?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Spelling fixes.
[wine/multimedia.git] / dlls / wined3d / arb_program_shader.c
bloba165b9c2a5e98379d2bf6fef5857a5a21569ffd3
1 /*
2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
4 *
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13 *
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2.1 of the License, or (at your option) any later version.
18 *
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
23 *
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
27 */
28
29 #include "config.h"
30
31 #include <math.h>
32 #include <stdio.h>
33
34 #include "wined3d_private.h"
35
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
39
40 #define GLINFO_LOCATION (*gl_info)
41
42 /********************************************************
43 * ARB_[vertex/fragment]_program helper functions follow
44 ********************************************************/
45
46 /**
47 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
48 * When constant_list == NULL, it will load all the constants.
49 *
50 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
51 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
52 */
53 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
54 unsigned int max_constants, float* constants, char *dirty_consts) {
55 local_constant* lconst;
56 DWORD i, j;
57 unsigned int ret;
58
59 if (TRACE_ON(d3d_shader)) {
60 for(i = 0; i < max_constants; i++) {
61 if(!dirty_consts[i]) continue;
62 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
63 constants[i * 4 + 0], constants[i * 4 + 1],
64 constants[i * 4 + 2], constants[i * 4 + 3]);
65 }
66 }
67 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
68 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
69 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
70 float lcl_const[4];
71 for(i = 0; i < max_constants; i++) {
72 if(!dirty_consts[i]) continue;
73 dirty_consts[i] = 0;
74
75 j = 4 * i;
76 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
77 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
78 else lcl_const[0] = constants[j + 0];
79
80 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
81 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
82 else lcl_const[1] = constants[j + 1];
83
84 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
85 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
86 else lcl_const[2] = constants[j + 2];
87
88 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
89 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
90 else lcl_const[3] = constants[j + 3];
91
92 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
93 }
94 } else {
95 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
96 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
97 * or just reloading *all* constants at once
98 *
99 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
100 */
101 for(i = 0; i < max_constants; i++) {
102 if(!dirty_consts[i]) continue;
103
104 /* Find the next block of dirty constants */
105 dirty_consts[i] = 0;
106 j = i;
107 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
108 dirty_consts[i] = 0;
109 }
110
111 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
112 }
113 } else {
114 for(i = 0; i < max_constants; i++) {
115 if(dirty_consts[i]) {
116 dirty_consts[i] = 0;
117 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
118 }
119 }
120 }
121 }
122 checkGLcall("glProgramEnvParameter4fvARB()");
123
124 /* Load immediate constants */
125 if(This->baseShader.load_local_constsF) {
126 if (TRACE_ON(d3d_shader)) {
127 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
128 GLfloat* values = (GLfloat*)lconst->value;
129 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
130 values[0], values[1], values[2], values[3]);
131 }
132 }
133 /* Immediate constants are clamped for 1.X shaders at loading times */
134 ret = 0;
135 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
136 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
137 ret = max(ret, lconst->idx);
138 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
139 }
140 checkGLcall("glProgramEnvParameter4fvARB()");
141 return ret; /* The loaded immediate constants need reloading for the next shader */
142 } else {
143 return 0; /* No constants are dirty now */
144 }
145 }
146
147 /**
148 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
149 *
150 * We only support float constants in ARB at the moment, so don't
151 * worry about the Integers or Booleans
152 */
153 void shader_arb_load_constants(
154 IWineD3DDevice* device,
155 char usePixelShader,
156 char useVertexShader) {
157
158 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
159 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
160 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
161 unsigned char i;
162
163 if (useVertexShader) {
164 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
165
166 /* Load DirectX 9 float constants for vertex shader */
167 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
168 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
169 deviceImpl->highest_dirty_vs_const,
170 stateBlock->vertexShaderConstantF,
171 deviceImpl->activeContext->vshader_const_dirty);
172
173 /* Upload the position fixup */
174 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
175 }
176
177 if (usePixelShader) {
178
179 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
180 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
181
182 /* Load DirectX 9 float constants for pixel shader */
183 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
184 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
185 deviceImpl->highest_dirty_ps_const,
186 stateBlock->pixelShaderConstantF,
187 deviceImpl->activeContext->pshader_const_dirty);
188
189 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
190 /* The state manager takes care that this function is always called if the bump env matrix changes
191 */
192 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
193 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
194 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
195
196 if(psi->luminanceconst[i].const_num != -1) {
197 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
198 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
199 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
200 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
201 */
202 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
203 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
204 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
205 }
206 }
207
208 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
209 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
210 float comparison[4];
211 float mul_low[4];
212
213 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
214 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
215 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
216
217 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
218 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
219 } else {
220 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
221 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
222
223 mul_low[0] = 1.0; mul_low[1] = 1.0;
224 mul_low[2] = 1.0; mul_low[3] = 1.0;
225 }
226 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
227 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
228 checkGLcall("Load sRGB correction constants\n");
229 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
230 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
231
232 }
233 }
234 }
235
236 /* Generate the variable & register declarations for the ARB_vertex_program output target */
237 void shader_generate_arb_declarations(
238 IWineD3DBaseShader *iface,
239 shader_reg_maps* reg_maps,
240 SHADER_BUFFER* buffer,
241 WineD3D_GL_Info* gl_info) {
242
243 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
244 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
245 DWORD i, cur;
246 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
247 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
248 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
249 UINT extra_constants_needed = 0;
250 local_constant* lconst;
251
252 /* Temporary Output register */
253 shader_addline(buffer, "TEMP TMP_OUT;\n");
254
255 for(i = 0; i < This->baseShader.limits.temporary; i++) {
256 if (reg_maps->temporary[i])
257 shader_addline(buffer, "TEMP R%u;\n", i);
258 }
259
260 for (i = 0; i < This->baseShader.limits.address; i++) {
261 if (reg_maps->address[i])
262 shader_addline(buffer, "ADDRESS A%d;\n", i);
263 }
264
265 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
266 if (reg_maps->texcoord[i])
267 shader_addline(buffer,"TEMP T%u;\n", i);
268 }
269
270 /* Texture coordinate registers must be pre-loaded */
271 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
272 if (reg_maps->texcoord[i])
273 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
274 }
275
276 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
277 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
278 if(!reg_maps->bumpmat[i]) continue;
279
280 cur = ps->numbumpenvmatconsts;
281 ps->bumpenvmatconst[cur].const_num = -1;
282 ps->bumpenvmatconst[cur].texunit = i;
283 ps->luminanceconst[cur].const_num = -1;
284 ps->luminanceconst[cur].texunit = i;
285
286 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
287 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
288 * bump mapping.
