2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
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
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.
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.
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
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
40 #define GLINFO_LOCATION (*gl_info)
42 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
43 * so upload them above that
45 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
46 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
48 /* ARB_program_shader private data */
49 struct shader_arb_priv {
50 GLuint current_vprogram_id;
51 GLuint current_fprogram_id;
52 GLuint depth_blt_vprogram_id;
53 GLuint depth_blt_fprogram_id[tex_type_count];
54 BOOL use_arbfp_fixed_func;
55 struct hash_table_t *fragment_shaders;
58 /********************************************************
59 * ARB_[vertex/fragment]_program helper functions follow
60 ********************************************************/
63 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
64 * When constant_list == NULL, it will load all the constants.
66 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
67 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
69 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
70 unsigned int max_constants, float* constants, char *dirty_consts) {
71 local_constant* lconst;
75 if (TRACE_ON(d3d_shader)) {
76 for(i = 0; i < max_constants; i++) {
77 if(!dirty_consts[i]) continue;
78 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
79 constants[i * 4 + 0], constants[i * 4 + 1],
80 constants[i * 4 + 2], constants[i * 4 + 3]);
83 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
84 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
85 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
87 for(i = 0; i < max_constants; i++) {
88 if(!dirty_consts[i]) continue;
92 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
93 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
94 else lcl_const[0] = constants[j + 0];
96 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
97 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
98 else lcl_const[1] = constants[j + 1];
100 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
101 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
102 else lcl_const[2] = constants[j + 2];
104 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
105 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
106 else lcl_const[3] = constants[j + 3];
108 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
111 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
112 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
113 * or just reloading *all* constants at once
115 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
117 for(i = 0; i < max_constants; i++) {
118 if(!dirty_consts[i]) continue;
120 /* Find the next block of dirty constants */
123 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
127 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
130 for(i = 0; i < max_constants; i++) {
131 if(dirty_consts[i]) {
133 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
138 checkGLcall("glProgramEnvParameter4fvARB()");
140 /* Load immediate constants */
141 if(This->baseShader.load_local_constsF) {
142 if (TRACE_ON(d3d_shader)) {
143 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
144 GLfloat* values = (GLfloat*)lconst->value;
145 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
146 values[0], values[1], values[2], values[3]);
149 /* Immediate constants are clamped for 1.X shaders at loading times */
151 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
152 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
153 ret = max(ret, lconst->idx + 1);
154 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
156 checkGLcall("glProgramEnvParameter4fvARB()");
157 return ret; /* The loaded immediate constants need reloading for the next shader */
159 return 0; /* No constants are dirty now */
164 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
166 * We only support float constants in ARB at the moment, so don't
167 * worry about the Integers or Booleans
169 static void shader_arb_load_constants(
170 IWineD3DDevice* device,
172 char useVertexShader) {
174 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
175 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
176 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
179 if (useVertexShader) {
180 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
182 /* Load DirectX 9 float constants for vertex shader */
183 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
184 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
185 deviceImpl->highest_dirty_vs_const,
186 stateBlock->vertexShaderConstantF,
187 deviceImpl->activeContext->vshader_const_dirty);
189 /* Upload the position fixup */
190 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
193 if (usePixelShader) {
195 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
196 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
198 /* Load DirectX 9 float constants for pixel shader */
199 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
200 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
201 deviceImpl->highest_dirty_ps_const,
202 stateBlock->pixelShaderConstantF,
203 deviceImpl->activeContext->pshader_const_dirty);
205 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
206 /* The state manager takes care that this function is always called if the bump env matrix changes
208 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
209 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
210 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
212 if(psi->luminanceconst[i].const_num != -1) {
213 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
214 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
215 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
216 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
218 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
219 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
220 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
224 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
225 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
229 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
230 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
231 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
233 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
234 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
236 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
237 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
239 mul_low[0] = 1.0; mul_low[1] = 1.0;
240 mul_low[2] = 1.0; mul_low[3] = 1.0;
242 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
243 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
244 checkGLcall("Load sRGB correction constants\n");
245 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
246 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
252 /* Generate the variable & register declarations for the ARB_vertex_program output target */
253 static void shader_generate_arb_declarations(
254 IWineD3DBaseShader *iface,
255 shader_reg_maps* reg_maps,
256 SHADER_BUFFER* buffer,
257 WineD3D_GL_Info* gl_info) {
259 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
260 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
262 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
263 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
264 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
265 UINT extra_constants_needed = 0;
266 local_constant* lconst;
268 /* Temporary Output register */
269 shader_addline(buffer, "TEMP TMP_OUT;\n");
271 for(i = 0; i < This->baseShader.limits.temporary; i++) {
272 if (reg_maps->temporary[i])
273 shader_addline(buffer, "TEMP R%u;\n", i);
276 for (i = 0; i < This->baseShader.limits.address; i++) {
277 if (reg_maps->address[i])
278 shader_addline(buffer, "ADDRESS A%d;\n", i);
281 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
282 if (reg_maps->texcoord[i])
283 shader_addline(buffer,"TEMP T%u;\n", i);
286 /* Texture coordinate registers must be pre-loaded */
287 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
288 if (reg_maps->texcoord[i])
289 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
292 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
293 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
294 if(!reg_maps->bumpmat[i]) continue;
296 cur = ps->numbumpenvmatconsts;
297 ps->bumpenvmatconst[cur].const_num = -1;
298 ps->bumpenvmatconst[cur].texunit = i;
299 ps->luminanceconst[cur].const_num = -1;
300 ps->luminanceconst[cur].texunit = i;
302 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
303 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
306 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
307 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
308 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
309 i, ps->bumpenvmatconst[cur].const_num);
310 extra_constants_needed++;
312 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
313 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
314 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
315 i, ps->luminanceconst[cur].const_num);
316 extra_constants_needed++;
317 } else if(reg_maps->luminanceparams) {
318 FIXME("No free constant to load the luminance parameters\n");
321 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
324 ps->numbumpenvmatconsts = cur + 1;
327 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
328 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
329 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
330 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
331 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
334 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
335 /* The idea is that if srgb is enabled, then disabled, the constant loading code
336 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
337 * and comparison constants. If it disables it that way, the shader won't be recompiled
338 * and the code will stay in, so sRGB writing can be turned on again by setting the
339 * constants from the spec
341 ps_impl->srgb_mode_hardcoded = 0;
342 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
343 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
344 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
345 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
347 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
348 srgb_mul_low, srgb_mul_low, srgb_mul_low);
349 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
350 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
351 ps_impl->srgb_mode_hardcoded = 1;
353 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
354 * path if the comparison value is set to INF
356 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
357 srgb_pow, srgb_pow, srgb_pow);
358 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
359 srgb_mul_high, srgb_mul_high, srgb_mul_high);
360 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
361 srgb_sub_high, srgb_sub_high, srgb_sub_high);
362 ps_impl->srgb_enabled = 1;
364 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
366 /* Do not write any srgb fixup into the shader to save shader size and processing time.
