?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
wined3d: Correct an off-by-one error in constant dirtification.
[wine/wine64.git] / dlls / wined3d / arb_program_shader.c
blob42b107a4c493d55b40d302f63a1cf6a952adca2c
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 + 1);
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 /* Load local constants using the program-local space,
358 * this avoids reloading them each time the shader is used
359 */
360 if(!This->baseShader.load_local_constsF) {
361 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
362 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
363 lconst->idx);
364 }
365 }
366
367 /* we use the array-based constants array if the local constants are marked for loading,
368 * because then we use indirect addressing, or when the local constant list is empty,
369 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
370 * local constants do not declare the loaded constants as an array because ARB compilers usually
371 * do not optimize unused constants away
372 */
373 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
374 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
375 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
376 max_constantsF, max_constantsF - 1);
377 } else {
378 for(i = 0; i < max_constantsF; i++) {
379 if(!shader_constant_is_local(This, i)) {
380 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
381 }
382 }
383 }
384 }
385
386 static const char * const shift_tab[] = {
387 "dummy", /* 0 (none) */
388 "coefmul.x", /* 1 (x2) */
389 "coefmul.y", /* 2 (x4) */
390 "coefmul.z", /* 3 (x8) */
391 "coefmul.w", /* 4 (x16) */
392 "dummy", /* 5 (x32) */
393 "dummy", /* 6 (x64) */
394 "dummy", /* 7 (x128) */
395 "dummy", /* 8 (d256) */
396 "dummy", /* 9 (d128) */
397 "dummy", /* 10 (d64) */
398 "dummy", /* 11 (d32) */
399 "coefdiv.w", /* 12 (d16) */
400 "coefdiv.z", /* 13 (d8) */
401 "coefdiv.y", /* 14 (d4) */
402 "coefdiv.x" /* 15 (d2) */
403 };
404
405 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
406 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
407 char *ptr = write_mask;
408 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
409
410 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
411 *ptr++ = '.';
412 *ptr++ = 'x';
413 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
414 *ptr++ = '.';
415 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
416 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
417 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
418 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
419 }
420
421 *ptr = '\0';
422 }
423
424 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
425 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
426 * but addressed as "rgba". To fix this we need to swap the register's x
427 * and z components. */
428 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
429 char *ptr = swizzle_str;
430
431 /* swizzle bits fields: wwzzyyxx */
432 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
433 DWORD swizzle_x = swizzle & 0x03;
434 DWORD swizzle_y = (swizzle >> 2) & 0x03;
435 DWORD swizzle_z = (swizzle >> 4) & 0x03;
436 DWORD swizzle_w = (swizzle >> 6) & 0x03;
437
438 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
439 * generate a swizzle string. Unless we need to our own swizzling. */
440 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
441 *ptr++ = '.';
442 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
443 *ptr++ = swizzle_chars[swizzle_x];
444 } else {
445 *ptr++ = swizzle_chars[swizzle_x];
446 *ptr++ = swizzle_chars[swizzle_y];
447 *ptr++ = swizzle_chars[swizzle_z];
448 *ptr++ = swizzle_chars[swizzle_w];
449 }
450 }
451
452 *ptr = '\0';
453 }
454
455 static void pshader_get_register_name(IWineD3DBaseShader* iface,
456 const DWORD param, char* regstr) {
457
458 DWORD reg = param & WINED3DSP_REGNUM_MASK;
459 DWORD regtype = shader_get_regtype(param);
460 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
461
462 switch (regtype) {
463 case WINED3DSPR_TEMP:
464 sprintf(regstr, "R%u", reg);
465 break;
466 case WINED3DSPR_INPUT:
467 if (reg==0) {
468 strcpy(regstr, "fragment.color.primary");
469 } else {
470 strcpy(regstr, "fragment.color.secondary");
471 }
472 break;
473 case WINED3DSPR_CONST:
474 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
475 sprintf(regstr, "C[%u]", reg);
476 } else {
477 sprintf(regstr, "C%u", reg);
478 }
479 break;
480 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
481 sprintf(regstr,"T%u", reg);
482 break;
483 case WINED3DSPR_COLOROUT:
484 if (reg == 0)
485 sprintf(regstr, "TMP_COLOR");
486 else {
487 /* TODO: See GL_ARB_draw_buffers */
488 FIXME("Unsupported write to render target %u\n", reg);
489 sprintf(regstr, "unsupported_register");
490 }
491 break;
492 case WINED3DSPR_DEPTHOUT:
493 sprintf(regstr, "result.depth");
494 break;
495 case WINED3DSPR_ATTROUT:
496 sprintf(regstr, "oD[%u]", reg);
497 break;
498 case WINED3DSPR_TEXCRDOUT:
499 sprintf(regstr, "oT[%u]", reg);
500 break;
501 default:
502 FIXME("Unhandled register name Type(%d)\n", regtype);
503 sprintf(regstr, "unrecognized_register");
504 break;
505 }
506 }
507
508 /* TODO: merge with pixel shader */
509 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
510
511 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
512
513 /* oPos, oFog and oPts in D3D */
514 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
515
516 DWORD reg = param & WINED3DSP_REGNUM_MASK;
517 DWORD regtype = shader_get_regtype(param);
518 char tmpReg[255];
519 BOOL is_color = FALSE;
520
521 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
522 strcat(hwLine, " -");
523 } else {
524 strcat(hwLine, " ");
525 }
526
527 switch (regtype) {
528 case WINED3DSPR_TEMP:
529 sprintf(tmpReg, "R%u", reg);
