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);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d);
41 #define GLINFO_LOCATION (*gl_info)
43 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
44 * so upload them above that
46 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
47 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
49 /* ARB_program_shader private data */
50 struct shader_arb_priv {
51 GLuint current_vprogram_id;
52 GLuint current_fprogram_id;
53 GLuint depth_blt_vprogram_id;
54 GLuint depth_blt_fprogram_id[tex_type_count];
55 BOOL use_arbfp_fixed_func;
56 struct hash_table_t *fragment_shaders;
59 /********************************************************
60 * ARB_[vertex/fragment]_program helper functions follow
61 ********************************************************/
64 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
65 * When constant_list == NULL, it will load all the constants.
67 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
68 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
70 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, const WineD3D_GL_Info *gl_info,
71 GLuint target_type, unsigned int max_constants, const float *constants, char *dirty_consts)
73 DWORD shader_version = This->baseShader.reg_maps.shader_version;
74 local_constant* lconst;
78 if (TRACE_ON(d3d_shader)) {
79 for(i = 0; i < max_constants; i++) {
80 if(!dirty_consts[i]) continue;
81 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
82 constants[i * 4 + 0], constants[i * 4 + 1],
83 constants[i * 4 + 2], constants[i * 4 + 3]);
86 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
87 if (target_type == GL_FRAGMENT_PROGRAM_ARB && WINED3DSHADER_VERSION_MAJOR(shader_version) == 1)
90 for(i = 0; i < max_constants; i++) {
91 if(!dirty_consts[i]) continue;
95 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
96 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
97 else lcl_const[0] = constants[j + 0];
99 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
100 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
101 else lcl_const[1] = constants[j + 1];
103 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
104 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
105 else lcl_const[2] = constants[j + 2];
107 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
108 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
109 else lcl_const[3] = constants[j + 3];
111 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
114 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
115 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
116 * or just reloading *all* constants at once
118 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
120 for(i = 0; i < max_constants; i++) {
121 if(!dirty_consts[i]) continue;
123 /* Find the next block of dirty constants */
126 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
130 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
133 for(i = 0; i < max_constants; i++) {
134 if(dirty_consts[i]) {
136 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
141 checkGLcall("glProgramEnvParameter4fvARB()");
143 /* Load immediate constants */
144 if(This->baseShader.load_local_constsF) {
145 if (TRACE_ON(d3d_shader)) {
146 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
147 GLfloat* values = (GLfloat*)lconst->value;
148 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
149 values[0], values[1], values[2], values[3]);
152 /* Immediate constants are clamped for 1.X shaders at loading times */
154 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
155 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
156 ret = max(ret, lconst->idx + 1);
157 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
159 checkGLcall("glProgramEnvParameter4fvARB()");
160 return ret; /* The loaded immediate constants need reloading for the next shader */
162 return 0; /* No constants are dirty now */
167 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
169 * We only support float constants in ARB at the moment, so don't
170 * worry about the Integers or Booleans
172 static void shader_arb_load_constants(
173 IWineD3DDevice* device,
175 char useVertexShader) {
177 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
178 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
179 const WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
182 if (useVertexShader) {
183 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
185 /* Load DirectX 9 float constants for vertex shader */
186 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
187 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
188 deviceImpl->highest_dirty_vs_const,
189 stateBlock->vertexShaderConstantF,
190 deviceImpl->activeContext->vshader_const_dirty);
192 /* Upload the position fixup */
193 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
196 if (usePixelShader) {
198 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
199 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
201 /* Load DirectX 9 float constants for pixel shader */
202 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
203 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
204 deviceImpl->highest_dirty_ps_const,
205 stateBlock->pixelShaderConstantF,
206 deviceImpl->activeContext->pshader_const_dirty);
208 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
209 /* The state manager takes care that this function is always called if the bump env matrix changes
211 const float *data = (const float *)&stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
212 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
213 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
215 if(psi->luminanceconst[i].const_num != -1) {
216 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
217 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
218 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
219 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
221 const float *scale = (const float *)&stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
222 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
223 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
229 static void shader_arb_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count)
231 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
233 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
234 * context. On a context switch the old context will be fully dirtified */
235 memset(This->activeContext->vshader_const_dirty + start, 1,
236 sizeof(*This->activeContext->vshader_const_dirty) * count);
237 This->highest_dirty_vs_const = max(This->highest_dirty_vs_const, start + count + 1);
240 static void shader_arb_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count)
242 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
244 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
245 * context. On a context switch the old context will be fully dirtified */
246 memset(This->activeContext->pshader_const_dirty + start, 1,
247 sizeof(*This->activeContext->pshader_const_dirty) * count);
248 This->highest_dirty_ps_const = max(This->highest_dirty_ps_const, start + count + 1);
251 /* Generate the variable & register declarations for the ARB_vertex_program output target */
252 static void shader_generate_arb_declarations(IWineD3DBaseShader *iface, const shader_reg_maps *reg_maps,
253 SHADER_BUFFER *buffer, const WineD3D_GL_Info *gl_info)
255 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
256 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
258 char pshader = shader_is_pshader_version(reg_maps->shader_version);
259 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
260 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
261 UINT extra_constants_needed = 0;
262 const local_constant *lconst;
264 /* Temporary Output register */
265 shader_addline(buffer, "TEMP TMP_OUT;\n");
267 for(i = 0; i < This->baseShader.limits.temporary; i++) {
268 if (reg_maps->temporary[i])
269 shader_addline(buffer, "TEMP R%u;\n", i);
272 for (i = 0; i < This->baseShader.limits.address; i++) {
273 if (reg_maps->address[i])
274 shader_addline(buffer, "ADDRESS A%d;\n", i);
277 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
278 if (reg_maps->texcoord[i])
279 shader_addline(buffer,"TEMP T%u;\n", i);
282 /* Texture coordinate registers must be pre-loaded */
283 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
284 if (reg_maps->texcoord[i])
285 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
288 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
289 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
290 if(!reg_maps->bumpmat[i]) continue;
292 cur = ps->numbumpenvmatconsts;
293 ps->bumpenvmatconst[cur].const_num = -1;
294 ps->bumpenvmatconst[cur].texunit = i;
295 ps->luminanceconst[cur].const_num = -1;
296 ps->luminanceconst[cur].texunit = i;
298 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
299 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
302 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
303 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
304 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
305 i, ps->bumpenvmatconst[cur].const_num);
306 extra_constants_needed++;
308 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
309 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
310 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
311 i, ps->luminanceconst[cur].const_num);
312 extra_constants_needed++;
313 } else if(reg_maps->luminanceparams) {
314 FIXME("No free constant to load the luminance parameters\n");
317 FIXME("No free constant found to load environment bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
320 ps->numbumpenvmatconsts = cur + 1;
323 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
324 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
325 srgb_mul_low, srgb_mul_low, srgb_mul_low);
326 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
327 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
328 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
329 srgb_pow, srgb_pow, srgb_pow);
330 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
331 srgb_mul_high, srgb_mul_high, srgb_mul_high);
332 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
333 srgb_sub_high, srgb_sub_high, srgb_sub_high);
336 /* Load local constants using the program-local space,
337 * this avoids reloading them each time the shader is used
339 if(!This->baseShader.load_local_constsF) {
340 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
341 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
346 /* we use the array-based constants array if the local constants are marked for loading,
347 * because then we use indirect addressing, or when the local constant list is empty,
