?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Pass the ps_compile_args structures to the shader generation code.
[wine/multimedia.git] / dlls / wined3d / arb_program_shader.c
blob4f37c9be07cf7a8f906d615c07af963312635079
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 != -1) {
216 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
217 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
218 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
219 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
220 */
221 const float *scale = (const float *)&stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
222 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
223 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
224 }
225 }
226 }
227 }
228
229 static void shader_arb_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count)
230 {
231 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
232
233 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
234 * context. On a context switch the old context will be fully dirtified */
235 memset(This->activeContext->vshader_const_dirty + start, 1,
236 sizeof(*This->activeContext->vshader_const_dirty) * count);
237 This->highest_dirty_vs_const = max(This->highest_dirty_vs_const, start + count + 1);
238 }
239
240 static void shader_arb_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count)
241 {
242 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
243
244 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
245 * context. On a context switch the old context will be fully dirtified */
246 memset(This->activeContext->pshader_const_dirty + start, 1,
247 sizeof(*This->activeContext->pshader_const_dirty) * count);
248 This->highest_dirty_ps_const = max(This->highest_dirty_ps_const, start + count + 1);
249 }
250
251 /* Generate the variable & register declarations for the ARB_vertex_program output target */
252 static void shader_generate_arb_declarations(IWineD3DBaseShader *iface, const shader_reg_maps *reg_maps,
253 SHADER_BUFFER *buffer, const WineD3D_GL_Info *gl_info)
254 {
255 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
256 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
257 DWORD i, cur;
258 char pshader = shader_is_pshader_version(reg_maps->shader_version);
259 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
260 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
261 UINT extra_constants_needed = 0;
262 const local_constant *lconst;
263
264 /* Temporary Output register */
265 shader_addline(buffer, "TEMP TMP_OUT;\n");
266
267 for(i = 0; i < This->baseShader.limits.temporary; i++) {
268 if (reg_maps->temporary[i])
269 shader_addline(buffer, "TEMP R%u;\n", i);
270 }
271
272 for (i = 0; i < This->baseShader.limits.address; i++) {
273 if (reg_maps->address[i])
274 shader_addline(buffer, "ADDRESS A%d;\n", i);
275 }
276
277 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
278 if (reg_maps->texcoord[i])
279 shader_addline(buffer,"TEMP T%u;\n", i);
280 }
281
282 /* Texture coordinate registers must be pre-loaded */
283 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
284 if (reg_maps->texcoord[i])
285 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
286 }
287
288 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
289 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
290 if(!reg_maps->bumpmat[i]) continue;
291
292 cur = ps->numbumpenvmatconsts;
293 ps->bumpenvmatconst[cur].const_num = -1;
294 ps->bumpenvmatconst[cur].texunit = i;
295 ps->luminanceconst[cur].const_num = -1;
296 ps->luminanceconst[cur].texunit = i;
297
298 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
299 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
300 * bump mapping.
301 */
302 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
303 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
304 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
305 i, ps->bumpenvmatconst[cur].const_num);
306 extra_constants_needed++;
307
308 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
309 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
310 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
311 i, ps->luminanceconst[cur].const_num);
312 extra_constants_needed++;
313 } else if(reg_maps->luminanceparams) {
314 FIXME("No free constant to load the luminance parameters\n");
315 }
316 } else {
317 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
318 }
319
320 ps->numbumpenvmatconsts = cur + 1;
321 }
322
323 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
324 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
325 srgb_mul_low, srgb_mul_low, srgb_mul_low);
326 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
327 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
328 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
329 srgb_pow, srgb_pow, srgb_pow);
330 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
331 srgb_mul_high, srgb_mul_high, srgb_mul_high);
332 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
333 srgb_sub_high, srgb_sub_high, srgb_sub_high);
334 }
335
336 /* Load local constants using the program-local space,
337 * this avoids reloading them each time the shader is used
338 */
339 if(!This->baseShader.load_local_constsF) {
340 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
341 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
342 lconst->idx);
343 }
344 }
345
346 /* we use the array-based constants array if the local constants are marked for loading,
347 * because then we use indirect addressing, or when the local constant list is empty,
348 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
349 * local constants do not declare the loaded constants as an array because ARB compilers usually
350 * do not optimize unused constants away
351 */
352 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
353 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
354 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
355 max_constantsF, max_constantsF - 1);
356 } else {
357 for(i = 0; i < max_constantsF; i++) {
358 if(!shader_constant_is_local(This, i)) {
359 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
360 }
361 }
362 }
363 }
364
365 static const char * const shift_tab[] = {
366 "dummy", /* 0 (none) */
367 "coefmul.x", /* 1 (x2) */
368 "coefmul.y", /* 2 (x4) */
369 "coefmul.z", /* 3 (x8) */
370 "coefmul.w", /* 4 (x16) */
371 "dummy", /* 5 (x32) */
372 "dummy", /* 6 (x64) */
373 "dummy", /* 7 (x128) */
374 "dummy", /* 8 (d256) */
375 "dummy", /* 9 (d128) */
376 "dummy", /* 10 (d64) */
377 "dummy", /* 11 (d32) */
378 "coefdiv.w", /* 12 (d16) */
379 "coefdiv.z", /* 13 (d8) */
380 "coefdiv.y", /* 14 (d4) */
381 "coefdiv.x" /* 15 (d2) */
382 };
383
384 static void shader_arb_get_write_mask(const SHADER_OPCODE_ARG *arg, const DWORD param, char *write_mask)
385 {
386 char *ptr = write_mask;
387 char vshader = shader_is_vshader_version(arg->reg_maps->shader_version);
388
389 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
390 *ptr++ = '.';
391 *ptr++ = 'x';
392 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
393 *ptr++ = '.';
394 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
395 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
396 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
397 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
398 }
399
400 *ptr = '\0';
401 }
402
403 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
404 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
405 * but addressed as "rgba". To fix this we need to swap the register's x
406 * and z components. */
407 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
408 char *ptr = swizzle_str;
409
410 /* swizzle bits fields: wwzzyyxx */
411 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
412 DWORD swizzle_x = swizzle & 0x03;
413 DWORD swizzle_y = (swizzle >> 2) & 0x03;
414 DWORD swizzle_z = (swizzle >> 4) & 0x03;
415 DWORD swizzle_w = (swizzle >> 6) & 0x03;
416
417 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
418 * generate a swizzle string. Unless we need to our own swizzling. */
419 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
420 *ptr++ = '.';
421 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
422 *ptr++ = swizzle_chars[swizzle_x];
423 } else {
424 *ptr++ = swizzle_chars[swizzle_x];
425 *ptr++ = swizzle_chars[swizzle_y];
426 *ptr++ = swizzle_chars[swizzle_z];
427 *ptr++ = swizzle_chars[swizzle_w];
428 }
429 }
430
431 *ptr = '\0';
432 }
433
434 static void pshader_get_register_name(IWineD3DBaseShader* iface,
435 const DWORD param, char* regstr) {
436
437 DWORD reg = param & WINED3DSP_REGNUM_MASK;
438 DWORD regtype = shader_get_regtype(param);
439 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
440
441 switch (regtype) {
442 case WINED3DSPR_TEMP:
443 sprintf(regstr, "R%u", reg);
444 break;
445 case WINED3DSPR_INPUT:
446 if (reg==0) {
447 strcpy(regstr, "fragment.color.primary");
448 } else {
449 strcpy(regstr, "fragment.color.secondary");
450 }
451 break;
452 case WINED3DSPR_CONST:
453 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
454 sprintf(regstr, "C[%u]", reg);
455 } else {
456 sprintf(regstr, "C%u", reg);
457 }
458 break;
459 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
460 sprintf(regstr,"T%u", reg);
461 break;
462 case WINED3DSPR_COLOROUT:
463 if (reg == 0)
464 sprintf(regstr, "TMP_COLOR");
465 else {
466 /* TODO: See GL_ARB_draw_buffers */
467 FIXME("Unsupported write to render target %u\n", reg);
468 sprintf(regstr, "unsupported_register");
469 }
470 break;
471 case WINED3DSPR_DEPTHOUT:
472 sprintf(regstr, "result.depth");
473 break;
474 case WINED3DSPR_ATTROUT:
475 sprintf(regstr, "oD[%u]", reg);
476 break;
477 case WINED3DSPR_TEXCRDOUT:
478 sprintf(regstr, "oT[%u]", reg);
479 break;
480 default:
481 FIXME("Unhandled register name Type(%d)\n", regtype);
482 sprintf(regstr, "unrecognized_register");
483 break;
484 }
485 }
486
487 /* TODO: merge with pixel shader */
488 static void vshader_program_add_param(const SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine)
489 {
490 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
491
492 /* oPos, oFog and oPts in D3D */
493 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
494
495 DWORD reg = param & WINED3DSP_REGNUM_MASK;
496 DWORD regtype = shader_get_regtype(param);
497 char tmpReg[255];
498 BOOL is_color = FALSE;
499
500 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
501 strcat(hwLine, " -");
502 } else {
503 strcat(hwLine, " ");
504 }
505
506 switch (regtype) {
507 case WINED3DSPR_TEMP:
508 sprintf(tmpReg, "R%u", reg);
509 strcat(hwLine, tmpReg);
510 break;
511 case WINED3DSPR_INPUT:
512
513 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
514 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 void shader_hw_sample(const SHADER_OPCODE_ARG *arg, DWORD sampler_idx, const char *dst_str,
572 const char *coord_reg, BOOL projected, BOOL bias)
573 {
574 SHADER_BUFFER* buffer = arg->buffer;
575 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
576 const char *tex_type;
577
578 switch(sampler_type) {
579 case WINED3DSTT_1D:
580 tex_type = "1D";
581 break;
582
583 case WINED3DSTT_2D:
584 {
585 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
586 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
587 if(device->stateBlock->textures[sampler_idx] &&
588 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
589 tex_type = "RECT";
590 } else {
591 tex_type = "2D";
592 }
593 break;
594 }
595
596 case WINED3DSTT_VOLUME:
597 tex_type = "3D";
598 break;
599
600 case WINED3DSTT_CUBE:
601 tex_type = "CUBE";
602 break;
603
604 default:
605 ERR("Unexpected texture type %d\n", sampler_type);
606 tex_type = "";
607 }
608
609 if (bias) {
610 /* Shouldn't be possible, but let's check for it */
611 if(projected) FIXME("Biased and Projected texture sampling\n");
612 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
613 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
614 } else if (projected) {
615 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
616 } else {
617 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
618 }
619 }
620
621 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source)
622 {
623 switch(channel_source)
624 {
625 case CHANNEL_SOURCE_ZERO: return "0";
626 case CHANNEL_SOURCE_ONE: return "1";
627 case CHANNEL_SOURCE_X: return "x";
628 case CHANNEL_SOURCE_Y: return "y";
629 case CHANNEL_SOURCE_Z: return "z";
630 case CHANNEL_SOURCE_W: return "w";
631 default:
632 FIXME("Unhandled channel source %#x\n", channel_source);
633 return "undefined";
634 }
635 }
636
637 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, DWORD dst_mask,
638 const char *one, const char *two, struct color_fixup_desc fixup)
639 {
640 DWORD mask;
641
642 if (is_yuv_fixup(fixup))
643 {
644 enum yuv_fixup yuv_fixup = get_yuv_fixup(fixup);
645 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup);
646 return;
647 }
648
649 mask = 0;
650 if (fixup.x_source != CHANNEL_SOURCE_X) mask |= WINED3DSP_WRITEMASK_0;
651 if (fixup.y_source != CHANNEL_SOURCE_Y) mask |= WINED3DSP_WRITEMASK_1;
652 if (fixup.z_source != CHANNEL_SOURCE_Z) mask |= WINED3DSP_WRITEMASK_2;
653 if (fixup.w_source != CHANNEL_SOURCE_W) mask |= WINED3DSP_WRITEMASK_3;
654 mask &= dst_mask;
655
656 if (mask)
657 {
658 shader_addline(buffer, "SWZ %s, %s, %s, %s, %s, %s;\n", reg, reg,
659 shader_arb_get_fixup_swizzle(fixup.x_source), shader_arb_get_fixup_swizzle(fixup.y_source),
660 shader_arb_get_fixup_swizzle(fixup.z_source), shader_arb_get_fixup_swizzle(fixup.w_source));
661 }
662
663 mask = 0;
664 if (fixup.x_sign_fixup) mask |= WINED3DSP_WRITEMASK_0;
665 if (fixup.y_sign_fixup) mask |= WINED3DSP_WRITEMASK_1;
666 if (fixup.z_sign_fixup) mask |= WINED3DSP_WRITEMASK_2;
667 if (fixup.w_sign_fixup) mask |= WINED3DSP_WRITEMASK_3;
