?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Make the state table a property of the shader backend.
[wine/multimedia.git] / dlls / wined3d / arb_program_shader.c
blob1331bf8b66ea0df109fa0f4a2d5831ee0ac8c558
1 /*
2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
4 *
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 *
13 * This library is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
17 *
18 * This library is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
22 *
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with this library; if not, write to the Free Software
25 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
26 */
27
28 #include "config.h"
29
30 #include <math.h>
31 #include <stdio.h>
32
33 #include "wined3d_private.h"
34
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
37
38 #define GLINFO_LOCATION (*gl_info)
39
40 /********************************************************
41 * ARB_[vertex/fragment]_program helper functions follow
42 ********************************************************/
43
44 /**
45 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
46 * When constant_list == NULL, it will load all the constants.
47 *
48 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
49 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
50 */
51 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
52 unsigned int max_constants, float* constants, char *dirty_consts) {
53 local_constant* lconst;
54 DWORD i, j;
55 unsigned int ret;
56
57 if (TRACE_ON(d3d_shader)) {
58 for(i = 0; i < max_constants; i++) {
59 if(!dirty_consts[i]) continue;
60 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
61 constants[i * 4 + 0], constants[i * 4 + 1],
62 constants[i * 4 + 2], constants[i * 4 + 3]);
63 }
64 }
65 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
66 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
67 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
68 float lcl_const[4];
69 for(i = 0; i < max_constants; i++) {
70 if(!dirty_consts[i]) continue;
71 dirty_consts[i] = 0;
72
73 j = 4 * i;
74 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
75 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
76 else lcl_const[0] = constants[j + 0];
77
78 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
79 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
80 else lcl_const[1] = constants[j + 1];
81
82 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
83 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
84 else lcl_const[2] = constants[j + 2];
85
86 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
87 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
88 else lcl_const[3] = constants[j + 3];
89
90 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
91 }
92 } else {
93 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
94 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
95 * or just reloading *all* constants at once
96 *
97 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
98 */
99 for(i = 0; i < max_constants; i++) {
100 if(!dirty_consts[i]) continue;
101
102 /* Find the next block of dirty constants */
103 dirty_consts[i] = 0;
104 j = i;
105 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
106 dirty_consts[i] = 0;
107 }
108
109 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
110 }
111 } else {
112 for(i = 0; i < max_constants; i++) {
113 if(dirty_consts[i]) {
114 dirty_consts[i] = 0;
115 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
116 }
117 }
118 }
119 }
120 checkGLcall("glProgramEnvParameter4fvARB()");
121
122 /* Load immediate constants */
123 if(This->baseShader.load_local_constsF) {
124 if (TRACE_ON(d3d_shader)) {
125 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
126 GLfloat* values = (GLfloat*)lconst->value;
127 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
128 values[0], values[1], values[2], values[3]);
129 }
130 }
131 /* Immediate constants are clamped for 1.X shaders at loading times */
132 ret = 0;
133 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
134 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
135 ret = max(ret, lconst->idx);
136 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
137 }
138 checkGLcall("glProgramEnvParameter4fvARB()");
139 return ret; /* The loaded immediate constants need reloading for the next shader */
140 } else {
141 return 0; /* No constants are dirty now */
142 }
143 }
144
145 /**
146 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
147 *
148 * We only support float constants in ARB at the moment, so don't
149 * worry about the Integers or Booleans
150 */
151 void shader_arb_load_constants(
152 IWineD3DDevice* device,
153 char usePixelShader,
154 char useVertexShader) {
155
156 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
157 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
158 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
159 unsigned char i;
160
161 if (useVertexShader) {
162 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
163
164 /* Load DirectX 9 float constants for vertex shader */
165 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
166 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
167 deviceImpl->highest_dirty_vs_const,
168 stateBlock->vertexShaderConstantF,
169 deviceImpl->activeContext->vshader_const_dirty);
170
171 /* Upload the position fixup */
172 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
173 }
174
175 if (usePixelShader) {
176
177 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
178 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
179
180 /* Load DirectX 9 float constants for pixel shader */
181 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
182 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
183 deviceImpl->highest_dirty_ps_const,
184 stateBlock->pixelShaderConstantF,
185 deviceImpl->activeContext->pshader_const_dirty);
186
187 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
188 /* The state manager takes care that this function is always called if the bump env matrix changes
189 */
190 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
191 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
192 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
193
194 if(psi->luminanceconst[i].const_num != -1) {
195 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
196 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
197 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
198 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
199 */
200 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
201 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
202 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
203 }
204 }
205
206 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
207 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
208 float comparison[4];
209 float mul_low[4];
210
211 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
212 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
213 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
214
215 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
216 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
217 } else {
218 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
219 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
220
221 mul_low[0] = 1.0; mul_low[1] = 1.0;
222 mul_low[2] = 1.0; mul_low[3] = 1.0;
223 }
224 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
225 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
226 checkGLcall("Load sRGB correction constants\n");
227 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
228 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
229
230 }
231 }
232 }
233
234 /* Generate the variable & register declarations for the ARB_vertex_program output target */
235 void shader_generate_arb_declarations(
236 IWineD3DBaseShader *iface,
237 shader_reg_maps* reg_maps,
238 SHADER_BUFFER* buffer,
239 WineD3D_GL_Info* gl_info) {
240
241 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
242 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
243 DWORD i, cur;
244 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
245 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
246 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
247 UINT extra_constants_needed = 0;
248 local_constant* lconst;
249
250 /* Temporary Output register */
251 shader_addline(buffer, "TEMP TMP_OUT;\n");
252
253 for(i = 0; i < This->baseShader.limits.temporary; i++) {
254 if (reg_maps->temporary[i])
255 shader_addline(buffer, "TEMP R%u;\n", i);
256 }
257
258 for (i = 0; i < This->baseShader.limits.address; i++) {
259 if (reg_maps->address[i])
260 shader_addline(buffer, "ADDRESS A%d;\n", i);
261 }
262
263 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
264 if (reg_maps->texcoord[i])
265 shader_addline(buffer,"TEMP T%u;\n", i);
266 }
267
268 /* Texture coordinate registers must be pre-loaded */
269 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
270 if (reg_maps->texcoord[i])
271 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
272 }
273
274 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
275 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
276 if(!reg_maps->bumpmat[i]) continue;
277
278 cur = ps->numbumpenvmatconsts;
279 ps->bumpenvmatconst[cur].const_num = -1;
280 ps->bumpenvmatconst[cur].texunit = i;
281 ps->luminanceconst[cur].const_num = -1;
282 ps->luminanceconst[cur].texunit = i;
283
284 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
285 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
286 * bump mapping.