348 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
349 * local constants do not declare the loaded constants as an array because ARB compilers usually
350 * do not optimize unused constants away
352 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
353 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
354 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
355 max_constantsF, max_constantsF - 1);
357 for(i = 0; i < max_constantsF; i++) {
358 if(!shader_constant_is_local(This, i)) {
359 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
365 static const char * const shift_tab[] = {
366 "dummy", /* 0 (none) */
367 "coefmul.x", /* 1 (x2) */
368 "coefmul.y", /* 2 (x4) */
369 "coefmul.z", /* 3 (x8) */
370 "coefmul.w", /* 4 (x16) */
371 "dummy", /* 5 (x32) */
372 "dummy", /* 6 (x64) */
373 "dummy", /* 7 (x128) */
374 "dummy", /* 8 (d256) */
375 "dummy", /* 9 (d128) */
376 "dummy", /* 10 (d64) */
377 "dummy", /* 11 (d32) */
378 "coefdiv.w", /* 12 (d16) */
379 "coefdiv.z", /* 13 (d8) */
380 "coefdiv.y", /* 14 (d4) */
381 "coefdiv.x" /* 15 (d2) */
384 static void shader_arb_get_write_mask(const SHADER_OPCODE_ARG *arg, const DWORD param, char *write_mask)
386 char *ptr = write_mask;
387 char vshader = shader_is_vshader_version(arg->reg_maps->shader_version);
389 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
392 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
394 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
395 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
396 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
397 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
403 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
404 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
405 * but addressed as "rgba". To fix this we need to swap the register's x
406 * and z components. */
407 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
408 char *ptr = swizzle_str;
410 /* swizzle bits fields: wwzzyyxx */
411 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
412 DWORD swizzle_x = swizzle & 0x03;
413 DWORD swizzle_y = (swizzle >> 2) & 0x03;
414 DWORD swizzle_z = (swizzle >> 4) & 0x03;
415 DWORD swizzle_w = (swizzle >> 6) & 0x03;
417 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
418 * generate a swizzle string. Unless we need to our own swizzling. */
419 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
421 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
422 *ptr++ = swizzle_chars[swizzle_x];
424 *ptr++ = swizzle_chars[swizzle_x];
425 *ptr++ = swizzle_chars[swizzle_y];
426 *ptr++ = swizzle_chars[swizzle_z];
427 *ptr++ = swizzle_chars[swizzle_w];
434 static void pshader_get_register_name(IWineD3DBaseShader* iface,
435 const DWORD param, char* regstr) {
437 DWORD reg = param & WINED3DSP_REGNUM_MASK;
438 DWORD regtype = shader_get_regtype(param);
439 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
442 case WINED3DSPR_TEMP:
443 sprintf(regstr, "R%u", reg);
445 case WINED3DSPR_INPUT:
447 strcpy(regstr, "fragment.color.primary");
449 strcpy(regstr, "fragment.color.secondary");
452 case WINED3DSPR_CONST:
453 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
454 sprintf(regstr, "C[%u]", reg);
456 sprintf(regstr, "C%u", reg);
459 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
460 sprintf(regstr,"T%u", reg);
462 case WINED3DSPR_COLOROUT:
464 sprintf(regstr, "TMP_COLOR");
466 /* TODO: See GL_ARB_draw_buffers */
467 FIXME("Unsupported write to render target %u\n", reg);
468 sprintf(regstr, "unsupported_register");
471 case WINED3DSPR_DEPTHOUT:
472 sprintf(regstr, "result.depth");
474 case WINED3DSPR_ATTROUT:
475 sprintf(regstr, "oD[%u]", reg);
477 case WINED3DSPR_TEXCRDOUT:
478 sprintf(regstr, "oT[%u]", reg);
481 FIXME("Unhandled register name Type(%d)\n", regtype);
482 sprintf(regstr, "unrecognized_register");
487 /* TODO: merge with pixel shader */
488 static void vshader_program_add_param(const SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine)
490 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
492 /* oPos, oFog and oPts in D3D */
493 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
495 DWORD reg = param & WINED3DSP_REGNUM_MASK;
496 DWORD regtype = shader_get_regtype(param);
498 BOOL is_color = FALSE;
500 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
501 strcat(hwLine, " -");
507 case WINED3DSPR_TEMP:
508 sprintf(tmpReg, "R%u", reg);
509 strcat(hwLine, tmpReg);
511 case WINED3DSPR_INPUT:
513 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
516 sprintf(tmpReg, "vertex.attrib[%u]", reg);
517 strcat(hwLine, tmpReg);
519 case WINED3DSPR_CONST:
520 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
521 if(reg >= This->rel_offset) {
522 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
524 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
527 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
528 sprintf(tmpReg, "C[%u]", reg);
530 sprintf(tmpReg, "C%u", reg);
533 strcat(hwLine, tmpReg);
535 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
536 sprintf(tmpReg, "A%u", reg);
537 strcat(hwLine, tmpReg);
539 case WINED3DSPR_RASTOUT:
540 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
541 strcat(hwLine, tmpReg);
543 case WINED3DSPR_ATTROUT:
545 strcat(hwLine, "result.color.primary");
547 strcat(hwLine, "result.color.secondary");
550 case WINED3DSPR_TEXCRDOUT:
551 sprintf(tmpReg, "result.texcoord[%u]", reg);
552 strcat(hwLine, tmpReg);
555 FIXME("Unknown reg type %d %d\n", regtype, reg);
556 strcat(hwLine, "unrecognized_register");
562 shader_arb_get_write_mask(arg, param, write_mask);
563 strcat(hwLine, write_mask);
566 shader_arb_get_swizzle(param, is_color, swizzle);
567 strcat(hwLine, swizzle);
571 static void shader_hw_sample(const SHADER_OPCODE_ARG *arg, DWORD sampler_idx, const char *dst_str,
572 const char *coord_reg, BOOL projected, BOOL bias)
574 SHADER_BUFFER* buffer = arg->buffer;
575 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
576 const char *tex_type;
578 switch(sampler_type) {
585 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
586 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
587 if(device->stateBlock->textures[sampler_idx] &&
588 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
596 case WINED3DSTT_VOLUME:
600 case WINED3DSTT_CUBE:
605 ERR("Unexpected texture type %d\n", sampler_type);
610 /* Shouldn't be possible, but let's check for it */
611 if(projected) FIXME("Biased and Projected texture sampling\n");
612 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
613 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
614 } else if (projected) {
615 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
617 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
621 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source)
623 switch(channel_source)
625 case CHANNEL_SOURCE_ZERO: return "0";
626 case CHANNEL_SOURCE_ONE: return "1";
627 case CHANNEL_SOURCE_X: return "x";
628 case CHANNEL_SOURCE_Y: return "y";
629 case CHANNEL_SOURCE_Z: return "z";
630 case CHANNEL_SOURCE_W: return "w";
632 FIXME("Unhandled channel source %#x\n", channel_source);
637 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, DWORD dst_mask,
638 const char *one, const char *two, struct color_fixup_desc fixup)
642 if (is_yuv_fixup(fixup))
644 enum yuv_fixup yuv_fixup = get_yuv_fixup(fixup);
645 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup);
650 if (fixup.x_source != CHANNEL_SOURCE_X) mask |= WINED3DSP_WRITEMASK_0;
651 if (fixup.y_source != CHANNEL_SOURCE_Y) mask |= WINED3DSP_WRITEMASK_1;
652 if (fixup.z_source != CHANNEL_SOURCE_Z) mask |= WINED3DSP_WRITEMASK_2;
653 if (fixup.w_source != CHANNEL_SOURCE_W) mask |= WINED3DSP_WRITEMASK_3;
658 shader_addline(buffer, "SWZ %s, %s, %s, %s, %s, %s;\n", reg, reg,
659 shader_arb_get_fixup_swizzle(fixup.x_source), shader_arb_get_fixup_swizzle(fixup.y_source),
660 shader_arb_get_fixup_swizzle(fixup.z_source), shader_arb_get_fixup_swizzle(fixup.w_source));
664 if (fixup.x_sign_fixup) mask |= WINED3DSP_WRITEMASK_0;
665 if (fixup.y_sign_fixup) mask |= WINED3DSP_WRITEMASK_1;
666 if (fixup.z_sign_fixup) mask |= WINED3DSP_WRITEMASK_2;
667 if (fixup.w_sign_fixup) mask |= WINED3DSP_WRITEMASK_3;
673 char *ptr = reg_mask;
675 if (mask != WINED3DSP_WRITEMASK_ALL)
678 if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
679 if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
680 if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
681 if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
685 shader_addline(buffer, "MAD %s%s, %s, %s, -%s;\n", reg, reg_mask, reg, two, one);
689 static void shader_arb_color_correction(const struct SHADER_OPCODE_ARG* arg, struct color_fixup_desc fixup)
692 pshader_get_register_name(arg->shader, arg->dst, reg);
693 gen_color_correction(arg->buffer, reg, arg->dst & WINED3DSP_WRITEMASK_ALL, "one", "coefmul.x", fixup);
696 static void pshader_gen_input_modifier_line (
697 IWineD3DBaseShader *iface,
698 SHADER_BUFFER* buffer,
703 /* Generate a line that does the input modifier computation and return the input register to use */
708 /* Assume a new line will be added */
711 /* Get register name */
712 pshader_get_register_name(iface, instr, regstr);
713 shader_arb_get_swizzle(instr, FALSE, swzstr);
715 switch (instr & WINED3DSP_SRCMOD_MASK) {
716 case WINED3DSPSM_NONE:
717 sprintf(outregstr, "%s%s", regstr, swzstr);
720 case WINED3DSPSM_NEG:
721 sprintf(outregstr, "-%s%s", regstr, swzstr);
724 case WINED3DSPSM_BIAS:
725 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
727 case WINED3DSPSM_BIASNEG:
728 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
730 case WINED3DSPSM_SIGN:
731 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
733 case WINED3DSPSM_SIGNNEG:
734 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
736 case WINED3DSPSM_COMP:
737 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
740 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
742 case WINED3DSPSM_X2NEG:
743 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
746 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
747 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
750 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
751 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
754 sprintf(outregstr, "%s%s", regstr, swzstr);
758 /* Return modified or original register, with swizzle */
760 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
763 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER *buffer, int saturate, const char *write_mask,
764 int shift, const char *regstr)
766 /* Generate a line that does the output modifier computation */
767 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
768 regstr, write_mask, regstr, shift_tab[shift]);
771 static void pshader_hw_bem(const SHADER_OPCODE_ARG *arg)