668 mask &= dst_mask;
669
670 if (mask)
671 {
672 char reg_mask[6];
673 char *ptr = reg_mask;
674
675 if (mask != WINED3DSP_WRITEMASK_ALL)
676 {
677 *ptr++ = '.';
678 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
679 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
680 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
681 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
682 }
683 *ptr = '\0';
684
685 shader_addline(buffer, "MAD %s%s, %s, %s, -%s;\n", reg, reg_mask, reg, two, one);
686 }
687 }
688
689 static void shader_arb_color_correction(const struct SHADER_OPCODE_ARG* arg, struct color_fixup_desc fixup)
690 {
691 char reg[256];
692 pshader_get_register_name(arg->shader, arg->dst, reg);
693 gen_color_correction(arg->buffer, reg, arg->dst & WINED3DSP_WRITEMASK_ALL, "one", "coefmul.x", fixup);
694 }
695
696 static void pshader_gen_input_modifier_line (
697 IWineD3DBaseShader *iface,
698 SHADER_BUFFER* buffer,
699 const DWORD instr,
700 int tmpreg,
701 char *outregstr) {
702
703 /* Generate a line that does the input modifier computation and return the input register to use */
704 char regstr[256];
705 char swzstr[20];
706 int insert_line;
707
708 /* Assume a new line will be added */
709 insert_line = 1;
710
711 /* Get register name */
712 pshader_get_register_name(iface, instr, regstr);
713 shader_arb_get_swizzle(instr, FALSE, swzstr);
714
715 switch (instr & WINED3DSP_SRCMOD_MASK) {
716 case WINED3DSPSM_NONE:
717 sprintf(outregstr, "%s%s", regstr, swzstr);
718 insert_line = 0;
719 break;
720 case WINED3DSPSM_NEG:
721 sprintf(outregstr, "-%s%s", regstr, swzstr);
722 insert_line = 0;
723 break;
724 case WINED3DSPSM_BIAS:
725 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
726 break;
727 case WINED3DSPSM_BIASNEG:
728 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
729 break;
730 case WINED3DSPSM_SIGN:
731 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
732 break;
733 case WINED3DSPSM_SIGNNEG:
734 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
735 break;
736 case WINED3DSPSM_COMP:
737 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
738 break;
739 case WINED3DSPSM_X2:
740 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
741 break;
742 case WINED3DSPSM_X2NEG:
743 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
744 break;
745 case WINED3DSPSM_DZ:
746 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
747 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
748 break;
749 case WINED3DSPSM_DW:
750 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
751 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
752 break;
753 default:
754 sprintf(outregstr, "%s%s", regstr, swzstr);
755 insert_line = 0;
756 }
757
758 /* Return modified or original register, with swizzle */
759 if (insert_line)
760 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
761 }
762
763 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER *buffer, int saturate, const char *write_mask,
764 int shift, const char *regstr)
765 {
766 /* Generate a line that does the output modifier computation */
767 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
768 regstr, write_mask, regstr, shift_tab[shift]);
769 }
770
771 static void pshader_hw_bem(const SHADER_OPCODE_ARG *arg)
772 {
773 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
774
775 SHADER_BUFFER* buffer = arg->buffer;
776 char dst_name[50];
777 char src_name[2][50];
778 char dst_wmask[20];
779 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
780 BOOL has_bumpmat = FALSE;
781 int i;
782
783 for(i = 0; i < This->numbumpenvmatconsts; i++) {
784 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
785 has_bumpmat = TRUE;
786 break;
787 }
788 }
789
790 pshader_get_register_name(arg->shader, arg->dst, dst_name);
791 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
792 strcat(dst_name, dst_wmask);
793
794 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
795 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
796
797 if(has_bumpmat) {
798 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
799 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
800 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
801 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
802 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
803
804 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
805 } else {
806 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
807 }
808 }
809
810 static void pshader_hw_cnd(const SHADER_OPCODE_ARG *arg)
811 {
812 SHADER_BUFFER* buffer = arg->buffer;
813 char dst_wmask[20];
814 char dst_name[50];
815 char src_name[3][50];
816 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
817 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
818
819 /* FIXME: support output modifiers */
820
821 /* Handle output register */
822 pshader_get_register_name(arg->shader, arg->dst, dst_name);
823 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
824
825 /* Generate input register names (with modifiers) */
826 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
827 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
828 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
829
830 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
831 if (arg->reg_maps->shader_version <= WINED3DPS_VERSION(1, 3) && arg->opcode_token & WINED3DSI_COISSUE)
832 {
833 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
834 } else {
835 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
836 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
837 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
838 }
839 if (shift != 0)
840 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
841 }
842
843 static void pshader_hw_cmp(const SHADER_OPCODE_ARG *arg)
844 {
845 SHADER_BUFFER* buffer = arg->buffer;
846 char dst_wmask[20];
847 char dst_name[50];
848 char src_name[3][50];
849 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
850 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
851
852 /* FIXME: support output modifiers */
853
854 /* Handle output register */
855 pshader_get_register_name(arg->shader, arg->dst, dst_name);
856 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
857
858 /* Generate input register names (with modifiers) */
859 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
860 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
861 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
862
863 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
864 src_name[0], src_name[2], src_name[1]);
865
866 if (shift != 0)
867 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
868 }
869
870 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
871 * dst = dot2(src0, src1) + src2 */
872 static void pshader_hw_dp2add(const SHADER_OPCODE_ARG *arg)
873 {
874 SHADER_BUFFER* buffer = arg->buffer;
875 char dst_wmask[20];
876 char dst_name[50];
877 char src_name[3][50];
878 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
879 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
880
881 pshader_get_register_name(arg->shader, arg->dst, dst_name);
882 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
883
884 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
885 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
886 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
887
888 /* Emulate a DP2 with a DP3 and 0.0 */
889 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
890 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
891 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
892 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
893
894 if (shift != 0)
895 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
896 }
897
898 /* Map the opcode 1-to-1 to the GL code */
899 static void shader_hw_map2gl(const SHADER_OPCODE_ARG *arg)
900 {
901 CONST SHADER_OPCODE* curOpcode = arg->opcode;
902 SHADER_BUFFER* buffer = arg->buffer;
903 DWORD dst = arg->dst;
904 const DWORD *src = arg->src;
905 char arguments[256];
906 unsigned int i;
907
908 if (shader_is_pshader_version(arg->reg_maps->shader_version))
909 {
910 /* Output token related */
911 char output_rname[256];
912 char output_wmask[20];
913 char operands[4][100];
914 BOOL saturate = FALSE;
915 BOOL centroid = FALSE;
916 BOOL partialprecision = FALSE;
917 const char *modifier;
918 DWORD shift;
919
920 if (!curOpcode->num_params)
921 {
922 ERR("Opcode \"%s\" has no parameters\n", curOpcode->name);
923 return;
924 }
925
926 /* Process modifiers */
927 if (dst & WINED3DSP_DSTMOD_MASK)
928 {
929 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
930
931 saturate = mask & WINED3DSPDM_SATURATE;
932 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
933 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
934 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
935 if (mask)
936 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
937
938 if (centroid)
939 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
940 }
941 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
942 modifier = (saturate && !shift) ? "_SAT" : "";
943
944 /* Generate input register names (with modifiers) */
945 for (i = 1; i < curOpcode->num_params; ++i)
946 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
947
948 /* Handle output register */
949 pshader_get_register_name(arg->shader, dst, output_rname);
950 strcpy(operands[0], output_rname);
951 shader_arb_get_write_mask(arg, dst, output_wmask);
952 strcat(operands[0], output_wmask);
953
954 arguments[0] = '\0';
955 strcat(arguments, operands[0]);
956 for (i = 1; i < curOpcode->num_params; i++)
957 {
958 strcat(arguments, ", ");
959 strcat(arguments, operands[i]);
960 }
961 shader_addline(buffer, "%s%s %s;\n", curOpcode->glname, modifier, arguments);
962
963 /* A shift requires another line. */
964 if (shift) pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
965 } else {
966 /* Note that vshader_program_add_param() adds spaces. */
967
968 arguments[0] = '\0';
969 if (curOpcode->num_params > 0)
970 {
971 vshader_program_add_param(arg, dst, FALSE, arguments);
972 for (i = 1; i < curOpcode->num_params; ++i)
973 {
974 strcat(arguments, ",");
975 vshader_program_add_param(arg, src[i-1], TRUE, arguments);
976 }
977 }
978 shader_addline(buffer, "%s%s;\n", curOpcode->glname, arguments);
979 }
980 }
981
982 static void shader_hw_mov(const SHADER_OPCODE_ARG *arg)
983 {
984 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
985
986 if ((WINED3DSHADER_VERSION_MAJOR(arg->reg_maps->shader_version) == 1
987 && !shader_is_pshader_version(arg->reg_maps->shader_version)
988 && shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)
989 || arg->opcode->opcode == WINED3DSIO_MOVA)
990 {
991 SHADER_BUFFER *buffer = arg->buffer;
992 char src0_param[256];
993
994 if (arg->opcode->opcode == WINED3DSIO_MOVA)
995 FIXME("mova should round\n");
996
997 src0_param[0] = '\0';
998 if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
999 {
1000 vshader_program_add_param(arg, arg->src[0], TRUE, src0_param);
1001 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", src0_param);
1002 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1003 }
1004 else
1005 {
1006 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1007 * with more than one component. Thus replicate the first source argument over all
1008 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1009 */
1010 DWORD parm = arg->src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1011 if((arg->src[0] & WINED3DVS_X_W) == WINED3DVS_X_W)
1012 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1013 else if((arg->src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z)
1014 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1015 else if((arg->src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y)
1016 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1017 else if((arg->src[0] & WINED3DVS_X_X) == WINED3DVS_X_X)
1018 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1019 vshader_program_add_param(arg, parm, TRUE, src0_param);
1020 shader_addline(buffer, "ARL A0.x, %s;\n", src0_param);
1021 }
1022 }
1023 else
1024 {
1025 shader_hw_map2gl(arg);
1026 }
1027 }
1028
1029 static void pshader_hw_texkill(const SHADER_OPCODE_ARG *arg)
1030 {
1031 DWORD shader_version = arg->reg_maps->shader_version;
1032 SHADER_BUFFER* buffer = arg->buffer;
1033 char reg_dest[40];
1034
1035 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1036 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1037 */
1038 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1039
1040 if (shader_version >= WINED3DPS_VERSION(2,0))
1041 {
1042 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1043 shader_addline(buffer, "KIL %s;\n", reg_dest);
1044 } else {
1045 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1046 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1047 */
1048 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1049 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1050 shader_addline(buffer, "KIL TMP;\n");
1051 }
1052 }
1053
1054 static void pshader_hw_tex(const SHADER_OPCODE_ARG *arg)
1055 {
1056 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1057 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1058
1059 DWORD dst = arg->dst;
1060 const DWORD *src = arg->src;
1061 SHADER_BUFFER* buffer = arg->buffer;
1062 DWORD shader_version = arg->reg_maps->shader_version;
1063 BOOL projected = FALSE, bias = FALSE;
1064
1065 char reg_dest[40];
1066 char reg_coord[40];
1067 DWORD reg_dest_code;
1068 DWORD reg_sampler_code;
1069
1070 /* All versions have a destination register */
1071 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1072 pshader_get_register_name(arg->shader, dst, reg_dest);
1073
1074 /* 1.0-1.3: Use destination register as coordinate source.