287 */
288 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
289 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
290 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
291 i, ps->bumpenvmatconst[cur].const_num);
292 extra_constants_needed++;
293
294 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
295 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
296 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
297 i, ps->luminanceconst[cur].const_num);
298 extra_constants_needed++;
299 } else if(reg_maps->luminanceparams) {
300 FIXME("No free constant to load the luminance parameters\n");
301 }
302 } else {
303 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
304 }
305
306 ps->numbumpenvmatconsts = cur + 1;
307 }
308
309 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
310 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
311 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
312 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
313 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
314 * off again
315 */
316 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
317 /* The idea is that if srgb is enabled, then disabled, the constant loading code
318 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
319 * and comparison constants. If it disables it that way, the shader won't be recompiled
320 * and the code will stay in, so sRGB writing can be turned on again by setting the
321 * constants from the spec
322 */
323 ps_impl->srgb_mode_hardcoded = 0;
324 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
325 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
326 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
327 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
328 } else {
329 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
330 srgb_mul_low, srgb_mul_low, srgb_mul_low);
331 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
332 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
333 ps_impl->srgb_mode_hardcoded = 1;
334 }
335 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
336 * path if the comparison value is set to INF
337 */
338 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
339 srgb_pow, srgb_pow, srgb_pow);
340 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
341 srgb_mul_high, srgb_mul_high, srgb_mul_high);
342 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
343 srgb_sub_high, srgb_sub_high, srgb_sub_high);
344 ps_impl->srgb_enabled = 1;
345 } else if(pshader) {
346 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
347
348 /* Do not write any srgb fixup into the shader to save shader size and processing time.
349 * As a consequence, we can't toggle srgb write on without recompilation
350 */
351 ps_impl->srgb_enabled = 0;
352 ps_impl->srgb_mode_hardcoded = 1;
353 }
354
355 /* Hardcodable local constants */
356 if(!This->baseShader.load_local_constsF) {
357 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
358 float *value = (float *) lconst->value;
359 shader_addline(buffer, "PARAM C%u = {%f, %f, %f, %f};\n", lconst->idx,
360 value[0], value[1], value[2], value[3]);
361 }
362 }
363
364 /* we use the array-based constants array if the local constants are marked for loading,
365 * because then we use indirect addressing, or when the local constant list is empty,
366 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
367 * local constants do not declare the loaded constants as an array because ARB compilers usually
368 * do not optimize unused constants away
369 */
370 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
371 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
372 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
373 max_constantsF, max_constantsF - 1);
374 } else {
375 for(i = 0; i < max_constantsF; i++) {
376 if(!shader_constant_is_local(This, i)) {
377 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
378 }
379 }
380 }
381 }
382
383 static const char * const shift_tab[] = {
384 "dummy", /* 0 (none) */
385 "coefmul.x", /* 1 (x2) */
386 "coefmul.y", /* 2 (x4) */
387 "coefmul.z", /* 3 (x8) */
388 "coefmul.w", /* 4 (x16) */
389 "dummy", /* 5 (x32) */
390 "dummy", /* 6 (x64) */
391 "dummy", /* 7 (x128) */
392 "dummy", /* 8 (d256) */
393 "dummy", /* 9 (d128) */
394 "dummy", /* 10 (d64) */
395 "dummy", /* 11 (d32) */
396 "coefdiv.w", /* 12 (d16) */
397 "coefdiv.z", /* 13 (d8) */
398 "coefdiv.y", /* 14 (d4) */
399 "coefdiv.x" /* 15 (d2) */
400 };
401
402 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
403 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
404 char *ptr = write_mask;
405 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
406
407 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
408 *ptr++ = '.';
409 *ptr++ = 'x';
410 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
411 *ptr++ = '.';
412 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
413 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
414 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
415 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
416 }
417
418 *ptr = '\0';
419 }
420
421 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
422 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
423 * but addressed as "rgba". To fix this we need to swap the register's x
424 * and z components. */
425 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
426 char *ptr = swizzle_str;
427
428 /* swizzle bits fields: wwzzyyxx */
429 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
430 DWORD swizzle_x = swizzle & 0x03;
431 DWORD swizzle_y = (swizzle >> 2) & 0x03;
432 DWORD swizzle_z = (swizzle >> 4) & 0x03;
433 DWORD swizzle_w = (swizzle >> 6) & 0x03;
434
435 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
436 * generate a swizzle string. Unless we need to our own swizzling. */
437 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
438 *ptr++ = '.';
439 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
440 *ptr++ = swizzle_chars[swizzle_x];
441 } else {
442 *ptr++ = swizzle_chars[swizzle_x];
443 *ptr++ = swizzle_chars[swizzle_y];
444 *ptr++ = swizzle_chars[swizzle_z];
445 *ptr++ = swizzle_chars[swizzle_w];
446 }
447 }
448
449 *ptr = '\0';
450 }
451
452 static void pshader_get_register_name(IWineD3DBaseShader* iface,
453 const DWORD param, char* regstr) {
454
455 DWORD reg = param & WINED3DSP_REGNUM_MASK;
456 DWORD regtype = shader_get_regtype(param);
457 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
458
459 switch (regtype) {
460 case WINED3DSPR_TEMP:
461 sprintf(regstr, "R%u", reg);
462 break;
463 case WINED3DSPR_INPUT:
464 if (reg==0) {
465 strcpy(regstr, "fragment.color.primary");
466 } else {
467 strcpy(regstr, "fragment.color.secondary");
468 }
469 break;
470 case WINED3DSPR_CONST:
471 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
472 sprintf(regstr, "C[%u]", reg);
473 } else {
474 sprintf(regstr, "C%u", reg);