773 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
775 SHADER_BUFFER* buffer = arg->buffer;
777 char src_name[2][50];
779 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
780 BOOL has_bumpmat = FALSE;
783 for(i = 0; i < This->numbumpenvmatconsts; i++) {
784 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
790 pshader_get_register_name(arg->shader, arg->dst, dst_name);
791 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
792 strcat(dst_name, dst_wmask);
794 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
795 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
798 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
799 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
800 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
801 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
802 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
804 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
806 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
810 static void pshader_hw_cnd(const SHADER_OPCODE_ARG *arg)
812 SHADER_BUFFER* buffer = arg->buffer;
815 char src_name[3][50];
816 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
817 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
819 /* FIXME: support output modifiers */
821 /* Handle output register */
822 pshader_get_register_name(arg->shader, arg->dst, dst_name);
823 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
825 /* Generate input register names (with modifiers) */
826 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
827 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
828 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
830 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
831 if (arg->reg_maps->shader_version <= WINED3DPS_VERSION(1, 3) && arg->opcode_token & WINED3DSI_COISSUE)
833 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
835 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
836 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
837 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
840 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
843 static void pshader_hw_cmp(const SHADER_OPCODE_ARG *arg)
845 SHADER_BUFFER* buffer = arg->buffer;
848 char src_name[3][50];
849 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
850 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
852 /* FIXME: support output modifiers */
854 /* Handle output register */
855 pshader_get_register_name(arg->shader, arg->dst, dst_name);
856 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
858 /* Generate input register names (with modifiers) */
859 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
860 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
861 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
863 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
864 src_name[0], src_name[2], src_name[1]);
867 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
870 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
871 * dst = dot2(src0, src1) + src2 */
872 static void pshader_hw_dp2add(const SHADER_OPCODE_ARG *arg)
874 SHADER_BUFFER* buffer = arg->buffer;
877 char src_name[3][50];
878 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
879 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
881 pshader_get_register_name(arg->shader, arg->dst, dst_name);
882 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
884 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
885 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
886 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
888 /* Emulate a DP2 with a DP3 and 0.0 */
889 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
890 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
891 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
892 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
895 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
898 /* Map the opcode 1-to-1 to the GL code */
899 static void shader_hw_map2gl(const SHADER_OPCODE_ARG *arg)
901 CONST SHADER_OPCODE* curOpcode = arg->opcode;
902 SHADER_BUFFER* buffer = arg->buffer;
903 DWORD dst = arg->dst;
904 const DWORD *src = arg->src;
905 const char *instruction;
909 switch (curOpcode->opcode)
911 case WINED3DSIO_ABS: instruction = "ABS"; break;
912 case WINED3DSIO_ADD: instruction = "ADD"; break;
913 case WINED3DSIO_CRS: instruction = "XPD"; break;
914 case WINED3DSIO_DP3: instruction = "DP3"; break;
915 case WINED3DSIO_DP4: instruction = "DP4"; break;
916 case WINED3DSIO_DST: instruction = "DST"; break;
917 case WINED3DSIO_EXP: instruction = "EX2"; break;
918 case WINED3DSIO_EXPP: instruction = "EXP"; break;
919 case WINED3DSIO_FRC: instruction = "FRC"; break;
920 case WINED3DSIO_LIT: instruction = "LIT"; break;
921 case WINED3DSIO_LOG: instruction = "LG2"; break;
922 case WINED3DSIO_LOGP: instruction = "LOG"; break;
923 case WINED3DSIO_LRP: instruction = "LRP"; break;
924 case WINED3DSIO_MAD: instruction = "MAD"; break;
925 case WINED3DSIO_MAX: instruction = "MAX"; break;
926 case WINED3DSIO_MIN: instruction = "MIN"; break;
927 case WINED3DSIO_MOV: instruction = "MOV"; break;
928 case WINED3DSIO_MUL: instruction = "MUL"; break;
929 case WINED3DSIO_NOP: instruction = "NOP"; break;
930 case WINED3DSIO_POW: instruction = "POW"; break;
931 case WINED3DSIO_SGE: instruction = "SGE"; break;
932 case WINED3DSIO_SLT: instruction = "SLT"; break;
933 case WINED3DSIO_SUB: instruction = "SUB"; break;
934 default: instruction = "";
935 FIXME("Unhandled opcode %s\n", curOpcode->name);
939 if (shader_is_pshader_version(arg->reg_maps->shader_version))
941 /* Output token related */
942 char output_rname[256];
943 char output_wmask[20];
944 char operands[4][100];
945 BOOL saturate = FALSE;
946 BOOL centroid = FALSE;
947 BOOL partialprecision = FALSE;
948 const char *modifier;
951 if (!curOpcode->num_params)
953 ERR("Opcode \"%s\" has no parameters\n", curOpcode->name);
957 /* Process modifiers */
958 if (dst & WINED3DSP_DSTMOD_MASK)
960 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
962 saturate = mask & WINED3DSPDM_SATURATE;
963 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
964 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
965 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
967 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
970 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
972 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
973 modifier = (saturate && !shift) ? "_SAT" : "";
975 /* Generate input register names (with modifiers) */
976 for (i = 1; i < curOpcode->num_params; ++i)
977 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
979 /* Handle output register */
980 pshader_get_register_name(arg->shader, dst, output_rname);
981 strcpy(operands[0], output_rname);
982 shader_arb_get_write_mask(arg, dst, output_wmask);
983 strcat(operands[0], output_wmask);
986 strcat(arguments, operands[0]);
987 for (i = 1; i < curOpcode->num_params; i++)
989 strcat(arguments, ", ");
990 strcat(arguments, operands[i]);
992 shader_addline(buffer, "%s%s %s;\n", instruction, modifier, arguments);
994 /* A shift requires another line. */
995 if (shift) pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
997 /* Note that vshader_program_add_param() adds spaces. */
1000 if (curOpcode->num_params > 0)
1002 vshader_program_add_param(arg, dst, FALSE, arguments);
1003 for (i = 1; i < curOpcode->num_params; ++i)
1005 strcat(arguments, ",");
1006 vshader_program_add_param(arg, src[i-1], TRUE, arguments);
1009 shader_addline(buffer, "%s%s;\n", instruction, arguments);
1013 static void shader_hw_mov(const SHADER_OPCODE_ARG *arg)
1015 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl*)arg->shader;
1017 if ((WINED3DSHADER_VERSION_MAJOR(arg->reg_maps->shader_version) == 1
1018 && !shader_is_pshader_version(arg->reg_maps->shader_version)
1019 && shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)
1020 || arg->opcode->opcode == WINED3DSIO_MOVA)
1022 SHADER_BUFFER *buffer = arg->buffer;
1023 char src0_param[256];
1025 if (arg->opcode->opcode == WINED3DSIO_MOVA)
1026 FIXME("mova should round\n");
1028 src0_param[0] = '\0';
1029 if (((IWineD3DVertexShaderImpl *)shader)->rel_offset)
1031 vshader_program_add_param(arg, arg->src[0], TRUE, src0_param);
1032 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", src0_param);
1033 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1037 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1038 * with more than one component. Thus replicate the first source argument over all
1039 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1041 DWORD parm = arg->src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1042 if((arg->src[0] & WINED3DVS_X_W) == WINED3DVS_X_W)
1043 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1044 else if((arg->src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z)
1045 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1046 else if((arg->src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y)
1047 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1048 else if((arg->src[0] & WINED3DVS_X_X) == WINED3DVS_X_X)
1049 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1050 vshader_program_add_param(arg, parm, TRUE, src0_param);
1051 shader_addline(buffer, "ARL A0.x, %s;\n", src0_param);
1056 shader_hw_map2gl(arg);
1060 static void pshader_hw_texkill(const SHADER_OPCODE_ARG *arg)
1062 DWORD shader_version = arg->reg_maps->shader_version;
1063 SHADER_BUFFER* buffer = arg->buffer;
1066 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1067 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1069 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1071 if (shader_version >= WINED3DPS_VERSION(2,0))
1073 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1074 shader_addline(buffer, "KIL %s;\n", reg_dest);
1076 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1077 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1079 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1080 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1081 shader_addline(buffer, "KIL TMP;\n");
1085 static void pshader_hw_tex(const SHADER_OPCODE_ARG *arg)
1087 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1088 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1090 DWORD dst = arg->dst;
1091 const DWORD *src = arg->src;
1092 SHADER_BUFFER* buffer = arg->buffer;
1093 DWORD shader_version = arg->reg_maps->shader_version;
1094 BOOL projected = FALSE, bias = FALSE;
1098 DWORD reg_dest_code;
1099 DWORD reg_sampler_code;
1101 /* All versions have a destination register */
1102 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1103 pshader_get_register_name(arg->shader, dst, reg_dest);
1105 /* 1.0-1.3: Use destination register as coordinate source.