367 * As a consequence, we can't toggle srgb write on without recompilation
369 ps_impl->srgb_enabled = 0;
370 ps_impl->srgb_mode_hardcoded = 1;
373 /* Load local constants using the program-local space,
374 * this avoids reloading them each time the shader is used
376 if(!This->baseShader.load_local_constsF) {
377 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
378 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
383 /* we use the array-based constants array if the local constants are marked for loading,
384 * because then we use indirect addressing, or when the local constant list is empty,
385 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
386 * local constants do not declare the loaded constants as an array because ARB compilers usually
387 * do not optimize unused constants away
389 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
390 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
391 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
392 max_constantsF, max_constantsF - 1);
394 for(i = 0; i < max_constantsF; i++) {
395 if(!shader_constant_is_local(This, i)) {
396 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
402 static const char * const shift_tab[] = {
403 "dummy", /* 0 (none) */
404 "coefmul.x", /* 1 (x2) */
405 "coefmul.y", /* 2 (x4) */
406 "coefmul.z", /* 3 (x8) */
407 "coefmul.w", /* 4 (x16) */
408 "dummy", /* 5 (x32) */
409 "dummy", /* 6 (x64) */
410 "dummy", /* 7 (x128) */
411 "dummy", /* 8 (d256) */
412 "dummy", /* 9 (d128) */
413 "dummy", /* 10 (d64) */
414 "dummy", /* 11 (d32) */
415 "coefdiv.w", /* 12 (d16) */
416 "coefdiv.z", /* 13 (d8) */
417 "coefdiv.y", /* 14 (d4) */
418 "coefdiv.x" /* 15 (d2) */
421 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
422 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
423 char *ptr = write_mask;
424 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
426 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
429 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
431 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
432 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
433 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
434 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
440 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
441 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
442 * but addressed as "rgba". To fix this we need to swap the register's x
443 * and z components. */
444 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
445 char *ptr = swizzle_str;
447 /* swizzle bits fields: wwzzyyxx */
448 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
449 DWORD swizzle_x = swizzle & 0x03;
450 DWORD swizzle_y = (swizzle >> 2) & 0x03;
451 DWORD swizzle_z = (swizzle >> 4) & 0x03;
452 DWORD swizzle_w = (swizzle >> 6) & 0x03;
454 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
455 * generate a swizzle string. Unless we need to our own swizzling. */
456 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
458 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
459 *ptr++ = swizzle_chars[swizzle_x];
461 *ptr++ = swizzle_chars[swizzle_x];
462 *ptr++ = swizzle_chars[swizzle_y];
463 *ptr++ = swizzle_chars[swizzle_z];
464 *ptr++ = swizzle_chars[swizzle_w];
471 static void pshader_get_register_name(IWineD3DBaseShader* iface,
472 const DWORD param, char* regstr) {
474 DWORD reg = param & WINED3DSP_REGNUM_MASK;
475 DWORD regtype = shader_get_regtype(param);
476 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
479 case WINED3DSPR_TEMP:
480 sprintf(regstr, "R%u", reg);
482 case WINED3DSPR_INPUT:
484 strcpy(regstr, "fragment.color.primary");
486 strcpy(regstr, "fragment.color.secondary");
489 case WINED3DSPR_CONST:
490 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
491 sprintf(regstr, "C[%u]", reg);
493 sprintf(regstr, "C%u", reg);
496 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
497 sprintf(regstr,"T%u", reg);
499 case WINED3DSPR_COLOROUT:
501 sprintf(regstr, "TMP_COLOR");
503 /* TODO: See GL_ARB_draw_buffers */
504 FIXME("Unsupported write to render target %u\n", reg);
505 sprintf(regstr, "unsupported_register");
508 case WINED3DSPR_DEPTHOUT:
509 sprintf(regstr, "result.depth");
511 case WINED3DSPR_ATTROUT:
512 sprintf(regstr, "oD[%u]", reg);
514 case WINED3DSPR_TEXCRDOUT:
515 sprintf(regstr, "oT[%u]", reg);
518 FIXME("Unhandled register name Type(%d)\n", regtype);
519 sprintf(regstr, "unrecognized_register");
524 /* TODO: merge with pixel shader */
525 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
527 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
529 /* oPos, oFog and oPts in D3D */
530 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
532 DWORD reg = param & WINED3DSP_REGNUM_MASK;
533 DWORD regtype = shader_get_regtype(param);
535 BOOL is_color = FALSE;
537 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
538 strcat(hwLine, " -");
544 case WINED3DSPR_TEMP:
545 sprintf(tmpReg, "R%u", reg);
546 strcat(hwLine, tmpReg);
548 case WINED3DSPR_INPUT:
550 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
553 sprintf(tmpReg, "vertex.attrib[%u]", reg);
554 strcat(hwLine, tmpReg);
556 case WINED3DSPR_CONST:
557 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
558 if(reg >= This->rel_offset) {
559 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
561 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
564 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
565 sprintf(tmpReg, "C[%u]", reg);
567 sprintf(tmpReg, "C%u", reg);
570 strcat(hwLine, tmpReg);
572 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
573 sprintf(tmpReg, "A%u", reg);
574 strcat(hwLine, tmpReg);
576 case WINED3DSPR_RASTOUT:
577 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
578 strcat(hwLine, tmpReg);
580 case WINED3DSPR_ATTROUT:
582 strcat(hwLine, "result.color.primary");
584 strcat(hwLine, "result.color.secondary");
587 case WINED3DSPR_TEXCRDOUT:
588 sprintf(tmpReg, "result.texcoord[%u]", reg);
589 strcat(hwLine, tmpReg);
592 FIXME("Unknown reg type %d %d\n", regtype, reg);
593 strcat(hwLine, "unrecognized_register");
599 shader_arb_get_write_mask(arg, param, write_mask);
600 strcat(hwLine, write_mask);
603 shader_arb_get_swizzle(param, is_color, swizzle);
604 strcat(hwLine, swizzle);
608 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
609 SHADER_BUFFER* buffer = arg->buffer;
610 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
611 const char *tex_type;
613 switch(sampler_type) {
620 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
621 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
622 if(device->stateBlock->textures[sampler_idx] &&
623 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
631 case WINED3DSTT_VOLUME:
635 case WINED3DSTT_CUBE:
640 ERR("Unexpected texture type %d\n", sampler_type);
645 /* Shouldn't be possible, but let's check for it */
646 if(projected) FIXME("Biased and Projected texture sampling\n");
647 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
648 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
649 } else if (projected) {
650 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
652 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
656 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, const char *writemask,
657 const char *one, const char *two, WINED3DFORMAT fmt,
658 WineD3D_GL_Info *gl_info) {
660 case WINED3DFMT_V8U8:
661 case WINED3DFMT_V16U16:
662 if(GL_SUPPORT(NV_TEXTURE_SHADER) && fmt == WINED3DFMT_V8U8) {
664 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
665 * disabled until an application that needs it is found because it causes unneeded
666 * shader recompilation in some game
668 if(strlen(writemask) >= 4) {
669 shader_addline(buffer, "MOV %s.%c, %s;\n", reg, writemask[3], one);
673 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
674 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
675 * all registers, do so, this saves an instruction.
677 if(strlen(writemask) >= 5) {
678 shader_addline(buffer, "MAD %s, %s, %s, -%s;\n", reg, reg, two, one);
679 } else if(strlen(writemask) >= 3) {
680 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
684 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
688 } else if(strlen(writemask) == 2) {
689 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n", reg, writemask[1],
690 reg, writemask[1], two, one);
695 case WINED3DFMT_X8L8V8U8:
696 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
697 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
698 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
700 if(strlen(writemask) >= 3) {
701 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
705 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
709 } else if(strlen(writemask) == 2) {
710 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
718 case WINED3DFMT_L6V5U5:
719 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
720 if(strlen(writemask) >= 4) {
721 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
722 shader_addline(buffer, "MOV TMP.g, %s.%c;\n",
724 shader_addline(buffer, "MAD %s.%c%c, %s.%c%c%c%c, %s, -%s;\n",
725 reg, writemask[1], writemask[2],
726 reg, writemask[3], writemask[1], writemask[3], writemask[1],
728 shader_addline(buffer, "MOV %s.%c, TMP.g;\n", reg,
730 } else if(strlen(writemask) == 3) {
731 /* This is bad: We have VL, but we need VU */
732 FIXME("2 components sampled from a converted L6V5U5 texture\n");
734 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
742 case WINED3DFMT_Q8W8V8U8:
743 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
744 /* Correct the sign in all channels */
745 switch(strlen(writemask)) {
747 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
752 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
757 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
762 /* Should not occur, since it's at minimum '.' and a letter */
764 ERR("Unexpected writemask: \"%s\"\n", writemask);
769 shader_addline(buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
774 case WINED3DFMT_ATI2N:
775 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
776 * which means the first one is replicated across .rgb, and the 2nd one is in
777 * .a. We need the 2nd in .g
779 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
780 * are swapped compared to d3d. So swap red and green.
