481 }
2482
2483 if(device->shader_backend == &arb_program_shader_backend) {
2484 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2485 return;
2486 }
2487 } else if(device->shader_backend == &arb_program_shader_backend) {
2488 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2489 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2490 }
2491
2492 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2493 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2494 param[2] = 0.0;
2495 param[3] = 0.0;
2496
2497 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2498 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2499 }
2500
2501 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2502 const char *ret;
2503
2504 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2505
2506 switch(arg & WINED3DTA_SELECTMASK) {
2507 case WINED3DTA_DIFFUSE:
2508 ret = "fragment.color.primary"; break;
2509
2510 case WINED3DTA_CURRENT:
2511 if(stage == 0) ret = "fragment.color.primary";
2512 else ret = "ret";
2513 break;
2514
2515 case WINED3DTA_TEXTURE:
2516 switch(stage) {
2517 case 0: ret = "tex0"; break;
2518 case 1: ret = "tex1"; break;
2519 case 2: ret = "tex2"; break;
2520 case 3: ret = "tex3"; break;
2521 case 4: ret = "tex4"; break;
2522 case 5: ret = "tex5"; break;
2523 case 6: ret = "tex6"; break;
2524 case 7: ret = "tex7"; break;
2525 default: ret = "unknown texture";
2526 }
2527 break;
2528
2529 case WINED3DTA_TFACTOR:
2530 ret = "tfactor"; break;
2531
2532 case WINED3DTA_SPECULAR:
2533 ret = "fragment.color.secondary"; break;
2534
2535 case WINED3DTA_TEMP:
2536 ret = "tempreg"; break;
2537
2538 case WINED3DTA_CONSTANT:
2539 FIXME("Implement perstage constants\n");
2540 switch(stage) {
2541 case 0: ret = "const0"; break;
2542 case 1: ret = "const1"; break;
2543 case 2: ret = "const2"; break;
2544 case 3: ret = "const3"; break;
2545 case 4: ret = "const4"; break;
2546 case 5: ret = "const5"; break;
2547 case 6: ret = "const6"; break;
2548 case 7: ret = "const7"; break;
2549 default: ret = "unknown constant";
2550 }
2551 break;
2552
2553 default:
2554 return "unknown";
2555 }
2556
2557 if(arg & WINED3DTA_COMPLEMENT) {
2558 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2559 if(argnum == 0) ret = "arg0";
2560 if(argnum == 1) ret = "arg1";
2561 if(argnum == 2) ret = "arg2";
2562 }
2563 if(arg & WINED3DTA_ALPHAREPLICATE) {
2564 shader_addline(buffer, "MOV arg%u, %s.w;\n", argnum, ret);
2565 if(argnum == 0) ret = "arg0";
2566 if(argnum == 1) ret = "arg1";
2567 if(argnum == 2) ret = "arg2";
2568 }
2569 return ret;
2570 }
2571
2572 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2573 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2574 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2575 unsigned int mul = 1;
2576 BOOL mul_final_dest = FALSE;
2577
2578 if(color && alpha) dstmask = "";
2579 else if(color) dstmask = ".xyz";
2580 else dstmask = ".w";
2581
2582 if(dst == tempreg) dstreg = "tempreg";
2583 else dstreg = "ret";
2584
2585 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2586 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2587 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2588
2589 switch(op) {
2590 case WINED3DTOP_DISABLE:
2591 if(stage == 0) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2592 break;
2593
2594 case WINED3DTOP_SELECTARG2:
2595 arg1 = arg2;
2596 case WINED3DTOP_SELECTARG1:
2597 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2598 break;
2599
2600 case WINED3DTOP_MODULATE4X:
2601 mul = 2;
2602 case WINED3DTOP_MODULATE2X:
2603 mul *= 2;
2604 if(strcmp(dstreg, "result.color") == 0) {
2605 dstreg = "ret";
2606 mul_final_dest = TRUE;
2607 }
2608 case WINED3DTOP_MODULATE:
2609 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2610 break;
2611
2612 case WINED3DTOP_ADDSIGNED2X:
2613 mul = 2;
2614 if(strcmp(dstreg, "result.color") == 0) {
2615 dstreg = "ret";
2616 mul_final_dest = TRUE;
2617 }
2618 case WINED3DTOP_ADDSIGNED:
2619 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2620 arg2 = "arg2";
2621 case WINED3DTOP_ADD:
2622 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2623 break;
2624
2625 case WINED3DTOP_SUBTRACT:
2626 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2627 break;
2628
2629 case WINED3DTOP_ADDSMOOTH:
2630 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2631 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2632 break;
2633
2634 case WINED3DTOP_BLENDCURRENTALPHA:
2635 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2636 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2637 break;
2638 case WINED3DTOP_BLENDFACTORALPHA:
2639 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2640 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2641 break;
2642 case WINED3DTOP_BLENDTEXTUREALPHA:
2643 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2644 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2645 break;
2646 case WINED3DTOP_BLENDDIFFUSEALPHA:
2647 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2648 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2649 break;
2650
2651 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2652 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2653 shader_addline(buffer, "SUB arg0.w, const.x, %s.w;\n", arg0);
2654 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg, dstmask, arg2, arg1);
2655 break;
2656
2657 /* D3DTOP_PREMODULATE ???? */
2658
2659 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2660 shader_addline(buffer, "SUB arg0.w, const.x, %s;\n", arg1);
2661 shader_addline(buffer, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2662 break;
2663 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2664 shader_addline(buffer, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2665 break;
2666 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2667 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2668 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg, dstmask, arg2, arg1);
2669 break;
2670 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2671 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg, dstmask, arg1, arg2, arg1);
2672 break;
2673
2674 case WINED3DTOP_DOTPRODUCT3:
2675 mul = 4;
2676 if(strcmp(dstreg, "result.color") == 0) {
2677 dstreg = "ret";
2678 mul_final_dest = TRUE;
2679 }
2680 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2681 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2682 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2683 break;
2684
2685 case WINED3DTOP_MULTIPLYADD:
2686 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2687 break;
2688
2689 case WINED3DTOP_LERP:
2690 /* The msdn is not quite right here */
2691 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2692 break;
2693
2694 case WINED3DTOP_BUMPENVMAP:
2695 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2696 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2697 break;
2698
2699 default:
2700 FIXME("Unhandled texture op %08x\n", op);
2701 }
2702
2703 if(mul == 2) {
2704 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2705 } else if(mul == 4) {
2706 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2707 }
2708 }
2709
2710 /* The stateblock is passed for GLINFO_LOCATION */
2711 static GLuint gen_arbfp_ffp_shader(const struct ffp_frag_settings *settings, IWineD3DStateBlockImpl *stateblock)
2712 {
2713 unsigned int stage;
2714 SHADER_BUFFER buffer;
2715 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2716 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2717 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2718 const char *textype;
2719 const char *instr, *sat;
2720 char colorcor_dst[8];
2721 GLuint ret;
2722 DWORD arg0, arg1, arg2;
2723 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2724 BOOL op_equal;
2725 const char *final_combiner_src = "ret";
2726
2727 /* Find out which textures are read */
2728 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2729 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2730 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2731 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2732 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2733 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2734 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2735 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2736
2737 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2738 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2739 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2740 bump_used[stage] = TRUE;
2741 tex_read[stage] = TRUE;
2742 }
2743 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2744 bump_used[stage] = TRUE;
2745 tex_read[stage] = TRUE;
2746 luminance_used[stage] = TRUE;
2747 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2748 tfactor_used = TRUE;
2749 }
2750
2751 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2752 tfactor_used = TRUE;
2753 }
2754
2755 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2756 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2757 tempreg_used = TRUE;
2758 }
2759
2760 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2761 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2762 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2763 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2764 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2765 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2766 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2767
2768 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2769 tempreg_used = TRUE;
2770 }
2771 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2772 tfactor_used = TRUE;
2773 }
2774 }
2775
2776 /* Shader header */
2777 shader_buffer_init(&buffer);
2778
2779 shader_addline(&buffer, "!!ARBfp1.0\n");
2780
2781 switch(settings->fog) {
2782 case FOG_OFF: break;
2783 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2784 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2785 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2786 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2787 }
2788
2789 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2790 shader_addline(&buffer, "TEMP TMP;\n");
2791 shader_addline(&buffer, "TEMP ret;\n");
2792 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2793 shader_addline(&buffer, "TEMP arg0;\n");
2794 shader_addline(&buffer, "TEMP arg1;\n");
2795 shader_addline(&buffer, "TEMP arg2;\n");
2796 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2797 if(!tex_read[stage]) continue;
2798 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2799 if(!bump_used[stage]) continue;
2800 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2801 if(!luminance_used[stage]) continue;
2802 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2803 }
2804 if(tfactor_used) {
2805 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2806 }
2807 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2808
2809 if(settings->sRGB_write) {
2810 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2811 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2812 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2813 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2814 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2815 srgb_pow, srgb_pow, srgb_pow);
2816 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2817 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2818 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2819 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2820 }
2821
2822 /* Generate texture sampling instructions) */
2823 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2824 if(!tex_read[stage]) continue;
2825
2826 switch(settings->op[stage].tex_type) {
2827 case tex_1d: textype = "1D"; break;
2828 case tex_2d: textype = "2D"; break;
2829 case tex_3d: textype = "3D"; break;
2830 case tex_cube: textype = "CUBE"; break;
2831 case tex_rect: textype = "RECT"; break;
2832 default: textype = "unexpected_textype"; break;
2833 }
2834
2835 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2836 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2837 sat = "";
2838 } else {
2839 sat = "_SAT";
2840 }
2841
2842 if(settings->op[stage].projected == proj_none) {
2843 instr = "TEX";
2844 } else if(settings->op[stage].projected == proj_count4 ||
2845 settings->op[stage].projected == proj_count3) {
2846 instr = "TXP";
2847 } else {
2848 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2849 instr = "TXP";
2850 }
2851
2852 if(stage > 0 &&
2853 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2854 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2855 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2856 shader_addline(&buffer, "DP3 ret.x, arg1, tex%u;\n", stage - 1);
2857 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2858 shader_addline(&buffer, "DP3 ret.y, arg1, tex%u;\n", stage - 1);
2859
2860 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2861 * so multiply the displacement with the dividing parameter before passing it to TXP
2862 */
2863 if (settings->op[stage].projected != proj_none) {
2864 if(settings->op[stage].projected == proj_count4) {
2865 shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].w;\n", stage);
2866 shader_addline(&buffer, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage, stage);
2867 } else {
2868 shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].z;\n", stage);
2869 shader_addline(&buffer, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage, stage);