530 strcat(hwLine, tmpReg);
531 break;
532 case WINED3DSPR_INPUT:
533
534 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
535 is_color = TRUE;
536
537 sprintf(tmpReg, "vertex.attrib[%u]", reg);
538 strcat(hwLine, tmpReg);
539 break;
540 case WINED3DSPR_CONST:
541 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
542 if(reg >= This->rel_offset) {
543 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
544 } else {
545 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
546 }
547 } else {
548 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
549 sprintf(tmpReg, "C[%u]", reg);
550 } else {
551 sprintf(tmpReg, "C%u", reg);
552 }
553 }
554 strcat(hwLine, tmpReg);
555 break;
556 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
557 sprintf(tmpReg, "A%u", reg);
558 strcat(hwLine, tmpReg);
559 break;
560 case WINED3DSPR_RASTOUT:
561 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
562 strcat(hwLine, tmpReg);
563 break;
564 case WINED3DSPR_ATTROUT:
565 if (reg==0) {
566 strcat(hwLine, "result.color.primary");
567 } else {
568 strcat(hwLine, "result.color.secondary");
569 }
570 break;
571 case WINED3DSPR_TEXCRDOUT:
572 sprintf(tmpReg, "result.texcoord[%u]", reg);
573 strcat(hwLine, tmpReg);
574 break;
575 default:
576 FIXME("Unknown reg type %d %d\n", regtype, reg);
577 strcat(hwLine, "unrecognized_register");
578 break;
579 }
580
581 if (!is_input) {
582 char write_mask[6];
583 shader_arb_get_write_mask(arg, param, write_mask);
584 strcat(hwLine, write_mask);
585 } else {
586 char swizzle[6];
587 shader_arb_get_swizzle(param, is_color, swizzle);
588 strcat(hwLine, swizzle);
589 }
590 }
591
592 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
593 SHADER_BUFFER* buffer = arg->buffer;
594 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
595 const char *tex_type;
596
597 switch(sampler_type) {
598 case WINED3DSTT_1D:
599 tex_type = "1D";
600 break;
601
602 case WINED3DSTT_2D:
603 {
604 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
605 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
606 if(device->stateBlock->textures[sampler_idx] &&
607 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
608 tex_type = "RECT";
609 } else {
610 tex_type = "2D";
611 }
612 break;
613 }
614
615 case WINED3DSTT_VOLUME:
616 tex_type = "3D";
617 break;
618
619 case WINED3DSTT_CUBE:
620 tex_type = "CUBE";
621 break;
622
623 default:
624 ERR("Unexpected texture type %d\n", sampler_type);
625 tex_type = "";
626 }
627
628 if (bias) {
629 /* Shouldn't be possible, but let's check for it */
630 if(projected) FIXME("Biased and Projected texture sampling\n");
631 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
632 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
633 } else if (projected) {
634 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
635 } else {
636 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
637 }
638 }
639
640 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
641 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
642 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
643 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
644 WINED3DFORMAT fmt;
645 WINED3DFORMAT conversion_group;
646 IWineD3DBaseTextureImpl *texture;
647 UINT i;
648 BOOL recorded = FALSE;
649 DWORD sampler_idx;
650 DWORD hex_version = shader->baseShader.hex_version;
651 char reg[256];
652 char writemask[6];
653
654 switch(arg->opcode->opcode) {
655 case WINED3DSIO_TEX:
656 if (hex_version < WINED3DPS_VERSION(2,0)) {
657 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
658 } else {
659 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
660 }
661 break;
662
663 case WINED3DSIO_TEXLDL:
664 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
665 return;
666
667 case WINED3DSIO_TEXDP3TEX:
668 case WINED3DSIO_TEXM3x3TEX:
669 case WINED3DSIO_TEXM3x3SPEC:
670 case WINED3DSIO_TEXM3x3VSPEC:
671 case WINED3DSIO_TEXBEM:
672 case WINED3DSIO_TEXREG2AR:
673 case WINED3DSIO_TEXREG2GB:
674 case WINED3DSIO_TEXREG2RGB:
675 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
676 break;
677
678 default:
679 /* Not a texture sampling instruction, nothing to do */
680 return;
681 };
682
683 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
684 if(texture) {
685 fmt = texture->resource.format;
686 conversion_group = texture->baseTexture.shader_conversion_group;
687 } else {
688 fmt = WINED3DFMT_UNKNOWN;
689 conversion_group = WINED3DFMT_UNKNOWN;
690 }
691
692 /* before doing anything, record the sampler with the format in the format conversion list,
693 * but check if it's not there already
694 */
695 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
696 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
697 recorded = TRUE;
698 }
699 }
700 if(!recorded) {
701 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
702 shader->baseShader.num_sampled_samplers++;
703 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
704 }
705
706 pshader_get_register_name(arg->shader, arg->dst, reg);
707 shader_arb_get_write_mask(arg, arg->dst, writemask);
708 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
709
710 switch(fmt) {
711 case WINED3DFMT_V8U8:
712 case WINED3DFMT_V16U16:
713 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
714 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
715 #if 0
716 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
717 * disabled until an application that needs it is found because it causes unneeded
718 * shader recompilation in some game
719 */
720 if(strlen(writemask) >= 4) {
721 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
722 }
723 #endif
724 } else {
725 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
726 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
727 * all registers, do so, this saves an instruction.