782 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC)) {
783 shader_addline(buffer, "SWZ %s, %s, %c, %c, 1, 0;\n",
784 reg, reg, writemask[2], writemask[1]);
786 if(strlen(writemask) == 5) {
787 shader_addline(buffer, "MOV %s.%c, %s.%c;\n",
788 reg, writemask[2], reg, writemask[4]);
789 } else if(strlen(writemask) == 2) {
792 /* This is bad: We have VL, but we need VU */
793 FIXME("2 or 3 components sampled from a converted ATI2N texture\n");
798 /* stupid compiler */
804 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
805 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
806 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
807 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
809 WINED3DFORMAT conversion_group;
810 IWineD3DBaseTextureImpl *texture;
812 BOOL recorded = FALSE;
814 DWORD hex_version = shader->baseShader.hex_version;
818 switch(arg->opcode->opcode) {
820 if (hex_version < WINED3DPS_VERSION(2,0)) {
821 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
823 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
827 case WINED3DSIO_TEXLDL:
828 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
831 case WINED3DSIO_TEXDP3TEX:
832 case WINED3DSIO_TEXM3x3TEX:
833 case WINED3DSIO_TEXM3x3SPEC:
834 case WINED3DSIO_TEXM3x3VSPEC:
835 case WINED3DSIO_TEXBEM:
836 case WINED3DSIO_TEXREG2AR:
837 case WINED3DSIO_TEXREG2GB:
838 case WINED3DSIO_TEXREG2RGB:
839 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
843 /* Not a texture sampling instruction, nothing to do */
847 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
849 fmt = texture->resource.format;
850 conversion_group = texture->baseTexture.shader_conversion_group;
852 fmt = WINED3DFMT_UNKNOWN;
853 conversion_group = WINED3DFMT_UNKNOWN;
856 /* before doing anything, record the sampler with the format in the format conversion list,
857 * but check if it's not there already
859 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
860 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
865 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
866 shader->baseShader.num_sampled_samplers++;
867 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
870 pshader_get_register_name(arg->shader, arg->dst, reg);
871 shader_arb_get_write_mask(arg, arg->dst, writemask);
872 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
874 gen_color_correction(arg->buffer, reg, writemask, "one", "coefmul.x", fmt, gl_info);
879 static void pshader_gen_input_modifier_line (
880 IWineD3DBaseShader *iface,
881 SHADER_BUFFER* buffer,
886 /* Generate a line that does the input modifier computation and return the input register to use */
891 /* Assume a new line will be added */
894 /* Get register name */
895 pshader_get_register_name(iface, instr, regstr);
896 shader_arb_get_swizzle(instr, FALSE, swzstr);
898 switch (instr & WINED3DSP_SRCMOD_MASK) {
899 case WINED3DSPSM_NONE:
900 sprintf(outregstr, "%s%s", regstr, swzstr);
903 case WINED3DSPSM_NEG:
904 sprintf(outregstr, "-%s%s", regstr, swzstr);
907 case WINED3DSPSM_BIAS:
908 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
910 case WINED3DSPSM_BIASNEG:
911 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
913 case WINED3DSPSM_SIGN:
914 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
916 case WINED3DSPSM_SIGNNEG:
917 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
919 case WINED3DSPSM_COMP:
920 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
923 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
925 case WINED3DSPSM_X2NEG:
926 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
929 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
930 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
933 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
934 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
937 sprintf(outregstr, "%s%s", regstr, swzstr);
941 /* Return modified or original register, with swizzle */
943 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
946 static inline void pshader_gen_output_modifier_line(
947 SHADER_BUFFER* buffer,
953 /* Generate a line that does the output modifier computation */
954 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
955 regstr, write_mask, regstr, shift_tab[shift]);
958 static void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
959 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
961 SHADER_BUFFER* buffer = arg->buffer;
963 char src_name[2][50];
965 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
966 BOOL has_bumpmat = FALSE;
969 for(i = 0; i < This->numbumpenvmatconsts; i++) {
970 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
976 pshader_get_register_name(arg->shader, arg->dst, dst_name);
977 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
978 strcat(dst_name, dst_wmask);
980 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
981 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
984 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
985 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
986 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
987 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
988 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
990 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
992 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
996 static void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
998 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
999 SHADER_BUFFER* buffer = arg->buffer;
1002 char src_name[3][50];
1003 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1004 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1006 /* FIXME: support output modifiers */
1008 /* Handle output register */
1009 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1010 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1012 /* Generate input register names (with modifiers) */
1013 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1014 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1015 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1017 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1018 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
1019 arg->opcode_token & WINED3DSI_COISSUE) {
1020 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
1022 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
1023 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
1024 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
1027 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1030 static void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
1032 SHADER_BUFFER* buffer = arg->buffer;
1035 char src_name[3][50];
1036 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1037 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1039 /* FIXME: support output modifiers */
1041 /* Handle output register */
1042 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1043 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1045 /* Generate input register names (with modifiers) */
1046 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1047 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1048 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1050 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1051 src_name[0], src_name[2], src_name[1]);
1054 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1057 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1058 * dst = dot2(src0, src1) + src2 */
1059 static void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1060 SHADER_BUFFER* buffer = arg->buffer;
1063 char src_name[3][50];
1064 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1065 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1067 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1068 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1070 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1071 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1072 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1074 /* Emulate a DP2 with a DP3 and 0.0 */
1075 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1076 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1077 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1078 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1081 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1084 /* Map the opcode 1-to-1 to the GL code */
1085 static void shader_hw_map2gl(SHADER_OPCODE_ARG* arg)
1087 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1088 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1089 SHADER_BUFFER* buffer = arg->buffer;
1090 DWORD dst = arg->dst;
1091 DWORD* src = arg->src;
1092 char arguments[256];
1095 if (shader_is_pshader_version(shader->baseShader.hex_version))
1097 /* Output token related */
1098 char output_rname[256];
1099 char output_wmask[20];
1100 char operands[4][100];
1101 BOOL saturate = FALSE;
1102 BOOL centroid = FALSE;
1103 BOOL partialprecision = FALSE;
1104 const char *modifier;
1107 if (!curOpcode->num_params)
1109 ERR("Opcode \"%s\" has no parameters\n", curOpcode->name);
1113 /* Process modifiers */
1114 if (dst & WINED3DSP_DSTMOD_MASK)
1116 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1118 saturate = mask & WINED3DSPDM_SATURATE;
1119 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1120 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1121 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1123 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1126 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1128 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1129 modifier = (saturate && !shift) ? "_SAT" : "";
1131 /* Generate input register names (with modifiers) */
1132 for (i = 1; i < curOpcode->num_params; ++i)
1133 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1135 /* Handle output register */
1136 pshader_get_register_name(arg->shader, dst, output_rname);
1137 strcpy(operands[0], output_rname);
1138 shader_arb_get_write_mask(arg, dst, output_wmask);
1139 strcat(operands[0], output_wmask);
1141 arguments[0] = '\0';
1142 strcat(arguments, operands[0]);
1143 for (i = 1; i < curOpcode->num_params; i++)
1145 strcat(arguments, ", ");
1146 strcat(arguments, operands[i]);
1148 shader_addline(buffer, "%s%s %s;\n", curOpcode->glname, modifier, arguments);
1150 /* A shift requires another line. */
1151 if (shift) pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1153 /* Note that vshader_program_add_param() adds spaces. */
1155 arguments[0] = '\0';
1156 if (curOpcode->num_params > 0)
1158 vshader_program_add_param(arg, dst, FALSE, arguments);
1159 for (i = 1; i < curOpcode->num_params; ++i)
1161 strcat(arguments, ",");
1162 vshader_program_add_param(arg, src[i-1], TRUE, arguments);
1165 shader_addline(buffer, "%s%s;\n", curOpcode->glname, arguments);
1169 static void shader_hw_mov(SHADER_OPCODE_ARG *arg)
1171 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1173 if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1
1174 && !shader_is_pshader_version(shader->baseShader.hex_version)
1175 && shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)
1176 || arg->opcode->opcode == WINED3DSIO_MOVA)
1178 SHADER_BUFFER *buffer = arg->buffer;
1179 char src0_param[256];
1181 if (arg->opcode->opcode == WINED3DSIO_MOVA)
1182 FIXME("mova should round\n");
1184 src0_param[0] = '\0';
1185 if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
1187 vshader_program_add_param(arg, arg->src[0], TRUE, src0_param);
1188 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", src0_param);
1189 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1193 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1194 * with more than one component. Thus replicate the first source argument over all
1195 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1197 DWORD parm = arg->src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1198 if((arg->src[0] & WINED3DVS_X_W) == WINED3DVS_X_W)
1199 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1200 else if((arg->src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z)
1201 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1202 else if((arg->src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y)
1203 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1204 else if((arg->src[0] & WINED3DVS_X_X) == WINED3DVS_X_X)
1205 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1206 vshader_program_add_param(arg, parm, TRUE, src0_param);
1207 shader_addline(buffer, "ARL A0.x, %s;\n", src0_param);
1212 shader_hw_map2gl(arg);
1216 static void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1217 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1218 DWORD hex_version = This->baseShader.hex_version;
1219 SHADER_BUFFER* buffer = arg->buffer;
1222 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1223 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1225 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1227 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1228 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1229 shader_addline(buffer, "KIL %s;\n", reg_dest);
1231 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1232 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1234 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1235 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1236 shader_addline(buffer, "KIL TMP;\n");
1240 static void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1241 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1242 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1244 DWORD dst = arg->dst;
1245 DWORD* src = arg->src;
1246 SHADER_BUFFER* buffer = arg->buffer;
1247 DWORD hex_version = This->baseShader.hex_version;
1248 BOOL projected = FALSE, bias = FALSE;
1252 DWORD reg_dest_code;
1253 DWORD reg_sampler_code;
1255 /* All versions have a destination register */
1256 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1257 pshader_get_register_name(arg->shader, dst, reg_dest);
1259 /* 1.0-1.3: Use destination register as coordinate source.
1260 1.4+: Use provided coordinate source register. */
1261 if (hex_version < WINED3DPS_VERSION(1,4))
1262 strcpy(reg_coord, reg_dest);
1264 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1266 /* 1.0-1.4: Use destination register number as texture code.
1267 2.0+: Use provided sampler number as texure code. */
1268 if (hex_version < WINED3DPS_VERSION(2,0))
1269 reg_sampler_code = reg_dest_code;
1271 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1274 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1275 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1276 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1278 if(hex_version < WINED3DPS_VERSION(1,4)) {
1280 if(reg_sampler_code < MAX_TEXTURES) {
1281 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1283 if (flags & WINED3DTTFF_PROJECTED) {
1286 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1287 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1288 if (src_mod == WINED3DSPSM_DZ) {
1290 } else if(src_mod == WINED3DSPSM_DW) {
1294 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1297 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1301 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1304 static void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1306 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1307 DWORD dst = arg->dst;
1308 SHADER_BUFFER* buffer = arg->buffer;
1309 DWORD hex_version = This->baseShader.hex_version;
1312 shader_arb_get_write_mask(arg, dst, tmp);
1313 if (hex_version != WINED3DPS_VERSION(1,4)) {
1314 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1315 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1317 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1320 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1321 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1325 static void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1327 SHADER_BUFFER* buffer = arg->buffer;
1328 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1329 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1332 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1336 sprintf(dst_str, "T%u", reg1);
1337 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1338 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1339 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1340 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1341 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1344 static void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1346 SHADER_BUFFER* buffer = arg->buffer;
1348 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1352 sprintf(dst_str, "T%u", reg1);
1353 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1354 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1355 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1356 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1359 static void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1361 SHADER_BUFFER* buffer = arg->buffer;
1362 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1366 sprintf(dst_str, "T%u", reg1);
1367 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1368 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1371 static void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1372 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1373 BOOL has_bumpmat = FALSE;
1374 BOOL has_luminance = FALSE;
1377 DWORD dst = arg->dst;
1378 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1379 SHADER_BUFFER* buffer = arg->buffer;
1382 DWORD reg_dest_code;
1384 /* All versions have a destination register */
1385 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1386 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1387 pshader_get_register_name(arg->shader, dst, reg_coord);
1389 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1390 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1395 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1396 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1397 has_luminance = TRUE;
1403 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1405 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1406 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1407 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1408 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1410 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1411 * so we can't let the GL handle this.