1075 1.4+: Use provided coordinate source register. */
1076 if (shader_version < WINED3DPS_VERSION(1,4))
1077 strcpy(reg_coord, reg_dest);
1078 else
1079 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1080
1081 /* 1.0-1.4: Use destination register number as texture code.
1082 2.0+: Use provided sampler number as texure code. */
1083 if (shader_version < WINED3DPS_VERSION(2,0))
1084 reg_sampler_code = reg_dest_code;
1085 else
1086 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1087
1088 /* projection flag:
1089 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1090 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1091 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1092 */
1093 if (shader_version < WINED3DPS_VERSION(1,4))
1094 {
1095 DWORD flags = 0;
1096 if(reg_sampler_code < MAX_TEXTURES) {
1097 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1098 }
1099 if (flags & WINED3DTTFF_PROJECTED) {
1100 projected = TRUE;
1101 }
1102 }
1103 else if (shader_version < WINED3DPS_VERSION(2,0))
1104 {
1105 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1106 if (src_mod == WINED3DSPSM_DZ) {
1107 projected = TRUE;
1108 } else if(src_mod == WINED3DSPSM_DW) {
1109 projected = TRUE;
1110 }
1111 } else {
1112 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1113 projected = TRUE;
1114 }
1115 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1116 bias = TRUE;
1117 }
1118 }
1119 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1120 }
1121
1122 static void pshader_hw_texcoord(const SHADER_OPCODE_ARG *arg)
1123 {
1124 DWORD dst = arg->dst;
1125 SHADER_BUFFER* buffer = arg->buffer;
1126
1127 char tmp[20];
1128 shader_arb_get_write_mask(arg, dst, tmp);
1129 if (arg->reg_maps->shader_version != WINED3DPS_VERSION(1,4))
1130 {
1131 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1132 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1133 } else {
1134 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1135 char reg_src[40];
1136
1137 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1138 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1139 }
1140 }
1141
1142 static void pshader_hw_texreg2ar(const SHADER_OPCODE_ARG *arg)
1143 {
1144 SHADER_BUFFER* buffer = arg->buffer;
1145 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1146 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1147 DWORD flags;
1148
1149 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1150 char dst_str[8];
1151 char src_str[50];
1152
1153 sprintf(dst_str, "T%u", reg1);
1154 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1155 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1156 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1157 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1158 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1159 }
1160
1161 static void pshader_hw_texreg2gb(const SHADER_OPCODE_ARG *arg)
1162 {
1163 SHADER_BUFFER* buffer = arg->buffer;
1164
1165 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1166 char dst_str[8];
1167 char src_str[50];
1168
1169 sprintf(dst_str, "T%u", reg1);
1170 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1171 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1172 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1173 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1174 }
1175
1176 static void pshader_hw_texreg2rgb(const SHADER_OPCODE_ARG *arg)
1177 {
1178 SHADER_BUFFER* buffer = arg->buffer;
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_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1186 }
1187
1188 static void pshader_hw_texbem(const SHADER_OPCODE_ARG *arg)
1189 {
1190 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1191 BOOL has_bumpmat = FALSE;
1192 BOOL has_luminance = FALSE;
1193 int i;
1194
1195 DWORD dst = arg->dst;
1196 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1197 SHADER_BUFFER* buffer = arg->buffer;
1198
1199 char reg_coord[40];
1200 DWORD reg_dest_code;
1201
1202 /* All versions have a destination register */
1203 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1204 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1205 pshader_get_register_name(arg->shader, dst, reg_coord);
1206
1207 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1208 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1209 has_bumpmat = TRUE;
1210 break;
1211 }
1212 }
1213 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1214 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1215 has_luminance = TRUE;
1216 break;
1217 }
1218 }
1219
1220 if(has_bumpmat) {
1221 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1222
1223 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1224 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1225 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1226 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1227
1228 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1229 * so we can't let the GL handle this.
1230 */
1231 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1232 & WINED3DTTFF_PROJECTED) {
1233 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1234 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1235 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1236 } else {
1237 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1238 }
1239
1240 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1241
1242 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1243 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1244 src, reg_dest_code, reg_dest_code);
1245 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1246 }
1247
1248 } else {
1249 DWORD tf;
1250 if(reg_dest_code < MAX_TEXTURES) {
1251 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1252 } else {
1253 tf = 0;
1254 }
1255 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1256 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1257 }
1258 }
1259
1260 static void pshader_hw_texm3x2pad(const SHADER_OPCODE_ARG *arg)
1261 {
1262 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1263 SHADER_BUFFER* buffer = arg->buffer;
1264 char src0_name[50];
1265
1266 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1267 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1268 }
1269
1270 static void pshader_hw_texm3x2tex(const SHADER_OPCODE_ARG *arg)
1271 {
1272 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1273 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1274 DWORD flags;
1275 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1276 SHADER_BUFFER* buffer = arg->buffer;
1277 char dst_str[8];
1278 char src0_name[50];
1279
1280 sprintf(dst_str, "T%u", reg);
1281 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1282 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1283 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1284 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1285 }
1286
1287 static void pshader_hw_texm3x3pad(const SHADER_OPCODE_ARG *arg)
1288 {
1289 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1290 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1291 SHADER_BUFFER* buffer = arg->buffer;
1292 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1293 char src0_name[50];
1294
1295 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1296 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1297 current_state->texcoord_w[current_state->current_row++] = reg;
1298 }
1299
1300 static void pshader_hw_texm3x3tex(const SHADER_OPCODE_ARG *arg)
1301 {
1302 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1303 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1304 DWORD flags;
1305 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1306 SHADER_BUFFER* buffer = arg->buffer;
1307 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1308 char dst_str[8];
1309 char src0_name[50];
1310
1311 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1312 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1313
1314 /* Sample the texture using the calculated coordinates */
1315 sprintf(dst_str, "T%u", reg);
1316 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1317 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1318 current_state->current_row = 0;
1319 }
1320
1321 static void pshader_hw_texm3x3vspec(const SHADER_OPCODE_ARG *arg)
1322 {
1323 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1324 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1325 DWORD flags;
1326 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1327 SHADER_BUFFER* buffer = arg->buffer;
1328 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1329 char dst_str[8];
1330 char src0_name[50];
1331
1332 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1333 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1334
1335 /* Construct the eye-ray vector from w coordinates */
1336 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1337 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1338 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1339
1340 /* Calculate reflection vector
1341 */
1342 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1343 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1344 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1345 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1346 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1347 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1348 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1349
1350 /* Sample the texture using the calculated coordinates */
1351 sprintf(dst_str, "T%u", reg);
1352 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1353 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1354 current_state->current_row = 0;
1355 }
1356
1357 static void pshader_hw_texm3x3spec(const SHADER_OPCODE_ARG *arg)
1358 {
1359 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1360 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1361 DWORD flags;
1362 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1363 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1364 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1365 SHADER_BUFFER* buffer = arg->buffer;
1366 char dst_str[8];
1367 char src0_name[50];
1368
1369 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1370 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1371
1372 /* Calculate reflection vector.