1106 1.4+: Use provided coordinate source register. */
1107 if (shader_version < WINED3DPS_VERSION(1,4))
1108 strcpy(reg_coord, reg_dest);
1110 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1112 /* 1.0-1.4: Use destination register number as texture code.
1113 2.0+: Use provided sampler number as texure code. */
1114 if (shader_version < WINED3DPS_VERSION(2,0))
1115 reg_sampler_code = reg_dest_code;
1117 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1120 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1121 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1122 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1124 if (shader_version < WINED3DPS_VERSION(1,4))
1127 if(reg_sampler_code < MAX_TEXTURES) {
1128 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1130 if (flags & WINED3DTTFF_PROJECTED) {
1134 else if (shader_version < WINED3DPS_VERSION(2,0))
1136 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1137 if (src_mod == WINED3DSPSM_DZ) {
1139 } else if(src_mod == WINED3DSPSM_DW) {
1143 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1146 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1150 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1153 static void pshader_hw_texcoord(const SHADER_OPCODE_ARG *arg)
1155 DWORD dst = arg->dst;
1156 SHADER_BUFFER* buffer = arg->buffer;
1159 shader_arb_get_write_mask(arg, dst, tmp);
1160 if (arg->reg_maps->shader_version != WINED3DPS_VERSION(1,4))
1162 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1163 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1165 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1168 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1169 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1173 static void pshader_hw_texreg2ar(const SHADER_OPCODE_ARG *arg)
1175 SHADER_BUFFER* buffer = arg->buffer;
1176 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1177 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1180 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1184 sprintf(dst_str, "T%u", reg1);
1185 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1186 shader_addline(buffer, "MOV TMP.x, %s.w;\n", src_str);
1187 shader_addline(buffer, "MOV TMP.y, %s.x;\n", src_str);
1188 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1189 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1192 static void pshader_hw_texreg2gb(const SHADER_OPCODE_ARG *arg)
1194 SHADER_BUFFER* buffer = arg->buffer;
1196 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1200 sprintf(dst_str, "T%u", reg1);
1201 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1202 shader_addline(buffer, "MOV TMP.x, %s.y;\n", src_str);
1203 shader_addline(buffer, "MOV TMP.y, %s.z;\n", src_str);
1204 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1207 static void pshader_hw_texreg2rgb(const SHADER_OPCODE_ARG *arg)
1209 SHADER_BUFFER* buffer = arg->buffer;
1210 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1214 sprintf(dst_str, "T%u", reg1);
1215 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1216 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1219 static void pshader_hw_texbem(const SHADER_OPCODE_ARG *arg)
1221 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1222 BOOL has_bumpmat = FALSE;
1223 BOOL has_luminance = FALSE;
1226 DWORD dst = arg->dst;
1227 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1228 SHADER_BUFFER* buffer = arg->buffer;
1231 DWORD reg_dest_code;
1233 /* All versions have a destination register */
1234 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1235 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1236 pshader_get_register_name(arg->shader, dst, reg_coord);
1238 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1239 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1244 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1245 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1246 has_luminance = TRUE;
1252 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1254 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1255 shader_addline(buffer, "DP3 TMP.x, TMP2, T%u;\n", src);
1256 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1257 shader_addline(buffer, "DP3 TMP.y, TMP2, T%u;\n", src);
1259 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1260 * so we can't let the GL handle this.
1262 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1263 & WINED3DTTFF_PROJECTED) {
1264 shader_addline(buffer, "RCP TMP2.w, %s.w;\n", reg_coord);
1265 shader_addline(buffer, "MUL TMP2.xy, %s, TMP2.w;\n", reg_coord);
1266 shader_addline(buffer, "ADD TMP.xy, TMP, TMP2;\n");
1268 shader_addline(buffer, "ADD TMP.xy, TMP, %s;\n", reg_coord);
1271 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1273 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1274 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1275 src, reg_dest_code, reg_dest_code);
1276 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1281 if(reg_dest_code < MAX_TEXTURES) {
1282 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1286 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1287 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1291 static void pshader_hw_texm3x2pad(const SHADER_OPCODE_ARG *arg)
1293 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1294 SHADER_BUFFER* buffer = arg->buffer;
1297 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1298 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1301 static void pshader_hw_texm3x2tex(const SHADER_OPCODE_ARG *arg)
1303 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1304 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1306 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1307 SHADER_BUFFER* buffer = arg->buffer;
1311 sprintf(dst_str, "T%u", reg);
1312 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1313 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1314 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1315 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1318 static void pshader_hw_texm3x3pad(const SHADER_OPCODE_ARG *arg)
1320 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1321 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1322 SHADER_BUFFER* buffer = arg->buffer;
1323 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1326 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1327 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1328 current_state->texcoord_w[current_state->current_row++] = reg;
1331 static void pshader_hw_texm3x3tex(const SHADER_OPCODE_ARG *arg)
1333 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1334 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1336 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1337 SHADER_BUFFER* buffer = arg->buffer;
1338 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1342 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1343 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1345 /* Sample the texture using the calculated coordinates */
1346 sprintf(dst_str, "T%u", reg);
1347 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1348 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1349 current_state->current_row = 0;
1352 static void pshader_hw_texm3x3vspec(const SHADER_OPCODE_ARG *arg)
1354 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1355 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1357 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1358 SHADER_BUFFER* buffer = arg->buffer;
1359 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1363 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1364 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1366 /* Construct the eye-ray vector from w coordinates */
1367 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1368 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1369 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1371 /* Calculate reflection vector
1373 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1374 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1375 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1376 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1377 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1378 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1379 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1381 /* Sample the texture using the calculated coordinates */
1382 sprintf(dst_str, "T%u", reg);
1383 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1384 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1385 current_state->current_row = 0;
1388 static void pshader_hw_texm3x3spec(const SHADER_OPCODE_ARG *arg)
1390 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1391 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1393 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1394 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1395 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1396 SHADER_BUFFER* buffer = arg->buffer;
1400 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1401 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1403 /* Calculate reflection vector.