289 */
290 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
291 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
292 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
293 i, ps->bumpenvmatconst[cur].const_num);
294 extra_constants_needed++;
295
296 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
297 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
298 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
299 i, ps->luminanceconst[cur].const_num);
300 extra_constants_needed++;
301 } else if(reg_maps->luminanceparams) {
302 FIXME("No free constant to load the luminance parameters\n");
303 }
304 } else {
305 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
306 }
307
308 ps->numbumpenvmatconsts = cur + 1;
309 }
310
311 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
312 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
313 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
314 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
315 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
316 * off again
317 */
318 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
319 /* The idea is that if srgb is enabled, then disabled, the constant loading code
320 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
321 * and comparison constants. If it disables it that way, the shader won't be recompiled
322 * and the code will stay in, so sRGB writing can be turned on again by setting the
323 * constants from the spec
324 */
325 ps_impl->srgb_mode_hardcoded = 0;
326 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
327 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
328 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
329 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
330 } else {
331 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
332 srgb_mul_low, srgb_mul_low, srgb_mul_low);
333 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
334 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
335 ps_impl->srgb_mode_hardcoded = 1;
336 }
337 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
338 * path if the comparison value is set to INF
339 */
340 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
341 srgb_pow, srgb_pow, srgb_pow);
342 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
343 srgb_mul_high, srgb_mul_high, srgb_mul_high);
344 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
345 srgb_sub_high, srgb_sub_high, srgb_sub_high);
346 ps_impl->srgb_enabled = 1;
347 } else if(pshader) {
348 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
349
350 /* Do not write any srgb fixup into the shader to save shader size and processing time.
351 * As a consequence, we can't toggle srgb write on without recompilation
352 */
353 ps_impl->srgb_enabled = 0;
354 ps_impl->srgb_mode_hardcoded = 1;
355 }
356
357 /* Hardcodable local constants */
358 if(!This->baseShader.load_local_constsF) {
359 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
360 float *value = (float *) lconst->value;
361 shader_addline(buffer, "PARAM C%u = {%f, %f, %f, %f};\n", lconst->idx,
362 value[0], value[1], value[2], value[3]);
363 }
364 }
365
366 /* we use the array-based constants array if the local constants are marked for loading,
367 * because then we use indirect addressing, or when the local constant list is empty,
368 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
369 * local constants do not declare the loaded constants as an array because ARB compilers usually
370 * do not optimize unused constants away
371 */
372 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
373 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
374 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
375 max_constantsF, max_constantsF - 1);
376 } else {
377 for(i = 0; i < max_constantsF; i++) {
378 if(!shader_constant_is_local(This, i)) {
379 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
380 }
381 }
382 }
383 }
384
385 static const char * const shift_tab[] = {
386 "dummy", /* 0 (none) */
387 "coefmul.x", /* 1 (x2) */
388 "coefmul.y", /* 2 (x4) */
389 "coefmul.z", /* 3 (x8) */
390 "coefmul.w", /* 4 (x16) */
391 "dummy", /* 5 (x32) */
392 "dummy", /* 6 (x64) */
393 "dummy", /* 7 (x128) */
394 "dummy", /* 8 (d256) */
395 "dummy", /* 9 (d128) */
396 "dummy", /* 10 (d64) */
397 "dummy", /* 11 (d32) */
398 "coefdiv.w", /* 12 (d16) */
399 "coefdiv.z", /* 13 (d8) */
400 "coefdiv.y", /* 14 (d4) */
401 "coefdiv.x" /* 15 (d2) */
402 };
403
404 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
405 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
406 char *ptr = write_mask;
407 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
408
409 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
410 *ptr++ = '.';
411 *ptr++ = 'x';
412 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
413 *ptr++ = '.';
414 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
415 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
416 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
417 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
418 }
419
420 *ptr = '\0';
421 }
422
423 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
424 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
425 * but addressed as "rgba". To fix this we need to swap the register's x
426 * and z components. */
427 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
428 char *ptr = swizzle_str;
429
430 /* swizzle bits fields: wwzzyyxx */
431 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
432 DWORD swizzle_x = swizzle & 0x03;
433 DWORD swizzle_y = (swizzle >> 2) & 0x03;
434 DWORD swizzle_z = (swizzle >> 4) & 0x03;
435 DWORD swizzle_w = (swizzle >> 6) & 0x03;
436
437 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
438 * generate a swizzle string. Unless we need to our own swizzling. */
439 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
440 *ptr++ = '.';
441 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
442 *ptr++ = swizzle_chars[swizzle_x];
443 } else {
444 *ptr++ = swizzle_chars[swizzle_x];
445 *ptr++ = swizzle_chars[swizzle_y];
446 *ptr++ = swizzle_chars[swizzle_z];
447 *ptr++ = swizzle_chars[swizzle_w];
448 }
449 }
450
451 *ptr = '\0';
452 }
453
454 static void pshader_get_register_name(IWineD3DBaseShader* iface,
455 const DWORD param, char* regstr) {
456
457 DWORD reg = param & WINED3DSP_REGNUM_MASK;
458 DWORD regtype = shader_get_regtype(param);
459 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
460
461 switch (regtype) {
462 case WINED3DSPR_TEMP:
463 sprintf(regstr, "R%u", reg);
464 break;
465 case WINED3DSPR_INPUT:
466 if (reg==0) {
467 strcpy(regstr, "fragment.color.primary");
468 } else {
469 strcpy(regstr, "fragment.color.secondary");
470 }
471 break;
472 case WINED3DSPR_CONST:
473 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
474 sprintf(regstr, "C[%u]", reg);
475 } else {
476 sprintf(regstr, "C%u", reg);
477 }
478 break;
479 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
480 sprintf(regstr,"T%u", reg);
481 break;
482 case WINED3DSPR_COLOROUT:
483 if (reg == 0)
484 sprintf(regstr, "TMP_COLOR");
485 else {
486 /* TODO: See GL_ARB_draw_buffers */
487 FIXME("Unsupported write to render target %u\n", reg);
488 sprintf(regstr, "unsupported_register");
489 }
490 break;
491 case WINED3DSPR_DEPTHOUT:
492 sprintf(regstr, "result.depth");
493 break;
494 case WINED3DSPR_ATTROUT:
495 sprintf(regstr, "oD[%u]", reg);
496 break;
497 case WINED3DSPR_TEXCRDOUT:
498 sprintf(regstr, "oT[%u]", reg);
499 break;
500 default:
501 FIXME("Unhandled register name Type(%d)\n", regtype);
502 sprintf(regstr, "unrecognized_register");
503 break;
504 }
505 }
506
507 /* TODO: merge with pixel shader */
508 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
509
510 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
511
512 /* oPos, oFog and oPts in D3D */
513 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
514
515 DWORD reg = param & WINED3DSP_REGNUM_MASK;
516 DWORD regtype = shader_get_regtype(param);
517 char tmpReg[255];
518 BOOL is_color = FALSE;
519
520 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
521 strcat(hwLine, " -");
522 } else {
523 strcat(hwLine, " ");
524 }
525
526 switch (regtype) {
527 case WINED3DSPR_TEMP:
528 sprintf(tmpReg, "R%u", reg);
529 strcat(hwLine, tmpReg);
530 break;
531 case WINED3DSPR_INPUT:
532
533 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
534 is_color = TRUE;
535
536 sprintf(tmpReg, "vertex.attrib[%u]", reg);
537 strcat(hwLine, tmpReg);
538 break;
539 case WINED3DSPR_CONST:
540 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
541 if(reg >= This->rel_offset) {
542 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
543 } else {
544 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
545 }
546 } else {
547 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
548 sprintf(tmpReg, "C[%u]", reg);
549 } else {
550 sprintf(tmpReg, "C%u", reg);
551 }
552 }
553 strcat(hwLine, tmpReg);
554 break;
555 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
556 sprintf(tmpReg, "A%u", reg);
557 strcat(hwLine, tmpReg);
558 break;
559 case WINED3DSPR_RASTOUT:
560 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
561 strcat(hwLine, tmpReg);
562 break;
563 case WINED3DSPR_ATTROUT:
564 if (reg==0) {
565 strcat(hwLine, "result.color.primary");
566 } else {
567 strcat(hwLine, "result.color.secondary");
568 }
569 break;
570 case WINED3DSPR_TEXCRDOUT:
571 sprintf(tmpReg, "result.texcoord[%u]", reg);
572 strcat(hwLine, tmpReg);
573 break;
574 default:
575 FIXME("Unknown reg type %d %d\n", regtype, reg);
576 strcat(hwLine, "unrecognized_register");
577 break;
578 }
579
580 if (!is_input) {
581 char write_mask[6];
582 shader_arb_get_write_mask(arg, param, write_mask);
583 strcat(hwLine, write_mask);
584 } else {
585 char swizzle[6];
586 shader_arb_get_swizzle(param, is_color, swizzle);
587 strcat(hwLine, swizzle);
588 }
589 }
590
591 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
592 SHADER_BUFFER* buffer = arg->buffer;
593 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
594 const char *tex_type;
595
596 switch(sampler_type) {
597 case WINED3DSTT_1D:
598 tex_type = "1D";
599 break;
600
601 case WINED3DSTT_2D:
602 {
603 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
604 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
605 if(device->stateBlock->textures[sampler_idx] &&
606 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
607 tex_type = "RECT";
608 } else {
609 tex_type = "2D";
610 }
611 break;
612 }
613
614 case WINED3DSTT_VOLUME:
615 tex_type = "3D";
616 break;
617
618 case WINED3DSTT_CUBE:
619 tex_type = "CUBE";
620 break;
621
622 default:
623 ERR("Unexpected texture type %d\n", sampler_type);
624 tex_type = "";
625 }
626
627 if (bias) {
628 /* Shouldn't be possible, but let's check for it */
629 if(projected) FIXME("Biased and Projected texture sampling\n");
630 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
631 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
632 } else if (projected) {
633 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
634 } else {
635 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
636 }
637 }
638
639 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
640 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
641 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
642 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
643 WINED3DFORMAT fmt;
644 WINED3DFORMAT conversion_group;
645 IWineD3DBaseTextureImpl *texture;
646 UINT i;
647 BOOL recorded = FALSE;
648 DWORD sampler_idx;
649 DWORD hex_version = shader->baseShader.hex_version;
650 char reg[256];
651 char writemask[6];
652
653 switch(arg->opcode->opcode) {
654 case WINED3DSIO_TEX:
655 if (hex_version < WINED3DPS_VERSION(2,0)) {
656 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
657 } else {
658 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
659 }
660 break;
661
662 case WINED3DSIO_TEXLDL:
663 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
664 return;
665
666 case WINED3DSIO_TEXDP3TEX:
667 case WINED3DSIO_TEXM3x3TEX:
668 case WINED3DSIO_TEXM3x3SPEC:
669 case WINED3DSIO_TEXM3x3VSPEC:
670 case WINED3DSIO_TEXBEM:
671 case WINED3DSIO_TEXREG2AR:
672 case WINED3DSIO_TEXREG2GB:
673 case WINED3DSIO_TEXREG2RGB:
674 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
675 break;
676
677 default:
678 /* Not a texture sampling instruction, nothing to do */
679 return;
680 };
681
682 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
683 if(texture) {
684 fmt = texture->resource.format;
685 conversion_group = texture->baseTexture.shader_conversion_group;
686 } else {
687 fmt = WINED3DFMT_UNKNOWN;
688 conversion_group = WINED3DFMT_UNKNOWN;
689 }
690
691 /* before doing anything, record the sampler with the format in the format conversion list,
692 * but check if it's not there already
693 */
694 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
695 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
696 recorded = TRUE;
697 }
698 }
699 if(!recorded) {
700 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
701 shader->baseShader.num_sampled_samplers++;
702 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
703 }
704
705 pshader_get_register_name(arg->shader, arg->dst, reg);
706 shader_arb_get_write_mask(arg, arg->dst, writemask);
707 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
708
709 switch(fmt) {
710 case WINED3DFMT_V8U8:
711 case WINED3DFMT_V16U16:
712 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
713 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
714 #if 0
715 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
716 * disabled until an application that needs it is found because it causes unneeded
717 * shader recompilation in some game
718 */
719 if(strlen(writemask) >= 4) {
720 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
721 }
722 #endif
723 } else {
724 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
725 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
726 * all registers, do so, this saves an instruction.