1413 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1414 & WINED3DTTFF_PROJECTED) {
1415 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1416 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1417 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1419 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1422 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1424 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1425 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1426 src, reg_dest_code, reg_dest_code);
1427 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1432 if(reg_dest_code < MAX_TEXTURES) {
1433 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1437 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1438 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1442 static void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1444 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1445 SHADER_BUFFER* buffer = arg->buffer;
1448 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1449 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1452 static void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1454 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1455 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1457 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1458 SHADER_BUFFER* buffer = arg->buffer;
1462 sprintf(dst_str, "T%u", reg);
1463 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1464 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1465 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1466 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1469 static void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1471 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1472 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1473 SHADER_BUFFER* buffer = arg->buffer;
1474 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1477 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1478 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1479 current_state->texcoord_w[current_state->current_row++] = reg;
1482 static void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1484 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1485 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1487 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1488 SHADER_BUFFER* buffer = arg->buffer;
1489 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1493 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1494 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1496 /* Sample the texture using the calculated coordinates */
1497 sprintf(dst_str, "T%u", reg);
1498 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1499 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1500 current_state->current_row = 0;
1503 static void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1505 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1506 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1508 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1509 SHADER_BUFFER* buffer = arg->buffer;
1510 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1514 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1515 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1517 /* Construct the eye-ray vector from w coordinates */
1518 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1519 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1520 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1522 /* Calculate reflection vector
1524 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1525 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1526 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1527 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1528 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1529 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1530 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1532 /* Sample the texture using the calculated coordinates */
1533 sprintf(dst_str, "T%u", reg);
1534 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1535 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1536 current_state->current_row = 0;
1539 static void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1541 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1542 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1544 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1545 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1546 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1547 SHADER_BUFFER* buffer = arg->buffer;
1551 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1552 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1554 /* Calculate reflection vector.
1557 * TMP.xyz = 2 * --------- * N - E
1560 * Which normalizes the normal vector
1562 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1563 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1564 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1565 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1566 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1567 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1569 /* Sample the texture using the calculated coordinates */
1570 sprintf(dst_str, "T%u", reg);
1571 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1572 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1573 current_state->current_row = 0;
1576 static void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1577 SHADER_BUFFER* buffer = arg->buffer;
1580 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1581 * which is essentially an input, is the destination register because it is the first
1582 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1585 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1587 /* According to the msdn, the source register(must be r5) is unusable after
1588 * the texdepth instruction, so we're free to modify it
1590 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1592 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1593 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1594 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1596 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1597 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1598 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1599 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1602 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1603 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1604 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1605 static void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1606 SHADER_BUFFER* buffer = arg->buffer;
1607 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1611 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1612 shader_addline(buffer, "MOV TMP, 0.0;\n");
1613 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1615 sprintf(dst_str, "T%u", sampler_idx);
1616 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1619 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1620 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1621 static void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1625 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1626 SHADER_BUFFER* buffer = arg->buffer;
1628 /* Handle output register */
1629 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1630 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1632 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1633 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1635 /* TODO: Handle output modifiers */
1638 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1639 * Perform the 3rd row of a 3x3 matrix multiply */
1640 static void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1641 SHADER_BUFFER* buffer = arg->buffer;
1645 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1647 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1648 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1650 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1651 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1652 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1654 /* TODO: Handle output modifiers */
1657 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1658 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1659 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1660 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1662 static void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1663 SHADER_BUFFER* buffer = arg->buffer;
1664 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1667 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1668 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1670 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1671 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1672 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1674 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1675 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1676 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1677 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1680 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1681 Vertex/Pixel shaders to ARB_vertex_program codes */
1682 static void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1685 int nComponents = 0;
1686 SHADER_OPCODE_ARG tmpArg;
1688 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1690 /* Set constants for the temporary argument */
1691 tmpArg.shader = arg->shader;
1692 tmpArg.buffer = arg->buffer;
1693 tmpArg.src[0] = arg->src[0];
1694 tmpArg.src_addr[0] = arg->src_addr[0];
1695 tmpArg.src_addr[1] = arg->src_addr[1];
1696 tmpArg.reg_maps = arg->reg_maps;
1698 switch(arg->opcode->opcode) {
1699 case WINED3DSIO_M4x4:
1701 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1703 case WINED3DSIO_M4x3:
1705 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1707 case WINED3DSIO_M3x4:
1709 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1711 case WINED3DSIO_M3x3:
1713 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1715 case WINED3DSIO_M3x2:
1717 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1723 for (i = 0; i < nComponents; i++) {
1724 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1725 tmpArg.src[1] = arg->src[1]+i;
1726 shader_hw_map2gl(&tmpArg);
1730 static void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1731 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1732 SHADER_BUFFER* buffer = arg->buffer;
1733 DWORD dst = arg->dst;
1734 DWORD src = arg->src[0];
1735 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1739 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1740 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1741 strcat(tmpLine, ",");
1742 vshader_program_add_param(arg, src, TRUE, tmpLine);
1743 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1744 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1747 strcat(tmpLine, ".w");
1750 shader_addline(buffer, "%s;\n", tmpLine);
1753 static void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1754 SHADER_BUFFER* buffer = arg->buffer;
1758 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1759 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1761 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1762 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1764 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1765 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1766 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1767 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1768 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1772 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1775 static void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1776 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1777 * must contain fixed constants. So we need a separate function to filter those constants and
1780 SHADER_BUFFER* buffer = arg->buffer;
1784 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1785 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1787 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1788 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1790 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1791 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1795 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1799 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1800 GLuint program_id = 0;
1801 const char *blt_vprogram =
1803 "PARAM c[1] = { { 1, 0.5 } };\n"
1804 "MOV result.position, vertex.position;\n"
1805 "MOV result.color, c[0].x;\n"
1806 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1809 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1810 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1811 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1813 if (glGetError() == GL_INVALID_OPERATION) {
1815 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1816 FIXME("Vertex program error at position %d: %s\n", pos,
1817 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1823 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1825 GLuint program_id = 0;
1826 const char *blt_fprograms[tex_type_count] =
1833 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1834 "MOV result.depth.z, R0.x;\n"
1841 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1842 "MOV result.depth.z, R0.x;\n"
1847 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1848 "MOV result.depth.z, R0.x;\n"
1852 if (!blt_fprograms[tex_type])
1854 FIXME("tex_type %#x not supported\n", tex_type);
1858 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1859 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1860 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1862 if (glGetError() == GL_INVALID_OPERATION) {
1864 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1865 FIXME("Fragment program error at position %d: %s\n", pos,
1866 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1872 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1873 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1874 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1875 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1878 TRACE("Using vertex shader\n");
1880 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId;
1882 /* Bind the vertex program */
1883 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1884 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1886 /* Enable OpenGL vertex programs */
1887 glEnable(GL_VERTEX_PROGRAM_ARB);
1888 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1889 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1890 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1891 priv->current_vprogram_id = 0;
1892 glDisable(GL_VERTEX_PROGRAM_ARB);
1893 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1897 TRACE("Using pixel shader\n");
1899 priv->current_fprogram_id = ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId;
1901 /* Bind the fragment program */
1902 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1903 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1905 if(!priv->use_arbfp_fixed_func) {
1906 /* Enable OpenGL fragment programs */
1907 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1908 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1910 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1911 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1912 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1913 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1914 * replacement shader
1916 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1917 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1918 priv->current_fprogram_id = 0;
1922 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1923 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1924 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1925 GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1926 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1928 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1929 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1930 glEnable(GL_VERTEX_PROGRAM_ARB);
1932 if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1933 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1934 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1937 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1938 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1939 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1940 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1942 if (priv->current_vprogram_id) {
1943 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1944 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1946 glEnable(GL_VERTEX_PROGRAM_ARB);
1947 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1949 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1951 glDisable(GL_VERTEX_PROGRAM_ARB);
1952 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1955 if (priv->current_fprogram_id) {
1956 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1957 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1959 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1960 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1962 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1964 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1965 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1969 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1970 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1971 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1972 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1973 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1976 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1977 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1978 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1981 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1982 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1984 This->baseShader.prgId = 0;
1985 This->baseShader.is_compiled = FALSE;
1988 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1989 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1990 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1994 static void shader_arb_free(IWineD3DDevice *iface) {
1995 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1996 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1997 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
2000 if(priv->depth_blt_vprogram_id) {
2001 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
2003 for (i = 0; i < tex_type_count; ++i) {
2004 if (priv->depth_blt_fprogram_id[i]) {
2005 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
2009 HeapFree(GetProcessHeap(), 0, This->shader_priv);
2012 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
2016 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
2017 const char *tmp2, const char *tmp3, const char *tmp4) {
2018 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
2020 /* Calculate the > 0.0031308 case */
2021 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
2022 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
2023 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
2024 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
2025 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
2026 /* Calculate the < case */
2027 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
2028 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
2029 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
2030 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
2031 /* Store the components > 0.0031308 in the destination */
2032 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
2033 /* Add the components that are < 0.0031308 */
2034 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
2035 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
2038 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
2039 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
2040 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2041 CONST DWORD *function = This->baseShader.function;
2042 const char *fragcolor;
2043 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
2044 local_constant* lconst;
2046 /* Create the hw ARB shader */
2047 shader_addline(buffer, "!!ARBfp1.0\n");
2049 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
2050 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
2051 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
2052 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
2053 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
2054 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2055 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2056 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2058 /* Base Declarations */
2059 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2061 /* We need two variables for fog blending */
2062 shader_addline(buffer, "TEMP TMP_FOG;\n");
2063 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
2064 shader_addline(buffer, "TEMP TMP_COLOR;\n");
2067 /* Base Shader Body */
2068 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2070 /* calculate fog and blend it
2071 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
2072 * -1/(e-s) and e/(e-s) respectively.