728 */
729 if(strlen(writemask) >= 5) {
730 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
731 } else if(strlen(writemask) >= 3) {
732 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
733 reg, writemask[1],
734 reg, writemask[1]);
735 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
736 reg, writemask[2],
737 reg, writemask[2]);
738 } else if(strlen(writemask) == 2) {
739 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
740 reg, writemask[1]);
741 }
742 }
743 break;
744
745 case WINED3DFMT_X8L8V8U8:
746 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
747 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
748 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
749 */
750 if(strlen(writemask) >= 3) {
751 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
752 reg, writemask[1],
753 reg, writemask[1]);
754 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
755 reg, writemask[2],
756 reg, writemask[2]);
757 } else if(strlen(writemask) == 2) {
758 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
759 reg, writemask[1],
760 reg, writemask[1]);
761 }
762 }
763 break;
764
765 case WINED3DFMT_L6V5U5:
766 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
767 if(strlen(writemask) >= 4) {
768 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
769 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
770 reg, writemask[2]);
771 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
772 reg, writemask[1], writemask[1],
773 reg, writemask[1], writemask[3]);
774 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
775 writemask[3]);
776 } else if(strlen(writemask) == 3) {
777 /* This is bad: We have VL, but we need VU */
778 FIXME("2 components sampled from a converted L6V5U5 texture\n");
779 } else {
780 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
781 reg, writemask[1],
782 reg, writemask[1]);
783 }
784 }
785 break;
786
787 case WINED3DFMT_Q8W8V8U8:
788 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
789 /* Correct the sign in all channels */
790 switch(strlen(writemask)) {
791 case 4:
792 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
793 reg, writemask[3],
794 reg, writemask[3]);
795 /* drop through */
796 case 3:
797 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
798 reg, writemask[2],
799 reg, writemask[2]);
800 /* drop through */
801 case 2:
802 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
803 reg, writemask[1],
804 reg, writemask[1]);
805 break;
806
807 /* Should not occur, since it's at minimum '.' and a letter */
808 case 1:
809 ERR("Unexpected writemask: \"%s\"\n", writemask);
810 break;
811
812 case 5:
813 default:
814 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
815 }
816 }
817 break;
818
819 /* stupid compiler */
820 default:
821 break;
822 }
823 }
824
825
826 static void pshader_gen_input_modifier_line (
827 IWineD3DBaseShader *iface,
828 SHADER_BUFFER* buffer,
829 const DWORD instr,
830 int tmpreg,
831 char *outregstr) {
832
833 /* Generate a line that does the input modifier computation and return the input register to use */
834 char regstr[256];
835 char swzstr[20];
836 int insert_line;
837
838 /* Assume a new line will be added */
839 insert_line = 1;
840
841 /* Get register name */
842 pshader_get_register_name(iface, instr, regstr);
843 shader_arb_get_swizzle(instr, FALSE, swzstr);
844
845 switch (instr & WINED3DSP_SRCMOD_MASK) {
846 case WINED3DSPSM_NONE:
847 sprintf(outregstr, "%s%s", regstr, swzstr);
848 insert_line = 0;
849 break;
850 case WINED3DSPSM_NEG:
851 sprintf(outregstr, "-%s%s", regstr, swzstr);
852 insert_line = 0;
853 break;
854 case WINED3DSPSM_BIAS:
855 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
856 break;
857 case WINED3DSPSM_BIASNEG:
858 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
859 break;
860 case WINED3DSPSM_SIGN:
861 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
862 break;
863 case WINED3DSPSM_SIGNNEG:
864 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
865 break;
866 case WINED3DSPSM_COMP:
867 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
868 break;
869 case WINED3DSPSM_X2:
870 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
871 break;
872 case WINED3DSPSM_X2NEG:
873 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
874 break;
875 case WINED3DSPSM_DZ:
876 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
877 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
878 break;
879 case WINED3DSPSM_DW:
880 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
881 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
882 break;
883 default:
884 sprintf(outregstr, "%s%s", regstr, swzstr);
885 insert_line = 0;
886 }
887
888 /* Return modified or original register, with swizzle */
889 if (insert_line)
890 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
891 }
892
893 static inline void pshader_gen_output_modifier_line(
894 SHADER_BUFFER* buffer,
895 int saturate,
896 char *write_mask,
897 int shift,
898 char *regstr) {
899
900 /* Generate a line that does the output modifier computation */
901 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
902 regstr, write_mask, regstr, shift_tab[shift]);
903 }
904
905 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
906 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
907
908 SHADER_BUFFER* buffer = arg->buffer;
909 char dst_name[50];
910 char src_name[2][50];
911 char dst_wmask[20];
912 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
913 BOOL has_bumpmat = FALSE;
914 int i;
915
916 for(i = 0; i < This->numbumpenvmatconsts; i++) {
917 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
918 has_bumpmat = TRUE;
919 break;
920 }
921 }
922
923 pshader_get_register_name(arg->shader, arg->dst, dst_name);
924 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
925 strcat(dst_name, dst_wmask);
926
927 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
928 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
929
930 if(has_bumpmat) {
931 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
932 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
933 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
934 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
935 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
936
937 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
938 } else {
939 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
940 }
941 }
942
943 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
944
945 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
946 SHADER_BUFFER* buffer = arg->buffer;
947 char dst_wmask[20];
948 char dst_name[50];
949 char src_name[3][50];
950 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
951 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
952
953 /* FIXME: support output modifiers */
954
955 /* Handle output register */
956 pshader_get_register_name(arg->shader, arg->dst, dst_name);