1406 * TMP.xyz = 2 * --------- * N - E
1409 * Which normalizes the normal vector
1411 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1412 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1413 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1414 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1415 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1416 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1418 /* Sample the texture using the calculated coordinates */
1419 sprintf(dst_str, "T%u", reg);
1420 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1421 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1422 current_state->current_row = 0;
1425 static void pshader_hw_texdepth(const SHADER_OPCODE_ARG *arg)
1427 SHADER_BUFFER* buffer = arg->buffer;
1430 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1431 * which is essentially an input, is the destination register because it is the first
1432 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1435 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1437 /* According to the msdn, the source register(must be r5) is unusable after
1438 * the texdepth instruction, so we're free to modify it
1440 shader_addline(buffer, "MIN %s.y, %s.y, one.y;\n", dst_name, dst_name);
1442 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1443 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1444 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1446 shader_addline(buffer, "RCP %s.y, %s.y;\n", dst_name, dst_name);
1447 shader_addline(buffer, "MUL TMP.x, %s.x, %s.y;\n", dst_name, dst_name);
1448 shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1449 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1452 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1453 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1454 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1455 static void pshader_hw_texdp3tex(const SHADER_OPCODE_ARG *arg)
1457 SHADER_BUFFER* buffer = arg->buffer;
1458 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1462 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1463 shader_addline(buffer, "MOV TMP, 0.0;\n");
1464 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1466 sprintf(dst_str, "T%u", sampler_idx);
1467 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1470 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1471 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1472 static void pshader_hw_texdp3(const SHADER_OPCODE_ARG *arg)
1477 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1478 SHADER_BUFFER* buffer = arg->buffer;
1480 /* Handle output register */
1481 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1482 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1484 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1485 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1487 /* TODO: Handle output modifiers */
1490 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1491 * Perform the 3rd row of a 3x3 matrix multiply */
1492 static void pshader_hw_texm3x3(const SHADER_OPCODE_ARG *arg)
1494 SHADER_BUFFER* buffer = arg->buffer;
1498 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1500 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1501 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1503 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1504 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1505 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1507 /* TODO: Handle output modifiers */
1510 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1511 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1512 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1513 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1515 static void pshader_hw_texm3x2depth(const SHADER_OPCODE_ARG *arg)
1517 SHADER_BUFFER* buffer = arg->buffer;
1518 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1521 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1522 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1524 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1525 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1526 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1528 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1529 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1530 shader_addline(buffer, "MIN TMP.x, TMP.x, one.x;\n");
1531 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1534 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1535 Vertex/Pixel shaders to ARB_vertex_program codes */
1536 static void shader_hw_mnxn(const SHADER_OPCODE_ARG *arg)
1539 int nComponents = 0;
1540 SHADER_OPCODE_ARG tmpArg;
1541 IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
1542 const SHADER_OPCODE *opcode_table = shader->baseShader.shader_ins;
1543 DWORD shader_version = arg->reg_maps->shader_version;
1545 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1547 /* Set constants for the temporary argument */
1548 tmpArg.shader = arg->shader;
1549 tmpArg.buffer = arg->buffer;
1550 tmpArg.src[0] = arg->src[0];
1551 tmpArg.src_addr[0] = arg->src_addr[0];
1552 tmpArg.src_addr[1] = arg->src_addr[1];
1553 tmpArg.reg_maps = arg->reg_maps;
1555 switch(arg->opcode->opcode) {
1556 case WINED3DSIO_M4x4:
1558 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1560 case WINED3DSIO_M4x3:
1562 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP4);
1564 case WINED3DSIO_M3x4:
1566 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1568 case WINED3DSIO_M3x3:
1570 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1572 case WINED3DSIO_M3x2:
1574 tmpArg.opcode = shader_get_opcode(opcode_table, shader_version, WINED3DSIO_DP3);
1580 for (i = 0; i < nComponents; i++) {
1581 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1582 tmpArg.src[1] = arg->src[1]+i;
1583 shader_hw_map2gl(&tmpArg);
1587 static void vshader_hw_rsq_rcp(const SHADER_OPCODE_ARG *arg)
1589 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1590 SHADER_BUFFER* buffer = arg->buffer;
1591 DWORD dst = arg->dst;
1592 DWORD src = arg->src[0];
1593 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1594 const char *instruction;
1598 switch(curOpcode->opcode)
1600 case WINED3DSIO_RSQ: instruction = "RSQ"; break;
1601 case WINED3DSIO_RCP: instruction = "RCP"; break;
1602 default: instruction = "";
1603 FIXME("Unhandled opcode %s\n", curOpcode->name);
1607 strcpy(tmpLine, instruction);
1608 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1609 strcat(tmpLine, ",");
1610 vshader_program_add_param(arg, src, TRUE, tmpLine);
1611 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1612 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1615 strcat(tmpLine, ".w");
1618 shader_addline(buffer, "%s;\n", tmpLine);
1621 static void shader_hw_nrm(const SHADER_OPCODE_ARG *arg)
1623 SHADER_BUFFER* buffer = arg->buffer;
1627 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1628 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1630 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1631 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1633 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1634 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1635 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1636 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1637 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1641 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1644 static void shader_hw_sincos(const SHADER_OPCODE_ARG *arg)
1646 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1647 * must contain fixed constants. So we need a separate function to filter those constants and
1650 SHADER_BUFFER* buffer = arg->buffer;
1654 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1655 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1657 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1658 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1660 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1661 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1665 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1669 static GLuint create_arb_blt_vertex_program(const WineD3D_GL_Info *gl_info)
1671 GLuint program_id = 0;
1672 const char *blt_vprogram =
1674 "PARAM c[1] = { { 1, 0.5 } };\n"
1675 "MOV result.position, vertex.position;\n"
1676 "MOV result.color, c[0].x;\n"
1677 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1680 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1681 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1682 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1684 if (glGetError() == GL_INVALID_OPERATION) {
1686 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1687 FIXME("Vertex program error at position %d: %s\n", pos,
1688 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1694 static GLuint create_arb_blt_fragment_program(const WineD3D_GL_Info *gl_info, enum tex_types tex_type)
1696 GLuint program_id = 0;
1697 static const char * const blt_fprograms[tex_type_count] =
1704 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1705 "MOV result.depth.z, R0.x;\n"
1712 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1713 "MOV result.depth.z, R0.x;\n"
1718 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1719 "MOV result.depth.z, R0.x;\n"
1723 if (!blt_fprograms[tex_type])
1725 FIXME("tex_type %#x not supported\n", tex_type);
1729 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1730 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1731 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprograms[tex_type]), blt_fprograms[tex_type]));
1733 if (glGetError() == GL_INVALID_OPERATION) {
1735 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1736 FIXME("Fragment program error at position %d: %s\n", pos,
1737 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1743 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1744 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1745 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1746 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1749 TRACE("Using vertex shader\n");
1750 IWineD3DVertexShaderImpl_CompileShader(This->stateBlock->vertexShader);
1752 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->prgId;
1754 /* Bind the vertex program */
1755 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1756 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1758 /* Enable OpenGL vertex programs */
1759 glEnable(GL_VERTEX_PROGRAM_ARB);
1760 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1761 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1762 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1763 priv->current_vprogram_id = 0;
1764 glDisable(GL_VERTEX_PROGRAM_ARB);
1765 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1769 struct ps_compile_args compile_args;
1770 TRACE("Using pixel shader\n");
1771 find_ps_compile_args((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader, This->stateBlock, &compile_args);
1772 priv->current_fprogram_id = find_gl_pshader((IWineD3DPixelShaderImpl *) This->stateBlock->pixelShader,
1775 /* Bind the fragment program */
1776 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1777 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1779 if(!priv->use_arbfp_fixed_func) {
1780 /* Enable OpenGL fragment programs */
1781 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1782 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1784 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1785 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1786 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1787 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1788 * replacement shader
1790 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1791 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1792 priv->current_fprogram_id = 0;
1796 static void shader_arb_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
1797 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1798 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1799 GLuint *blt_fprogram = &priv->depth_blt_fprogram_id[tex_type];
1800 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1802 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1803 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1804 glEnable(GL_VERTEX_PROGRAM_ARB);
1806 if (!*blt_fprogram) *blt_fprogram = create_arb_blt_fragment_program(gl_info, tex_type);
1807 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, *blt_fprogram));
1808 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1811 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1812 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1813 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1814 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1816 if (priv->current_vprogram_id) {
1817 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1818 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1820 glEnable(GL_VERTEX_PROGRAM_ARB);
1821 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1823 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1825 glDisable(GL_VERTEX_PROGRAM_ARB);
1826 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1829 if (priv->current_fprogram_id) {
1830 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1831 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1833 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1834 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1836 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1838 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1839 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1843 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1844 IWineD3DBaseShaderImpl *baseShader = (IWineD3DBaseShaderImpl *) iface;
1845 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)baseShader->baseShader.device)->adapter->gl_info;
1847 if (shader_is_pshader_version(baseShader->baseShader.reg_maps.shader_version))
1849 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
1853 for(i = 0; i < This->num_gl_shaders; i++) {
1854 GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId));
1855 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1858 HeapFree(GetProcessHeap(), 0, This->gl_shaders);
1859 This->gl_shaders = NULL;
1860 This->num_gl_shaders = 0;
1862 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *) iface;
1865 GL_EXTCALL(glDeleteProgramsARB(1, &This->prgId));
1866 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->prgId))");
1867 ((IWineD3DVertexShaderImpl *) This)->prgId = 0;
1870 baseShader->baseShader.is_compiled = FALSE;
1873 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1874 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1875 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1879 static void shader_arb_free(IWineD3DDevice *iface) {
1880 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1881 const WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1882 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1885 if(priv->depth_blt_vprogram_id) {
1886 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1888 for (i = 0; i < tex_type_count; ++i) {
1889 if (priv->depth_blt_fprogram_id[i]) {
1890 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id[i]));
1894 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1897 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1901 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1902 const char *tmp2, const char *tmp3, const char *tmp4) {
1903 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1905 /* Calculate the > 0.0031308 case */
1906 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1907 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1908 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1909 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1910 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1911 /* Calculate the < case */
1912 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1913 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1914 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1915 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1916 /* Store the components > 0.0031308 in the destination */
1917 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1918 /* Add the components that are < 0.0031308 */
1919 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1920 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1923 static GLuint shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer, const struct ps_compile_args *args) {
1924 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1925 const shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1926 CONST DWORD *function = This->baseShader.function;
1927 DWORD shader_version = reg_maps->shader_version;
1928 const char *fragcolor;
1929 const WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1930 const local_constant *lconst;
1933 /* Create the hw ARB shader */
1934 shader_addline(buffer, "!!ARBfp1.0\n");
1936 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1937 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1938 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1939 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1940 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1941 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1942 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1943 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1945 /* Base Declarations */
1946 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
1948 /* We need two variables for fog blending */
1949 if(args->fog != FOG_OFF) shader_addline(buffer, "TEMP TMP_FOG;\n");
1950 if (shader_version >= WINED3DPS_VERSION(2,0)) shader_addline(buffer, "TEMP TMP_COLOR;\n");
1952 /* Base Shader Body */
1953 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
1955 if (shader_version < WINED3DPS_VERSION(2,0)) {
1958 fragcolor = "TMP_COLOR";
1960 if(args->srgb_correction) {
1961 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
1963 if (shader_version < WINED3DPS_VERSION(3,0)) {
1964 /* calculate fog and blend it
1965 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
1966 * -1/(e-s) and e/(e-s) respectively.