2074 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
2076 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
2079 fragcolor = "TMP_COLOR";
2081 if(This->srgb_enabled) {
2082 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
2084 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
2085 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
2086 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
2089 shader_addline(buffer, "END\n");
2091 /* TODO: change to resource.glObjectHandle or something like that */
2092 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2094 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
2095 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
2097 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
2098 /* Create the program and check for errors */
2099 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2100 buffer->bsize, buffer->buffer));
2102 if (glGetError() == GL_INVALID_OPERATION) {
2104 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2105 FIXME("HW PixelShader Error at position %d: %s\n",
2106 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2107 This->baseShader.prgId = -1;
2110 /* Load immediate constants */
2111 if(!This->baseShader.load_local_constsF) {
2112 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2113 float *value = (float *) lconst->value;
2114 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2115 checkGLcall("glProgramLocalParameter4fvARB");
2120 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
2121 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2122 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2123 CONST DWORD *function = This->baseShader.function;
2124 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2125 WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2126 local_constant* lconst;
2128 /* Create the hw ARB shader */
2129 shader_addline(buffer, "!!ARBvp1.0\n");
2130 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2132 /* Mesa supports only 95 constants */
2133 if (GL_VEND(MESA) || GL_VEND(WINE))
2134 This->baseShader.limits.constant_float =
2135 min(95, This->baseShader.limits.constant_float);
2137 shader_addline(buffer, "TEMP TMP;\n");
2139 /* Base Declarations */
2140 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2142 /* We need a constant to fixup the final position */
2143 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2145 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2146 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2147 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2148 * a replacement shader depend on the texcoord.w being set properly.
2150 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2151 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2152 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2153 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2154 * this can eat a number of instructions, so skip it unless this cap is set as well
2156 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2157 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2159 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2161 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2162 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2163 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2164 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2170 /* Base Shader Body */
2171 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2173 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2175 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2177 /* Write the final position.
2179 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2180 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2181 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2182 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2184 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2185 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2186 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2188 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2189 * and the glsl equivalent
2191 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2193 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2195 shader_addline(buffer, "END\n");
2197 /* TODO: change to resource.glObjectHandle or something like that */
2198 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2200 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2201 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2203 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2204 /* Create the program and check for errors */
2205 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2206 buffer->bsize, buffer->buffer));
2208 if (glGetError() == GL_INVALID_OPERATION) {
2210 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2211 FIXME("HW VertexShader Error at position %d: %s\n",
2212 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2213 This->baseShader.prgId = -1;
2216 /* Load immediate constants */
2217 if(!This->baseShader.load_local_constsF) {
2218 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2219 float *value = (float *) lconst->value;
2220 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2225 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2226 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2227 * then overwrite the shader specific ones
2229 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2231 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2232 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2233 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2234 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2237 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2238 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2239 pCaps->PixelShader1xMaxValue = 8.0;
2240 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2244 static BOOL shader_arb_conv_supported(WINED3DFORMAT fmt) {
2245 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt));
2247 case WINED3DFMT_V8U8:
2248 case WINED3DFMT_V16U16:
2249 case WINED3DFMT_X8L8V8U8:
2250 case WINED3DFMT_L6V5U5:
2251 case WINED3DFMT_Q8W8V8U8:
2252 case WINED3DFMT_ATI2N:
2261 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2263 /* WINED3DSIH_ABS */ shader_hw_map2gl,
2264 /* WINED3DSIH_ADD */ shader_hw_map2gl,
2265 /* WINED3DSIH_BEM */ pshader_hw_bem,
2266 /* WINED3DSIH_BREAK */ NULL,
2267 /* WINED3DSIH_BREAKC */ NULL,
2268 /* WINED3DSIH_BREAKP */ NULL,
2269 /* WINED3DSIH_CALL */ NULL,
2270 /* WINED3DSIH_CALLNZ */ NULL,
2271 /* WINED3DSIH_CMP */ pshader_hw_cmp,
2272 /* WINED3DSIH_CND */ pshader_hw_cnd,
2273 /* WINED3DSIH_CRS */ shader_hw_map2gl,
2274 /* WINED3DSIH_DCL */ NULL,
2275 /* WINED3DSIH_DEF */ NULL,
2276 /* WINED3DSIH_DEFB */ NULL,
2277 /* WINED3DSIH_DEFI */ NULL,
2278 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add,
2279 /* WINED3DSIH_DP3 */ shader_hw_map2gl,
2280 /* WINED3DSIH_DP4 */ shader_hw_map2gl,
2281 /* WINED3DSIH_DST */ shader_hw_map2gl,
2282 /* WINED3DSIH_DSX */ NULL,
2283 /* WINED3DSIH_DSY */ NULL,
2284 /* WINED3DSIH_ELSE */ NULL,
2285 /* WINED3DSIH_ENDIF */ NULL,
2286 /* WINED3DSIH_ENDLOOP */ NULL,
2287 /* WINED3DSIH_ENDREP */ NULL,
2288 /* WINED3DSIH_EXP */ shader_hw_map2gl,
2289 /* WINED3DSIH_EXPP */ shader_hw_map2gl,
2290 /* WINED3DSIH_FRC */ shader_hw_map2gl,
2291 /* WINED3DSIH_IF */ NULL,
2292 /* WINED3DSIH_IFC */ NULL,
2293 /* WINED3DSIH_LABEL */ NULL,
2294 /* WINED3DSIH_LIT */ shader_hw_map2gl,
2295 /* WINED3DSIH_LOG */ shader_hw_map2gl,
2296 /* WINED3DSIH_LOGP */ shader_hw_map2gl,
2297 /* WINED3DSIH_LOOP */ NULL,
2298 /* WINED3DSIH_LRP */ shader_hw_map2gl,
2299 /* WINED3DSIH_M3x2 */ shader_hw_mnxn,
2300 /* WINED3DSIH_M3x3 */ shader_hw_mnxn,
2301 /* WINED3DSIH_M3x4 */ shader_hw_mnxn,
2302 /* WINED3DSIH_M4x3 */ shader_hw_mnxn,
2303 /* WINED3DSIH_M4x4 */ shader_hw_mnxn,
2304 /* WINED3DSIH_MAD */ shader_hw_map2gl,
2305 /* WINED3DSIH_MAX */ shader_hw_map2gl,
2306 /* WINED3DSIH_MIN */ shader_hw_map2gl,
2307 /* WINED3DSIH_MOV */ shader_hw_mov,
2308 /* WINED3DSIH_MOVA */ shader_hw_mov,
2309 /* WINED3DSIH_MUL */ shader_hw_map2gl,
2310 /* WINED3DSIH_NOP */ shader_hw_map2gl,
2311 /* WINED3DSIH_NRM */ shader_hw_nrm,
2312 /* WINED3DSIH_PHASE */ NULL,
2313 /* WINED3DSIH_POW */ shader_hw_map2gl,
2314 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp,
2315 /* WINED3DSIH_REP */ NULL,
2316 /* WINED3DSIH_RET */ NULL,
2317 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp,
2318 /* WINED3DSIH_SETP */ NULL,
2319 /* WINED3DSIH_SGE */ shader_hw_map2gl,
2320 /* WINED3DSIH_SGN */ NULL,
2321 /* WINED3DSIH_SINCOS */ shader_hw_sincos,
2322 /* WINED3DSIH_SLT */ shader_hw_map2gl,
2323 /* WINED3DSIH_SUB */ shader_hw_map2gl,
2324 /* WINED3DSIH_TEX */ pshader_hw_tex,
2325 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem,
2326 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem,
2327 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord,
2328 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth,
2329 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3,
2330 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex,
2331 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill,
2332 /* WINED3DSIH_TEXLDD */ NULL,
2333 /* WINED3DSIH_TEXLDL */ NULL,
2334 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth,
2335 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad,
2336 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex,
2337 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3,
2338 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
2339 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad,
2340 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec,
2341 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex,
2342 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec,
2343 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar,
2344 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb,
2345 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb,
2348 const shader_backend_t arb_program_shader_backend = {
2349 shader_arb_instruction_handler_table,
2351 shader_arb_select_depth_blt,
2352 shader_arb_deselect_depth_blt,
2353 shader_arb_load_constants,
2355 shader_arb_color_correction,
2359 shader_arb_dirty_const,
2360 shader_arb_generate_pshader,
2361 shader_arb_generate_vshader,
2362 shader_arb_get_caps,
2363 shader_arb_conv_supported,
2366 /* ARB_fragment_program fixed function pipeline replacement definitions */
2367 #define ARB_FFP_CONST_TFACTOR 0
2368 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2369 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2370 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2371 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2373 struct arbfp_ffp_desc
2375 struct ffp_desc parent;
2377 unsigned int num_textures_used;
2380 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2382 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2383 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2385 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2386 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2390 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2391 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2392 struct shader_arb_priv *priv;
2393 /* Share private data between the shader backend and the pipeline replacement, if both
2394 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2395 * if no pixel shader is bound or not
2397 if(This->shader_backend == &arb_program_shader_backend) {
2398 This->fragment_priv = This->shader_priv;
2400 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2401 if(!This->fragment_priv) return E_OUTOFMEMORY;
2403 priv = (struct shader_arb_priv *) This->fragment_priv;
2404 priv->fragment_shaders = hash_table_create(ffp_program_key_hash, ffp_program_key_compare);
2405 priv->use_arbfp_fixed_func = TRUE;
2409 static void arbfp_free_ffpshader(void *value, void *gli) {
2410 WineD3D_GL_Info *gl_info = gli;
2411 struct arbfp_ffp_desc *entry_arb = value;
2414 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2415 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2416 HeapFree(GetProcessHeap(), 0, entry_arb);
2420 static void arbfp_free(IWineD3DDevice *iface) {
2421 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2422 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2424 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2425 priv->use_arbfp_fixed_func = FALSE;
2427 if(This->shader_backend != &arb_program_shader_backend) {
2428 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2432 static void arbfp_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct fragment_caps *caps) {
2433 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2434 WINED3DTEXOPCAPS_SELECTARG1 |
2435 WINED3DTEXOPCAPS_SELECTARG2 |
2436 WINED3DTEXOPCAPS_MODULATE4X |
2437 WINED3DTEXOPCAPS_MODULATE2X |
2438 WINED3DTEXOPCAPS_MODULATE |
2439 WINED3DTEXOPCAPS_ADDSIGNED2X |
2440 WINED3DTEXOPCAPS_ADDSIGNED |
2441 WINED3DTEXOPCAPS_ADD |
2442 WINED3DTEXOPCAPS_SUBTRACT |
2443 WINED3DTEXOPCAPS_ADDSMOOTH |
2444 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2445 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2446 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2447 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2448 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2449 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2450 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2451 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2452 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2453 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2454 WINED3DTEXOPCAPS_MULTIPLYADD |
2455 WINED3DTEXOPCAPS_LERP |
2456 WINED3DTEXOPCAPS_BUMPENVMAP |
2457 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2459 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2461 caps->MaxTextureBlendStages = 8;
2462 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2464 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2466 #undef GLINFO_LOCATION
2468 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2469 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2471 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2473 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2474 * application provided constants
2476 if(device->shader_backend == &arb_program_shader_backend) {
2477 if(use_ps(device)) return;
2479 device = stateblock->wineD3DDevice;
2480 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2481 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2484 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2485 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2486 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2490 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2492 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2494 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2495 * application provided constants
2497 if(device->shader_backend == &arb_program_shader_backend) {
2498 if(use_ps(device)) return;
2500 device = stateblock->wineD3DDevice;
2501 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2502 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2505 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2506 /* The specular color has no alpha */
2507 col[0] = 1.0; col[1] = 1.0;
2508 col[2] = 1.0; col[3] = 0.0;
2510 col[0] = 0.0; col[1] = 0.0;
2511 col[2] = 0.0; col[3] = 0.0;
2513 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2514 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2517 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2518 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2519 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2522 if(use_ps(device)) {