2870 }
2871 } else {
2872 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2873 }
2874
2875 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2876 instr, sat, stage, stage, textype);
2877 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2878 shader_addline(&buffer, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
2879 stage - 1, stage - 1, stage - 1);
2880 shader_addline(&buffer, "MUL tex%u, tex%u, ret.x;\n", stage, stage);
2881 }
2882 } else if(settings->op[stage].projected == proj_count3) {
2883 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2884 shader_addline(&buffer, "MOV ret.w, ret.z;\n");
2885 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2886 instr, sat, stage, stage, textype);
2887 } else {
2888 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2889 instr, sat, stage, stage, stage, textype);
2890 }
2891
2892 sprintf(colorcor_dst, "tex%u", stage);
2893 gen_color_correction(&buffer, colorcor_dst, WINED3DSP_WRITEMASK_ALL, "const.x", "const.y",
2894 settings->op[stage].color_fixup);
2895 }
2896
2897 /* Generate the main shader */
2898 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2899 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2900 if(stage == 0) {
2901 final_combiner_src = "fragment.color.primary";
2902 }
2903 break;
2904 }
2905
2906 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2907 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2908 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2909 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2910 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2911 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2912 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2913 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2914 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2915 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2916 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2917 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2918 } else {
2919 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
2920 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2921 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2922 settings->op[stage].carg2 == settings->op[stage].aarg2;
2923 }
2924
2925 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2926 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2927 settings->op[stage].cop, settings->op[stage].carg0,
2928 settings->op[stage].carg1, settings->op[stage].carg2);
2929 if(stage == 0) {
2930 shader_addline(&buffer, "MOV ret.w, fragment.color.primary.w;\n");
2931 }
2932 } else if(op_equal) {
2933 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2934 settings->op[stage].cop, settings->op[stage].carg0,
2935 settings->op[stage].carg1, settings->op[stage].carg2);
2936 } else {
2937 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2938 settings->op[stage].cop, settings->op[stage].carg0,
2939 settings->op[stage].carg1, settings->op[stage].carg2);
2940 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
2941 settings->op[stage].aop, settings->op[stage].aarg0,
2942 settings->op[stage].aarg1, settings->op[stage].aarg2);
2943 }
2944 }
2945
2946 if(settings->sRGB_write) {
2947 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2948 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2949 shader_addline(&buffer, "MOV result.color.w, ret.w;\n");
2950 } else {
2951 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2952 }
2953
2954 /* Footer */
2955 shader_addline(&buffer, "END\n");
2956
2957 /* Generate the shader */
2958 GL_EXTCALL(glGenProgramsARB(1, &ret));
2959 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2960 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2961
2962 if (glGetError() == GL_INVALID_OPERATION) {
2963 GLint pos;
2964 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2965 FIXME("Fragment program error at position %d: %s\n", pos,
2966 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2967 }
2968 shader_buffer_free(&buffer);
2969 return ret;
2970 }
2971
2972 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2973 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2974 struct shader_arb_priv *priv = device->fragment_priv;
2975 BOOL use_pshader = use_ps(stateblock);
2976 BOOL use_vshader = use_vs(stateblock);
2977 struct ffp_frag_settings settings;
2978 const struct arbfp_ffp_desc *desc;
2979 unsigned int i;
2980
2981 TRACE("state %#x, stateblock %p, context %p\n", state, stateblock, context);
2982
2983 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2984 if(!use_pshader && device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2985 /* Reload fixed function constants since they collide with the pixel shader constants */
2986 for(i = 0; i < MAX_TEXTURES; i++) {
2987 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2988 }
2989 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2990 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2991 } else if(use_pshader && !isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2992 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2993 }
2994 return;
2995 }
2996
2997 if(!use_pshader) {
2998 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2999 gen_ffp_frag_op(stateblock, &settings, FALSE);
3000 desc = (const struct arbfp_ffp_desc *)find_ffp_frag_shader(priv->fragment_shaders, &settings);
3001 if(!desc) {
3002 struct arbfp_ffp_desc *new_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc));
3003 if (!new_desc)
3004 {
3005 ERR("Out of memory\n");
3006 return;
3007 }
3008 new_desc->num_textures_used = 0;
3009 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
3010 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
3011 new_desc->num_textures_used = i;
3012 }
3013
3014 memcpy(&new_desc->parent.settings, &settings, sizeof(settings));
3015 new_desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
3016 add_ffp_frag_shader(priv->fragment_shaders, &new_desc->parent);
3017 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc);
3018 desc = new_desc;
3019 }
3020
3021 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3022 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3023 * deactivate it.
3024 */
3025 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
3026 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3027 priv->current_fprogram_id = desc->shader;
3028
3029 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3030 /* Reload fixed function constants since they collide with the pixel shader constants */
3031 for(i = 0; i < MAX_TEXTURES; i++) {
3032 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3033 }
3034 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3035 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3036 }