475 }
476 break;
477 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
478 sprintf(regstr,"T%u", reg);
479 break;
480 case WINED3DSPR_COLOROUT:
481 if (reg == 0)
482 sprintf(regstr, "TMP_COLOR");
483 else {
484 /* TODO: See GL_ARB_draw_buffers */
485 FIXME("Unsupported write to render target %u\n", reg);
486 sprintf(regstr, "unsupported_register");
487 }
488 break;
489 case WINED3DSPR_DEPTHOUT:
490 sprintf(regstr, "result.depth");
491 break;
492 case WINED3DSPR_ATTROUT:
493 sprintf(regstr, "oD[%u]", reg);
494 break;
495 case WINED3DSPR_TEXCRDOUT:
496 sprintf(regstr, "oT[%u]", reg);
497 break;
498 default:
499 FIXME("Unhandled register name Type(%d)\n", regtype);
500 sprintf(regstr, "unrecognized_register");
501 break;
502 }
503 }
504
505 /* TODO: merge with pixel shader */
506 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
507
508 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
509
510 /* oPos, oFog and oPts in D3D */
511 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
512
513 DWORD reg = param & WINED3DSP_REGNUM_MASK;
514 DWORD regtype = shader_get_regtype(param);
515 char tmpReg[255];
516 BOOL is_color = FALSE;
517
518 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
519 strcat(hwLine, " -");
520 } else {
521 strcat(hwLine, " ");
522 }
523
524 switch (regtype) {
525 case WINED3DSPR_TEMP:
526 sprintf(tmpReg, "R%u", reg);
527 strcat(hwLine, tmpReg);
528 break;
529 case WINED3DSPR_INPUT:
530
531 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
532 is_color = TRUE;
533
534 sprintf(tmpReg, "vertex.attrib[%u]", reg);
535 strcat(hwLine, tmpReg);
536 break;
537 case WINED3DSPR_CONST:
538 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
539 if(reg >= This->rel_offset) {
540 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
541 } else {
542 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
543 }
544 } else {
545 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
546 sprintf(tmpReg, "C[%u]", reg);
547 } else {
548 sprintf(tmpReg, "C%u", reg);
549 }
550 }
551 strcat(hwLine, tmpReg);
552 break;
553 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
554 sprintf(tmpReg, "A%u", reg);
555 strcat(hwLine, tmpReg);
556 break;
557 case WINED3DSPR_RASTOUT:
558 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
559 strcat(hwLine, tmpReg);
560 break;
561 case WINED3DSPR_ATTROUT:
562 if (reg==0) {
563 strcat(hwLine, "result.color.primary");
564 } else {
565 strcat(hwLine, "result.color.secondary");
566 }
567 break;
568 case WINED3DSPR_TEXCRDOUT:
569 sprintf(tmpReg, "result.texcoord[%u]", reg);
570 strcat(hwLine, tmpReg);
571 break;
572 default:
573 FIXME("Unknown reg type %d %d\n", regtype, reg);
574 strcat(hwLine, "unrecognized_register");
575 break;
576 }
577
578 if (!is_input) {
579 char write_mask[6];
580 shader_arb_get_write_mask(arg, param, write_mask);
581 strcat(hwLine, write_mask);
582 } else {
583 char swizzle[6];
584 shader_arb_get_swizzle(param, is_color, swizzle);
585 strcat(hwLine, swizzle);
586 }
587 }
588
589 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
590 SHADER_BUFFER* buffer = arg->buffer;
591 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
592 const char *tex_type;
593
594 switch(sampler_type) {
595 case WINED3DSTT_1D:
596 tex_type = "1D";
597 break;
598
599 case WINED3DSTT_2D:
600 {
601 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
602 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
603 if(device->stateBlock->textures[sampler_idx] &&
604 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
605 tex_type = "RECT";
606 } else {
607 tex_type = "2D";
608 }
609 break;
610 }
611
612 case WINED3DSTT_VOLUME:
613 tex_type = "3D";
614 break;
615
616 case WINED3DSTT_CUBE:
617 tex_type = "CUBE";
618 break;
619
620 default:
621 ERR("Unexpected texture type %d\n", sampler_type);
622 tex_type = "";
623 }
624
625 if (bias) {
626 /* Shouldn't be possible, but let's check for it */
627 if(projected) FIXME("Biased and Projected texture sampling\n");
628 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
629 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
630 } else if (projected) {
631 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
632 } else {
633 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
634 }
635 }
636
637 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
638 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
639 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
640 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
641 WINED3DFORMAT fmt;
642 WINED3DFORMAT conversion_group;
643 IWineD3DBaseTextureImpl *texture;
644 UINT i;
645 BOOL recorded = FALSE;
646 DWORD sampler_idx;
647 DWORD hex_version = shader->baseShader.hex_version;
648 char reg[256];
649 char writemask[6];
650
651 switch(arg->opcode->opcode) {
652 case WINED3DSIO_TEX:
653 if (hex_version < WINED3DPS_VERSION(2,0)) {
654 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
655 } else {
656 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
657 }
658 break;
659
660 case WINED3DSIO_TEXLDL:
661 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
662 return;
663
664 case WINED3DSIO_TEXDP3TEX:
665 case WINED3DSIO_TEXM3x3TEX:
666 case WINED3DSIO_TEXM3x3SPEC:
667 case WINED3DSIO_TEXM3x3VSPEC:
668 case WINED3DSIO_TEXBEM:
669 case WINED3DSIO_TEXREG2AR:
670 case WINED3DSIO_TEXREG2GB:
671 case WINED3DSIO_TEXREG2RGB:
672 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
673 break;
674
675 default:
676 /* Not a texture sampling instruction, nothing to do */
677 return;
678 };
679
680 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
681 if(texture) {
682 fmt = texture->resource.format;
683 conversion_group = texture->baseTexture.shader_conversion_group;
684 } else {
685 fmt = WINED3DFMT_UNKNOWN;
686 conversion_group = WINED3DFMT_UNKNOWN;
687 }
688
689 /* before doing anything, record the sampler with the format in the format conversion list,
690 * but check if it's not there already
691 */
692 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
693 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
694 recorded = TRUE;
695 }
696 }
697 if(!recorded) {
698 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
699 shader->baseShader.num_sampled_samplers++;
700 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
701 }
702
703 pshader_get_register_name(arg->shader, arg->dst, reg);
704 shader_arb_get_write_mask(arg, arg->dst, writemask);
705 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
706
707 switch(fmt) {
708 case WINED3DFMT_V8U8:
709 case WINED3DFMT_V16U16:
710 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
711 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
712 #if 0
713 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
714 * disabled until an application that needs it is found because it causes unneeded
715 * shader recompilation in some game
716 */