1373 *
1374 * dot(N, E)
1375 * TMP.xyz = 2 * --------- * N - E
1376 * dot(N, N)
1377 *
1378 * Which normalizes the normal vector
1379 */
1380 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1381 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1382 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1383 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1384 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1385 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1386
1387 /* Sample the texture using the calculated coordinates */
1388 sprintf(dst_str, "T%u", reg);
1389 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1390 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1391 current_state->current_row = 0;
1392 }
1393
1394 static void pshader_hw_texdepth(const SHADER_OPCODE_ARG *arg)
1395 {
1396 SHADER_BUFFER* buffer = arg->buffer;
1397 char dst_name[50];
1398
1399 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1400 * which is essentially an input, is the destination register because it is the first
1401 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1402 * here
1403 */
1404 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1405
1406 /* According to the msdn, the source register(must be r5) is unusable after
1407 * the texdepth instruction, so we're free to modify it
1408 */
1409 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1410
1411 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1412 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1413 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1414 */
1415 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1416 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1417 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1418 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1419 }
1420
1421 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1422 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1423 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1424 static void pshader_hw_texdp3tex(const SHADER_OPCODE_ARG *arg)
1425 {
1426 SHADER_BUFFER* buffer = arg->buffer;
1427 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1428 char src0[50];
1429 char dst_str[8];
1430
1431 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1432 shader_addline(buffer, "MOV TMP, 0.0;\n");
1433 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1434
1435 sprintf(dst_str, "T%u", sampler_idx);
1436 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1437 }
1438
1439 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1440 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1441 static void pshader_hw_texdp3(const SHADER_OPCODE_ARG *arg)
1442 {
1443 char src0[50];
1444 char dst_str[50];
1445 char dst_mask[6];
1446 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1447 SHADER_BUFFER* buffer = arg->buffer;
1448
1449 /* Handle output register */
1450 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1451 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1452
1453 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1454 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1455
1456 /* TODO: Handle output modifiers */
1457 }
1458
1459 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1460 * Perform the 3rd row of a 3x3 matrix multiply */
1461 static void pshader_hw_texm3x3(const SHADER_OPCODE_ARG *arg)
1462 {
1463 SHADER_BUFFER* buffer = arg->buffer;
1464 char dst_str[50];
1465 char dst_mask[6];
1466 char src0[50];
1467 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1468
1469 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1470 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1471
1472 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1473 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1474 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1475
1476 /* TODO: Handle output modifiers */
1477 }
1478
1479 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1480 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1481 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1482 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1483 */
1484 static void pshader_hw_texm3x2depth(const SHADER_OPCODE_ARG *arg)
1485 {
1486 SHADER_BUFFER* buffer = arg->buffer;
1487 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1488 char src0[50];
1489
1490 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1491 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1492
1493 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1494 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1495 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1496 */
1497 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1498 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1499 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1500 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1501 }
1502
1503 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1504 Vertex/Pixel shaders to ARB_vertex_program codes */
1505 static void shader_hw_mnxn(const SHADER_OPCODE_ARG *arg)
1506 {
1507 int i;
1508 int nComponents = 0;
1509 SHADER_OPCODE_ARG tmpArg;
1510 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1511 const SHADER_OPCODE *opcode_table = shader->baseShader.shader_ins;
1512 DWORD shader_version = arg->reg_maps->shader_version;
1513
1514 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1515
1516 /* Set constants for the temporary argument */
1517 tmpArg.shader = arg->shader;
1518 tmpArg.buffer = arg->buffer;
1519 tmpArg.src[0] = arg->src[0];
1520 tmpArg.src_addr[0] = arg->src_addr[0];
1521 tmpArg.src_addr[1] = arg->src_addr[1];
1522 tmpArg.reg_maps = arg->reg_maps;
1523
1524 switch(arg->opcode->opcode) {
1525 case WINED3DSIO_M4x4:
1526 nComponents = 4;
1527 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1528 break;
1529 case WINED3DSIO_M4x3:
1530 nComponents = 3;
1531 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1532 break;
1533 case WINED3DSIO_M3x4:
1534 nComponents = 4;
1535 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1536 break;
1537 case WINED3DSIO_M3x3:
1538 nComponents = 3;
1539 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1540 break;
1541 case WINED3DSIO_M3x2:
1542 nComponents = 2;
1543 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1544 break;
1545 default:
1546 break;
1547 }
1548
1549 for (i = 0; i < nComponents; i++) {
1550 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1551 tmpArg.src[1] = arg->src[1]+i;
1552 shader_hw_map2gl(&tmpArg);
1553 }
1554 }
1555
1556 static void vshader_hw_rsq_rcp(const SHADER_OPCODE_ARG *arg)
1557 {
1558 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1559 SHADER_BUFFER* buffer = arg->buffer;
1560 DWORD dst = arg->dst;
1561 DWORD src = arg->src[0];
1562 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1563
1564 char tmpLine[256];
1565
1566 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1567 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1568 strcat(tmpLine, ",");
1569 vshader_program_add_param(arg, src, TRUE, tmpLine);
1570 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1571 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1572 * .w is used
1573 */
1574 strcat(tmpLine, ".w");
1575 }
1576
1577 shader_addline(buffer, "%s;\n", tmpLine);
1578 }
1579
1580 static void shader_hw_nrm(const SHADER_OPCODE_ARG *arg)
1581 {
1582 SHADER_BUFFER* buffer = arg->buffer;
1583 char dst_name[50];
1584 char src_name[50];
1585 char dst_wmask[20];
1586 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1587 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1588
1589 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1590 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1591
1592 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1593 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1594 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1595 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1596 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1597 src_name);
1598
1599 if (shift != 0)
1600 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1601 }
1602
1603 static void shader_hw_sincos(const SHADER_OPCODE_ARG *arg)
1604 {
1605 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1606 * must contain fixed constants. So we need a separate function to filter those constants and
1607 * can't use map2gl
1608 */
1609 SHADER_BUFFER* buffer = arg->buffer;
1610 char dst_name[50];
1611 char src_name[50];
1612 char dst_wmask[20];
1613 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1614 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1615
1616 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1617 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1618
1619 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1620 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1621 src_name);
1622
1623 if (shift != 0)
1624 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1625
1626 }
1627
1628 static GLuint create_arb_blt_vertex_program(const WineD3D_GL_Info *gl_info)
1629 {
1630 GLuint program_id = 0;
1631 const char *blt_vprogram =
1632 "!!ARBvp1.0\n"
1633 "PARAM c[1] = { { 1, 0.5 } };\n"
1634 "MOV result.position, vertex.position;\n"
1635 "MOV result.color, c[0].x;\n"
1636 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1637 "END\n";
1638
1639 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1640 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1641 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1642
1643 if (glGetError() == GL_INVALID_OPERATION) {
1644 GLint pos;
1645 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1646 FIXME("Vertex program error at position %d: %s\n", pos,
1647 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1648 }
1649
1650 return program_id;
1651 }
1652
1653 static GLuint create_arb_blt_fragment_program(const WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1654 {
1655 GLuint program_id = 0;
1656 static const char * const blt_fprograms[tex_type_count] =
1657 {
1658 /* tex_1d */
1659 NULL,
1660 /* tex_2d */
1661 "!!ARBfp1.0\n"
1662 "TEMP R0;\n"
1663 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1664 "MOV result.depth.z, R0.x;\n"
1665 "END\n",
1666 /* tex_3d */
1667 NULL,
1668 /* tex_cube */
1669 "!!ARBfp1.0\n"
1670 "TEMP R0;\n"
1671 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1672 "MOV result.depth.z, R0.x;\n"
1673 "END\n",
1674 /* tex_rect */
1675 "!!ARBfp1.0\n"
1676 "TEMP R0;\n"
1677 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1678 "MOV result.depth.z, R0.x;\n"
1679 "END\n",
1680 };
1681
1682 if (!blt_fprograms[tex_type])
1683 {
1684 FIXME("tex_type %#x not supported\n", tex_type);
1685 tex_type = tex_2d;
1686 }
1687
1688 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1689 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1690 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1691
1692 if (glGetError() == GL_INVALID_OPERATION) {
1693 GLint pos;
1694 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1695 FIXME("Fragment program error at position %d: %s\n", pos,
1696 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1697 }
1698
1699 return program_id;
1700 }
1701
1702 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1703 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1704 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1705 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1706
1707 if (useVS) {
1708 TRACE("Using vertex shader\n");
1709 IWineD3DVertexShaderImpl_CompileShader(This->stateBlock->vertexShader);
1710
1711 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->prgId;
1712
1713 /* Bind the vertex program */
1714 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1715 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1716
1717 /* Enable OpenGL vertex programs */
1718 glEnable(GL_VERTEX_PROGRAM_ARB);
1719 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1720 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1721 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1722 priv->current_vprogram_id = 0;
1723 glDisable(GL_VERTEX_PROGRAM_ARB);
1724 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1725 }
1726
1727 if (usePS) {
1728 struct ps_compile_args compile_args;
1729 TRACE("Using pixel shader\n");
1730 find_ps_compile_args((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader, This->stateBlock, &compile_args);
1731 priv->current_fprogram_id = find_gl_pshader((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader,
1732 &compile_args);
1733
1734 /* Bind the fragment program */
1735 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1736 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1737
1738 if(!priv->use_arbfp_fixed_func) {
1739 /* Enable OpenGL fragment programs */
1740 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1741 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1742 }
1743 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1744 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1745 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1746 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1747 * replacement shader
1748 */
1749 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1750 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1751 priv->current_fprogram_id = 0;
1752 }
1753 }
1754
1755 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1756 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1757 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1758 GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1759 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1760
1761 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1762 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1763 glEnable(GL_VERTEX_PROGRAM_ARB);
1764
1765 if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1766 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1767 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1768 }
1769
1770 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1771 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1772 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1773 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1774
1775 if (priv->current_vprogram_id) {
1776 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1777 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1778
1779 glEnable(GL_VERTEX_PROGRAM_ARB);
1780 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1781
1782 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1783 } else {
1784 glDisable(GL_VERTEX_PROGRAM_ARB);
1785 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1786 }
1787
1788 if (priv->current_fprogram_id) {
1789 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1790 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1791
1792 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1793 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1794
1795 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1796 } else {
1797 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1798 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1799 }
1800 }
1801
1802 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1803 IWineD3DBaseShaderImpl *baseShader = (IWineD3DBaseShaderImpl *) iface;
1804 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)baseShader->baseShader.device)->adapter->gl_info;
1805
1806 if (shader_is_pshader_version(baseShader->baseShader.reg_maps.shader_version))
1807 {
1808 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
1809 UINT i;
1810
1811 ENTER_GL();
1812 for(i = 0; i < This->num_gl_shaders; i++) {
1813 GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1814 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1815 }
1816 LEAVE_GL();
1817 HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1818 This->gl_shaders = NULL;
1819 This->num_gl_shaders = 0;
1820 } else {
1821 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *) iface;
1822
1823 ENTER_GL();
1824 GL_EXTCALL(glDeleteProgramsARB(1, &This->prgId));
1825 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->prgId))");
1826 ((IWineD3DVertexShaderImpl *) This)->prgId = 0;
1827 LEAVE_GL();
1828 }
1829 baseShader->baseShader.is_compiled = FALSE;
1830 }
1831
1832 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1833 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1834 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1835 return WINED3D_OK;
1836 }
1837
1838 static void shader_arb_free(IWineD3DDevice *iface) {
1839 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1840 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1841 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1842 int i;
1843
1844 if(priv->depth_blt_vprogram_id) {
1845 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1846 }
1847 for (i = 0; i < tex_type_count; ++i) {
1848 if (priv->depth_blt_fprogram_id[i]) {
1849 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
1850 }
1851 }
1852
1853 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1854 }
1855
1856 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1857 return TRUE;
1858 }
1859
1860 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1861 const char *tmp2, const char *tmp3, const char *tmp4) {
1862 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1863
1864 /* Calculate the > 0.0031308 case */
1865 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1866 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1867 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1868 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1869 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1870 /* Calculate the < case */
1871 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1872 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1873 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1874 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1875 /* Store the components > 0.0031308 in the destination */
1876 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1877 /* Add the components that are < 0.0031308 */
1878 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1879 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1880 }
1881
1882 static GLuint shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer, const struct ps_compile_args *args) {
1883 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1884 const shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1885 CONST DWORD *function = This->baseShader.function;
1886 DWORD shader_version = reg_maps->shader_version;
1887 const char *fragcolor;
1888 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1889 const local_constant *lconst;
1890 GLuint retval;
1891
1892 /* Create the hw ARB shader */
1893 shader_addline(buffer, "!!ARBfp1.0\n");
1894
1895 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1896 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1897 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1898 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1899 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1900 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1901 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1902 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1903
1904 /* Base Declarations */
1905 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1906
1907 /* We need two variables for fog blending */
1908 shader_addline(buffer, "TEMP TMP_FOG;\n");
1909 if (shader_version >= WINED3DPS_VERSION(2,0)) shader_addline(buffer, "TEMP TMP_COLOR;\n");
1910
1911 /* Base Shader Body */
1912 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1913
1914 /* calculate fog and blend it
1915 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
1916 * -1/(e-s) and e/(e-s) respectively.
1917 */
1918 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
1919
1920 fragcolor = (shader_version < WINED3DPS_VERSION(2,0)) ? "R0" : "TMP_COLOR";
1921
1922 if(((IWineD3DDeviceImpl *)This->baseShader.device)->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
1923 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
1924 }
1925 if (shader_version < WINED3DPS_VERSION(3,0))
1926 {
1927 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
1928 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
1929 }
1930
1931 shader_addline(buffer, "END\n");
1932
1933 /* TODO: change to resource.glObjectHandle or something like that */
1934 GL_EXTCALL(glGenProgramsARB(1, &retval));
1935
1936 TRACE("Creating a hw pixel shader, prg=%d\n", retval);
1937 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, retval));
1938
1939 TRACE("Created hw pixel shader, prg=%d\n", retval);
1940 /* Create the program and check for errors */
1941 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1942 buffer->bsize, buffer->buffer));
1943
1944 if (glGetError() == GL_INVALID_OPERATION) {
1945 GLint errPos;
1946 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
1947 FIXME("HW PixelShader Error at position %d: %s\n",
1948 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1949 retval = 0;
1950 }
1951
1952 /* Load immediate constants */
1953 if(!This->baseShader.load_local_constsF) {
1954 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
1955 const float *value = (const float *)lconst->value;
1956 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
1957 checkGLcall("glProgramLocalParameter4fvARB");
1958 }
1959 }
1960
1961 return retval;
1962 }
1963
1964 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
1965 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
1966 const shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
1967 CONST DWORD *function = This->baseShader.function;
1968 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
1969 const WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
1970 const local_constant *lconst;
1971
1972 /* Create the hw ARB shader */
1973 shader_addline(buffer, "!!ARBvp1.0\n");
1974 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
1975
1976 /* Mesa supports only 95 constants */
1977 if (GL_VEND(MESA) || GL_VEND(WINE))
1978 This->baseShader.limits.constant_float =
1979 min(95, This->baseShader.limits.constant_float);
1980
1981 shader_addline(buffer, "TEMP TMP;\n");
1982
1983 /* Base Declarations */
1984 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1985
1986 /* We need a constant to fixup the final position */
1987 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
1988
1989 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
1990 * for output parameters. D3D in theory does not do that either, but some applications depend on a
1991 * proper initialization of the secondary color, and programs using the fixed function pipeline without
1992 * a replacement shader depend on the texcoord.w being set properly.