3037 context->last_was_pshader = FALSE;
3038 } else {
3039 context->last_was_pshader = TRUE;
3040 }
3041
3042 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3043 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3044 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3045 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3046 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3047 *
3048 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3049 * to be compiled before activating them(needs some cleanups in the shader backend interface)
3050 */
3051 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
3052 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
3053
3054 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
3055 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
3056 }
3057 }
3058 if(use_pshader) {
3059 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
3060 }
3061 }
3062
3063 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3064 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3065 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3066 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3067 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3068 */
3069 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3070 enum fogsource new_source;
3071
3072 TRACE("state %#x, stateblock %p, context %p\n", state, stateblock, context);
3073
3074 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3075 fragment_prog_arbfp(state, stateblock, context);
3076 }
3077
3078 if(!stateblock->renderState[WINED3DRS_FOGENABLE]) return;
3079
3080 if(stateblock->renderState[WINED3DRS_FOGTABLEMODE] == WINED3DFOG_NONE) {
3081 if(use_vs(stateblock)) {
3082 new_source = FOGSOURCE_VS;
3083 } else {
3084 if(stateblock->renderState[WINED3DRS_FOGVERTEXMODE] == WINED3DFOG_NONE || context->last_was_rhw) {
3085 new_source = FOGSOURCE_COORD;
3086 } else {
3087 new_source = FOGSOURCE_FFP;
3088 }
3089 }
3090 } else {
3091 new_source = FOGSOURCE_FFP;
3092 }
3093 if(new_source != context->fog_source) {
3094 context->fog_source = new_source;
3095 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART), stateblock, context);
3096 }
3097 }
3098
3099 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3100 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3101 fragment_prog_arbfp(state, stateblock, context);
3102 }
3103 }
3104
3105 #undef GLINFO_LOCATION
3106
3107 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3108 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3109 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3110 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3111 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3112 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3113 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3114 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3115 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3116 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3117 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3118 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3119 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3120 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3121 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3122 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3123 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3124 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3125 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3126 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3127 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3128 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3129 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3130 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3131 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3132 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3133 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3134 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3135 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3136 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3137 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3138 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3139 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3140 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3141 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3142 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3143 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3144 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3145 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3146 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3147 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3148 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3149 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3150 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3151 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3152 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3153 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3154 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3155 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3156 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3157 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3158 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3159 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3160 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3161 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3162 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3163 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3164 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3165 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3166 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3167 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3168 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3169 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3170 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3171 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3172 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3173 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3174 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3175 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3176 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3177 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3178 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3179 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3180 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3181 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3182 