957 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
958
959 /* Generate input register names (with modifiers) */
960 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
961 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
962 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
963
964 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
965 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
966 arg->opcode_token & WINED3DSI_COISSUE) {
967 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
968 } else {
969 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
970 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
971 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
972 }
973 if (shift != 0)
974 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
975 }
976
977 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
978
979 SHADER_BUFFER* buffer = arg->buffer;
980 char dst_wmask[20];
981 char dst_name[50];
982 char src_name[3][50];
983 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
984 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
985
986 /* FIXME: support output modifiers */
987
988 /* Handle output register */
989 pshader_get_register_name(arg->shader, arg->dst, dst_name);
990 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
991
992 /* Generate input register names (with modifiers) */
993 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
994 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
995 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
996
997 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
998 src_name[0], src_name[2], src_name[1]);
999
1000 if (shift != 0)
1001 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1002 }
1003
1004 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1005 * dst = dot2(src0, src1) + src2 */
1006 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1007 SHADER_BUFFER* buffer = arg->buffer;
1008 char dst_wmask[20];
1009 char dst_name[50];
1010 char src_name[3][50];
1011 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1012 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1013
1014 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1015 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1016
1017 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1018 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1019 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1020
1021 /* Emulate a DP2 with a DP3 and 0.0 */
1022 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1023 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1024 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1025 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1026
1027 if (shift != 0)
1028 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1029 }
1030
1031 /* Map the opcode 1-to-1 to the GL code */
1032 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1033
1034 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1035 SHADER_BUFFER* buffer = arg->buffer;
1036 DWORD dst = arg->dst;
1037 DWORD* src = arg->src;
1038
1039 unsigned int i;
1040 char tmpLine[256];
1041
1042 /* Output token related */
1043 char output_rname[256];
1044 char output_wmask[20];
1045 BOOL saturate = FALSE;
1046 BOOL centroid = FALSE;
1047 BOOL partialprecision = FALSE;
1048 DWORD shift;
1049
1050 strcpy(tmpLine, curOpcode->glname);
1051
1052 /* Process modifiers */
1053 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
1054 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1055
1056 saturate = mask & WINED3DSPDM_SATURATE;
1057 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1058 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1059 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1060 if (mask)
1061 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1062
1063 if (centroid)
1064 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1065 }
1066 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1067
1068 /* Generate input and output registers */
1069 if (curOpcode->num_params > 0) {
1070 char operands[4][100];
1071
1072 /* Generate input register names (with modifiers) */
1073 for (i = 1; i < curOpcode->num_params; ++i)
1074 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1075
1076 /* Handle output register */
1077 pshader_get_register_name(arg->shader, dst, output_rname);
1078 strcpy(operands[0], output_rname);
1079 shader_arb_get_write_mask(arg, dst, output_wmask);
1080 strcat(operands[0], output_wmask);
1081
1082 if (saturate && (shift == 0))
1083 strcat(tmpLine, "_SAT");
1084 strcat(tmpLine, " ");
1085 strcat(tmpLine, operands[0]);
1086 for (i = 1; i < curOpcode->num_params; i++) {
1087 strcat(tmpLine, ", ");
1088 strcat(tmpLine, operands[i]);
1089 }
1090 strcat(tmpLine,";\n");
1091 shader_addline(buffer, tmpLine);
1092
1093 /* A shift requires another line. */
1094 if (shift != 0)
1095 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1096 }
1097 }
1098
1099 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1100 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1101 DWORD hex_version = This->baseShader.hex_version;
1102 SHADER_BUFFER* buffer = arg->buffer;
1103 char reg_dest[40];
1104
1105 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1106 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1107 */
1108 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1109
1110 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1111 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1112 shader_addline(buffer, "KIL %s;\n", reg_dest);
1113 } else {
1114 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1115 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1116 */
1117 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1118 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1119 shader_addline(buffer, "KIL TMP;\n");
1120 }
1121 }
1122
1123 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1124 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1125 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1126
1127 DWORD dst = arg->dst;
1128 DWORD* src = arg->src;
1129 SHADER_BUFFER* buffer = arg->buffer;
1130 DWORD hex_version = This->baseShader.hex_version;
1131 BOOL projected = FALSE, bias = FALSE;
1132
1133 char reg_dest[40];
1134 char reg_coord[40];
1135 DWORD reg_dest_code;
1136 DWORD reg_sampler_code;
1137
1138 /* All versions have a destination register */
1139 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1140 pshader_get_register_name(arg->shader, dst, reg_dest);
1141
1142 /* 1.0-1.3: Use destination register as coordinate source.
1143 1.4+: Use provided coordinate source register. */
1144 if (hex_version < WINED3DPS_VERSION(1,4))
1145 strcpy(reg_coord, reg_dest);
1146 else
1147 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1148
1149 /* 1.0-1.4: Use destination register number as texture code.