1970 shader_addline(buffer, "MOV result.color, %s;\n", fragcolor);
1973 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
1974 shader_addline(buffer, "LRP result.color.xyz, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
1975 shader_addline(buffer, "MOV result.color.w, %s.w;\n", fragcolor);
1978 FIXME("Implement EXP fog in ARB\n");
1979 shader_addline(buffer, "MOV result.color, %s;\n", fragcolor);
1982 FIXME("Implement EXP2 fog in ARB\n");
1983 shader_addline(buffer, "MOV result.color, %s;\n", fragcolor);
1988 shader_addline(buffer, "END\n");
1990 /* TODO: change to resource.glObjectHandle or something like that */
1991 GL_EXTCALL(glGenProgramsARB(1, &retval));
1993 TRACE("Creating a hw pixel shader, prg=%d\n", retval);
1994 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, retval));
1996 TRACE("Created hw pixel shader, prg=%d\n", retval);
1997 /* Create the program and check for errors */
1998 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
1999 buffer->bsize, buffer->buffer));
2001 if (glGetError() == GL_INVALID_OPERATION) {
2003 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2004 FIXME("HW PixelShader Error at position %d: %s\n",
2005 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2009 /* Load immediate constants */
2010 if(!This->baseShader.load_local_constsF) {
2011 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2012 const float *value = (const float *)lconst->value;
2013 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2014 checkGLcall("glProgramLocalParameter4fvARB");
2021 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
2022 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2023 const shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
2024 CONST DWORD *function = This->baseShader.function;
2025 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2026 const WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2027 const local_constant *lconst;
2029 /* Create the hw ARB shader */
2030 shader_addline(buffer, "!!ARBvp1.0\n");
2031 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2033 /* Mesa supports only 95 constants */
2034 if (GL_VEND(MESA) || GL_VEND(WINE))
2035 This->baseShader.limits.constant_float =
2036 min(95, This->baseShader.limits.constant_float);
2038 shader_addline(buffer, "TEMP TMP;\n");
2040 /* Base Declarations */
2041 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2043 /* We need a constant to fixup the final position */
2044 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2046 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2047 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2048 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2049 * a replacement shader depend on the texcoord.w being set properly.
2051 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2052 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2053 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2054 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2055 * this can eat a number of instructions, so skip it unless this cap is set as well
2057 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2058 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2060 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2062 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2063 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2064 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2065 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2071 /* Base Shader Body */
2072 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2074 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2076 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2078 /* Write the final position.
2080 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2081 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2082 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2083 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2085 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2086 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2087 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2089 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2090 * and the glsl equivalent
2092 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2094 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2096 shader_addline(buffer, "END\n");
2098 /* TODO: change to resource.glObjectHandle or something like that */
2099 GL_EXTCALL(glGenProgramsARB(1, &This->prgId));
2101 TRACE("Creating a hw vertex shader, prg=%d\n", This->prgId);
2102 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->prgId));
2104 TRACE("Created hw vertex shader, prg=%d\n", This->prgId);
2105 /* Create the program and check for errors */
2106 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2107 buffer->bsize, buffer->buffer));
2109 if (glGetError() == GL_INVALID_OPERATION) {
2111 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2112 FIXME("HW VertexShader Error at position %d: %s\n",
2113 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2117 /* Load immediate constants */
2118 if(!This->baseShader.load_local_constsF) {
2119 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2120 const float *value = (const float *)lconst->value;
2121 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2126 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
2128 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2129 * then overwrite the shader specific ones
2131 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2133 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2134 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2135 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2136 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2139 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2140 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2141 pCaps->PixelShader1xMaxValue = 8.0;
2142 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2146 static BOOL shader_arb_color_fixup_supported(struct color_fixup_desc fixup)
2148 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
2150 TRACE("Checking support for color_fixup:\n");
2151 dump_color_fixup_desc(fixup);
2154 /* We support everything except YUV conversions. */
2155 if (!is_yuv_fixup(fixup))
2161 TRACE("[FAILED]\n");
2165 static const SHADER_HANDLER shader_arb_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
2167 /* WINED3DSIH_ABS */ shader_hw_map2gl,
2168 /* WINED3DSIH_ADD */ shader_hw_map2gl,
2169 /* WINED3DSIH_BEM */ pshader_hw_bem,
2170 /* WINED3DSIH_BREAK */ NULL,
2171 /* WINED3DSIH_BREAKC */ NULL,
2172 /* WINED3DSIH_BREAKP */ NULL,
2173 /* WINED3DSIH_CALL */ NULL,
2174 /* WINED3DSIH_CALLNZ */ NULL,
2175 /* WINED3DSIH_CMP */ pshader_hw_cmp,
2176 /* WINED3DSIH_CND */ pshader_hw_cnd,
2177 /* WINED3DSIH_CRS */ shader_hw_map2gl,
2178 /* WINED3DSIH_DCL */ NULL,
2179 /* WINED3DSIH_DEF */ NULL,
2180 /* WINED3DSIH_DEFB */ NULL,
2181 /* WINED3DSIH_DEFI */ NULL,
2182 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add,
2183 /* WINED3DSIH_DP3 */ shader_hw_map2gl,
2184 /* WINED3DSIH_DP4 */ shader_hw_map2gl,
2185 /* WINED3DSIH_DST */ shader_hw_map2gl,
2186 /* WINED3DSIH_DSX */ NULL,
2187 /* WINED3DSIH_DSY */ NULL,
2188 /* WINED3DSIH_ELSE */ NULL,
2189 /* WINED3DSIH_ENDIF */ NULL,
2190 /* WINED3DSIH_ENDLOOP */ NULL,
2191 /* WINED3DSIH_ENDREP */ NULL,
2192 /* WINED3DSIH_EXP */ shader_hw_map2gl,
2193 /* WINED3DSIH_EXPP */ shader_hw_map2gl,
2194 /* WINED3DSIH_FRC */ shader_hw_map2gl,
2195 /* WINED3DSIH_IF */ NULL,
2196 /* WINED3DSIH_IFC */ NULL,
2197 /* WINED3DSIH_LABEL */ NULL,
2198 /* WINED3DSIH_LIT */ shader_hw_map2gl,
2199 /* WINED3DSIH_LOG */ shader_hw_map2gl,
2200 /* WINED3DSIH_LOGP */ shader_hw_map2gl,
2201 /* WINED3DSIH_LOOP */ NULL,
2202 /* WINED3DSIH_LRP */ shader_hw_map2gl,
2203 /* WINED3DSIH_M3x2 */ shader_hw_mnxn,
2204 /* WINED3DSIH_M3x3 */ shader_hw_mnxn,
2205 /* WINED3DSIH_M3x4 */ shader_hw_mnxn,
2206 /* WINED3DSIH_M4x3 */ shader_hw_mnxn,
2207 /* WINED3DSIH_M4x4 */ shader_hw_mnxn,
2208 /* WINED3DSIH_MAD */ shader_hw_map2gl,
2209 /* WINED3DSIH_MAX */ shader_hw_map2gl,
2210 /* WINED3DSIH_MIN */ shader_hw_map2gl,
2211 /* WINED3DSIH_MOV */ shader_hw_mov,
2212 /* WINED3DSIH_MOVA */ shader_hw_mov,
2213 /* WINED3DSIH_MUL */ shader_hw_map2gl,
2214 /* WINED3DSIH_NOP */ shader_hw_map2gl,
2215 /* WINED3DSIH_NRM */ shader_hw_nrm,
2216 /* WINED3DSIH_PHASE */ NULL,
2217 /* WINED3DSIH_POW */ shader_hw_map2gl,
2218 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp,
2219 /* WINED3DSIH_REP */ NULL,
2220 /* WINED3DSIH_RET */ NULL,
2221 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp,
2222 /* WINED3DSIH_SETP */ NULL,
2223 /* WINED3DSIH_SGE */ shader_hw_map2gl,
2224 /* WINED3DSIH_SGN */ NULL,
2225 /* WINED3DSIH_SINCOS */ shader_hw_sincos,
2226 /* WINED3DSIH_SLT */ shader_hw_map2gl,
2227 /* WINED3DSIH_SUB */ shader_hw_map2gl,
2228 /* WINED3DSIH_TEX */ pshader_hw_tex,
2229 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem,
2230 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem,
2231 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord,
2232 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth,
2233 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3,
2234 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex,
2235 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill,
2236 /* WINED3DSIH_TEXLDD */ NULL,
2237 /* WINED3DSIH_TEXLDL */ NULL,
2238 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth,
2239 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad,
2240 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex,
2241 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3,
2242 /* WINED3DSIH_TEXM3x3DIFF */ NULL,
2243 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad,
2244 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec,
2245 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex,
2246 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec,
2247 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar,
2248 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb,
2249 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb,
2252 const shader_backend_t arb_program_shader_backend = {
2253 shader_arb_instruction_handler_table,
2255 shader_arb_select_depth_blt,
2256 shader_arb_deselect_depth_blt,
2257 shader_arb_update_float_vertex_constants,
2258 shader_arb_update_float_pixel_constants,
2259 shader_arb_load_constants,
2260 shader_arb_color_correction,
2264 shader_arb_dirty_const,
2265 shader_arb_generate_pshader,
2266 shader_arb_generate_vshader,
2267 shader_arb_get_caps,
2268 shader_arb_color_fixup_supported,
2271 /* ARB_fragment_program fixed function pipeline replacement definitions */
2272 #define ARB_FFP_CONST_TFACTOR 0
2273 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2274 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2275 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2276 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2278 struct arbfp_ffp_desc
2280 struct ffp_frag_desc parent;
2282 unsigned int num_textures_used;
2285 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2287 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2288 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2290 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2291 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2295 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2296 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2297 struct shader_arb_priv *priv;
2298 /* Share private data between the shader backend and the pipeline replacement, if both
2299 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2300 * if no pixel shader is bound or not
2302 if(This->shader_backend == &arb_program_shader_backend) {
2303 This->fragment_priv = This->shader_priv;
2305 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2306 if(!This->fragment_priv) return E_OUTOFMEMORY;
2308 priv = (struct shader_arb_priv *) This->fragment_priv;
2309 priv->fragment_shaders = hash_table_create(ffp_frag_program_key_hash, ffp_frag_program_key_compare);
2310 priv->use_arbfp_fixed_func = TRUE;
2314 static void arbfp_free_ffpshader(void *value, void *gli) {
2315 const WineD3D_GL_Info *gl_info = gli;
2316 struct arbfp_ffp_desc *entry_arb = value;
2319 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2320 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2321 HeapFree(GetProcessHeap(), 0, entry_arb);