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3183 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3184 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3185 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3186 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3187 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3188 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3189 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3190 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3191 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3192 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3193 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3194 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3195 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3196 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3197 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3198 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3199 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3200 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3201 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3202 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3203 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3204 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3205 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3206 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3207 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3208 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3209 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3210 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3211 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3212 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3213 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3214 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3215 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3216 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3217 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3218 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3219 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3220 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3221 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3222 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3223 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3224 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3225 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3226 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3227 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3228 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3229 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3230 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3231 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3232 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3233 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3234 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3235 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3236 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3237 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3238 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3239 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3240 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3241 { STATE_RENDER(WINED3DRS_FOGSTART), { STATE_RENDER(WINED3DRS_FOGSTART), state_fogstartend }, 0 },
3242 { STATE_RENDER(WINED3DRS_FOGEND), { STATE_RENDER(WINED3DRS_FOGSTART), state_fogstartend }, 0 },
3243 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3244 { STATE_RENDER(WINED3DRS_FOGCOLOR), { STATE_RENDER(WINED3DRS_FOGCOLOR), state_fogcolor }, 0 },
3245 { STATE_RENDER(WINED3DRS_FOGDENSITY), { STATE_RENDER(WINED3DRS_FOGDENSITY), state_fogdensity }, 0 },
3246 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3247 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3248 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3249 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3250 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3251 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3252 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3253 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3254 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3255 {0 /* Terminate */, { 0, 0 }, 0 },
3256 };
3257
3258 const struct fragment_pipeline arbfp_fragment_pipeline = {
3259 arbfp_enable,
3260 arbfp_get_caps,
3261 arbfp_alloc,
3262 arbfp_free,
3263 shader_arb_color_fixup_supported,
3264 arbfp_fragmentstate_template,
3265 TRUE /* We can disable projected textures */
3266 };
3267
3268 #define GLINFO_LOCATION device->adapter->gl_info
3269
3270 struct arbfp_blit_priv {
3271 GLenum yuy2_rect_shader, yuy2_2d_shader;
3272 GLenum uyvy_rect_shader, uyvy_2d_shader;
3273 GLenum yv12_rect_shader, yv12_2d_shader;
3274 };
3275
3276 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3277 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3278 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3279 if(!device->blit_priv) {
3280 ERR("Out of memory\n");
3281 return E_OUTOFMEMORY;
3282 }
3283 return WINED3D_OK;
3284 }
3285 static void arbfp_blit_free(IWineD3DDevice *iface) {
3286 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3287 struct arbfp_blit_priv *priv = device->blit_priv;
3288
3289 ENTER_GL();
3290 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3291 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3292 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3293 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3294 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3295 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3296 checkGLcall("Delete yuv programs\n");
3297 LEAVE_GL();
3298 }
3299
3300 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, enum yuv_fixup yuv_fixup, GLenum textype, char *luminance)
3301 {
3302 char chroma;
3303 const char *tex, *texinstr;
3304
3305 if (yuv_fixup == YUV_FIXUP_UYVY) {
3306 chroma = 'x';
3307 *luminance = 'w';
3308 } else {
3309 chroma = 'w';
3310 *luminance = 'x';
3311 }
3312 switch(textype) {
3313 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3314 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3315 default:
3316 /* This is more tricky than just replacing the texture type - we have to navigate
3317 * properly in the texture to find the correct chroma values
3318 */
3319 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3320 return FALSE;
3321 }
3322
3323 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3324 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3325 * filtering when we sample the texture.