717 if(strlen(writemask) >= 4) {
718 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
719 }
720 #endif
721 } else {
722 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
723 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
724 * all registers, do so, this saves an instruction.
725 */
726 if(strlen(writemask) >= 5) {
727 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
728 } else if(strlen(writemask) >= 3) {
729 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
730 reg, writemask[1],
731 reg, writemask[1]);
732 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
733 reg, writemask[2],
734 reg, writemask[2]);
735 } else if(strlen(writemask) == 2) {
736 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
737 reg, writemask[1]);
738 }
739 }
740 break;
741
742 case WINED3DFMT_X8L8V8U8:
743 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
744 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
745 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
746 */
747 if(strlen(writemask) >= 3) {
748 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
749 reg, writemask[1],
750 reg, writemask[1]);
751 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
752 reg, writemask[2],
753 reg, writemask[2]);
754 } else if(strlen(writemask) == 2) {
755 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
756 reg, writemask[1],
757 reg, writemask[1]);
758 }
759 }
760 break;
761
762 case WINED3DFMT_L6V5U5:
763 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
764 if(strlen(writemask) >= 4) {
765 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
766 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
767 reg, writemask[2]);
768 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
769 reg, writemask[1], writemask[1],
770 reg, writemask[1], writemask[3]);
771 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
772 writemask[3]);
773 } else if(strlen(writemask) == 3) {
774 /* This is bad: We have VL, but we need VU */
775 FIXME("2 components sampled from a converted L6V5U5 texture\n");
776 } else {
777 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
778 reg, writemask[1],
779 reg, writemask[1]);
780 }
781 }
782 break;
783
784 case WINED3DFMT_Q8W8V8U8:
785 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
786 /* Correct the sign in all channels */
787 switch(strlen(writemask)) {
788 case 4:
789 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
790 reg, writemask[3],
791 reg, writemask[3]);
792 /* drop through */
793 case 3:
794 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
795 reg, writemask[2],
796 reg, writemask[2]);
797 /* drop through */
798 case 2:
799 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
800 reg, writemask[1],
801 reg, writemask[1]);
802 break;
803
804 /* Should not occur, since it's at minimum '.' and a letter */
805 case 1:
806 ERR("Unexpected writemask: \"%s\"\n", writemask);
807 break;
808
809 case 5:
810 default:
811 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
812 }
813 }
814 break;
815
816 /* stupid compiler */
817 default:
818 break;
819 }
820 }
821
822
823 static void pshader_gen_input_modifier_line (
824 IWineD3DBaseShader *iface,
825 SHADER_BUFFER* buffer,
826 const DWORD instr,
827 int tmpreg,
828 char *outregstr) {
829
830 /* Generate a line that does the input modifier computation and return the input register to use */
831 char regstr[256];
832 char swzstr[20];
833 int insert_line;
834
835 /* Assume a new line will be added */
836 insert_line = 1;
837
838 /* Get register name */
839 pshader_get_register_name(iface, instr, regstr);
840 shader_arb_get_swizzle(instr, FALSE, swzstr);
841
842 switch (instr & WINED3DSP_SRCMOD_MASK) {
843 case WINED3DSPSM_NONE:
844 sprintf(outregstr, "%s%s", regstr, swzstr);
845 insert_line = 0;
846 break;
847 case WINED3DSPSM_NEG:
848 sprintf(outregstr, "-%s%s", regstr, swzstr);
849 insert_line = 0;
850 break;
851 case WINED3DSPSM_BIAS:
852 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
853 break;
854 case WINED3DSPSM_BIASNEG:
855 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
856 break;
857 case WINED3DSPSM_SIGN:
858 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
859 break;
860 case WINED3DSPSM_SIGNNEG:
861 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
862 break;
863 case WINED3DSPSM_COMP:
864 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
865 break;
866 case WINED3DSPSM_X2:
867 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
868 break;
869 case WINED3DSPSM_X2NEG:
870 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
871 break;
872 case WINED3DSPSM_DZ:
873 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
874 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
875 break;
876 case WINED3DSPSM_DW:
877 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
878 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
879 break;
880 default:
881 sprintf(outregstr, "%s%s", regstr, swzstr);
882 insert_line = 0;
883 }
884
885 /* Return modified or original register, with swizzle */
886 if (insert_line)
887 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
888 }
889
890 static inline void pshader_gen_output_modifier_line(
891 SHADER_BUFFER* buffer,
892 int saturate,
893 char *write_mask,
894 int shift,
895 char *regstr) {
896
897 /* Generate a line that does the output modifier computation */
898 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
899 regstr, write_mask, regstr, shift_tab[shift]);
900 }
901
902 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
903 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
904
905 SHADER_BUFFER* buffer = arg->buffer;
906 char dst_name[50];
907 char src_name[2][50];
908 char dst_wmask[20];
909 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
910 BOOL has_bumpmat = FALSE;
911 int i;
912
913 for(i = 0; i < This->numbumpenvmatconsts; i++) {
914 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
915 has_bumpmat = TRUE;
916 break;
917 }
918 }
919
920 pshader_get_register_name(arg->shader, arg->dst, dst_name);
921 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
922 strcat(dst_name, dst_wmask);
923
924 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
925 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
926
927 if(has_bumpmat) {
928 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
929 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
930 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
931 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
932 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
933
934 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
935 } else {
936 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
937 }
938 }
939
940 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
941
942 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
943 SHADER_BUFFER* buffer = arg->buffer;
944 char dst_wmask[20];