727 */
728 if(strlen(writemask) >= 5) {
729 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
730 } else if(strlen(writemask) >= 3) {
731 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
732 reg, writemask[1],
733 reg, writemask[1]);
734 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
735 reg, writemask[2],
736 reg, writemask[2]);
737 } else if(strlen(writemask) == 2) {
738 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
739 reg, writemask[1]);
740 }
741 }
742 break;
743
744 case WINED3DFMT_X8L8V8U8:
745 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
746 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
747 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
748 */
749 if(strlen(writemask) >= 3) {
750 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
751 reg, writemask[1],
752 reg, writemask[1]);
753 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
754 reg, writemask[2],
755 reg, writemask[2]);
756 } else if(strlen(writemask) == 2) {
757 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
758 reg, writemask[1],
759 reg, writemask[1]);
760 }
761 }
762 break;
763
764 case WINED3DFMT_L6V5U5:
765 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
766 if(strlen(writemask) >= 4) {
767 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
768 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
769 reg, writemask[2]);
770 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
771 reg, writemask[1], writemask[1],
772 reg, writemask[1], writemask[3]);
773 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
774 writemask[3]);
775 } else if(strlen(writemask) == 3) {
776 /* This is bad: We have VL, but we need VU */
777 FIXME("2 components sampled from a converted L6V5U5 texture\n");
778 } else {
779 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
780 reg, writemask[1],
781 reg, writemask[1]);
782 }
783 }
784 break;
785
786 case WINED3DFMT_Q8W8V8U8:
787 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
788 /* Correct the sign in all channels */
789 switch(strlen(writemask)) {
790 case 4:
791 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
792 reg, writemask[3],
793 reg, writemask[3]);
794 /* drop through */
795 case 3:
796 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
797 reg, writemask[2],
798 reg, writemask[2]);
799 /* drop through */
800 case 2:
801 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
802 reg, writemask[1],
803 reg, writemask[1]);
804 break;
805
806 /* Should not occur, since it's at minimum '.' and a letter */
807 case 1:
808 ERR("Unexpected writemask: \"%s\"\n", writemask);
809 break;
810
811 case 5:
812 default:
813 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
814 }
815 }
816 break;
817
818 /* stupid compiler */
819 default:
820 break;
821 }
822 }
823
824
825 static void pshader_gen_input_modifier_line (
826 IWineD3DBaseShader *iface,
827 SHADER_BUFFER* buffer,
828 const DWORD instr,
829 int tmpreg,
830 char *outregstr) {
831
832 /* Generate a line that does the input modifier computation and return the input register to use */
833 char regstr[256];
834 char swzstr[20];
835 int insert_line;
836
837 /* Assume a new line will be added */
838 insert_line = 1;
839
840 /* Get register name */
841 pshader_get_register_name(iface, instr, regstr);
842 shader_arb_get_swizzle(instr, FALSE, swzstr);
843
844 switch (instr & WINED3DSP_SRCMOD_MASK) {
845 case WINED3DSPSM_NONE:
846 sprintf(outregstr, "%s%s", regstr, swzstr);
847 insert_line = 0;
848 break;
849 case WINED3DSPSM_NEG:
850 sprintf(outregstr, "-%s%s", regstr, swzstr);
851 insert_line = 0;
852 break;
853 case WINED3DSPSM_BIAS:
854 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
855 break;
856 case WINED3DSPSM_BIASNEG:
857 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
858 break;
859 case WINED3DSPSM_SIGN:
860 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
861 break;
862 case WINED3DSPSM_SIGNNEG:
863 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
864 break;
865 case WINED3DSPSM_COMP:
866 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
867 break;
868 case WINED3DSPSM_X2:
869 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
870 break;
871 case WINED3DSPSM_X2NEG:
872 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
873 break;
874 case WINED3DSPSM_DZ:
875 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
876 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
877 break;
878 case WINED3DSPSM_DW:
879 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
880 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
881 break;
882 default:
883 sprintf(outregstr, "%s%s", regstr, swzstr);
884 insert_line = 0;
885 }
886
887 /* Return modified or original register, with swizzle */
888 if (insert_line)
889 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
890 }
891
892 static inline void pshader_gen_output_modifier_line(
893 SHADER_BUFFER* buffer,
894 int saturate,
895 char *write_mask,
896 int shift,
897 char *regstr) {
898
899 /* Generate a line that does the output modifier computation */
900 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
901 regstr, write_mask, regstr, shift_tab[shift]);
902 }
903
904 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
905 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
906
907 SHADER_BUFFER* buffer = arg->buffer;
908 char dst_name[50];
909 char src_name[2][50];
910 char dst_wmask[20];
911 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
912 BOOL has_bumpmat = FALSE;
913 int i;
914
915 for(i = 0; i < This->numbumpenvmatconsts; i++) {
916 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
917 has_bumpmat = TRUE;
918 break;
919 }
920 }
921
922 pshader_get_register_name(arg->shader, arg->dst, dst_name);
923 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
924 strcat(dst_name, dst_wmask);
925
926 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
927 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
928
929 if(has_bumpmat) {
930 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
931 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
932 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
933 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
934 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
935
936 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
937 } else {
938 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
939 }
940 }
941
942 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
943
944 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
945 SHADER_BUFFER* buffer = arg->buffer;
946 char dst_wmask[20];
947 char dst_name[50];
948 char src_name[3][50];
949 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
950 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
951
952 /* FIXME: support output modifiers */
953
954 /* Handle output register */
955 pshader_get_register_name(arg->shader, arg->dst, dst_name);
956 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
957
958 /* Generate input register names (with modifiers) */
959 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
960 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
961 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
962
963 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
964 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
965 arg->opcode_token & WINED3DSI_COISSUE) {
966 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
967 } else {
968 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
969 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
970 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
971 }
972 if (shift != 0)
973 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
974 }
975
976 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
977
978 SHADER_BUFFER* buffer = arg->buffer;
979 char dst_wmask[20];
980 char dst_name[50];
981 char src_name[3][50];
982 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
983 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
984
985 /* FIXME: support output modifiers */
986
987 /* Handle output register */
988 pshader_get_register_name(arg->shader, arg->dst, dst_name);
989 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
990
991 /* Generate input register names (with modifiers) */
992 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
993 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
994 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
995
996 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
997 src_name[0], src_name[2], src_name[1]);
998
999 if (shift != 0)
1000 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1001 }
1002
1003 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1004 * dst = dot2(src0, src1) + src2 */
1005 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1006 SHADER_BUFFER* buffer = arg->buffer;
1007 char dst_wmask[20];
1008 char dst_name[50];
1009 char src_name[3][50];
1010 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1011 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1012
1013 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1014 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1015
1016 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1017 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1018 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1019
1020 /* Emulate a DP2 with a DP3 and 0.0 */
1021 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1022 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1023 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1024 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1025