3326 *
3327 * These are the rules for reading the chroma:
3328 *
3329 * Even pixel: Cr
3330 * Even pixel: U
3331 * Odd pixel: V
3332 *
3333 * So we have to get the sampling x position in non-normalized coordinates in integers
3334 */
3335 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3336 shader_addline(buffer, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
3337 shader_addline(buffer, "MOV texcrd.w, size.x;\n");
3338 } else {
3339 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3340 }
3341 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3342 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3343 * 0.5, so add 0.5.
3344 */
3345 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3346 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3347
3348 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3349 * even and odd pixels respectively
3350 */
3351 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3352 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3353
3354 /* Sample Pixel 1 */
3355 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3356
3357 /* Put the value into either of the chroma values */
3358 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3359 shader_addline(buffer, "MUL chroma.x, luminance.%c, temp.x;\n", chroma);
3360 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3361 shader_addline(buffer, "MUL chroma.y, luminance.%c, temp.x;\n", chroma);
3362
3363 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3364 * the pixel right to the current one. Otherwise, sample the left pixel.
3365 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3366 */
3367 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3368 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3369 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3370
3371 /* Put the value into the other chroma */
3372 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3373 shader_addline(buffer, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma);
3374 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3375 shader_addline(buffer, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma);
3376
3377 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3378 * the current one and lerp the two U and V values
3379 */
3380
3381 /* This gives the correctly filtered luminance value */
3382 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3383
3384 return TRUE;
3385 }
3386
3387 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, GLenum textype, char *luminance)
3388 {
3389 const char *tex;
3390
3391 switch(textype) {
3392 case GL_TEXTURE_2D: tex = "2D"; break;
3393 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3394 default:
3395 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3396 return FALSE;
3397 }
3398
3399 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3400 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3401 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3402 * pitch of the luminance plane, the packing into the gl texture is a bit
3403 * unfortunate. If the whole texture is interpreted as luminance data it looks
3404 * approximately like this:
3405 *
3406 * +----------------------------------+----
3407 * | |
3408 * | |
3409 * | |
3410 * | |
3411 * | | 2
3412 * | LUMINANCE | -
3413 * | | 3
3414 * | |
3415 * | |
3416 * | |
3417 * | |
3418 * +----------------+-----------------+----
3419 * | | |
3420 * | U even rows | U odd rows |
3421 * | | | 1
3422 * +----------------+------------------ -
3423 * | | | 3
3424 * | V even rows | V odd rows |
3425 * | | |
3426 * +----------------+-----------------+----
3427 * | | |
3428 * | 0.5 | 0.5 |
3429 *
3430 * So it appears as if there are 4 chroma images, but in fact the odd rows
3431 * in the chroma images are in the same row as the even ones. So its is
3432 * kinda tricky to read
3433 *
3434 * When reading from rectangle textures, keep in mind that the input y coordinates
3435 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3436 */
3437 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3438 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3439
3440 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3441 /* the chroma planes have only half the width */
3442 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3443
3444 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3445 * the coordinate. Also read the right side of the image when reading odd lines
3446 *
3447 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3448 * bleeding
3449 */
3450 if(textype == GL_TEXTURE_2D) {
3451
3452 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3453
3454 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3455
3456 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3457 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3458
3459 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3460 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3461 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3462 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3463 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3464
3465 /* clamp, keep the half pixel origin in mind */
3466 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3467 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3468 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3469 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3470 } else {
3471 /* Read from [size - size+size/4] */
3472 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3473 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3474
3475 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3476 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3477 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3478 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3479 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3480 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3481
3482 /* Make sure to read exactly from the pixel center */
3483 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3484 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3485
3486 /* Clamp */
3487 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3488 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3489 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3490 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3491 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3492 }
3493 /* Read the texture, put the result into the output register */
3494 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3495 shader_addline(buffer, "MOV chroma.x, temp.w;\n");
3496
3497 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3498 * No need to clamp because we're just reusing the already clamped value from above
3499 */
3500 if(textype == GL_TEXTURE_2D) {
3501 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3502 } else {
3503 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3504 }
3505 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3506 shader_addline(buffer, "MOV chroma.y, temp.w;\n");
3507
3508 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3509 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3510 * values due to filtering
3511 */
3512 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3513 if(textype == GL_TEXTURE_2D) {
3514 /* Multiply the y coordinate by 2/3 and clamp it */
3515 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3516 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3517 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3518 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3519 } else {
3520 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
3521 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3522 * is bigger
3523 */
3524 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3525 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3526 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3527 }
3528 *luminance = 'a';
3529
3530 return TRUE;
3531 }
3532
3533 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, enum yuv_fixup yuv_fixup, GLenum textype)
3534 {
3535 GLenum shader;
3536 SHADER_BUFFER buffer;
3537 char luminance_component;
3538 struct arbfp_blit_priv *priv = device->blit_priv;
3539
3540 /* Shader header */
3541 shader_buffer_init(&buffer);
3542
3543 ENTER_GL();
3544 GL_EXTCALL(glGenProgramsARB(1, &shader));
3545 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3546 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3547 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3548 LEAVE_GL();
3549 if(!shader) {
3550 shader_buffer_free(&buffer);
3551 return 0;
3552 }
3553
3554 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3555 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3556 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3557 * each single pixel it contains, and one U and one V value shared between both
3558 * pixels.