945 char dst_name[50];
946 char src_name[3][50];
947 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
948 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
949
950 /* FIXME: support output modifiers */
951
952 /* Handle output register */
953 pshader_get_register_name(arg->shader, arg->dst, dst_name);
954 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
955
956 /* Generate input register names (with modifiers) */
957 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
958 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
959 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
960
961 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
962 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
963 arg->opcode_token & WINED3DSI_COISSUE) {
964 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
965 } else {
966 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
967 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
968 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
969 }
970 if (shift != 0)
971 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
972 }
973
974 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
975
976 SHADER_BUFFER* buffer = arg->buffer;
977 char dst_wmask[20];
978 char dst_name[50];
979 char src_name[3][50];
980 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
981 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
982
983 /* FIXME: support output modifiers */
984
985 /* Handle output register */
986 pshader_get_register_name(arg->shader, arg->dst, dst_name);
987 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
988
989 /* Generate input register names (with modifiers) */
990 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
991 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
992 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
993
994 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
995 src_name[0], src_name[2], src_name[1]);
996
997 if (shift != 0)
998 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
999 }
1000
1001 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1002 * dst = dot2(src0, src1) + src2 */
1003 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1004 SHADER_BUFFER* buffer = arg->buffer;
1005 char dst_wmask[20];
1006 char dst_name[50];
1007 char src_name[3][50];
1008 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1009 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1010
1011 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1012 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1013
1014 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1015 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1016 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1017
1018 /* Emulate a DP2 with a DP3 and 0.0 */
1019 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1020 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1021 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1022 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1023
1024 if (shift != 0)
1025 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1026 }
1027
1028 /* Map the opcode 1-to-1 to the GL code */
1029 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1030
1031 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1032 SHADER_BUFFER* buffer = arg->buffer;
1033 DWORD dst = arg->dst;
1034 DWORD* src = arg->src;
1035
1036 unsigned int i;
1037 char tmpLine[256];
1038
1039 /* Output token related */
1040 char output_rname[256];
1041 char output_wmask[20];
1042 BOOL saturate = FALSE;
1043 BOOL centroid = FALSE;
1044 BOOL partialprecision = FALSE;
1045 DWORD shift;
1046
1047 strcpy(tmpLine, curOpcode->glname);
1048
1049 /* Process modifiers */
1050 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
1051 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1052
1053 saturate = mask & WINED3DSPDM_SATURATE;
1054 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1055 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1056 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1057 if (mask)
1058 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1059
1060 if (centroid)
1061 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1062 }
1063 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1064
1065 /* Generate input and output registers */
1066 if (curOpcode->num_params > 0) {
1067 char operands[4][100];
1068
1069 /* Generate input register names (with modifiers) */
1070 for (i = 1; i < curOpcode->num_params; ++i)
1071 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1072
1073 /* Handle output register */
1074 pshader_get_register_name(arg->shader, dst, output_rname);
1075 strcpy(operands[0], output_rname);
1076 shader_arb_get_write_mask(arg, dst, output_wmask);
1077 strcat(operands[0], output_wmask);
1078
1079 if (saturate && (shift == 0))
1080 strcat(tmpLine, "_SAT");
1081 strcat(tmpLine, " ");
1082 strcat(tmpLine, operands[0]);
1083 for (i = 1; i < curOpcode->num_params; i++) {
1084 strcat(tmpLine, ", ");
1085 strcat(tmpLine, operands[i]);
1086 }
1087 strcat(tmpLine,";\n");
1088 shader_addline(buffer, tmpLine);
1089
1090 /* A shift requires another line. */
1091 if (shift != 0)
1092 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1093 }
1094 }
1095
1096 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1097 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1098 DWORD hex_version = This->baseShader.hex_version;
1099 SHADER_BUFFER* buffer = arg->buffer;
1100 char reg_dest[40];
1101
1102 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1103 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1104 */
1105 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1106
1107 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1108 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1109 shader_addline(buffer, "KIL %s;\n", reg_dest);
1110 } else {
1111 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1112 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1113 */
1114 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1115 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1116 shader_addline(buffer, "KIL TMP;\n");
1117 }
1118 }
1119
1120 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1121 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1122 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1123
1124 DWORD dst = arg->dst;
1125 DWORD* src = arg->src;
1126 SHADER_BUFFER* buffer = arg->buffer;
1127 DWORD hex_version = This->baseShader.hex_version;
1128 BOOL projected = FALSE, bias = FALSE;
1129
1130 char reg_dest[40];
1131 char reg_coord[40];
1132 DWORD reg_dest_code;
1133 DWORD reg_sampler_code;
1134
1135 /* All versions have a destination register */
1136 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1137 pshader_get_register_name(arg->shader, dst, reg_dest);
1138
1139 /* 1.0-1.3: Use destination register as coordinate source.