1026 if (shift != 0)
1027 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1028 }
1029
1030 /* Map the opcode 1-to-1 to the GL code */
1031 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1032
1033 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1034 SHADER_BUFFER* buffer = arg->buffer;
1035 DWORD dst = arg->dst;
1036 DWORD* src = arg->src;
1037
1038 unsigned int i;
1039 char tmpLine[256];
1040
1041 /* Output token related */
1042 char output_rname[256];
1043 char output_wmask[20];
1044 BOOL saturate = FALSE;
1045 BOOL centroid = FALSE;
1046 BOOL partialprecision = FALSE;
1047 DWORD shift;
1048
1049 strcpy(tmpLine, curOpcode->glname);
1050
1051 /* Process modifiers */
1052 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
1053 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1054
1055 saturate = mask & WINED3DSPDM_SATURATE;
1056 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1057 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1058 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1059 if (mask)
1060 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1061
1062 if (centroid)
1063 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1064 }
1065 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1066
1067 /* Generate input and output registers */
1068 if (curOpcode->num_params > 0) {
1069 char operands[4][100];
1070
1071 /* Generate input register names (with modifiers) */
1072 for (i = 1; i < curOpcode->num_params; ++i)
1073 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1074
1075 /* Handle output register */
1076 pshader_get_register_name(arg->shader, dst, output_rname);
1077 strcpy(operands[0], output_rname);
1078 shader_arb_get_write_mask(arg, dst, output_wmask);
1079 strcat(operands[0], output_wmask);
1080
1081 if (saturate && (shift == 0))
1082 strcat(tmpLine, "_SAT");
1083 strcat(tmpLine, " ");
1084 strcat(tmpLine, operands[0]);
1085 for (i = 1; i < curOpcode->num_params; i++) {
1086 strcat(tmpLine, ", ");
1087 strcat(tmpLine, operands[i]);
1088 }
1089 strcat(tmpLine,";\n");
1090 shader_addline(buffer, tmpLine);
1091
1092 /* A shift requires another line. */
1093 if (shift != 0)
1094 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1095 }
1096 }
1097
1098 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1099 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1100 DWORD hex_version = This->baseShader.hex_version;
1101 SHADER_BUFFER* buffer = arg->buffer;
1102 char reg_dest[40];
1103
1104 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1105 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1106 */
1107 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1108
1109 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1110 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1111 shader_addline(buffer, "KIL %s;\n", reg_dest);
1112 } else {
1113 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1114 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1115 */
1116 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1117 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1118 shader_addline(buffer, "KIL TMP;\n");
1119 }
1120 }
1121
1122 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1123 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1124 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1125
1126 DWORD dst = arg->dst;
1127 DWORD* src = arg->src;
1128 SHADER_BUFFER* buffer = arg->buffer;
1129 DWORD hex_version = This->baseShader.hex_version;
1130 BOOL projected = FALSE, bias = FALSE;
1131
1132 char reg_dest[40];
1133 char reg_coord[40];
1134 DWORD reg_dest_code;
1135 DWORD reg_sampler_code;
1136
1137 /* All versions have a destination register */
1138 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1139 pshader_get_register_name(arg->shader, dst, reg_dest);
1140
1141 /* 1.0-1.3: Use destination register as coordinate source.
1142 1.4+: Use provided coordinate source register. */
1143 if (hex_version < WINED3DPS_VERSION(1,4))
1144 strcpy(reg_coord, reg_dest);
1145 else
1146 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1147
1148 /* 1.0-1.4: Use destination register number as texture code.
1149 2.0+: Use provided sampler number as texure code. */
1150 if (hex_version < WINED3DPS_VERSION(2,0))
1151 reg_sampler_code = reg_dest_code;
1152 else
1153 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1154
1155 /* projection flag:
1156 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1157 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1158 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1159 */
1160 if(hex_version < WINED3DPS_VERSION(1,4)) {
1161 DWORD flags = 0;
1162 if(reg_sampler_code < MAX_TEXTURES) {
1163 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1164 }
1165 if (flags & WINED3DTTFF_PROJECTED) {
1166 projected = TRUE;
1167 }
1168 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1169 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1170 if (src_mod == WINED3DSPSM_DZ) {
1171 projected = TRUE;
1172 } else if(src_mod == WINED3DSPSM_DW) {
1173 projected = TRUE;
1174 }
1175 } else {
1176 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1177 projected = TRUE;
1178 }
1179 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1180 bias = TRUE;
1181 }
1182 }
1183 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1184 }
1185
1186 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1187
1188 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1189 DWORD dst = arg->dst;
1190 SHADER_BUFFER* buffer = arg->buffer;
1191 DWORD hex_version = This->baseShader.hex_version;
1192
1193 char tmp[20];
1194 shader_arb_get_write_mask(arg, dst, tmp);
1195 if (hex_version != WINED3DPS_VERSION(1,4)) {
1196 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1197 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1198 } else {
1199 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1200 char reg_src[40];
1201
1202 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1203 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1204 }
1205 }
1206
1207 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1208
1209 SHADER_BUFFER* buffer = arg->buffer;
1210 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1211 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1212 DWORD flags;
1213
1214 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1215 char dst_str[8];
1216 char src_str[50];
1217
1218 sprintf(dst_str, "T%u", reg1);
1219 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1220 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1221 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1222 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1223 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1224 }
1225
1226 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1227
1228 SHADER_BUFFER* buffer = arg->buffer;
1229 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1230 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1231 DWORD flags;
1232
1233 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1234 char dst_str[8];
1235 char src_str[50];
1236
1237 sprintf(dst_str, "T%u", reg1);
1238 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1239 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1240 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1241 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1242 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1243 }
1244
1245 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1246
1247 SHADER_BUFFER* buffer = arg->buffer;
1248 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1249 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1250 DWORD flags;
1251 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1252 char dst_str[8];
1253 char src_str[50];
1254
1255 sprintf(dst_str, "T%u", reg1);
1256 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1257 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1258 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1259 }
1260
1261 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1262 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1263 BOOL has_bumpmat = FALSE;
1264 BOOL has_luminance = FALSE;
1265 int i;
1266
1267 DWORD dst = arg->dst;
1268 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1269 SHADER_BUFFER* buffer = arg->buffer;
1270
1271 char reg_coord[40];
1272 DWORD reg_dest_code;
1273
1274 /* All versions have a destination register */
1275 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1276 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1277 pshader_get_register_name(arg->shader, dst, reg_coord);
1278
1279 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1280 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1281 has_bumpmat = TRUE;
1282 break;
1283 }
1284 }
1285 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1286 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1287 has_luminance = TRUE;
1288 break;
1289 }
1290 }
1291
1292 if(has_bumpmat) {
1293 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1294
1295 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1296 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1297 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1298 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1299
1300 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1301 * so we can't let the GL handle this.