2325 static void arbfp_free(IWineD3DDevice *iface) {
2326 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2327 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2329 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2330 priv->use_arbfp_fixed_func = FALSE;
2332 if(This->shader_backend != &arb_program_shader_backend) {
2333 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2337 static void arbfp_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct fragment_caps *caps)
2339 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2340 WINED3DTEXOPCAPS_SELECTARG1 |
2341 WINED3DTEXOPCAPS_SELECTARG2 |
2342 WINED3DTEXOPCAPS_MODULATE4X |
2343 WINED3DTEXOPCAPS_MODULATE2X |
2344 WINED3DTEXOPCAPS_MODULATE |
2345 WINED3DTEXOPCAPS_ADDSIGNED2X |
2346 WINED3DTEXOPCAPS_ADDSIGNED |
2347 WINED3DTEXOPCAPS_ADD |
2348 WINED3DTEXOPCAPS_SUBTRACT |
2349 WINED3DTEXOPCAPS_ADDSMOOTH |
2350 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2351 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2352 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2353 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2354 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2355 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2356 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2357 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2358 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2359 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2360 WINED3DTEXOPCAPS_MULTIPLYADD |
2361 WINED3DTEXOPCAPS_LERP |
2362 WINED3DTEXOPCAPS_BUMPENVMAP |
2363 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2365 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2367 caps->MaxTextureBlendStages = 8;
2368 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2370 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2372 #undef GLINFO_LOCATION
2374 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2375 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2377 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2379 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2380 * application provided constants
2382 if(device->shader_backend == &arb_program_shader_backend) {
2383 if (use_ps(stateblock)) return;
2385 device = stateblock->wineD3DDevice;
2386 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2387 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2390 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2391 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2392 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2396 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2398 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2400 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2401 * application provided constants
2403 if(device->shader_backend == &arb_program_shader_backend) {
2404 if (use_ps(stateblock)) return;
2406 device = stateblock->wineD3DDevice;
2407 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2408 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2411 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2412 /* The specular color has no alpha */
2413 col[0] = 1.0; col[1] = 1.0;
2414 col[2] = 1.0; col[3] = 0.0;
2416 col[0] = 0.0; col[1] = 0.0;
2417 col[2] = 0.0; col[3] = 0.0;
2419 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2420 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2423 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2424 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2425 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2428 if (use_ps(stateblock))
2431 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2432 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2435 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2436 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2440 if(device->shader_backend == &arb_program_shader_backend) {
2441 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2444 } else if(device->shader_backend == &arb_program_shader_backend) {
2445 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2446 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2449 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2450 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2451 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2452 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2454 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2455 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2458 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2459 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2460 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2463 if (use_ps(stateblock))
2466 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2467 /* The pixel shader has to know the luminance offset. Do a constants update if it
2468 * isn't scheduled anyway
2470 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2471 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2475 if(device->shader_backend == &arb_program_shader_backend) {
2476 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2479 } else if(device->shader_backend == &arb_program_shader_backend) {
2480 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2481 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2484 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2485 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2489 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2490 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2493 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2496 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2498 switch(arg & WINED3DTA_SELECTMASK) {
2499 case WINED3DTA_DIFFUSE:
2500 ret = "fragment.color.primary"; break;
2502 case WINED3DTA_CURRENT:
2503 if(stage == 0) ret = "fragment.color.primary";
2507 case WINED3DTA_TEXTURE:
2509 case 0: ret = "tex0"; break;
2510 case 1: ret = "tex1"; break;
2511 case 2: ret = "tex2"; break;
2512 case 3: ret = "tex3"; break;
2513 case 4: ret = "tex4"; break;
2514 case 5: ret = "tex5"; break;
2515 case 6: ret = "tex6"; break;
2516 case 7: ret = "tex7"; break;
2517 default: ret = "unknown texture";
2521 case WINED3DTA_TFACTOR:
2522 ret = "tfactor"; break;
2524 case WINED3DTA_SPECULAR:
2525 ret = "fragment.color.secondary"; break;
2527 case WINED3DTA_TEMP:
2528 ret = "tempreg"; break;
2530 case WINED3DTA_CONSTANT:
2531 FIXME("Implement perstage constants\n");
2533 case 0: ret = "const0"; break;
2534 case 1: ret = "const1"; break;
2535 case 2: ret = "const2"; break;
2536 case 3: ret = "const3"; break;
2537 case 4: ret = "const4"; break;
2538 case 5: ret = "const5"; break;
2539 case 6: ret = "const6"; break;
2540 case 7: ret = "const7"; break;
2541 default: ret = "unknown constant";
2549 if(arg & WINED3DTA_COMPLEMENT) {
2550 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2551 if(argnum == 0) ret = "arg0";
2552 if(argnum == 1) ret = "arg1";
2553 if(argnum == 2) ret = "arg2";
2555 if(arg & WINED3DTA_ALPHAREPLICATE) {
2556 shader_addline(buffer, "MOV arg%u, %s.w;\n", argnum, ret);
2557 if(argnum == 0) ret = "arg0";
2558 if(argnum == 1) ret = "arg1";
2559 if(argnum == 2) ret = "arg2";
2564 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2565 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2566 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2567 unsigned int mul = 1;
2568 BOOL mul_final_dest = FALSE;
2570 if(color && alpha) dstmask = "";
2571 else if(color) dstmask = ".xyz";
2572 else dstmask = ".w";
2574 if(dst == tempreg) dstreg = "tempreg";
2575 else dstreg = "ret";
2577 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2578 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2579 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2582 case WINED3DTOP_DISABLE:
2583 if(stage == 0) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2586 case WINED3DTOP_SELECTARG2:
2588 case WINED3DTOP_SELECTARG1:
2589 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2592 case WINED3DTOP_MODULATE4X:
2594 case WINED3DTOP_MODULATE2X:
2596 if(strcmp(dstreg, "result.color") == 0) {
2598 mul_final_dest = TRUE;
2600 case WINED3DTOP_MODULATE:
2601 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2604 case WINED3DTOP_ADDSIGNED2X:
2606 if(strcmp(dstreg, "result.color") == 0) {
2608 mul_final_dest = TRUE;
2610 case WINED3DTOP_ADDSIGNED:
2611 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2613 case WINED3DTOP_ADD:
2614 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2617 case WINED3DTOP_SUBTRACT:
2618 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2621 case WINED3DTOP_ADDSMOOTH:
2622 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2623 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2626 case WINED3DTOP_BLENDCURRENTALPHA:
2627 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2628 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2630 case WINED3DTOP_BLENDFACTORALPHA:
2631 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2632 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2634 case WINED3DTOP_BLENDTEXTUREALPHA:
2635 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2636 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2638 case WINED3DTOP_BLENDDIFFUSEALPHA:
2639 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2640 shader_addline(buffer, "LRP %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2643 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2644 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2645 shader_addline(buffer, "SUB arg0.w, const.x, %s.w;\n", arg0);
2646 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg, dstmask, arg2, arg1);
2649 /* D3DTOP_PREMODULATE ???? */
2651 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2652 shader_addline(buffer, "SUB arg0.w, const.x, %s;\n", arg1);
2653 shader_addline(buffer, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2655 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2656 shader_addline(buffer, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2658 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2659 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2660 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg, dstmask, arg2, arg1);
2662 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2663 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg, dstmask, arg1, arg2, arg1);
2666 case WINED3DTOP_DOTPRODUCT3:
2668 if(strcmp(dstreg, "result.color") == 0) {
2670 mul_final_dest = TRUE;
2672 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2673 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2674 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2677 case WINED3DTOP_MULTIPLYADD:
2678 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2681 case WINED3DTOP_LERP:
2682 /* The msdn is not quite right here */
2683 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2686 case WINED3DTOP_BUMPENVMAP:
2687 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2688 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2692 FIXME("Unhandled texture op %08x\n", op);
2696 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2697 } else if(mul == 4) {
2698 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2702 /* The stateblock is passed for GLINFO_LOCATION */
2703 static GLuint gen_arbfp_ffp_shader(const struct ffp_frag_settings *settings, IWineD3DStateBlockImpl *stateblock)
2706 SHADER_BUFFER buffer;
2707 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2708 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2709 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2710 const char *textype;
2711 const char *instr, *sat;
2712 char colorcor_dst[8];
2714 DWORD arg0, arg1, arg2;
2715 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2717 const char *final_combiner_src = "ret";
2719 /* Find out which textures are read */
2720 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2721 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2722 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2723 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2724 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2725 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2726 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2727 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2729 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2730 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2731 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2732 bump_used[stage] = TRUE;