2524 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2525 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2528 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2529 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2533 if(device->shader_backend == &arb_program_shader_backend) {
2534 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2537 } else if(device->shader_backend == &arb_program_shader_backend) {
2538 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2539 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2542 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2543 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2544 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2545 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2547 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2548 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2551 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2552 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2553 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2556 if(use_ps(device)) {
2558 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2559 /* The pixel shader has to know the luminance offset. Do a constants update if it
2560 * isn't scheduled anyway
2562 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2563 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2567 if(device->shader_backend == &arb_program_shader_backend) {
2568 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2571 } else if(device->shader_backend == &arb_program_shader_backend) {
2572 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2573 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2576 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2577 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2581 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2582 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2585 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2588 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2590 switch(arg & WINED3DTA_SELECTMASK) {
2591 case WINED3DTA_DIFFUSE:
2592 ret = "fragment.color.primary"; break;
2594 case WINED3DTA_CURRENT:
2595 if(stage == 0) ret = "fragment.color.primary";
2599 case WINED3DTA_TEXTURE:
2601 case 0: ret = "tex0"; break;
2602 case 1: ret = "tex1"; break;
2603 case 2: ret = "tex2"; break;
2604 case 3: ret = "tex3"; break;
2605 case 4: ret = "tex4"; break;
2606 case 5: ret = "tex5"; break;
2607 case 6: ret = "tex6"; break;
2608 case 7: ret = "tex7"; break;
2609 default: ret = "unknown texture";
2613 case WINED3DTA_TFACTOR:
2614 ret = "tfactor"; break;
2616 case WINED3DTA_SPECULAR:
2617 ret = "fragment.color.secondary"; break;
2619 case WINED3DTA_TEMP:
2620 ret = "tempreg"; break;
2622 case WINED3DTA_CONSTANT:
2623 FIXME("Implement perstage constants\n");
2625 case 0: ret = "const0"; break;
2626 case 1: ret = "const1"; break;
2627 case 2: ret = "const2"; break;
2628 case 3: ret = "const3"; break;
2629 case 4: ret = "const4"; break;
2630 case 5: ret = "const5"; break;
2631 case 6: ret = "const6"; break;
2632 case 7: ret = "const7"; break;
2633 default: ret = "unknown constant";
2641 if(arg & WINED3DTA_COMPLEMENT) {
2642 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2643 if(argnum == 0) ret = "arg0";
2644 if(argnum == 1) ret = "arg1";
2645 if(argnum == 2) ret = "arg2";
2647 if(arg & WINED3DTA_ALPHAREPLICATE) {
2648 shader_addline(buffer, "MOV arg%u, %s.a;\n", argnum, ret);
2649 if(argnum == 0) ret = "arg0";
2650 if(argnum == 1) ret = "arg1";
2651 if(argnum == 2) ret = "arg2";
2656 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2657 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2658 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2659 unsigned int mul = 1;
2660 BOOL mul_final_dest = FALSE;
2662 if(color && alpha) dstmask = "";
2663 else if(color) dstmask = ".rgb";
2664 else dstmask = ".a";
2666 if(dst == tempreg) dstreg = "tempreg";
2667 else dstreg = "ret";
2669 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2670 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2671 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2674 case WINED3DTOP_DISABLE:
2675 if(stage == 1) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2678 case WINED3DTOP_SELECTARG2:
2680 case WINED3DTOP_SELECTARG1:
2681 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2684 case WINED3DTOP_MODULATE4X:
2686 case WINED3DTOP_MODULATE2X:
2688 if(strcmp(dstreg, "result.color") == 0) {
2690 mul_final_dest = TRUE;
2692 case WINED3DTOP_MODULATE:
2693 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2696 case WINED3DTOP_ADDSIGNED2X:
2698 if(strcmp(dstreg, "result.color") == 0) {
2700 mul_final_dest = TRUE;
2702 case WINED3DTOP_ADDSIGNED:
2703 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2705 case WINED3DTOP_ADD:
2706 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2709 case WINED3DTOP_SUBTRACT:
2710 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2713 case WINED3DTOP_ADDSMOOTH:
2714 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2715 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2718 case WINED3DTOP_BLENDCURRENTALPHA:
2719 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2720 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2722 case WINED3DTOP_BLENDFACTORALPHA:
2723 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2724 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2726 case WINED3DTOP_BLENDTEXTUREALPHA:
2727 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2728 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2730 case WINED3DTOP_BLENDDIFFUSEALPHA:
2731 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2732 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2735 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2736 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2737 shader_addline(buffer, "SUB arg0.a, const.x, %s.a;\n", arg0);
2738 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg, dstmask, arg2, arg1);
2741 /* D3DTOP_PREMODULATE ???? */
2743 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2744 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2745 shader_addline(buffer, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2747 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2748 shader_addline(buffer, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2750 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2751 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2752 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg, dstmask, arg2, arg1);
2754 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2755 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg, dstmask, arg1, arg2, arg1);
2758 case WINED3DTOP_DOTPRODUCT3:
2760 if(strcmp(dstreg, "result.color") == 0) {
2762 mul_final_dest = TRUE;
2764 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2765 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2766 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2769 case WINED3DTOP_MULTIPLYADD:
2770 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2773 case WINED3DTOP_LERP:
2774 /* The msdn is not quite right here */
2775 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2778 case WINED3DTOP_BUMPENVMAP:
2779 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2780 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2784 FIXME("Unhandled texture op %08x\n", op);
2788 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2789 } else if(mul == 4) {
2790 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2794 /* The stateblock is passed for GLINFO_LOCATION */
2795 static GLuint gen_arbfp_ffp_shader(struct ffp_settings *settings, IWineD3DStateBlockImpl *stateblock) {
2797 SHADER_BUFFER buffer;
2798 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2799 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2800 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2801 const char *textype;
2802 const char *instr, *sat;
2803 char colorcor_dst[8];
2805 DWORD arg0, arg1, arg2;
2806 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2808 const char *final_combiner_src = "ret";
2810 /* Find out which textures are read */
2811 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2812 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2813 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2814 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2815 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2816 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2817 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2818 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2820 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2821 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2822 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2823 bump_used[stage] = TRUE;
2824 tex_read[stage] = TRUE;
2826 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2827 bump_used[stage] = TRUE;
2828 tex_read[stage] = TRUE;
2829 luminance_used[stage] = TRUE;
2830 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2831 tfactor_used = TRUE;
2834 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2835 tfactor_used = TRUE;
2838 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2839 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2840 tempreg_used = TRUE;
2843 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2844 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2845 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2846 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2847 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2848 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2849 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2851 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2852 tempreg_used = TRUE;
2854 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2855 tfactor_used = TRUE;
2862 buffer.newline = TRUE;
2863 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
2865 shader_addline(&buffer, "!!ARBfp1.0\n");
2867 switch(settings->fog) {
2868 case FOG_OFF: break;
2869 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2870 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2871 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2872 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2875 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2876 shader_addline(&buffer, "TEMP TMP;\n");
2877 shader_addline(&buffer, "TEMP ret;\n");
2878 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2879 shader_addline(&buffer, "TEMP arg0;\n");
2880 shader_addline(&buffer, "TEMP arg1;\n");
2881 shader_addline(&buffer, "TEMP arg2;\n");
2882 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2883 if(!tex_read[stage]) continue;
2884 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2885 if(!bump_used[stage]) continue;
2886 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2887 if(!luminance_used[stage]) continue;
2888 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2891 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2893 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2895 if(settings->sRGB_write) {
2896 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2897 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2898 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2899 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2900 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2901 srgb_pow, srgb_pow, srgb_pow);
2902 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2903 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2904 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2905 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2908 /* Generate texture sampling instructions) */
2909 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2910 if(!tex_read[stage]) continue;
2912 switch(settings->op[stage].tex_type) {
2913 case tex_1d: textype = "1D"; break;
2914 case tex_2d: textype = "2D"; break;
2915 case tex_3d: textype = "3D"; break;
2916 case tex_cube: textype = "CUBE"; break;
2917 case tex_rect: textype = "RECT"; break;
2918 default: textype = "unexpected_textype"; break;
2921 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2922 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2928 if(settings->op[stage].projected == proj_none) {
2930 } else if(settings->op[stage].projected == proj_count4 ||
2931 settings->op[stage].projected == proj_count3) {
2934 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2939 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2940 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2941 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2942 shader_addline(&buffer, "DP3 ret.r, arg1, tex%u;\n", stage - 1);
2943 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2944 shader_addline(&buffer, "DP3 ret.g, arg1, tex%u;\n", stage - 1);
2946 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2947 * so multiply the displacement with the dividing parameter before passing it to TXP
2949 if (settings->op[stage].projected != proj_none) {
2950 if(settings->op[stage].projected == proj_count4) {
2951 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].a;\n", stage);
2952 shader_addline(&buffer, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage, stage);
2954 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].b;\n", stage);
2955 shader_addline(&buffer, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage, stage);
2958 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2961 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2962 instr, sat, stage, stage, textype);
2963 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2964 shader_addline(&buffer, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2965 stage - 1, stage - 1, stage - 1);
2966 shader_addline(&buffer, "MUL tex%u, tex%u, ret.r;\n", stage, stage);
2968 } else if(settings->op[stage].projected == proj_count3) {
2969 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2970 shader_addline(&buffer, "MOV ret.a, ret.b;\n");
2971 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2972 instr, sat, stage, stage, textype);
2974 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2975 instr, sat, stage, stage, stage, textype);
2978 sprintf(colorcor_dst, "tex%u", stage);
2979 gen_color_correction(&buffer, colorcor_dst, ".rgba", "const.x", "const.y",
2980 settings->op[stage].color_correction, &GLINFO_LOCATION);
2983 /* Generate the main shader */
2984 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2985 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2987 final_combiner_src = "fragment.color.primary";
2992 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2993 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2994 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2995 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2996 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2997 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2998 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2999 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
3000 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
3001 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
3002 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
3003 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
3005 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
3006 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