1150 2.0+: Use provided sampler number as texure code. */
1151 if (hex_version < WINED3DPS_VERSION(2,0))
1152 reg_sampler_code = reg_dest_code;
1153 else
1154 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1155
1156 /* projection flag:
1157 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1158 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1159 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1160 */
1161 if(hex_version < WINED3DPS_VERSION(1,4)) {
1162 DWORD flags = 0;
1163 if(reg_sampler_code < MAX_TEXTURES) {
1164 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1165 }
1166 if (flags & WINED3DTTFF_PROJECTED) {
1167 projected = TRUE;
1168 }
1169 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1170 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1171 if (src_mod == WINED3DSPSM_DZ) {
1172 projected = TRUE;
1173 } else if(src_mod == WINED3DSPSM_DW) {
1174 projected = TRUE;
1175 }
1176 } else {
1177 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1178 projected = TRUE;
1179 }
1180 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1181 bias = TRUE;
1182 }
1183 }
1184 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1185 }
1186
1187 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1188
1189 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1190 DWORD dst = arg->dst;
1191 SHADER_BUFFER* buffer = arg->buffer;
1192 DWORD hex_version = This->baseShader.hex_version;
1193
1194 char tmp[20];
1195 shader_arb_get_write_mask(arg, dst, tmp);
1196 if (hex_version != WINED3DPS_VERSION(1,4)) {
1197 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1198 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1199 } else {
1200 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1201 char reg_src[40];
1202
1203 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1204 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1205 }
1206 }
1207
1208 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1209
1210 SHADER_BUFFER* buffer = arg->buffer;
1211 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1212 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1213 DWORD flags;
1214
1215 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1216 char dst_str[8];
1217 char src_str[50];
1218
1219 sprintf(dst_str, "T%u", reg1);
1220 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1221 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1222 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1223 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1224 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1225 }
1226
1227 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1228
1229 SHADER_BUFFER* buffer = arg->buffer;
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 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1240 }
1241
1242 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1243
1244 SHADER_BUFFER* buffer = arg->buffer;
1245 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1246 char dst_str[8];
1247 char src_str[50];
1248
1249 sprintf(dst_str, "T%u", reg1);
1250 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1251 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1252 }
1253
1254 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1255 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1256 BOOL has_bumpmat = FALSE;
1257 BOOL has_luminance = FALSE;
1258 int i;
1259
1260 DWORD dst = arg->dst;
1261 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1262 SHADER_BUFFER* buffer = arg->buffer;
1263
1264 char reg_coord[40];
1265 DWORD reg_dest_code;
1266
1267 /* All versions have a destination register */
1268 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1269 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1270 pshader_get_register_name(arg->shader, dst, reg_coord);
1271
1272 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1273 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1274 has_bumpmat = TRUE;
1275 break;
1276 }
1277 }
1278 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1279 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1280 has_luminance = TRUE;
1281 break;
1282 }
1283 }
1284
1285 if(has_bumpmat) {
1286 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1287
1288 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1289 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1290 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1291 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1292
1293 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1294 * so we can't let the GL handle this.
1295 */
1296 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1297 & WINED3DTTFF_PROJECTED) {
1298 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1299 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1300 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1301 } else {
1302 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1303 }
1304
1305 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1306
1307 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1308 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1309 src, reg_dest_code, reg_dest_code);
1310 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1311 }
1312
1313 } else {
1314 DWORD tf;
1315 if(reg_dest_code < MAX_TEXTURES) {
1316 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1317 } else {
1318 tf = 0;
1319 }
1320 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1321 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1322 }
1323 }
1324
1325 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1326
1327 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1328 SHADER_BUFFER* buffer = arg->buffer;
1329 char src0_name[50];
1330
1331 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1332 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1333 }
1334
1335 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1336
1337 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1338 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1339 DWORD flags;
1340 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1341 SHADER_BUFFER* buffer = arg->buffer;
1342 char dst_str[8];
1343 char src0_name[50];
1344
1345 sprintf(dst_str, "T%u", reg);
1346 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1347 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1348 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1349 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1350 }
1351
1352 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1353
1354 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1355 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1356 SHADER_BUFFER* buffer = arg->buffer;
1357 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1358 char src0_name[50];
1359
1360 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1361 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1362 current_state->texcoord_w[current_state->current_row++] = reg;
1363 }
1364
1365 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1366
1367 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1368 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1369 DWORD flags;
1370 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1371 SHADER_BUFFER* buffer = arg->buffer;
1372 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1373 char dst_str[8];
1374 char src0_name[50];
1375
1376 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1377 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1378
1379 /* Sample the texture using the calculated coordinates */
1380 sprintf(dst_str, "T%u", reg);
1381 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1382 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1383 current_state->current_row = 0;
1384 }
1385
1386 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1387
1388 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1389 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1390 DWORD flags;
1391 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1392 SHADER_BUFFER* buffer = arg->buffer;
1393 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1394 char dst_str[8];
1395 char src0_name[50];
1396
1397 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1398 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1399
1400 /* Construct the eye-ray vector from w coordinates */
1401 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1402 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1403 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1404
1405 /* Calculate reflection vector
1406 */
1407 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1408 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1409 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1410 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1411 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1412 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1413 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1414
1415 /* Sample the texture using the calculated coordinates */
1416 sprintf(dst_str, "T%u", reg);
1417 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1418 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1419 current_state->current_row = 0;
1420 }
1421
1422 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1423
1424 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1425 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1426 DWORD flags;
1427 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1428 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1429 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1430 SHADER_BUFFER* buffer = arg->buffer;
1431 char dst_str[8];
1432 char src0_name[50];
1433
1434 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1435 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1436
1437 /* Calculate reflection vector.