1302 */
1303 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1304 & WINED3DTTFF_PROJECTED) {
1305 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1306 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1307 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1308 } else {
1309 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1310 }
1311
1312 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1313
1314 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1315 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1316 src, reg_dest_code, reg_dest_code);
1317 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1318 }
1319
1320 } else {
1321 DWORD tf;
1322 if(reg_dest_code < MAX_TEXTURES) {
1323 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1324 } else {
1325 tf = 0;
1326 }
1327 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1328 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1329 }
1330 }
1331
1332 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1333
1334 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1335 SHADER_BUFFER* buffer = arg->buffer;
1336 char src0_name[50];
1337
1338 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1339 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1340 }
1341
1342 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1343
1344 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1345 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1346 DWORD flags;
1347 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1348 SHADER_BUFFER* buffer = arg->buffer;
1349 char dst_str[8];
1350 char src0_name[50];
1351
1352 sprintf(dst_str, "T%u", reg);
1353 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1354 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1355 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1356 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1357 }
1358
1359 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1360
1361 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1362 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1363 SHADER_BUFFER* buffer = arg->buffer;
1364 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1365 char src0_name[50];
1366
1367 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1368 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1369 current_state->texcoord_w[current_state->current_row++] = reg;
1370 }
1371
1372 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1373
1374 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1375 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1376 DWORD flags;
1377 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1378 SHADER_BUFFER* buffer = arg->buffer;
1379 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1380 char dst_str[8];
1381 char src0_name[50];
1382
1383 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1384 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1385
1386 /* Sample the texture using the calculated coordinates */
1387 sprintf(dst_str, "T%u", reg);
1388 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1389 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1390 current_state->current_row = 0;
1391 }
1392
1393 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1394
1395 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1396 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1397 DWORD flags;
1398 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1399 SHADER_BUFFER* buffer = arg->buffer;
1400 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1401 char dst_str[8];
1402 char src0_name[50];
1403
1404 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1405 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1406
1407 /* Construct the eye-ray vector from w coordinates */
1408 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1409 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1410 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1411
1412 /* Calculate reflection vector
1413 */
1414 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1415 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1416 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1417 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1418 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1419 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1420 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1421
1422 /* Sample the texture using the calculated coordinates */
1423 sprintf(dst_str, "T%u", reg);
1424 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1425 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1426 current_state->current_row = 0;
1427 }
1428
1429 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1430
1431 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1432 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1433 DWORD flags;
1434 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1435 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1436 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1437 SHADER_BUFFER* buffer = arg->buffer;
1438 char dst_str[8];
1439 char src0_name[50];
1440
1441 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1442 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1443
1444 /* Calculate reflection vector.
1445 *
1446 * dot(N, E)
1447 * TMP.xyz = 2 * --------- * N - E
1448 * dot(N, N)
1449 *
1450 * Which normalizes the normal vector
1451 */
1452 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1453 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1454 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1455 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1456 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1457 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1458
1459 /* Sample the texture using the calculated coordinates */
1460 sprintf(dst_str, "T%u", reg);
1461 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1462 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1463 current_state->current_row = 0;
1464 }
1465
1466 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1467 SHADER_BUFFER* buffer = arg->buffer;
1468 char dst_name[50];
1469
1470 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1471 * which is essentially an input, is the destination register because it is the first
1472 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1473 * here
1474 */
1475 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1476
1477 /* According to the msdn, the source register(must be r5) is unusable after
1478 * the texdepth instruction, so we're free to modify it
1479 */
1480 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1481
1482 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1483 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1484 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1485 */
1486 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1487 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1488 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1489 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1490 }
1491
1492 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1493 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1494 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1495 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1496 SHADER_BUFFER* buffer = arg->buffer;
1497 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1498 char src0[50];
1499 char dst_str[8];
1500
1501 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1502 shader_addline(buffer, "MOV TMP, 0.0;\n");
1503 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1504
1505 sprintf(dst_str, "T%u", sampler_idx);
1506 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1507 }
1508
1509 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1510 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1511 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1512 char src0[50];
1513 char dst_str[50];
1514 char dst_mask[6];
1515 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1516 SHADER_BUFFER* buffer = arg->buffer;
1517
1518 /* Handle output register */
1519 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1520 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1521
1522 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1523 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1524
1525 /* TODO: Handle output modifiers */
1526 }
1527
1528 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1529 * Perform the 3rd row of a 3x3 matrix multiply */
1530 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1531 SHADER_BUFFER* buffer = arg->buffer;
1532 char dst_str[50];
1533 char dst_mask[6];
1534 char src0[50];
1535 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1536
1537 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1538 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1539
1540 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1541 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1542 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1543
1544 /* TODO: Handle output modifiers */
1545 }
1546
1547 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1548 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1549 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1550 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1551 */
1552 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1553 SHADER_BUFFER* buffer = arg->buffer;
1554 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1555 char src0[50];
1556
1557 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1558 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1559
1560 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1561 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1562 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1563 */
1564 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1565 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1566 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1567 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1568 }
1569
1570 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1571 Vertex/Pixel shaders to ARB_vertex_program codes */
1572 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1573
1574 int i;
1575 int nComponents = 0;
1576 SHADER_OPCODE_ARG tmpArg;
1577
1578 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1579
1580 /* Set constants for the temporary argument */
1581 tmpArg.shader = arg->shader;
1582 tmpArg.buffer = arg->buffer;
1583 tmpArg.src[0] = arg->src[0];
1584 tmpArg.src_addr[0] = arg->src_addr[0];
1585 tmpArg.src_addr[1] = arg->src_addr[1];
1586 tmpArg.reg_maps = arg->reg_maps;
1587
1588 switch(arg->opcode->opcode) {
1589 case WINED3DSIO_M4x4:
1590 nComponents = 4;
1591 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1592 break;
1593 case WINED3DSIO_M4x3:
1594 nComponents = 3;
1595 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1596 break;
1597 case WINED3DSIO_M3x4:
1598 nComponents = 4;
1599 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1600 break;