3007 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
3008 settings->op[stage].carg2 == settings->op[stage].aarg2;
3011 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
3012 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
3013 settings->op[stage].cop, settings->op[stage].carg0,
3014 settings->op[stage].carg1, settings->op[stage].carg2);
3016 shader_addline(&buffer, "MOV ret.a, fragment.color.primary.a;\n");
3018 } else if(op_equal) {
3019 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
3020 settings->op[stage].cop, settings->op[stage].carg0,
3021 settings->op[stage].carg1, settings->op[stage].carg2);
3023 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
3024 settings->op[stage].cop, settings->op[stage].carg0,
3025 settings->op[stage].carg1, settings->op[stage].carg2);
3026 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
3027 settings->op[stage].aop, settings->op[stage].aarg0,
3028 settings->op[stage].aarg1, settings->op[stage].aarg2);
3032 if(settings->sRGB_write) {
3033 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
3034 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
3035 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
3037 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
3041 shader_addline(&buffer, "END\n");
3043 /* Generate the shader */
3044 GL_EXTCALL(glGenProgramsARB(1, &ret));
3045 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
3046 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3048 if (glGetError() == GL_INVALID_OPERATION) {
3050 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3051 FIXME("Fragment program error at position %d: %s\n", pos,
3052 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3054 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3058 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3059 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
3060 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
3061 BOOL use_pshader = use_ps(device);
3062 BOOL use_vshader = use_vs(device);
3063 struct ffp_settings settings;
3064 struct arbfp_ffp_desc *desc;
3067 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
3069 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
3070 } else if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3071 /* Reload fixed function constants since they collide with the pixel shader constants */
3072 for(i = 0; i < MAX_TEXTURES; i++) {
3073 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3075 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3076 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3082 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
3084 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
3085 gen_ffp_op(stateblock, &settings, FALSE);
3086 desc = (struct arbfp_ffp_desc *) find_ffp_shader(priv->fragment_shaders, &settings);
3088 desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*desc));
3090 ERR("Out of memory\n");
3093 desc->num_textures_used = 0;
3094 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
3095 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
3096 desc->num_textures_used = i;
3099 memcpy(&desc->parent.settings, &settings, sizeof(settings));
3100 desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
3101 add_ffp_shader(priv->fragment_shaders, &desc->parent);
3102 TRACE("Allocated fixed function replacement shader descriptor %p\n", desc);
3105 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3106 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3109 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
3110 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3111 priv->current_fprogram_id = desc->shader;
3113 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3114 /* Reload fixed function constants since they collide with the pixel shader constants */
3115 for(i = 0; i < MAX_TEXTURES; i++) {
3116 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3118 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3119 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3123 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3124 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3125 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3126 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3127 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3129 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3130 * to be compiled before activating them(needs some cleanups in the shader backend interface)
3132 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
3133 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
3135 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
3136 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
3140 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
3144 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3145 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3146 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3147 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3148 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3150 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3151 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3152 fragment_prog_arbfp(state, stateblock, context);
3156 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3157 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3158 fragment_prog_arbfp(state, stateblock, context);
3162 #undef GLINFO_LOCATION
3164 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3165 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3166 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3167 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3168 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3169 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3170 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3171 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3172 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3173 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3174 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3175 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3176 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3177 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3178 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3179 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3180 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3181 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3182 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3183 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3184 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3185 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3186 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3187 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3188 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3189 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3190 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3191 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3192 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3193 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3194 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3195 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3196 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3197 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3198 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3199 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3200 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3201 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3202 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3203 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3204 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3205 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3206 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3207 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3208 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3209 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3210 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3211 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3212 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3213 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3214 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3215 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3216 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3217 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3218 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3219 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3220 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3221 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3222 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3223 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3224 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3225 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3226 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3227 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3228 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3229 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3230 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3231 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3232 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3233 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3234 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3235 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3236 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3237 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3238 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3239 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3240 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3241 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3242 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3243 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3244 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3245 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3246 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3247 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3248 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3249 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3250 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3251 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3252 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3253 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3254 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3255 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3256 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3257 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3258 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3259 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3260 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3261 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3262 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3263 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3264 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3265 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3266 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3267 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3268 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3269 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3270 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3271 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3272 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3273 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3274 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3275 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3276 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3277 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3278 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3279 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3280 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3281 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3282 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3283 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3284 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3285 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3286 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3287 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3288 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3289 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3290 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3291 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3292 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3293 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3294 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3295 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3296 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3297 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3298 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3299 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3300 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3301 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3302 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3303 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3304 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3305 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3306 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3307 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3308 {0 /* Terminate */, { 0, 0 }, 0 },
3311 const struct fragment_pipeline arbfp_fragment_pipeline = {
3316 shader_arb_conv_supported,
3317 arbfp_fragmentstate_template,
3318 TRUE /* We can disable projected textures */
3321 #define GLINFO_LOCATION device->adapter->gl_info
3323 struct arbfp_blit_priv {
3324 GLenum yuy2_rect_shader, yuy2_2d_shader;
3325 GLenum uyvy_rect_shader, uyvy_2d_shader;
3326 GLenum yv12_rect_shader, yv12_2d_shader;
3329 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3330 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3331 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3332 if(!device->blit_priv) {
3333 ERR("Out of memory\n");
3334 return E_OUTOFMEMORY;
3338 static void arbfp_blit_free(IWineD3DDevice *iface) {
3339 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3340 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3343 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3344 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3345 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3346 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3347 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3348 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3349 checkGLcall("Delete yuv programs\n");
3353 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3355 const char *tex, *texinstr;
3357 if(fmt == WINED3DFMT_UYVY) {
3365 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3366 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3368 /* This is more tricky than just replacing the texture type - we have to navigate
3369 * properly in the texture to find the correct chroma values
3371 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3375 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3376 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3377 * filtering when we sample the texture.