1438 *
1439 * dot(N, E)
1440 * TMP.xyz = 2 * --------- * N - E
1441 * dot(N, N)
1442 *
1443 * Which normalizes the normal vector
1444 */
1445 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1446 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1447 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1448 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1449 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1450 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1451
1452 /* Sample the texture using the calculated coordinates */
1453 sprintf(dst_str, "T%u", reg);
1454 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1455 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1456 current_state->current_row = 0;
1457 }
1458
1459 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1460 SHADER_BUFFER* buffer = arg->buffer;
1461 char dst_name[50];
1462
1463 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1464 * which is essentially an input, is the destination register because it is the first
1465 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1466 * here
1467 */
1468 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1469
1470 /* According to the msdn, the source register(must be r5) is unusable after
1471 * the texdepth instruction, so we're free to modify it
1472 */
1473 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1474
1475 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1476 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1477 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1478 */
1479 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1480 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1481 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1482 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1483 }
1484
1485 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1486 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1487 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1488 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1489 SHADER_BUFFER* buffer = arg->buffer;
1490 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1491 char src0[50];
1492 char dst_str[8];
1493
1494 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1495 shader_addline(buffer, "MOV TMP, 0.0;\n");
1496 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1497
1498 sprintf(dst_str, "T%u", sampler_idx);
1499 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1500 }
1501
1502 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1503 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1504 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1505 char src0[50];
1506 char dst_str[50];
1507 char dst_mask[6];
1508 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1509 SHADER_BUFFER* buffer = arg->buffer;
1510
1511 /* Handle output register */
1512 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1513 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1514
1515 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1516 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1517
1518 /* TODO: Handle output modifiers */
1519 }
1520
1521 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1522 * Perform the 3rd row of a 3x3 matrix multiply */
1523 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1524 SHADER_BUFFER* buffer = arg->buffer;
1525 char dst_str[50];
1526 char dst_mask[6];
1527 char src0[50];
1528 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1529
1530 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1531 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1532
1533 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1534 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1535 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1536
1537 /* TODO: Handle output modifiers */
1538 }
1539
1540 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1541 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1542 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1543 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1544 */
1545 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1546 SHADER_BUFFER* buffer = arg->buffer;
1547 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1548 char src0[50];
1549
1550 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1551 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1552
1553 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1554 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1555 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1556 */
1557 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1558 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1559 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1560 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1561 }
1562
1563 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1564 Vertex/Pixel shaders to ARB_vertex_program codes */
1565 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1566
1567 int i;
1568 int nComponents = 0;
1569 SHADER_OPCODE_ARG tmpArg;
1570
1571 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1572
1573 /* Set constants for the temporary argument */
1574 tmpArg.shader = arg->shader;
1575 tmpArg.buffer = arg->buffer;
1576 tmpArg.src[0] = arg->src[0];
1577 tmpArg.src_addr[0] = arg->src_addr[0];
1578 tmpArg.src_addr[1] = arg->src_addr[1];
1579 tmpArg.reg_maps = arg->reg_maps;
1580
1581 switch(arg->opcode->opcode) {
1582 case WINED3DSIO_M4x4:
1583 nComponents = 4;
1584 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1585 break;
1586 case WINED3DSIO_M4x3:
1587 nComponents = 3;
1588 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1589 break;
1590 case WINED3DSIO_M3x4:
1591 nComponents = 4;
1592 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1593 break;
1594 case WINED3DSIO_M3x3:
1595 nComponents = 3;
1596 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1597 break;
1598 case WINED3DSIO_M3x2:
1599 nComponents = 2;
1600 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1601 break;
1602 default:
1603 break;
1604 }
1605
1606 for (i = 0; i < nComponents; i++) {
1607 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1608 tmpArg.src[1] = arg->src[1]+i;
1609 vshader_hw_map2gl(&tmpArg);
1610 }
1611 }
1612
1613 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1614 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1615 SHADER_BUFFER* buffer = arg->buffer;
1616 DWORD dst = arg->dst;
1617 DWORD src = arg->src[0];
1618 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1619
1620 char tmpLine[256];
1621
1622 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1623 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1624 strcat(tmpLine, ",");
1625 vshader_program_add_param(arg, src, TRUE, tmpLine);
1626 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1627 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1628 * .w is used
1629 */
1630 strcat(tmpLine, ".w");
1631 }
1632
1633 shader_addline(buffer, "%s;\n", tmpLine);
1634 }
1635
1636 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1637 SHADER_BUFFER* buffer = arg->buffer;
1638 char dst_name[50];
1639 char src_name[50];
1640 char dst_wmask[20];
1641 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1642 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1643
1644 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1645 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1646
1647 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1648 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1649 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1650 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1651 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1652 src_name);
1653
1654 if (shift != 0)
1655 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1656 }
1657
1658 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1659 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1660 * must contain fixed constants. So we need a separate function to filter those constants and
1661 * can't use map2gl
1662 */
1663 SHADER_BUFFER* buffer = arg->buffer;
1664 char dst_name[50];
1665 char src_name[50];
1666 char dst_wmask[20];
1667 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1668 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1669
1670 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1671 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1672