1140 1.4+: Use provided coordinate source register. */
1141 if (hex_version < WINED3DPS_VERSION(1,4))
1142 strcpy(reg_coord, reg_dest);
1143 else
1144 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1145
1146 /* 1.0-1.4: Use destination register number as texture code.
1147 2.0+: Use provided sampler number as texure code. */
1148 if (hex_version < WINED3DPS_VERSION(2,0))
1149 reg_sampler_code = reg_dest_code;
1150 else
1151 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1152
1153 /* projection flag:
1154 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1155 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1156 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1157 */
1158 if(hex_version < WINED3DPS_VERSION(1,4)) {
1159 DWORD flags = 0;
1160 if(reg_sampler_code < MAX_TEXTURES) {
1161 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1162 }
1163 if (flags & WINED3DTTFF_PROJECTED) {
1164 projected = TRUE;
1165 }
1166 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1167 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1168 if (src_mod == WINED3DSPSM_DZ) {
1169 projected = TRUE;
1170 } else if(src_mod == WINED3DSPSM_DW) {
1171 projected = TRUE;
1172 }
1173 } else {
1174 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1175 projected = TRUE;
1176 }
1177 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1178 bias = TRUE;
1179 }
1180 }
1181 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1182 }
1183
1184 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1185
1186 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1187 DWORD dst = arg->dst;
1188 SHADER_BUFFER* buffer = arg->buffer;
1189 DWORD hex_version = This->baseShader.hex_version;
1190
1191 char tmp[20];
1192 shader_arb_get_write_mask(arg, dst, tmp);
1193 if (hex_version != WINED3DPS_VERSION(1,4)) {
1194 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1195 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1196 } else {
1197 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1198 char reg_src[40];
1199
1200 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1201 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1202 }
1203 }
1204
1205 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1206
1207 SHADER_BUFFER* buffer = arg->buffer;
1208 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1209 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1210 DWORD flags;
1211
1212 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1213 char dst_str[8];
1214 char src_str[50];
1215
1216 sprintf(dst_str, "T%u", reg1);
1217 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1218 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1219 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1220 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1221 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1222 }
1223
1224 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1225
1226 SHADER_BUFFER* buffer = arg->buffer;
1227 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1228 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1229 DWORD flags;
1230
1231 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1232 char dst_str[8];
1233 char src_str[50];
1234
1235 sprintf(dst_str, "T%u", reg1);
1236 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1237 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1238 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1239 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1240 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1241 }
1242
1243 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1244
1245 SHADER_BUFFER* buffer = arg->buffer;
1246 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1247 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1248 DWORD flags;
1249 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1250 char dst_str[8];
1251 char src_str[50];
1252
1253 sprintf(dst_str, "T%u", reg1);
1254 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1255 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1256 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1257 }
1258
1259 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1260 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1261 BOOL has_bumpmat = FALSE;
1262 BOOL has_luminance = FALSE;
1263 int i;
1264
1265 DWORD dst = arg->dst;
1266 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1267 SHADER_BUFFER* buffer = arg->buffer;
1268
1269 char reg_coord[40];
1270 DWORD reg_dest_code;
1271
1272 /* All versions have a destination register */
1273 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1274 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1275 pshader_get_register_name(arg->shader, dst, reg_coord);
1276
1277 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1278 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1279 has_bumpmat = TRUE;
1280 break;
1281 }
1282 }
1283 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1284 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1285 has_luminance = TRUE;
1286 break;
1287 }
1288 }
1289
1290 if(has_bumpmat) {
1291 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1292
1293 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1294 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1295 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1296 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1297
1298 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1299 * so we can't let the GL handle this.