1993 *
1994 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
1995 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
1996 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
1997 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
1998 * this can eat a number of instructions, so skip it unless this cap is set as well
1999 */
2000 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2001 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2002
2003 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2004 int i;
2005 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2006 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2007 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2008 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2009 }
2010 }
2011 }
2012 }
2013
2014 /* Base Shader Body */
2015 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2016
2017 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2018 if (!reg_maps->fog)
2019 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2020
2021 /* Write the final position.
2022 *
2023 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2024 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2025 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2026 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2027 */
2028 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2029 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2030 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2031
2032 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2033 * and the glsl equivalent
2034 */
2035 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2036
2037 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2038
2039 shader_addline(buffer, "END\n");
2040
2041 /* TODO: change to resource.glObjectHandle or something like that */
2042 GL_EXTCALL(glGenProgramsARB(1, &This->prgId));
2043
2044 TRACE("Creating a hw vertex shader, prg=%d\n", This->prgId);
2045 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->prgId));
2046
2047 TRACE("Created hw vertex shader, prg=%d\n", This->prgId);
2048 /* Create the program and check for errors */
2049 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2050 buffer->bsize, buffer->buffer));
2051
2052 if (glGetError() == GL_INVALID_OPERATION) {
2053 GLint errPos;
2054 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2055 FIXME("HW VertexShader Error at position %d: %s\n",
2056 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2057 This->prgId = -1;
2058 }
2059
2060 /* Load immediate constants */
2061 if(!This->baseShader.load_local_constsF) {
2062 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2063 const float *value = (const float *)lconst->value;
2064 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2065 }
2066 }
2067 }
2068
2069 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
2070 {
2071 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2072 * then overwrite the shader specific ones
2073 */
2074 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2075
2076 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2077 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2078 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2079 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2080 }
2081
2082 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2083 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2084 pCaps->PixelShader1xMaxValue = 8.0;
2085 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2086 }
2087 }
2088
2089 static BOOL shader_arb_color_fixup_supported(struct color_fixup_desc fixup)
2090 {
2091 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
2092 {
2093 TRACE("Checking support for color_fixup:\n");
2094 dump_color_fixup_desc(fixup);
2095 }
2096
2097 /* We support everything except YUV conversions. */
2098 if (!is_yuv_fixup(fixup))
2099 {
2100 TRACE("[OK]\n");
2101 return TRUE;
2102 }
2103
2104 TRACE("[FAILED]\n");
2105 return FALSE;
2106 }
2107
2108 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2109 {
2110 /* WINED3DSIH_ABS */ shader_hw_map2gl,
2111 /* WINED3DSIH_ADD */ shader_hw_map2gl,
2112 /* WINED3DSIH_BEM */ pshader_hw_bem,
2113 /* WINED3DSIH_BREAK */ NULL,
2114 /* WINED3DSIH_BREAKC */ NULL,
2115 /* WINED3DSIH_BREAKP */ NULL,
2116 /* WINED3DSIH_CALL */ NULL,
2117 /* WINED3DSIH_CALLNZ */ NULL,
2118 /* WINED3DSIH_CMP */ pshader_hw_cmp,
2119 /* WINED3DSIH_CND */ pshader_hw_cnd,
2120 /* WINED3DSIH_CRS */ shader_hw_map2gl,
2121 /* WINED3DSIH_DCL */ NULL,
2122 /* WINED3DSIH_DEF */ NULL,
2123 /* WINED3DSIH_DEFB */ NULL,
2124 /* WINED3DSIH_DEFI */ NULL,
2125 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add,
2126 /* WINED3DSIH_DP3 */ shader_hw_map2gl,
2127 /* WINED3DSIH_DP4 */ shader_hw_map2gl,
2128 /* WINED3DSIH_DST */ shader_hw_map2gl,
2129 /* WINED3DSIH_DSX */ NULL,
2130 /* WINED3DSIH_DSY */ NULL,
2131 /* WINED3DSIH_ELSE */ NULL,
2132 /* WINED3DSIH_ENDIF */ NULL,
2133 /* WINED3DSIH_ENDLOOP */ NULL,
2134 /* WINED3DSIH_ENDREP */ NULL,
2135 /* WINED3DSIH_EXP */ shader_hw_map2gl,
2136 /* WINED3DSIH_EXPP */ shader_hw_map2gl,
2137 /* WINED3DSIH_FRC */ shader_hw_map2gl,
2138 /* WINED3DSIH_IF */ NULL,
2139 /* WINED3DSIH_IFC */ NULL,
2140 /* WINED3DSIH_LABEL */ NULL,
2141 /* WINED3DSIH_LIT */ shader_hw_map2gl,
2142 /* WINED3DSIH_LOG */ shader_hw_map2gl,
2143 /* WINED3DSIH_LOGP */ shader_hw_map2gl,
2144 /* WINED3DSIH_LOOP */ NULL,
2145 /* WINED3DSIH_LRP */ shader_hw_map2gl,
2146 /* WINED3DSIH_M3x2 */ shader_hw_mnxn,
2147 /* WINED3DSIH_M3x3 */ shader_hw_mnxn,
2148 /* WINED3DSIH_M3x4 */ shader_hw_mnxn,
2149 /* WINED3DSIH_M4x3 */ shader_hw_mnxn,
2150 /* WINED3DSIH_M4x4 */ shader_hw_mnxn,
2151 /* WINED3DSIH_MAD */ shader_hw_map2gl,
2152 /* WINED3DSIH_MAX */ shader_hw_map2gl,
2153 /* WINED3DSIH_MIN */ shader_hw_map2gl,
2154 /* WINED3DSIH_MOV */ shader_hw_mov,
2155 /* WINED3DSIH_MOVA */ shader_hw_mov,
2156 /* WINED3DSIH_MUL */ shader_hw_map2gl,
2157 /* WINED3DSIH_NOP */ shader_hw_map2gl,
2158 /* WINED3DSIH_NRM */ shader_hw_nrm,
2159 /* WINED3DSIH_PHASE */ NULL,
2160 /* WINED3DSIH_POW */ shader_hw_map2gl,
2161 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp,
2162 /* WINED3DSIH_REP */ NULL,
2163 /* WINED3DSIH_RET */ NULL,
2164 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp,
2165 /* WINED3DSIH_SETP */ NULL,
2166 /* WINED3DSIH_SGE */ shader_hw_map2gl,
2167 /* WINED3DSIH_SGN */ NULL,
2168 /* WINED3DSIH_SINCOS */ shader_hw_sincos,
2169 /* WINED3DSIH_SLT */ shader_hw_map2gl,
2170 /* WINED3DSIH_SUB */ shader_hw_map2gl,
2171 /* WINED3DSIH_TEX */ pshader_hw_tex,
2172 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem,
2173 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem,
2174 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord,
2175 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth,
2176 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3,
2177 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex,
2178 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill,
2179 /* WINED3DSIH_TEXLDD */ NULL,
2180 /* WINED3DSIH_TEXLDL */ NULL,
2181 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth,
2182 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad,
2183 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex,
2184 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3,
2185 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
2186 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad,
2187 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec,
2188 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex,
2189 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec,
2190 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar,
2191 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb,
2192 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb,
2193 };
2194
2195 const shader_backend_t arb_program_shader_backend = {
2196 shader_arb_instruction_handler_table,
2197 shader_arb_select,
2198 shader_arb_select_depth_blt,
2199 shader_arb_deselect_depth_blt,
2200 shader_arb_update_float_vertex_constants,
2201 shader_arb_update_float_pixel_constants,
2202 shader_arb_load_constants,
2203 shader_arb_color_correction,
2204 shader_arb_destroy,
2205 shader_arb_alloc,
2206 shader_arb_free,
2207 shader_arb_dirty_const,
2208 shader_arb_generate_pshader,
2209 shader_arb_generate_vshader,
2210 shader_arb_get_caps,
2211 shader_arb_color_fixup_supported,
2212 };
2213
2214 /* ARB_fragment_program fixed function pipeline replacement definitions */
2215 #define ARB_FFP_CONST_TFACTOR 0
2216 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2217 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2218 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2219 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2220
2221 struct arbfp_ffp_desc
2222 {
2223 struct ffp_frag_desc parent;
2224 GLuint shader;
2225 unsigned int num_textures_used;
2226 };
2227
2228 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2229 if(enable) {
2230 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2231 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2232 } else {
2233 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2234 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2235 }
2236 }
2237
2238 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2239 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2240 struct shader_arb_priv *priv;
2241 /* Share private data between the shader backend and the pipeline replacement, if both
2242 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2243 * if no pixel shader is bound or not
2244 */
2245 if(This->shader_backend == &arb_program_shader_backend) {
2246 This->fragment_priv = This->shader_priv;
2247 } else {
2248 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2249 if(!This->fragment_priv) return E_OUTOFMEMORY;
2250 }
2251 priv = (struct shader_arb_priv *) This->fragment_priv;
2252 priv->fragment_shaders = hash_table_create(ffp_frag_program_key_hash, ffp_frag_program_key_compare);
2253 priv->use_arbfp_fixed_func = TRUE;
2254 return WINED3D_OK;
2255 }
2256
2257 static void arbfp_free_ffpshader(void *value, void *gli) {
2258 const WineD3D_GL_Info *gl_info = gli;
2259 struct arbfp_ffp_desc *entry_arb = value;
2260
2261 ENTER_GL();
2262 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2263 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2264 HeapFree(GetProcessHeap(), 0, entry_arb);
2265 LEAVE_GL();
2266 }
2267
2268 static void arbfp_free(IWineD3DDevice *iface) {
2269 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2270 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2271
2272 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2273 priv->use_arbfp_fixed_func = FALSE;
2274
2275 if(This->shader_backend != &arb_program_shader_backend) {
2276 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2277 }
2278 }
2279
2280 static void arbfp_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct fragment_caps *caps)
2281 {
2282 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2283 WINED3DTEXOPCAPS_SELECTARG1 |
2284 WINED3DTEXOPCAPS_SELECTARG2 |
2285 WINED3DTEXOPCAPS_MODULATE4X |
2286 WINED3DTEXOPCAPS_MODULATE2X |
2287 WINED3DTEXOPCAPS_MODULATE |
2288 WINED3DTEXOPCAPS_ADDSIGNED2X |
2289 WINED3DTEXOPCAPS_ADDSIGNED |
2290 WINED3DTEXOPCAPS_ADD |
2291 WINED3DTEXOPCAPS_SUBTRACT |
2292 WINED3DTEXOPCAPS_ADDSMOOTH |
2293 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2294 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2295 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2296 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2297 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2298 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2299 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2300 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2301 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2302 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2303 WINED3DTEXOPCAPS_MULTIPLYADD |
2304 WINED3DTEXOPCAPS_LERP |
2305 WINED3DTEXOPCAPS_BUMPENVMAP |
2306 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2307
2308 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2309
2310 caps->MaxTextureBlendStages = 8;
2311 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2312
2313 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2314 }
2315 #undef GLINFO_LOCATION
2316
2317 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2318 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2319 float col[4];
2320 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2321
2322 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2323 * application provided constants
2324 */
2325 if(device->shader_backend == &arb_program_shader_backend) {
2326 if(use_ps(device)) return;
2327
2328 device = stateblock->wineD3DDevice;
2329 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2330 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2331 }
2332
2333 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2334 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2335 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2336
2337 }
2338
2339 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2340 float col[4];
2341 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2342
2343 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2344 * application provided constants
2345 */
2346 if(device->shader_backend == &arb_program_shader_backend) {
2347 if(use_ps(device)) return;
2348
2349 device = stateblock->wineD3DDevice;
2350 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2351 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2352 }
2353
2354 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2355 /* The specular color has no alpha */
2356 col[0] = 1.0; col[1] = 1.0;
2357 col[2] = 1.0; col[3] = 0.0;
2358 } else {
2359 col[0] = 0.0; col[1] = 0.0;
2360 col[2] = 0.0; col[3] = 0.0;
2361 }
2362 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2363 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2364 }
2365
2366 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2367 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2368 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2369 float mat[2][2];
2370
2371 if(use_ps(device)) {
2372 if(stage != 0 &&
2373 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2374 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2375 * anyway
2376 */
2377 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2378 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2379 }
2380 }
2381
2382 if(device->shader_backend == &arb_program_shader_backend) {
2383 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2384 return;
2385 }
2386 } else if(device->shader_backend == &arb_program_shader_backend) {
2387 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2388 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2389 }
2390
2391 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2392 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2393 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2394 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2395
2396 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2397 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2398 }
2399
2400 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2401 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2402 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2403 float param[4];
2404
2405 if(use_ps(device)) {
2406 if(stage != 0 &&
2407 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2408 /* The pixel shader has to know the luminance offset. Do a constants update if it
2409 * isn't scheduled anyway
2410 */
2411 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2412 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2413 }
2414 }
2415
2416 if(device->shader_backend == &arb_program_shader_backend) {
2417 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2418 return;
2419 }
2420 } else if(device->shader_backend == &arb_program_shader_backend) {
2421 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2422 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2423 }
2424
2425 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2426 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2427 param[2] = 0.0;
2428 param[3] = 0.0;
2429
2430 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2431 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2432 }
2433
2434 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2435 const char *ret;
2436
2437 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2438