1673 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1674 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1675 src_name);
1676
1677 if (shift != 0)
1678 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1679
1680 }
1681
1682 /* TODO: merge with pixel shader */
1683 /* Map the opcode 1-to-1 to the GL code */
1684 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1685
1686 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1687 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1688 SHADER_BUFFER* buffer = arg->buffer;
1689 DWORD dst = arg->dst;
1690 DWORD* src = arg->src;
1691
1692 DWORD dst_regtype = shader_get_regtype(dst);
1693 char tmpLine[256];
1694 unsigned int i;
1695
1696 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1697 if(shader->rel_offset) {
1698 memset(tmpLine, 0, sizeof(tmpLine));
1699 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1700 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1701 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1702 return;
1703 } else {
1704 strcpy(tmpLine, "ARL");
1705 }
1706 } else
1707 strcpy(tmpLine, curOpcode->glname);
1708
1709 if (curOpcode->num_params > 0) {
1710 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1711 for (i = 1; i < curOpcode->num_params; ++i) {
1712 strcat(tmpLine, ",");
1713 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1714 }
1715 }
1716 shader_addline(buffer, "%s;\n", tmpLine);
1717 }
1718
1719 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1720 GLuint program_id = 0;
1721 const char *blt_vprogram =
1722 "!!ARBvp1.0\n"
1723 "PARAM c[1] = { { 1, 0.5 } };\n"
1724 "MOV result.position, vertex.position;\n"
1725 "MOV result.color, c[0].x;\n"
1726 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1727 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1728 "END\n";
1729
1730 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1731 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1732 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1733
1734 if (glGetError() == GL_INVALID_OPERATION) {
1735 GLint pos;
1736 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1737 FIXME("Vertex program error at position %d: %s\n", pos,
1738 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1739 }
1740
1741 return program_id;
1742 }
1743
1744 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1745 GLuint program_id = 0;
1746 const char *blt_fprogram =
1747 "!!ARBfp1.0\n"
1748 "TEMP R0;\n"
1749 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1750 "MOV result.depth.z, R0.x;\n"
1751 "END\n";
1752
1753 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1754 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1755 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1756
1757 if (glGetError() == GL_INVALID_OPERATION) {
1758 GLint pos;
1759 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1760 FIXME("Fragment program error at position %d: %s\n", pos,
1761 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1762 }
1763
1764 return program_id;
1765 }
1766
1767 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1768 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1769 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1770
1771 if (useVS) {
1772 TRACE("Using vertex shader\n");
1773
1774 /* Bind the vertex program */
1775 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1776 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1777 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1778
1779 /* Enable OpenGL vertex programs */
1780 glEnable(GL_VERTEX_PROGRAM_ARB);
1781 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1782 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1783 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1784 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1785 glDisable(GL_VERTEX_PROGRAM_ARB);
1786 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1787 }
1788
1789 if (usePS) {
1790 TRACE("Using pixel shader\n");
1791
1792 /* Bind the fragment program */
1793 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1794 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1795 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1796
1797 /* Enable OpenGL fragment programs */
1798 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1799 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1800 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1801 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1802 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1803 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1804 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1805 }
1806 }
1807
1808 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1809 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1810 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1811 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1812
1813 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1814 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1815 glEnable(GL_VERTEX_PROGRAM_ARB);
1816
1817 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1818 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1819 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1820 }
1821
1822 static void shader_arb_destroy_depth_blt(IWineD3DDevice *iface) {
1823 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1824 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1825 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1826
1827 if(priv->depth_blt_vprogram_id) {
1828 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1829 priv->depth_blt_vprogram_id = 0;
1830 }
1831 if(priv->depth_blt_fprogram_id) {
1832 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1833 priv->depth_blt_fprogram_id = 0;
1834 }
1835 }
1836
1837 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1838 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1839 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1840 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1841 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1842 }
1843
1844 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1845 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1846 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1847
1848 ENTER_GL();
1849 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1850 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1851 LEAVE_GL();
1852 This->baseShader.prgId = 0;
1853 This->baseShader.is_compiled = FALSE;
1854 }
1855
1856 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1857 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1858 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1859 return WINED3D_OK;
1860 }
1861
1862 static void shader_arb_free(IWineD3DDevice *iface) {
1863 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1864 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1865 }
1866
1867 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1868 return TRUE;
1869 }
1870
1871 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
1872 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1873 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1874 CONST DWORD *function = This->baseShader.function;
1875 const char *fragcolor;
1876 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1877 local_constant* lconst;
1878
1879 /* Create the hw ARB shader */
1880 shader_addline(buffer, "!!ARBfp1.0\n");
1881
1882 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1883 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1884 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1885 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1886 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1887 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1888 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1889 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1890
1891 /* Base Declarations */
1892 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1893
1894 /* We need two variables for fog blending */
1895 shader_addline(buffer, "TEMP TMP_FOG;\n");
1896 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
1897 shader_addline(buffer, "TEMP TMP_COLOR;\n");
1898 }
1899
1900 /* Base Shader Body */
1901 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1902
1903 /* calculate fog and blend it
1904 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
1905 * -1/(e-s) and e/(e-s) respectively.