1300 */
1301 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1302 & WINED3DTTFF_PROJECTED) {
1303 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1304 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1305 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1306 } else {
1307 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1308 }
1309
1310 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1311
1312 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1313 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1314 src, reg_dest_code, reg_dest_code);
1315 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1316 }
1317
1318 } else {
1319 DWORD tf;
1320 if(reg_dest_code < MAX_TEXTURES) {
1321 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1322 } else {
1323 tf = 0;
1324 }
1325 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1326 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1327 }
1328 }
1329
1330 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1331
1332 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1333 SHADER_BUFFER* buffer = arg->buffer;
1334 char src0_name[50];
1335
1336 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1337 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1338 }
1339
1340 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1341
1342 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1343 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1344 DWORD flags;
1345 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1346 SHADER_BUFFER* buffer = arg->buffer;
1347 char dst_str[8];
1348 char src0_name[50];
1349
1350 sprintf(dst_str, "T%u", reg);
1351 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1352 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1353 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1354 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1355 }
1356
1357 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1358
1359 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1360 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1361 SHADER_BUFFER* buffer = arg->buffer;
1362 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1363 char src0_name[50];
1364
1365 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1366 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1367 current_state->texcoord_w[current_state->current_row++] = reg;
1368 }
1369
1370 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1371
1372 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1373 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1374 DWORD flags;
1375 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1376 SHADER_BUFFER* buffer = arg->buffer;
1377 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1378 char dst_str[8];
1379 char src0_name[50];
1380
1381 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1382 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1383
1384 /* Sample the texture using the calculated coordinates */
1385 sprintf(dst_str, "T%u", reg);
1386 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1387 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1388 current_state->current_row = 0;
1389 }
1390
1391 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1392
1393 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1394 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1395 DWORD flags;
1396 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1397 SHADER_BUFFER* buffer = arg->buffer;
1398 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1399 char dst_str[8];
1400 char src0_name[50];
1401
1402 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1403 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1404
1405 /* Construct the eye-ray vector from w coordinates */
1406 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1407 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1408 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1409
1410 /* Calculate reflection vector
1411 */
1412 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1413 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1414 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1415 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1416 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1417 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1418 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1419
1420 /* Sample the texture using the calculated coordinates */
1421 sprintf(dst_str, "T%u", reg);
1422 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1423 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1424 current_state->current_row = 0;
1425 }
1426
1427 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1428
1429 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1430 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1431 DWORD flags;
1432 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1433 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1434 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1435 SHADER_BUFFER* buffer = arg->buffer;
1436 char dst_str[8];
1437 char src0_name[50];
1438
1439 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1440 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1441
1442 /* Calculate reflection vector.
1443 *
1444 * dot(N, E)
1445 * TMP.xyz = 2 * --------- * N - E
1446 * dot(N, N)
1447 *
1448 * Which normalizes the normal vector
1449 */
1450 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1451 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1452 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1453 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1454 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1455 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1456
1457 /* Sample the texture using the calculated coordinates */
1458 sprintf(dst_str, "T%u", reg);
1459 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1460 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1461 current_state->current_row = 0;
1462 }
1463
1464 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1465 SHADER_BUFFER* buffer = arg->buffer;
1466 char dst_name[50];
1467
1468 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1469 * which is essentially an input, is the destination register because it is the first
1470 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1471 * here
1472 */
1473 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1474
1475 /* According to the msdn, the source register(must be r5) is unusable after
1476 * the texdepth instruction, so we're free to modify it
1477 */
1478 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1479
1480 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1481 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1482 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1483 */
1484 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1485 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1486 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1487 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1488 }
1489
1490 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1491 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1492 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1493 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1494 SHADER_BUFFER* buffer = arg->buffer;
1495 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1496 char src0[50];
1497 char dst_str[8];
1498
1499 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1500 shader_addline(buffer, "MOV TMP, 0.0;\n");
1501 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1502
1503 sprintf(dst_str, "T%u", sampler_idx);
1504 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1505 }
1506
1507 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1508 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1509 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1510 char src0[50];
1511 char dst_str[50];
1512 char dst_mask[6];
1513 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1514 SHADER_BUFFER* buffer = arg->buffer;
1515
1516 /* Handle output register */
1517 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1518 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1519
1520 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1521 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1522
1523 /* TODO: Handle output modifiers */
1524 }
1525
1526 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1527 * Perform the 3rd row of a 3x3 matrix multiply */
1528 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1529 SHADER_BUFFER* buffer = arg->buffer;
1530 char dst_str[50];
1531 char dst_mask[6];
1532 char src0[50];
1533 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1534
1535 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1536 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1537
1538 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1539 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1540 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1541
1542 /* TODO: Handle output modifiers */
1543 }
1544
1545 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1546 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1547 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1548 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1549 */
1550 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1551 SHADER_BUFFER* buffer = arg->buffer;
1552 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1553 char src0[50];
1554
1555 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1556 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1557
1558 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1559 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1560 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1561 */
1562 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1563 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1564 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1565 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1566 }
1567
1568 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1569 Vertex/Pixel shaders to ARB_vertex_program codes */
1570 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1571
1572 int i;
1573 int nComponents = 0;
1574 SHADER_OPCODE_ARG tmpArg;
1575
1576 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1577
1578 /* Set constants for the temporary argument */
1579 tmpArg.shader = arg->shader;
1580 tmpArg.buffer = arg->buffer;
1581 tmpArg.src[0] = arg->src[0];