2439 switch(arg & WINED3DTA_SELECTMASK) {
2440 case WINED3DTA_DIFFUSE:
2441 ret = "fragment.color.primary"; break;
2442
2443 case WINED3DTA_CURRENT:
2444 if(stage == 0) ret = "fragment.color.primary";
2445 else ret = "ret";
2446 break;
2447
2448 case WINED3DTA_TEXTURE:
2449 switch(stage) {
2450 case 0: ret = "tex0"; break;
2451 case 1: ret = "tex1"; break;
2452 case 2: ret = "tex2"; break;
2453 case 3: ret = "tex3"; break;
2454 case 4: ret = "tex4"; break;
2455 case 5: ret = "tex5"; break;
2456 case 6: ret = "tex6"; break;
2457 case 7: ret = "tex7"; break;
2458 default: ret = "unknown texture";
2459 }
2460 break;
2461
2462 case WINED3DTA_TFACTOR:
2463 ret = "tfactor"; break;
2464
2465 case WINED3DTA_SPECULAR:
2466 ret = "fragment.color.secondary"; break;
2467
2468 case WINED3DTA_TEMP:
2469 ret = "tempreg"; break;
2470
2471 case WINED3DTA_CONSTANT:
2472 FIXME("Implement perstage constants\n");
2473 switch(stage) {
2474 case 0: ret = "const0"; break;
2475 case 1: ret = "const1"; break;
2476 case 2: ret = "const2"; break;
2477 case 3: ret = "const3"; break;
2478 case 4: ret = "const4"; break;
2479 case 5: ret = "const5"; break;
2480 case 6: ret = "const6"; break;
2481 case 7: ret = "const7"; break;
2482 default: ret = "unknown constant";
2483 }
2484 break;
2485
2486 default:
2487 return "unknown";
2488 }
2489
2490 if(arg & WINED3DTA_COMPLEMENT) {
2491 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2492 if(argnum == 0) ret = "arg0";
2493 if(argnum == 1) ret = "arg1";
2494 if(argnum == 2) ret = "arg2";
2495 }
2496 if(arg & WINED3DTA_ALPHAREPLICATE) {
2497 shader_addline(buffer, "MOV arg%u, %s.a;\n", argnum, ret);
2498 if(argnum == 0) ret = "arg0";
2499 if(argnum == 1) ret = "arg1";
2500 if(argnum == 2) ret = "arg2";
2501 }
2502 return ret;
2503 }
2504
2505 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2506 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2507 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2508 unsigned int mul = 1;
2509 BOOL mul_final_dest = FALSE;
2510
2511 if(color && alpha) dstmask = "";
2512 else if(color) dstmask = ".rgb";
2513 else dstmask = ".a";
2514
2515 if(dst == tempreg) dstreg = "tempreg";
2516 else dstreg = "ret";
2517
2518 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2519 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2520 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2521
2522 switch(op) {
2523 case WINED3DTOP_DISABLE:
2524 if(stage == 0) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2525 break;
2526
2527 case WINED3DTOP_SELECTARG2:
2528 arg1 = arg2;
2529 case WINED3DTOP_SELECTARG1:
2530 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2531 break;
2532
2533 case WINED3DTOP_MODULATE4X:
2534 mul = 2;
2535 case WINED3DTOP_MODULATE2X:
2536 mul *= 2;
2537 if(strcmp(dstreg, "result.color") == 0) {
2538 dstreg = "ret";
2539 mul_final_dest = TRUE;
2540 }
2541 case WINED3DTOP_MODULATE:
2542 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2543 break;
2544
2545 case WINED3DTOP_ADDSIGNED2X:
2546 mul = 2;
2547 if(strcmp(dstreg, "result.color") == 0) {
2548 dstreg = "ret";
2549 mul_final_dest = TRUE;
2550 }
2551 case WINED3DTOP_ADDSIGNED:
2552 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2553 arg2 = "arg2";
2554 case WINED3DTOP_ADD:
2555 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2556 break;
2557
2558 case WINED3DTOP_SUBTRACT:
2559 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2560 break;
2561
2562 case WINED3DTOP_ADDSMOOTH:
2563 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2564 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2565 break;
2566
2567 case WINED3DTOP_BLENDCURRENTALPHA:
2568 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2569 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2570 break;
2571 case WINED3DTOP_BLENDFACTORALPHA:
2572 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2573 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2574 break;
2575 case WINED3DTOP_BLENDTEXTUREALPHA:
2576 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2577 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2578 break;
2579 case WINED3DTOP_BLENDDIFFUSEALPHA:
2580 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2581 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2582 break;
2583
2584 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2585 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2586 shader_addline(buffer, "SUB arg0.a, const.x, %s.a;\n", arg0);
2587 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg, dstmask, arg2, arg1);
2588 break;
2589
2590 /* D3DTOP_PREMODULATE ???? */
2591
2592 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2593 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2594 shader_addline(buffer, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2595 break;
2596 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2597 shader_addline(buffer, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2598 break;
2599 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2600 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2601 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg, dstmask, arg2, arg1);
2602 break;
2603 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2604 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg, dstmask, arg1, arg2, arg1);
2605 break;
2606
2607 case WINED3DTOP_DOTPRODUCT3:
2608 mul = 4;
2609 if(strcmp(dstreg, "result.color") == 0) {
2610 dstreg = "ret";
2611 mul_final_dest = TRUE;
2612 }
2613 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2614 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2615 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2616 break;
2617
2618 case WINED3DTOP_MULTIPLYADD:
2619 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2620 break;
2621
2622 case WINED3DTOP_LERP:
2623 /* The msdn is not quite right here */
2624 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2625 break;
2626
2627 case WINED3DTOP_BUMPENVMAP:
2628 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2629 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2630 break;
2631
2632 default:
2633 FIXME("Unhandled texture op %08x\n", op);
2634 }
2635
2636 if(mul == 2) {
2637 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2638 } else if(mul == 4) {
2639 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2640 }
2641 }
2642
2643 /* The stateblock is passed for GLINFO_LOCATION */
2644 static GLuint gen_arbfp_ffp_shader(const struct ffp_frag_settings *settings, IWineD3DStateBlockImpl *stateblock)
2645 {
2646 unsigned int stage;
2647 SHADER_BUFFER buffer;
2648 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2649 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2650 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2651 const char *textype;
2652 const char *instr, *sat;
2653 char colorcor_dst[8];
2654 GLuint ret;
2655 DWORD arg0, arg1, arg2;
2656 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2657 BOOL op_equal;
2658 const char *final_combiner_src = "ret";
2659
2660 /* Find out which textures are read */
2661 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2662 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2663 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2664 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2665 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2666 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2667 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2668 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2669
2670 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2671 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2672 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2673 bump_used[stage] = TRUE;
2674 tex_read[stage] = TRUE;
2675 }
2676 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2677 bump_used[stage] = TRUE;
2678 tex_read[stage] = TRUE;
2679 luminance_used[stage] = TRUE;
2680 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2681 tfactor_used = TRUE;
2682 }
2683
2684 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2685 tfactor_used = TRUE;
2686 }
2687
2688 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2689 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2690 tempreg_used = TRUE;
2691 }
2692
2693 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2694 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2695 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2696 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2697 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2698 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2699 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2700
2701 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2702 tempreg_used = TRUE;
2703 }
2704 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2705 tfactor_used = TRUE;
2706 }
2707 }
2708
2709 /* Shader header */
2710 shader_buffer_init(&buffer);
2711
2712 shader_addline(&buffer, "!!ARBfp1.0\n");
2713
2714 switch(settings->fog) {
2715 case FOG_OFF: break;
2716 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2717 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2718 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2719 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2720 }
2721
2722 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2723 shader_addline(&buffer, "TEMP TMP;\n");
2724 shader_addline(&buffer, "TEMP ret;\n");
2725 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2726 shader_addline(&buffer, "TEMP arg0;\n");
2727 shader_addline(&buffer, "TEMP arg1;\n");
2728 shader_addline(&buffer, "TEMP arg2;\n");
2729 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2730 if(!tex_read[stage]) continue;
2731 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2732 if(!bump_used[stage]) continue;
2733 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2734 if(!luminance_used[stage]) continue;
2735 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2736 }
2737 if(tfactor_used) {
2738 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2739 }
2740 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2741
2742 if(settings->sRGB_write) {
2743 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2744 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2745 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2746 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2747 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2748 srgb_pow, srgb_pow, srgb_pow);
2749 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2750 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2751 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2752 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2753 }
2754
2755 /* Generate texture sampling instructions) */
2756 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2757 if(!tex_read[stage]) continue;
2758
2759 switch(settings->op[stage].tex_type) {
2760 case tex_1d: textype = "1D"; break;
2761 case tex_2d: textype = "2D"; break;
2762 case tex_3d: textype = "3D"; break;
2763 case tex_cube: textype = "CUBE"; break;
2764 case tex_rect: textype = "RECT"; break;
2765 default: textype = "unexpected_textype"; break;
2766 }
2767
2768 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2769 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2770 sat = "";
2771 } else {
2772 sat = "_SAT";
2773 }
2774
2775 if(settings->op[stage].projected == proj_none) {
2776 instr = "TEX";
2777 } else if(settings->op[stage].projected == proj_count4 ||
2778 settings->op[stage].projected == proj_count3) {
2779 instr = "TXP";
2780 } else {
2781 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2782 instr = "TXP";
2783 }
2784
2785 if(stage > 0 &&
2786 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2787 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2788 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2789 shader_addline(&buffer, "DP3 ret.r, arg1, tex%u;\n", stage - 1);
2790 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2791 shader_addline(&buffer, "DP3 ret.g, arg1, tex%u;\n", stage - 1);
2792
2793 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2794 * so multiply the displacement with the dividing parameter before passing it to TXP
2795 */
2796 if (settings->op[stage].projected != proj_none) {
2797 if(settings->op[stage].projected == proj_count4) {
2798 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].a;\n", stage);
2799 shader_addline(&buffer, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage, stage);
2800 } else {
2801 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].b;\n", stage);
2802 shader_addline(&buffer, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage, stage);
2803 }
2804 } else {
2805 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2806 }
2807
2808 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2809 instr, sat, stage, stage, textype);
2810 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2811 shader_addline(&buffer, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2812 stage - 1, stage - 1, stage - 1);
2813 shader_addline(&buffer, "MUL tex%u, tex%u, ret.r;\n", stage, stage);
2814 }
2815 } else if(settings->op[stage].projected == proj_count3) {
2816 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2817 shader_addline(&buffer, "MOV ret.a, ret.b;\n");
2818 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2819 instr, sat, stage, stage, textype);
2820 } else {
2821 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2822 instr, sat, stage, stage, stage, textype);
2823 }
2824
2825 sprintf(colorcor_dst, "tex%u", stage);
2826 gen_color_correction(&buffer, colorcor_dst, WINED3DSP_WRITEMASK_ALL, "const.x", "const.y",
2827 settings->op[stage].color_fixup);
2828 }
2829
2830 /* Generate the main shader */
2831 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2832 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2833 if(stage == 0) {
2834 final_combiner_src = "fragment.color.primary";
2835 }
2836 break;
2837 }
2838
2839 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2840 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2841 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2842 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2843 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2844 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2845 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2846 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2847 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2848 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2849 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2850 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2851 } else {
2852 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
2853 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2854 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2855 settings->op[stage].carg2 == settings->op[stage].aarg2;
2856 }
2857
2858 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2859 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2860 settings->op[stage].cop, settings->op[stage].carg0,
2861 settings->op[stage].carg1, settings->op[stage].carg2);
2862 if(stage == 0) {
2863 shader_addline(&buffer, "MOV ret.a, fragment.color.primary.a;\n");
2864 }
2865 } else if(op_equal) {
2866 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2867 settings->op[stage].cop, settings->op[stage].carg0,
2868 settings->op[stage].carg1, settings->op[stage].carg2);
2869 } else {
2870 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2871 settings->op[stage].cop, settings->op[stage].carg0,
2872 settings->op[stage].carg1, settings->op[stage].carg2);
2873 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
2874 settings->op[stage].aop, settings->op[stage].aarg0,
2875 settings->op[stage].aarg1, settings->op[stage].aarg2);
2876 }
2877 }
2878
2879 if(settings->sRGB_write) {
2880 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2881 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2882 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
2883 } else {
2884 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2885 }
2886
2887 /* Footer */
2888 shader_addline(&buffer, "END\n");
2889
2890 /* Generate the shader */
2891 GL_EXTCALL(glGenProgramsARB(1, &ret));
2892 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2893 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2894
2895 if (glGetError() == GL_INVALID_OPERATION) {
2896 GLint pos;
2897 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2898 FIXME("Fragment program error at position %d: %s\n", pos,