3559 *
3560 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3561 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3562 * take the format into account when generating the read swizzles
3563 *
3564 * Reading the Y value is straightforward - just sample the texture. The hardware
3565 * takes care of filtering in the horizontal and vertical direction.
3566 *
3567 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3568 * because that would mix the U and V values of one pixel or two adjacent pixels.
3569 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3570 * regardless of the filtering setting. Vertical filtering works automatically
3571 * though - the U and V values of two rows are mixed nicely.
3572 *
3573 * Appart of avoiding filtering issues, the code has to know which value it just
3574 * read, and where it can find the other one. To determine this, it checks if
3575 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3576 *
3577 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3578 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3579 *
3580 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3581 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3582 * in an unfiltered situation. Finding the luminance on the other hand requires
3583 * finding out if it is an odd or even pixel. The real drawback of this approach
3584 * is filtering. This would have to be emulated completely in the shader, reading
3585 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3586 * vertically. Beyond that it would require adjustments to the texture handling
3587 * code to deal with the width scaling
3588 */
3589 shader_addline(&buffer, "!!ARBfp1.0\n");
3590 shader_addline(&buffer, "TEMP luminance;\n");
3591 shader_addline(&buffer, "TEMP temp;\n");
3592 shader_addline(&buffer, "TEMP chroma;\n");
3593 shader_addline(&buffer, "TEMP texcrd;\n");
3594 shader_addline(&buffer, "TEMP texcrd2;\n");
3595 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3596 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3597 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3598
3599 switch (yuv_fixup)
3600 {
3601 case YUV_FIXUP_UYVY:
3602 case YUV_FIXUP_YUY2:
3603 if (!gen_planar_yuv_read(&buffer, yuv_fixup, textype, &luminance_component))
3604 {
3605 shader_buffer_free(&buffer);
3606 return 0;
3607 }
3608 break;
3609
3610 case YUV_FIXUP_YV12:
3611 if (!gen_yv12_read(&buffer, textype, &luminance_component))
3612 {
3613 shader_buffer_free(&buffer);
3614 return 0;
3615 }
3616 break;
3617
3618 default:
3619 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3620 shader_buffer_free(&buffer);
3621 return 0;
3622 }
3623
3624 /* Calculate the final result. Formula is taken from
3625 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3626 * ranges from -0.5 to 0.5
3627 */
3628 shader_addline(&buffer, "SUB chroma.xy, chroma, coef.y;\n");
3629
3630 shader_addline(&buffer, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component);
3631 shader_addline(&buffer, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component);
3632 shader_addline(&buffer, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
3633 shader_addline(&buffer, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component);
3634 shader_addline(&buffer, "END\n");
3635
3636 ENTER_GL();
3637 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3638
3639 if (glGetError() == GL_INVALID_OPERATION) {
3640 GLint pos;
3641 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3642 FIXME("Fragment program error at position %d: %s\n", pos,
3643 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3644 }
3645 shader_buffer_free(&buffer);
3646 LEAVE_GL();
3647
3648 switch (yuv_fixup)
3649 {
3650 case YUV_FIXUP_YUY2:
3651 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yuy2_rect_shader = shader;
3652 else priv->yuy2_2d_shader = shader;
3653 break;
3654
3655 case YUV_FIXUP_UYVY:
3656 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->uyvy_rect_shader = shader;