1601 case WINED3DSIO_M3x3:
1602 nComponents = 3;
1603 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1604 break;
1605 case WINED3DSIO_M3x2:
1606 nComponents = 2;
1607 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1608 break;
1609 default:
1610 break;
1611 }
1612
1613 for (i = 0; i < nComponents; i++) {
1614 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1615 tmpArg.src[1] = arg->src[1]+i;
1616 vshader_hw_map2gl(&tmpArg);
1617 }
1618 }
1619
1620 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1621 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1622 SHADER_BUFFER* buffer = arg->buffer;
1623 DWORD dst = arg->dst;
1624 DWORD src = arg->src[0];
1625 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1626
1627 char tmpLine[256];
1628
1629 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1630 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1631 strcat(tmpLine, ",");
1632 vshader_program_add_param(arg, src, TRUE, tmpLine);
1633 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1634 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1635 * .w is used
1636 */
1637 strcat(tmpLine, ".w");
1638 }
1639
1640 shader_addline(buffer, "%s;\n", tmpLine);
1641 }
1642
1643 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1644 SHADER_BUFFER* buffer = arg->buffer;
1645 char dst_name[50];
1646 char src_name[50];
1647 char dst_wmask[20];
1648 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1649 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1650
1651 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1652 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1653
1654 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1655 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1656 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1657 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1658 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1659 src_name);
1660
1661 if (shift != 0)
1662 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1663 }
1664
1665 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1666 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1667 * must contain fixed constants. So we need a separate function to filter those constants and
1668 * can't use map2gl
1669 */
1670 SHADER_BUFFER* buffer = arg->buffer;
1671 char dst_name[50];
1672 char src_name[50];
1673 char dst_wmask[20];
1674 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1675 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1676
1677 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1678 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1679
1680 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1681 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1682 src_name);
1683
1684 if (shift != 0)
1685 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1686
1687 }
1688
1689 /* TODO: merge with pixel shader */
1690 /* Map the opcode 1-to-1 to the GL code */
1691 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1692
1693 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1694 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1695 SHADER_BUFFER* buffer = arg->buffer;
1696 DWORD dst = arg->dst;
1697 DWORD* src = arg->src;
1698
1699 DWORD dst_regtype = shader_get_regtype(dst);
1700 char tmpLine[256];
1701 unsigned int i;
1702
1703 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1704 if(shader->rel_offset) {
1705 memset(tmpLine, 0, sizeof(tmpLine));
1706 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1707 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1708 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1709 return;
1710 } else {
1711 strcpy(tmpLine, "ARL");
1712 }
1713 } else
1714 strcpy(tmpLine, curOpcode->glname);
1715
1716 if (curOpcode->num_params > 0) {
1717 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1718 for (i = 1; i < curOpcode->num_params; ++i) {
1719 strcat(tmpLine, ",");
1720 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1721 }
1722 }
1723 shader_addline(buffer, "%s;\n", tmpLine);
1724 }
1725
1726 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1727 GLuint program_id = 0;
1728 const char *blt_vprogram =
1729 "!!ARBvp1.0\n"
1730 "PARAM c[1] = { { 1, 0.5 } };\n"
1731 "MOV result.position, vertex.position;\n"
1732 "MOV result.color, c[0].x;\n"
1733 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1734 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1735 "END\n";
1736
1737 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1738 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1739 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1740
1741 if (glGetError() == GL_INVALID_OPERATION) {
1742 GLint pos;
1743 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1744 FIXME("Vertex program error at position %d: %s\n", pos,
1745 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1746 }
1747
1748 return program_id;
1749 }
1750
1751 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1752 GLuint program_id = 0;
1753 const char *blt_fprogram =
1754 "!!ARBfp1.0\n"
1755 "TEMP R0;\n"
1756 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1757 "MOV result.depth.z, R0.x;\n"
1758 "END\n";
1759
1760 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1761 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1762 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1763
1764 if (glGetError() == GL_INVALID_OPERATION) {
1765 GLint pos;
1766 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1767 FIXME("Fragment program error at position %d: %s\n", pos,
1768 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1769 }
1770
1771 return program_id;
1772 }
1773
1774 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1775 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1776 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1777
1778 if (useVS) {
1779 TRACE("Using vertex shader\n");
1780
1781 /* Bind the vertex program */
1782 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1783 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1784 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1785
1786 /* Enable OpenGL vertex programs */
1787 glEnable(GL_VERTEX_PROGRAM_ARB);
1788 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1789 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1790 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1791 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1792 glDisable(GL_VERTEX_PROGRAM_ARB);
1793 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1794 }
1795
1796 if (usePS) {
1797 TRACE("Using pixel shader\n");
1798
1799 /* Bind the fragment program */
1800 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1801 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1802 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1803
1804 /* Enable OpenGL fragment programs */
1805 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1806 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1807 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1808 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1809 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1810 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1811 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1812 }
1813 }
1814
1815 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1816 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1817 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1818 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1819
1820 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1821 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1822 glEnable(GL_VERTEX_PROGRAM_ARB);
1823
1824 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1825 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1826 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1827 }
1828
1829 static void shader_arb_destroy_depth_blt(IWineD3DDevice *iface) {
1830 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1831 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1832 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1833
1834 if(priv->depth_blt_vprogram_id) {
1835 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1836 priv->depth_blt_vprogram_id = 0;
1837 }
1838 if(priv->depth_blt_fprogram_id) {
1839 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1840 priv->depth_blt_fprogram_id = 0;
1841 }
1842 }
1843
1844 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1845 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1846 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1847 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1848 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1849 }
1850
1851 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1852 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1853 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1854
1855 ENTER_GL();
1856 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1857 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1858 LEAVE_GL();
1859 This->baseShader.prgId = 0;
1860 This->baseShader.is_compiled = FALSE;
1861 }
1862
1863 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1864 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1865 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1866 return WINED3D_OK;
1867 }
1868
1869 static void shader_arb_free(IWineD3DDevice *iface) {
1870 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1871 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1872 }
1873
1874 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1875 return TRUE;
1876 }
1877
1878 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
1879 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1880 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1881 CONST DWORD *function = This->baseShader.function;
1882 const char *fragcolor;
1883 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1884
1885 /* Create the hw ARB shader */
1886 shader_addline(buffer, "!!ARBfp1.0\n");
1887
1888 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1889 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1890 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1891 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1892 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1893 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1894 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1895 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1896
1897 /* Base Declarations */
1898 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1899
1900 /* We need two variables for fog blending */
1901 shader_addline(buffer, "TEMP TMP_FOG;\n");
1902 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
1903 shader_addline(buffer, "TEMP TMP_COLOR;\n");
1904 }
1905
1906 /* Base Shader Body */
1907 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1908
1909 /* calculate fog and blend it
1910 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
1911 * -1/(e-s) and e/(e-s) respectively.