2899 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2900 }
2901 shader_buffer_free(&buffer);
2902 return ret;
2903 }
2904
2905 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2906 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2907 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
2908 BOOL use_pshader = use_ps(device);
2909 BOOL use_vshader = use_vs(device);
2910 struct ffp_frag_settings settings;
2911 const struct arbfp_ffp_desc *desc;
2912 unsigned int i;
2913
2914 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2915 if(!use_pshader && device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2916 /* Reload fixed function constants since they collide with the pixel shader constants */
2917 for(i = 0; i < MAX_TEXTURES; i++) {
2918 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2919 }
2920 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2921 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2922 } else if(use_pshader && !isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2923 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2924 }
2925 return;
2926 }
2927
2928 if(!use_pshader) {
2929 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2930 gen_ffp_frag_op(stateblock, &settings, FALSE);
2931 desc = (const struct arbfp_ffp_desc *)find_ffp_frag_shader(priv->fragment_shaders, &settings);
2932 if(!desc) {
2933 struct arbfp_ffp_desc *new_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc));
2934 if (!new_desc)
2935 {
2936 ERR("Out of memory\n");
2937 return;
2938 }
2939 new_desc->num_textures_used = 0;
2940 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
2941 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
2942 new_desc->num_textures_used = i;
2943 }
2944
2945 memcpy(&new_desc->parent.settings, &settings, sizeof(settings));
2946 new_desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
2947 add_ffp_frag_shader(priv->fragment_shaders, &new_desc->parent);
2948 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc);
2949 desc = new_desc;
2950 }
2951
2952 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
2953 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
2954 * deactivate it.
2955 */
2956 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
2957 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
2958 priv->current_fprogram_id = desc->shader;
2959
2960 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2961 /* Reload fixed function constants since they collide with the pixel shader constants */
2962 for(i = 0; i < MAX_TEXTURES; i++) {
2963 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2964 }
2965 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2966 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2967 }
2968 }
2969
2970 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
2971 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
2972 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
2973 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
2974 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
2975 *
2976 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
2977 * to be compiled before activating them(needs some cleanups in the shader backend interface)
2978 */
2979 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2980 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2981
2982 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
2983 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
2984 }
2985 }
2986 if(use_pshader) {
2987 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2988 }
2989 }
2990
2991 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
2992 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
2993 * state table, so we need to handle that with a forwarding function. The other invisible side effect
2994 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
2995 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
2996 */
2997 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2998 if(!isStateDirty(context, STATE_PIXELSHADER)) {
2999 fragment_prog_arbfp(state, stateblock, context);
3000 }
3001 }
3002
3003 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3004 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3005 fragment_prog_arbfp(state, stateblock, context);
3006 }
3007 }
3008
3009 #undef GLINFO_LOCATION
3010
3011 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3012 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3013 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3014 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3015 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3016 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3017 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3018 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3019 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3020 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3021 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3022 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3023 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3024 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3025 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3026 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3027 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3028 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3029 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3030 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3031 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3032 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3033 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3034 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3035 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3036 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3037 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3038 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3039 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3040 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3041 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3042 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3043 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3044 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3045 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3046 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3047 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3048 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3049 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3050 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3051 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3052 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3053 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3054 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3055 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3056 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3057 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3058 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3059 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3060 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3061 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3062 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3063 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3064 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3065 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3066 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3067 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3068 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3069 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3070 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3071 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3072 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3073 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3074 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3075 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3076 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3077 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3078 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3079 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3080 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3081 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3082 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3083 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3084 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3085 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3086 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3087 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3088 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3089 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3090 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3091 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3092 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3093 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3094 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3095 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3096 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3097 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3098 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3099 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3100 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3101 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3102 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3103 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3104 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3105 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3106 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3107 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3108 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3109 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3110 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3111 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3112 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3113 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3114 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3115 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3116 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3117 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3118 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3119 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3120 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3121 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3122 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3123 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3124 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3125 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3126 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3127 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3128 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3129 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3130 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3131 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3132 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3133 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3134 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3135 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3136 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3137 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3138 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3139 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3140 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3141 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3142 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3143 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3144 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3145 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3146 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3147 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3148 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3149 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3150 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3151 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3152 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3153 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3154 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3155 {0 /* Terminate */, { 0, 0 }, 0 },
3156 };
3157
3158 const struct fragment_pipeline arbfp_fragment_pipeline = {
3159 arbfp_enable,
3160 arbfp_get_caps,
3161 arbfp_alloc,
3162 arbfp_free,
3163 shader_arb_color_fixup_supported,
3164 arbfp_fragmentstate_template,
3165 TRUE /* We can disable projected textures */
3166 };
3167
3168 #define GLINFO_LOCATION device->adapter->gl_info
3169
3170 struct arbfp_blit_priv {
3171 GLenum yuy2_rect_shader, yuy2_2d_shader;
3172 GLenum uyvy_rect_shader, uyvy_2d_shader;
3173 GLenum yv12_rect_shader, yv12_2d_shader;
3174 };
3175
3176 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3177 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3178 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3179 if(!device->blit_priv) {
3180 ERR("Out of memory\n");
3181 return E_OUTOFMEMORY;
3182 }
3183 return WINED3D_OK;
3184 }
3185 static void arbfp_blit_free(IWineD3DDevice *iface) {
3186 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3187 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3188
3189 ENTER_GL();
3190 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3191 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3192 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3193 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3194 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3195 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3196 checkGLcall("Delete yuv programs\n");
3197 LEAVE_GL();
3198 }
3199
3200 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, enum yuv_fixup yuv_fixup, GLenum textype, char *luminance)
3201 {
3202 char chroma;
3203 const char *tex, *texinstr;
3204
3205 if (yuv_fixup == YUV_FIXUP_UYVY) {
3206 chroma = 'r';
3207 *luminance = 'a';
3208 } else {
3209 chroma = 'a';
3210 *luminance = 'r';
3211 }
3212 switch(textype) {
3213 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3214 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3215 default:
3216 /* This is more tricky than just replacing the texture type - we have to navigate
3217 * properly in the texture to find the correct chroma values
3218 */
3219 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3220 return FALSE;
3221 }
3222
3223 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3224 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3225 * filtering when we sample the texture.
3226 *
3227 * These are the rules for reading the chroma:
3228 *
3229 * Even pixel: Cr
3230 * Even pixel: U
3231 * Odd pixel: V
3232 *
3233 * So we have to get the sampling x position in non-normalized coordinates in integers
3234 */
3235 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3236 shader_addline(buffer, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3237 shader_addline(buffer, "MOV texcrd.a, size.x;\n");
3238 } else {
3239 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3240 }
3241 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3242 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3243 * 0.5, so add 0.5.
3244 */
3245 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3246 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3247
3248 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3249 * even and odd pixels respectively
3250 */
3251 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3252 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3253
3254 /* Sample Pixel 1 */
3255 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3256
3257 /* Put the value into either of the chroma values */
3258 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3259 shader_addline(buffer, "MUL chroma.r, luminance.%c, temp.x;\n", chroma);
3260 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3261 shader_addline(buffer, "MUL chroma.g, luminance.%c, temp.x;\n", chroma);
3262
3263 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3264 * the pixel right to the current one. Otherwise, sample the left pixel.