1582 tmpArg.src_addr[0] = arg->src_addr[0];
1583 tmpArg.src_addr[1] = arg->src_addr[1];
1584 tmpArg.reg_maps = arg->reg_maps;
1585
1586 switch(arg->opcode->opcode) {
1587 case WINED3DSIO_M4x4:
1588 nComponents = 4;
1589 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1590 break;
1591 case WINED3DSIO_M4x3:
1592 nComponents = 3;
1593 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1594 break;
1595 case WINED3DSIO_M3x4:
1596 nComponents = 4;
1597 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1598 break;
1599 case WINED3DSIO_M3x3:
1600 nComponents = 3;
1601 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1602 break;
1603 case WINED3DSIO_M3x2:
1604 nComponents = 2;
1605 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1606 break;
1607 default:
1608 break;
1609 }
1610
1611 for (i = 0; i < nComponents; i++) {
1612 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1613 tmpArg.src[1] = arg->src[1]+i;
1614 vshader_hw_map2gl(&tmpArg);
1615 }
1616 }
1617
1618 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1619 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1620 SHADER_BUFFER* buffer = arg->buffer;
1621 DWORD dst = arg->dst;
1622 DWORD src = arg->src[0];
1623 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1624
1625 char tmpLine[256];
1626
1627 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1628 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1629 strcat(tmpLine, ",");
1630 vshader_program_add_param(arg, src, TRUE, tmpLine);
1631 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1632 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1633 * .w is used
1634 */
1635 strcat(tmpLine, ".w");
1636 }
1637
1638 shader_addline(buffer, "%s;\n", tmpLine);
1639 }
1640
1641 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1642 SHADER_BUFFER* buffer = arg->buffer;
1643 char dst_name[50];
1644 char src_name[50];
1645 char dst_wmask[20];
1646 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1647 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1648
1649 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1650 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1651
1652 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1653 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1654 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1655 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1656 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1657 src_name);
1658
1659 if (shift != 0)
1660 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1661 }
1662
1663 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1664 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1665 * must contain fixed constants. So we need a separate function to filter those constants and
1666 * can't use map2gl
1667 */
1668 SHADER_BUFFER* buffer = arg->buffer;
1669 char dst_name[50];
1670 char src_name[50];
1671 char dst_wmask[20];
1672 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1673 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1674
1675 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1676 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1677
1678 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1679 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1680 src_name);
1681
1682 if (shift != 0)
1683 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1684
1685 }
1686
1687 /* TODO: merge with pixel shader */
1688 /* Map the opcode 1-to-1 to the GL code */
1689 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1690
1691 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1692 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1693 SHADER_BUFFER* buffer = arg->buffer;
1694 DWORD dst = arg->dst;
1695 DWORD* src = arg->src;
1696
1697 DWORD dst_regtype = shader_get_regtype(dst);
1698 char tmpLine[256];
1699 unsigned int i;
1700
1701 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1702 if(shader->rel_offset) {
1703 memset(tmpLine, 0, sizeof(tmpLine));
1704 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1705 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1706 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1707 return;
1708 } else {
1709 strcpy(tmpLine, "ARL");
1710 }
1711 } else
1712 strcpy(tmpLine, curOpcode->glname);
1713
1714 if (curOpcode->num_params > 0) {
1715 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1716 for (i = 1; i < curOpcode->num_params; ++i) {
1717 strcat(tmpLine, ",");
1718 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1719 }
1720 }
1721 shader_addline(buffer, "%s;\n", tmpLine);
1722 }
1723
1724 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1725 GLuint program_id = 0;
1726 const char *blt_vprogram =
1727 "!!ARBvp1.0\n"
1728 "PARAM c[1] = { { 1, 0.5 } };\n"
1729 "MOV result.position, vertex.position;\n"
1730 "MOV result.color, c[0].x;\n"
1731 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1732 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1733 "END\n";
1734
1735 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1736 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1737 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1738
1739 if (glGetError() == GL_INVALID_OPERATION) {
1740 GLint pos;
1741 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1742 FIXME("Vertex program error at position %d: %s\n", pos,
1743 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1744 }
1745
1746 return program_id;
1747 }
1748
1749 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1750 GLuint program_id = 0;
1751 const char *blt_fprogram =
1752 "!!ARBfp1.0\n"
1753 "TEMP R0;\n"
1754 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1755 "MOV result.depth.z, R0.x;\n"
1756 "END\n";
1757
1758 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1759 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1760 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1761
1762 if (glGetError() == GL_INVALID_OPERATION) {
1763 GLint pos;
1764 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1765 FIXME("Fragment program error at position %d: %s\n", pos,
1766 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1767 }
1768
1769 return program_id;
1770 }
1771
1772 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1773 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1774 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1775
1776 if (useVS) {
1777 TRACE("Using vertex shader\n");
1778
1779 /* Bind the vertex program */
1780 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1781 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1782 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1783
1784 /* Enable OpenGL vertex programs */
1785 glEnable(GL_VERTEX_PROGRAM_ARB);
1786 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1787 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1788 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1789 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1790 glDisable(GL_VERTEX_PROGRAM_ARB);
1791 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1792 }
1793
1794 if (usePS) {
1795 TRACE("Using pixel shader\n");
1796
1797 /* Bind the fragment program */
1798 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1799 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1800 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1801
1802 /* Enable OpenGL fragment programs */
1803 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1804 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1805 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1806 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1807 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1808 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1809 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1810 }
1811 }
1812
1813 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1814 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1815 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1816 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1817
1818 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1819 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1820 glEnable(GL_VERTEX_PROGRAM_ARB);
1821
1822 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1823 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1824 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1825 }
1826
1827 static void shader_arb_destroy_depth_blt(IWineD3DDevice *iface) {
1828 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1829 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1830 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1831
1832 if(priv->depth_blt_vprogram_id) {
1833 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1834 priv->depth_blt_vprogram_id = 0;
1835 }
1836 if(priv->depth_blt_fprogram_id) {
1837 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1838 priv->depth_blt_fprogram_id = 0;
1839 }
1840 }
1841
1842 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1843 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1844 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1845 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1846 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1847 }
1848
1849 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1850 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1851 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1852
1853 ENTER_GL();
1854 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1855 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1856 LEAVE_GL();
1857 This->baseShader.prgId = 0;
1858 This->baseShader.is_compiled = FALSE;
1859 }
1860
1861 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1862 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1863 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1864 return WINED3D_OK;
1865 }
1866
1867 static void shader_arb_free(IWineD3DDevice *iface) {
1868 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1869 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1870 }
1871
1872 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1873 return TRUE;
1874 }
1875
1876 const shader_backend_t arb_program_shader_backend = {
1877 &shader_arb_select,
1878 &shader_arb_select_depth_blt,
1879 &shader_arb_destroy_depth_blt,
1880 &shader_arb_load_constants,
1881 &shader_arb_cleanup,
1882 &shader_arb_color_correction,
1883 &shader_arb_destroy,
1884 &shader_arb_alloc,
1885 &shader_arb_free,
1886 &shader_arb_dirty_const,
1887 FFPStateTable
1888 };
Maarten's multimedia branch