3379 * These are the rules for reading the chroma:
3385 * So we have to get the sampling x position in non-normalized coordinates in integers
3387 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3388 shader_addline(buffer, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3389 shader_addline(buffer, "MOV texcrd.a, size.x;\n");
3391 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3393 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3394 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3397 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3398 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3400 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3401 * even and odd pixels respectively
3403 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3404 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3406 /* Sample Pixel 1 */
3407 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3409 /* Put the value into either of the chroma values */
3410 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3411 shader_addline(buffer, "MUL chroma.r, luminance.%c, temp.x;\n", chroma);
3412 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3413 shader_addline(buffer, "MUL chroma.g, luminance.%c, temp.x;\n", chroma);
3415 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3416 * the pixel right to the current one. Otherwise, sample the left pixel.
3417 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3419 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3420 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3421 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3423 /* Put the value into the other chroma */
3424 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3425 shader_addline(buffer, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma);
3426 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3427 shader_addline(buffer, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma);
3429 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3430 * the current one and lerp the two U and V values
3433 /* This gives the correctly filtered luminance value */
3434 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3439 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3443 case GL_TEXTURE_2D: tex = "2D"; break;
3444 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3446 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3450 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3451 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3452 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3453 * pitch of the luminance plane, the packing into the gl texture is a bit
3454 * unfortunate. If the whole texture is interpreted as luminance data it looks
3455 * approximately like this:
3457 * +----------------------------------+----
3469 * +----------------+-----------------+----
3471 * | U even rows | U odd rows |
3473 * +----------------+------------------ -
3475 * | V even rows | V odd rows |
3477 * +----------------+-----------------+----
3481 * So it appears as if there are 4 chroma images, but in fact the odd rows
3482 * in the chroma images are in the same row as the even ones. So its is
3483 * kinda tricky to read
3485 * When reading from rectangle textures, keep in mind that the input y coordinates
3486 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3488 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3489 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3491 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3492 /* the chroma planes have only half the width */
3493 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3495 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3496 * the coordinate. Also read the right side of the image when reading odd lines
3498 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3501 if(textype == GL_TEXTURE_2D) {
3503 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3505 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3507 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3508 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3510 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3511 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3512 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3513 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3514 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3516 /* clamp, keep the half pixel origin in mind */
3517 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3518 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3519 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3520 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3522 /* Read from [size - size+size/4] */
3523 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3524 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3526 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3527 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3528 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3529 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3530 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3531 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3533 /* Make sure to read exactly from the pixel center */
3534 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3535 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3538 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3539 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3540 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3541 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3542 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3544 /* Read the texture, put the result into the output register */
3545 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3546 shader_addline(buffer, "MOV chroma.r, temp.a;\n");
3548 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3549 * No need to clamp because we're just reusing the already clamped value from above
3551 if(textype == GL_TEXTURE_2D) {
3552 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3554 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3556 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3557 shader_addline(buffer, "MOV chroma.g, temp.a;\n");
3559 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3560 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3561 * values due to filtering
3563 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3564 if(textype == GL_TEXTURE_2D) {
3565 /* Multiply the y coordinate by 2/3 and clamp it */
3566 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3567 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3568 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3569 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3571 /* Reading from texture_rectangles is pretty streightforward, just use the unmodified
3572 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3575 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3576 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3577 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3584 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, WINED3DFORMAT fmt, GLenum textype) {
3586 SHADER_BUFFER buffer;
3587 char luminance_component;
3588 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3593 buffer.newline = TRUE;
3594 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
3597 GL_EXTCALL(glGenProgramsARB(1, &shader));
3598 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3599 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3600 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3603 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3607 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3608 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3609 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3610 * each single pixel it contains, and one U and one V value shared between both
3613 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3614 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3615 * take the format into account when generating the read swizzles
3617 * Reading the Y value is streightforward - just sample the texture. The hardware
3618 * takes care of filtering in the horizontal and vertical direction.
3620 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3621 * because that would mix the U and V values of one pixel or two adjacent pixels.
3622 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3623 * regardless of the filtering setting. Vertical filtering works automatically
3624 * though - the U and V values of two rows are mixed nicely.
3626 * Appart of avoiding filtering issues, the code has to know which value it just
3627 * read, and where it can find the other one. To determine this, it checks if
3628 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3630 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3631 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3633 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3634 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3635 * in an unfiltered situation. Finding the luminance on the other hand requires
3636 * finding out if it is an odd or even pixel. The real drawback of this approach
3637 * is filtering. This would have to be emulated completely in the shader, reading
3638 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3639 * vertically. Beyond that it would require adjustments to the texture handling
3640 * code to deal with the width scaling
3642 shader_addline(&buffer, "!!ARBfp1.0\n");
3643 shader_addline(&buffer, "TEMP luminance;\n");
3644 shader_addline(&buffer, "TEMP temp;\n");
3645 shader_addline(&buffer, "TEMP chroma;\n");
3646 shader_addline(&buffer, "TEMP texcrd;\n");
3647 shader_addline(&buffer, "TEMP texcrd2;\n");
3648 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3649 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3650 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3652 if(fmt == WINED3DFMT_UYVY || fmt ==WINED3DFMT_YUY2) {
3653 if(gen_planar_yuv_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3654 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3658 if(gen_yv12_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3659 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3664 /* Calculate the final result. Formula is taken from
3665 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3666 * ranges from -0.5 to 0.5
3668 shader_addline(&buffer, "SUB chroma.rg, chroma, coef.y;\n");
3670 shader_addline(&buffer, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component);
3671 shader_addline(&buffer, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component);
3672 shader_addline(&buffer, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3673 shader_addline(&buffer, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component);
3674 shader_addline(&buffer, "END\n");
3677 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3679 if (glGetError() == GL_INVALID_OPERATION) {
3681 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3682 FIXME("Fragment program error at position %d: %s\n", pos,
3683 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3685 HeapFree(GetProcessHeap(), 0, buffer.buffer);
3688 if(fmt == WINED3DFMT_YUY2) {
3689 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3690 priv->yuy2_rect_shader = shader;
3692 priv->yuy2_2d_shader = shader;
3694 } else if(fmt == WINED3DFMT_UYVY) {
3695 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3696 priv->uyvy_rect_shader = shader;
3698 priv->uyvy_2d_shader = shader;
3701 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3702 priv->yv12_rect_shader = shader;
3704 priv->yv12_2d_shader = shader;
3710 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3712 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3713 float size[4] = {width, height, 1, 1};
3714 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3715 const GlPixelFormatDesc *glDesc;
3717 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3719 if(glDesc->conversion_group != WINED3DFMT_YUY2 && glDesc->conversion_group != WINED3DFMT_UYVY &&
3720 glDesc->conversion_group != WINED3DFMT_YV12) {
3721 TRACE("Format: %s\n", debug_d3dformat(glDesc->conversion_group));
3722 /* Don't bother setting up a shader for unconverted formats */
3725 checkGLcall("glEnable(textype)");
3730 if(glDesc->conversion_group == WINED3DFMT_YUY2) {
3731 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3732 shader = priv->yuy2_rect_shader;
3734 shader = priv->yuy2_2d_shader;
3736 } else if(glDesc->conversion_group == WINED3DFMT_UYVY) {
3737 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3738 shader = priv->uyvy_rect_shader;
3740 shader = priv->uyvy_2d_shader;
3743 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3744 shader = priv->yv12_rect_shader;
3746 shader = priv->yv12_2d_shader;
3751 shader = gen_yuv_shader(device, glDesc->conversion_group, textype);
3755 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3756 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3757 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3758 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3759 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3760 checkGLcall("glProgramLocalParameter4fvARB");
3766 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3767 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3770 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3771 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3772 glDisable(GL_TEXTURE_2D);
3773 checkGLcall("glDisable(GL_TEXTURE_2D)");
3774 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3775 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3776 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3778 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3779 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3780 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3785 static BOOL arbfp_blit_conv_supported(WINED3DFORMAT fmt) {
3786 TRACE("Checking blit format support for format %s:", debug_d3dformat(fmt));
3788 case WINED3DFMT_YUY2:
3789 case WINED3DFMT_UYVY:
3790 case WINED3DFMT_YV12:
3794 TRACE("[FAILED]\n");
3799 const struct blit_shader arbfp_blit = {
3804 arbfp_blit_conv_supported
3807 #undef GLINFO_LOCATION