3265 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3266 */
3267 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3268 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3269 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3270
3271 /* Put the value into the other chroma */
3272 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3273 shader_addline(buffer, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma);
3274 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3275 shader_addline(buffer, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma);
3276
3277 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3278 * the current one and lerp the two U and V values
3279 */
3280
3281 /* This gives the correctly filtered luminance value */
3282 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3283
3284 return TRUE;
3285 }
3286
3287 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, GLenum textype, char *luminance)
3288 {
3289 const char *tex;
3290
3291 switch(textype) {
3292 case GL_TEXTURE_2D: tex = "2D"; break;
3293 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3294 default:
3295 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3296 return FALSE;
3297 }
3298
3299 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3300 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3301 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3302 * pitch of the luminance plane, the packing into the gl texture is a bit
3303 * unfortunate. If the whole texture is interpreted as luminance data it looks
3304 * approximately like this:
3305 *
3306 * +----------------------------------+----
3307 * | |
3308 * | |
3309 * | |
3310 * | |
3311 * | | 2
3312 * | LUMINANCE | -
3313 * | | 3
3314 * | |
3315 * | |
3316 * | |
3317 * | |
3318 * +----------------+-----------------+----
3319 * | | |
3320 * | U even rows | U odd rows |
3321 * | | | 1
3322 * +----------------+------------------ -
3323 * | | | 3
3324 * | V even rows | V odd rows |
3325 * | | |
3326 * +----------------+-----------------+----
3327 * | | |
3328 * | 0.5 | 0.5 |
3329 *
3330 * So it appears as if there are 4 chroma images, but in fact the odd rows
3331 * in the chroma images are in the same row as the even ones. So its is
3332 * kinda tricky to read
3333 *
3334 * When reading from rectangle textures, keep in mind that the input y coordinates
3335 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3336 */
3337 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3338 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3339
3340 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3341 /* the chroma planes have only half the width */
3342 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3343
3344 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3345 * the coordinate. Also read the right side of the image when reading odd lines
3346 *
3347 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3348 * bleeding
3349 */
3350 if(textype == GL_TEXTURE_2D) {
3351
3352 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3353
3354 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3355
3356 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3357 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3358
3359 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3360 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3361 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3362 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3363 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3364
3365 /* clamp, keep the half pixel origin in mind */
3366 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3367 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3368 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3369 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3370 } else {
3371 /* Read from [size - size+size/4] */
3372 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3373 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3374
3375 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3376 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3377 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3378 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3379 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3380 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3381
3382 /* Make sure to read exactly from the pixel center */
3383 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3384 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3385
3386 /* Clamp */
3387 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3388 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3389 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3390 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3391 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3392 }
3393 /* Read the texture, put the result into the output register */
3394 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3395 shader_addline(buffer, "MOV chroma.r, temp.a;\n");
3396
3397 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3398 * No need to clamp because we're just reusing the already clamped value from above
3399 */
3400 if(textype == GL_TEXTURE_2D) {
3401 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3402 } else {
3403 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3404 }
3405 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3406 shader_addline(buffer, "MOV chroma.g, temp.a;\n");
3407
3408 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3409 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3410 * values due to filtering
3411 */
3412 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3413 if(textype == GL_TEXTURE_2D) {
3414 /* Multiply the y coordinate by 2/3 and clamp it */
3415 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3416 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3417 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3418 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3419 } else {
3420 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
3421 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3422 * is bigger
3423 */
3424 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3425 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3426 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3427 }
3428 *luminance = 'a';
3429
3430 return TRUE;
3431 }
3432
3433 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, enum yuv_fixup yuv_fixup, GLenum textype)
3434 {
3435 GLenum shader;
3436 SHADER_BUFFER buffer;
3437 char luminance_component;
3438 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3439
3440 /* Shader header */
3441 shader_buffer_init(&buffer);
3442
3443 ENTER_GL();
3444 GL_EXTCALL(glGenProgramsARB(1, &shader));
3445 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3446 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3447 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3448 LEAVE_GL();
3449 if(!shader) {
3450 shader_buffer_free(&buffer);
3451 return 0;
3452 }
3453
3454 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3455 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3456 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3457 * each single pixel it contains, and one U and one V value shared between both
3458 * pixels.
3459 *
3460 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3461 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3462 * take the format into account when generating the read swizzles
3463 *
3464 * Reading the Y value is straightforward - just sample the texture. The hardware
3465 * takes care of filtering in the horizontal and vertical direction.
3466 *
3467 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3468 * because that would mix the U and V values of one pixel or two adjacent pixels.
3469 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3470 * regardless of the filtering setting. Vertical filtering works automatically
3471 * though - the U and V values of two rows are mixed nicely.
3472 *
3473 * Appart of avoiding filtering issues, the code has to know which value it just
3474 * read, and where it can find the other one. To determine this, it checks if
3475 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3476 *
3477 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3478 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3479 *
3480 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3481 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3482 * in an unfiltered situation. Finding the luminance on the other hand requires
3483 * finding out if it is an odd or even pixel. The real drawback of this approach
3484 * is filtering. This would have to be emulated completely in the shader, reading
3485 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3486 * vertically. Beyond that it would require adjustments to the texture handling
3487 * code to deal with the width scaling
3488 */
3489 shader_addline(&buffer, "!!ARBfp1.0\n");
3490 shader_addline(&buffer, "TEMP luminance;\n");
3491 shader_addline(&buffer, "TEMP temp;\n");
3492 shader_addline(&buffer, "TEMP chroma;\n");
3493 shader_addline(&buffer, "TEMP texcrd;\n");
3494 shader_addline(&buffer, "TEMP texcrd2;\n");
3495 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3496 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3497 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3498
3499 switch (yuv_fixup)
3500 {
3501 case YUV_FIXUP_UYVY:
3502 case YUV_FIXUP_YUY2:
3503 if (!gen_planar_yuv_read(&buffer, yuv_fixup, textype, &luminance_component))
3504 {
3505 shader_buffer_free(&buffer);
3506 return 0;
3507 }
3508 break;
3509
3510 case YUV_FIXUP_YV12:
3511 if (!gen_yv12_read(&buffer, textype, &luminance_component))
3512 {
3513 shader_buffer_free(&buffer);
3514 return 0;
3515 }
3516 break;
3517
3518 default:
3519 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3520 shader_buffer_free(&buffer);
3521 return 0;
3522 }
3523
3524 /* Calculate the final result. Formula is taken from
3525 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3526 * ranges from -0.5 to 0.5
3527 */
3528 shader_addline(&buffer, "SUB chroma.rg, chroma, coef.y;\n");
3529
3530 shader_addline(&buffer, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component);
3531 shader_addline(&buffer, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component);
3532 shader_addline(&buffer, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3533 shader_addline(&buffer, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component);
3534 shader_addline(&buffer, "END\n");
3535
3536 ENTER_GL();
3537 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3538
3539 if (glGetError() == GL_INVALID_OPERATION) {
3540 GLint pos;
3541 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3542 FIXME("Fragment program error at position %d: %s\n", pos,
3543 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3544 }
3545 shader_buffer_free(&buffer);
3546 LEAVE_GL();
3547
3548 switch (yuv_fixup)
3549 {
3550 case YUV_FIXUP_YUY2:
3551 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yuy2_rect_shader = shader;
3552 else priv->yuy2_2d_shader = shader;
3553 break;
3554
3555 case YUV_FIXUP_UYVY:
3556 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->uyvy_rect_shader = shader;
3557 else priv->uyvy_2d_shader = shader;
3558 break;
3559
3560 case YUV_FIXUP_YV12:
3561 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yv12_rect_shader = shader;
3562 else priv->yv12_2d_shader = shader;
3563 break;
3564 }
3565
3566 return shader;
3567 }
3568
3569 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3570 GLenum shader;
3571 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3572 float size[4] = {width, height, 1, 1};
3573 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3574 const struct GlPixelFormatDesc *glDesc;
3575 enum yuv_fixup yuv_fixup;
3576
3577 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3578
3579 if (!is_yuv_fixup(glDesc->color_fixup))
3580 {
3581 TRACE("Fixup:\n");
3582 dump_color_fixup_desc(glDesc->color_fixup);
3583 /* Don't bother setting up a shader for unconverted formats */
3584 ENTER_GL();
3585 glEnable(textype);
3586 checkGLcall("glEnable(textype)");
3587 LEAVE_GL();
3588 return WINED3D_OK;
3589 }
3590
3591 yuv_fixup = get_yuv_fixup(glDesc->color_fixup);
3592
3593 switch(yuv_fixup)
3594 {
3595 case YUV_FIXUP_YUY2:
3596 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yuy2_rect_shader : priv->yuy2_2d_shader;
3597 break;
3598
3599 case YUV_FIXUP_UYVY:
3600 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->uyvy_rect_shader : priv->uyvy_2d_shader;
3601 break;
3602
3603 case YUV_FIXUP_YV12:
3604 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yv12_rect_shader : priv->yv12_2d_shader;
3605 break;
3606
3607 default:
3608 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup);
3609 ENTER_GL();
3610 glEnable(textype);
3611 checkGLcall("glEnable(textype)");
3612 LEAVE_GL();
3613 return E_NOTIMPL;
3614 }
3615
3616 if (!shader) shader = gen_yuv_shader(device, yuv_fixup, textype);
3617
3618 ENTER_GL();
3619 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3620 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3621 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3622 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3623 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3624 checkGLcall("glProgramLocalParameter4fvARB");
3625 LEAVE_GL();
3626
3627 return WINED3D_OK;
3628 }
3629
3630 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3631 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3632
3633 ENTER_GL();
3634 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3635 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3636 glDisable(GL_TEXTURE_2D);
3637 checkGLcall("glDisable(GL_TEXTURE_2D)");
3638 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3639 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3640 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3641 }
3642 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3643 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3644 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3645 }
3646 LEAVE_GL();
3647 }
3648
3649 static BOOL arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup)
3650 {
3651 enum yuv_fixup yuv_fixup;
3652
3653 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
3654 {
3655 TRACE("Checking support for fixup:\n");
3656 dump_color_fixup_desc(fixup);
3657 }
3658
3659 if (is_identity_fixup(fixup))
3660 {
3661 TRACE("[OK]\n");
3662 return TRUE;
3663 }
3664
3665 /* We only support YUV conversions. */
3666 if (!is_yuv_fixup(fixup))
3667 {
3668 TRACE("[FAILED]\n");
3669 return FALSE;
3670 }
3671
3672 yuv_fixup = get_yuv_fixup(fixup);
3673 switch(yuv_fixup)
3674 {
3675 case YUV_FIXUP_YUY2:
3676 case YUV_FIXUP_UYVY:
3677 case YUV_FIXUP_YV12:
3678 TRACE("[OK]\n");
3679 return TRUE;
3680
3681 default:
3682 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3683 TRACE("[FAILED]\n");
3684 return FALSE;
3685 }
3686 }
3687
3688 const struct blit_shader arbfp_blit = {
3689 arbfp_blit_alloc,
3690 arbfp_blit_free,
3691 arbfp_blit_set,
3692 arbfp_blit_unset,
3693 arbfp_blit_color_fixup_supported,
3694 };
3695
3696 #undef GLINFO_LOCATION
Maarten's multimedia branch