3657 else priv->uyvy_2d_shader = shader;
3658 break;
3659
3660 case YUV_FIXUP_YV12:
3661 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yv12_rect_shader = shader;
3662 else priv->yv12_2d_shader = shader;
3663 break;
3664 }
3665
3666 return shader;
3667 }
3668
3669 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3670 GLenum shader;
3671 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3672 float size[4] = {width, height, 1, 1};
3673 struct arbfp_blit_priv *priv = device->blit_priv;
3674 const struct GlPixelFormatDesc *glDesc;
3675 enum yuv_fixup yuv_fixup;
3676
3677 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3678
3679 if (!is_yuv_fixup(glDesc->color_fixup))
3680 {
3681 TRACE("Fixup:\n");
3682 dump_color_fixup_desc(glDesc->color_fixup);
3683 /* Don't bother setting up a shader for unconverted formats */
3684 ENTER_GL();
3685 glEnable(textype);
3686 checkGLcall("glEnable(textype)");
3687 LEAVE_GL();
3688 return WINED3D_OK;
3689 }
3690
3691 yuv_fixup = get_yuv_fixup(glDesc->color_fixup);
3692
3693 switch(yuv_fixup)
3694 {
3695 case YUV_FIXUP_YUY2:
3696 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yuy2_rect_shader : priv->yuy2_2d_shader;
3697 break;
3698
3699 case YUV_FIXUP_UYVY:
3700 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->uyvy_rect_shader : priv->uyvy_2d_shader;
3701 break;
3702
3703 case YUV_FIXUP_YV12:
3704 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yv12_rect_shader : priv->yv12_2d_shader;
3705 break;
3706
3707 default:
3708 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup);
3709 ENTER_GL();
3710 glEnable(textype);
3711 checkGLcall("glEnable(textype)");
3712 LEAVE_GL();
3713 return E_NOTIMPL;
3714 }
3715
3716 if (!shader) shader = gen_yuv_shader(device, yuv_fixup, textype);
3717
3718 ENTER_GL();
3719 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3720 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3721 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3722 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3723 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3724 checkGLcall("glProgramLocalParameter4fvARB");
3725 LEAVE_GL();
3726
3727 return WINED3D_OK;
3728 }
3729
3730 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3731 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3732
3733 ENTER_GL();
3734 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3735 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3736 glDisable(GL_TEXTURE_2D);
3737 checkGLcall("glDisable(GL_TEXTURE_2D)");
3738 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3739 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3740 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3741 }
3742 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3743 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3744 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3745 }
3746 LEAVE_GL();
3747 }
3748
3749 static BOOL arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup)
3750 {
3751 enum yuv_fixup yuv_fixup;
3752
3753 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
3754 {
3755 TRACE("Checking support for fixup:\n");
3756 dump_color_fixup_desc(fixup);
3757 }
3758
3759 if (is_identity_fixup(fixup))
3760 {
3761 TRACE("[OK]\n");
3762 return TRUE;
3763 }
3764
3765 /* We only support YUV conversions. */
3766 if (!is_yuv_fixup(fixup))
3767 {
3768 TRACE("[FAILED]\n");
3769 return FALSE;
3770 }
3771
3772 yuv_fixup = get_yuv_fixup(fixup);
3773 switch(yuv_fixup)
3774 {
3775 case YUV_FIXUP_YUY2:
3776 case YUV_FIXUP_UYVY:
3777 case YUV_FIXUP_YV12:
3778 TRACE("[OK]\n");
3779 return TRUE;
3780
3781 default:
3782 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3783 TRACE("[FAILED]\n");
3784 return FALSE;
3785 }
3786 }
3787
3788 const struct blit_shader arbfp_blit = {
3789 arbfp_blit_alloc,
3790 arbfp_blit_free,
3791 arbfp_blit_set,
3792 arbfp_blit_unset,
3793 arbfp_blit_color_fixup_supported,
3794 };
3795
3796 #undef GLINFO_LOCATION
useful hacks and other patches not (yet) suitable for mainline