1912 */
1913 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
1914
1915 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1916 fragcolor = "R0";
1917 } else {
1918 fragcolor = "TMP_COLOR";
1919 }
1920 if(This->srgb_enabled) {
1921 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1922
1923 /* Calculate the > 0.0031308 case */
1924 shader_addline(buffer, "POW TMP.x, %s.x, srgb_pow.x;\n", fragcolor);
1925 shader_addline(buffer, "POW TMP.y, %s.y, srgb_pow.y;\n", fragcolor);
1926 shader_addline(buffer, "POW TMP.z, %s.z, srgb_pow.z;\n", fragcolor);
1927 shader_addline(buffer, "MUL TMP, TMP, srgb_mul_hi;\n");
1928 shader_addline(buffer, "SUB TMP, TMP, srgb_sub_hi;\n");
1929 /* Calculate the < case */
1930 shader_addline(buffer, "MUL TMP2, srgb_mul_low, %s;\n", fragcolor);
1931 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1932 shader_addline(buffer, "SLT TA, srgb_comparison, %s;\n", fragcolor);
1933 shader_addline(buffer, "SGE TB, srgb_comparison, %s;\n", fragcolor);
1934 /* Store the components > 0.0031308 in the destination */
1935 shader_addline(buffer, "MUL %s, TMP, TA;\n", fragcolor);
1936 /* Add the components that are < 0.0031308 */
1937 shader_addline(buffer, "MAD result.color.xyz, TMP2, TB, %s;\n", fragcolor);
1938 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1939 }
1940 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
1941 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
1942 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
1943 }
1944
1945 shader_addline(buffer, "END\n");
1946
1947 /* TODO: change to resource.glObjectHandle or something like that */
1948 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
1949
1950 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
1951 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
1952
1953 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
1954 /* Create the program and check for errors */
1955 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1956 buffer->bsize, buffer->buffer));
1957
1958 if (glGetError() == GL_INVALID_OPERATION) {
1959 GLint errPos;
1960 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
1961 FIXME("HW PixelShader Error at position %d: %s\n",
1962 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1963 This->baseShader.prgId = -1;
1964 }
1965 }
1966
1967 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
1968 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
1969 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1970 CONST DWORD *function = This->baseShader.function;
1971 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1972
1973 /* Create the hw ARB shader */
1974 shader_addline(buffer, "!!ARBvp1.0\n");
1975 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
1976
1977 /* Mesa supports only 95 constants */
1978 if (GL_VEND(MESA) || GL_VEND(WINE))
1979 This->baseShader.limits.constant_float =
1980 min(95, This->baseShader.limits.constant_float);
1981
1982 shader_addline(buffer, "TEMP TMP;\n");
1983
1984 /* Base Declarations */
1985 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1986
1987 /* We need a constant to fixup the final position */
1988 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
1989
1990 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
1991 * for output parameters. D3D in theory does not do that either, but some applications depend on a
1992 * proper initialization of the secondary color, and programs using the fixed function pipeline without
1993 * a replacement shader depend on the texcoord.w being set properly.
1994 *
1995 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
1996 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
1997 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
1998 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
1999 * this can eat a number of instructions, so skip it unless this cap is set as well
2000 */
2001 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2002 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2003
2004 if((GLINFO_LOCATION).set_texcoord_w) {
2005 int i;
2006 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2007 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2008 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2009 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2010 }
2011 }
2012 }
2013 }
2014
2015 /* Base Shader Body */
2016 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2017
2018 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2019 if (!reg_maps->fog)
2020 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2021
2022 /* Write the final position.
2023 *
2024 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2025 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2026 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2027 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2028 */
2029 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2030 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2031 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2032
2033 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2034 * and the glsl equivalent
2035 */
2036 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2037
2038 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2039
2040 shader_addline(buffer, "END\n");
2041
2042 /* TODO: change to resource.glObjectHandle or something like that */
2043 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2044
2045 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2046 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2047
2048 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2049 /* Create the program and check for errors */
2050 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2051 buffer->bsize, buffer->buffer));
2052
2053 if (glGetError() == GL_INVALID_OPERATION) {
2054 GLint errPos;
2055 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2056 FIXME("HW VertexShader Error at position %d: %s\n",
2057 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2058 This->baseShader.prgId = -1;
2059 }
2060 }
2061
2062 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2063 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2064 * then overwrite the shader specific ones
2065 */
2066 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2067
2068 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2069 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2070 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2071 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2072 }
2073
2074 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2075 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2076 pCaps->PixelShader1xMaxValue = 8.0;
2077 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2078 }
2079 }
2080
2081 static void shader_arb_load_init(void) {
2082 }
2083
2084 const shader_backend_t arb_program_shader_backend = {
2085 &shader_arb_select,
2086 &shader_arb_select_depth_blt,
2087 &shader_arb_destroy_depth_blt,
2088 &shader_arb_load_constants,
2089 &shader_arb_cleanup,
2090 &shader_arb_color_correction,
2091 &shader_arb_destroy,
2092 &shader_arb_alloc,
2093 &shader_arb_free,
2094 &shader_arb_dirty_const,
2095 &shader_arb_generate_pshader,
2096 &shader_arb_generate_vshader,
2097 &shader_arb_get_caps,
2098 &shader_arb_load_init,
2099 FFPStateTable
2100 };
Maarten's multimedia branch