1906 */
1907 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
1908
1909 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
1910 fragcolor = "R0";
1911 } else {
1912 fragcolor = "TMP_COLOR";
1913 }
1914 if(This->srgb_enabled) {
1915 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1916
1917 /* Calculate the > 0.0031308 case */
1918 shader_addline(buffer, "POW TMP.x, %s.x, srgb_pow.x;\n", fragcolor);
1919 shader_addline(buffer, "POW TMP.y, %s.y, srgb_pow.y;\n", fragcolor);
1920 shader_addline(buffer, "POW TMP.z, %s.z, srgb_pow.z;\n", fragcolor);
1921 shader_addline(buffer, "MUL TMP, TMP, srgb_mul_hi;\n");
1922 shader_addline(buffer, "SUB TMP, TMP, srgb_sub_hi;\n");
1923 /* Calculate the < case */
1924 shader_addline(buffer, "MUL TMP2, srgb_mul_low, %s;\n", fragcolor);
1925 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1926 shader_addline(buffer, "SLT TA, srgb_comparison, %s;\n", fragcolor);
1927 shader_addline(buffer, "SGE TB, srgb_comparison, %s;\n", fragcolor);
1928 /* Store the components > 0.0031308 in the destination */
1929 shader_addline(buffer, "MUL %s, TMP, TA;\n", fragcolor);
1930 /* Add the components that are < 0.0031308 */
1931 shader_addline(buffer, "MAD result.color.xyz, TMP2, TB, %s;\n", fragcolor);
1932 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1933 }
1934 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
1935 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
1936 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
1937 }
1938
1939 shader_addline(buffer, "END\n");
1940
1941 /* TODO: change to resource.glObjectHandle or something like that */
1942 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
1943
1944 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
1945 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
1946
1947 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
1948 /* Create the program and check for errors */
1949 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1950 buffer->bsize, buffer->buffer));
1951
1952 if (glGetError() == GL_INVALID_OPERATION) {
1953 GLint errPos;
1954 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
1955 FIXME("HW PixelShader Error at position %d: %s\n",
1956 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1957 This->baseShader.prgId = -1;
1958 }
1959
1960 /* Load immediate constants */
1961 if(!This->baseShader.load_local_constsF) {
1962 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
1963 float *value = (float *) lconst->value;
1964 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
1965 checkGLcall("glProgramLocalParameter4fvARB");
1966 }
1967 }
1968 }
1969
1970 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
1971 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
1972 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1973 CONST DWORD *function = This->baseShader.function;
1974 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1975 local_constant* lconst;
1976
1977 /* Create the hw ARB shader */
1978 shader_addline(buffer, "!!ARBvp1.0\n");
1979 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
1980
1981 /* Mesa supports only 95 constants */
1982 if (GL_VEND(MESA) || GL_VEND(WINE))
1983 This->baseShader.limits.constant_float =
1984 min(95, This->baseShader.limits.constant_float);
1985
1986 shader_addline(buffer, "TEMP TMP;\n");
1987
1988 /* Base Declarations */
1989 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1990
1991 /* We need a constant to fixup the final position */
1992 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
1993
1994 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
1995 * for output parameters. D3D in theory does not do that either, but some applications depend on a
1996 * proper initialization of the secondary color, and programs using the fixed function pipeline without
1997 * a replacement shader depend on the texcoord.w being set properly.
1998 *
1999 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2000 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2001 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2002 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2003 * this can eat a number of instructions, so skip it unless this cap is set as well
2004 */
2005 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2006 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2007
2008 if((GLINFO_LOCATION).set_texcoord_w) {
2009 int i;
2010 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2011 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2012 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2013 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2014 }
2015 }
2016 }
2017 }
2018
2019 /* Base Shader Body */
2020 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2021
2022 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2023 if (!reg_maps->fog)
2024 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2025
2026 /* Write the final position.
2027 *
2028 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2029 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2030 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2031 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2032 */
2033 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2034 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2035 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2036
2037 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2038 * and the glsl equivalent
2039 */
2040 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2041
2042 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2043
2044 shader_addline(buffer, "END\n");
2045
2046 /* TODO: change to resource.glObjectHandle or something like that */
2047 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2048
2049 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2050 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2051
2052 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2053 /* Create the program and check for errors */
2054 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2055 buffer->bsize, buffer->buffer));
2056
2057 if (glGetError() == GL_INVALID_OPERATION) {
2058 GLint errPos;
2059 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2060 FIXME("HW VertexShader Error at position %d: %s\n",
2061 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2062 This->baseShader.prgId = -1;
2063 }
2064
2065 /* Load immediate constants */
2066 if(!This->baseShader.load_local_constsF) {
2067 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2068 float *value = (float *) lconst->value;
2069 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2070 }
2071 }
2072 }
2073
2074 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2075 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2076 * then overwrite the shader specific ones
2077 */
2078 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2079
2080 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2081 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2082 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2083 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2084 }
2085
2086 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2087 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2088 pCaps->PixelShader1xMaxValue = 8.0;
2089 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2090 }
2091 }
2092
2093 static void shader_arb_load_init(void) {
2094 }
2095
2096 static void shader_arb_fragment_enable(IWineD3DDevice *iface, BOOL enable) {
2097 none_shader_backend.shader_fragment_enable(iface, enable);
2098 }
2099
2100 const shader_backend_t arb_program_shader_backend = {
2101 shader_arb_select,
2102 shader_arb_select_depth_blt,
2103 shader_arb_destroy_depth_blt,
2104 shader_arb_load_constants,
2105 shader_arb_cleanup,
2106 shader_arb_color_correction,
2107 shader_arb_destroy,
2108 shader_arb_alloc,
2109 shader_arb_free,
2110 shader_arb_dirty_const,
2111 shader_arb_generate_pshader,
2112 shader_arb_generate_vshader,
2113 shader_arb_get_caps,
2114 shader_arb_load_init,
2115 shader_arb_fragment_enable,
2116 FFPStateTable
2117 };