2733 tex_read[stage] = TRUE;
2735 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2736 bump_used[stage] = TRUE;
2737 tex_read[stage] = TRUE;
2738 luminance_used[stage] = TRUE;
2739 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2740 tfactor_used = TRUE;
2743 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2744 tfactor_used = TRUE;
2747 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2748 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2749 tempreg_used = TRUE;
2752 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2753 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2754 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2755 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2756 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2757 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2758 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2760 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2761 tempreg_used = TRUE;
2763 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2764 tfactor_used = TRUE;
2769 shader_buffer_init(&buffer);
2771 shader_addline(&buffer, "!!ARBfp1.0\n");
2773 switch(settings->fog) {
2774 case FOG_OFF: break;
2775 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2776 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2777 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2778 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2781 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2782 shader_addline(&buffer, "TEMP TMP;\n");
2783 shader_addline(&buffer, "TEMP ret;\n");
2784 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2785 shader_addline(&buffer, "TEMP arg0;\n");
2786 shader_addline(&buffer, "TEMP arg1;\n");
2787 shader_addline(&buffer, "TEMP arg2;\n");
2788 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2789 if(!tex_read[stage]) continue;
2790 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2791 if(!bump_used[stage]) continue;
2792 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2793 if(!luminance_used[stage]) continue;
2794 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2797 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2799 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2801 if(settings->sRGB_write) {
2802 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2803 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2804 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2805 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2806 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2807 srgb_pow, srgb_pow, srgb_pow);
2808 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2809 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2810 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2811 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2814 /* Generate texture sampling instructions) */
2815 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2816 if(!tex_read[stage]) continue;
2818 switch(settings->op[stage].tex_type) {
2819 case tex_1d: textype = "1D"; break;
2820 case tex_2d: textype = "2D"; break;
2821 case tex_3d: textype = "3D"; break;
2822 case tex_cube: textype = "CUBE"; break;
2823 case tex_rect: textype = "RECT"; break;
2824 default: textype = "unexpected_textype"; break;
2827 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2828 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2834 if(settings->op[stage].projected == proj_none) {
2836 } else if(settings->op[stage].projected == proj_count4 ||
2837 settings->op[stage].projected == proj_count3) {
2840 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2845 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2846 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2847 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2848 shader_addline(&buffer, "DP3 ret.x, arg1, tex%u;\n", stage - 1);
2849 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2850 shader_addline(&buffer, "DP3 ret.y, arg1, tex%u;\n", stage - 1);
2852 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2853 * so multiply the displacement with the dividing parameter before passing it to TXP
2855 if (settings->op[stage].projected != proj_none) {
2856 if(settings->op[stage].projected == proj_count4) {
2857 shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].w;\n", stage);
2858 shader_addline(&buffer, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage, stage);
2860 shader_addline(&buffer, "MOV ret.w, fragment.texcoord[%u].z;\n", stage);
2861 shader_addline(&buffer, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage, stage);
2864 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2867 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2868 instr, sat, stage, stage, textype);
2869 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2870 shader_addline(&buffer, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
2871 stage - 1, stage - 1, stage - 1);
2872 shader_addline(&buffer, "MUL tex%u, tex%u, ret.x;\n", stage, stage);
2874 } else if(settings->op[stage].projected == proj_count3) {
2875 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2876 shader_addline(&buffer, "MOV ret.w, ret.z;\n");
2877 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2878 instr, sat, stage, stage, textype);
2880 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2881 instr, sat, stage, stage, stage, textype);
2884 sprintf(colorcor_dst, "tex%u", stage);
2885 gen_color_correction(&buffer, colorcor_dst, WINED3DSP_WRITEMASK_ALL, "const.x", "const.y",
2886 settings->op[stage].color_fixup);
2889 /* Generate the main shader */
2890 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2891 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2893 final_combiner_src = "fragment.color.primary";
2898 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2899 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2900 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2901 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2902 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2903 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2904 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2905 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2906 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2907 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2908 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2909 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2911 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
2912 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2913 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2914 settings->op[stage].carg2 == settings->op[stage].aarg2;
2917 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2918 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2919 settings->op[stage].cop, settings->op[stage].carg0,
2920 settings->op[stage].carg1, settings->op[stage].carg2);
2922 shader_addline(&buffer, "MOV ret.w, fragment.color.primary.w;\n");
2924 } else if(op_equal) {
2925 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2926 settings->op[stage].cop, settings->op[stage].carg0,
2927 settings->op[stage].carg1, settings->op[stage].carg2);
2929 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2930 settings->op[stage].cop, settings->op[stage].carg0,
2931 settings->op[stage].carg1, settings->op[stage].carg2);
2932 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
2933 settings->op[stage].aop, settings->op[stage].aarg0,
2934 settings->op[stage].aarg1, settings->op[stage].aarg2);
2938 if(settings->sRGB_write) {
2939 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2940 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2941 shader_addline(&buffer, "MOV result.color.w, ret.w;\n");
2943 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2947 shader_addline(&buffer, "END\n");
2949 /* Generate the shader */
2950 GL_EXTCALL(glGenProgramsARB(1, &ret));
2951 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2952 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2954 if (glGetError() == GL_INVALID_OPERATION) {
2956 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2957 FIXME("Fragment program error at position %d: %s\n", pos,
2958 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2960 shader_buffer_free(&buffer);
2964 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2965 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2966 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
2967 BOOL use_pshader = use_ps(stateblock);
2968 BOOL use_vshader = use_vs(stateblock);
2969 struct ffp_frag_settings settings;
2970 const struct arbfp_ffp_desc *desc;
2973 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2974 if(!use_pshader && device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2975 /* Reload fixed function constants since they collide with the pixel shader constants */
2976 for(i = 0; i < MAX_TEXTURES; i++) {
2977 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2979 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2980 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2981 } else if(use_pshader && !isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2982 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2988 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2989 gen_ffp_frag_op(stateblock, &settings, FALSE);
2990 desc = (const struct arbfp_ffp_desc *)find_ffp_frag_shader(priv->fragment_shaders, &settings);
2992 struct arbfp_ffp_desc *new_desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc));
2995 ERR("Out of memory\n");
2998 new_desc->num_textures_used = 0;
2999 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
3000 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
3001 new_desc->num_textures_used = i;
3004 memcpy(&new_desc->parent.settings, &settings, sizeof(settings));
3005 new_desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
3006 add_ffp_frag_shader(priv->fragment_shaders, &new_desc->parent);
3007 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc);
3011 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3012 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3015 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
3016 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3017 priv->current_fprogram_id = desc->shader;
3019 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
3020 /* Reload fixed function constants since they collide with the pixel shader constants */
3021 for(i = 0; i < MAX_TEXTURES; i++) {
3022 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
3024 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
3025 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
3029 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3030 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3031 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3032 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3033 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3035 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3036 * to be compiled before activating them(needs some cleanups in the shader backend interface)
3038 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
3039 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
3041 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
3042 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
3046 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
3050 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3051 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3052 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3053 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3054 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3056 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3057 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3058 fragment_prog_arbfp(state, stateblock, context);
3062 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3063 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3064 fragment_prog_arbfp(state, stateblock, context);
3068 #undef GLINFO_LOCATION
3070 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3071 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3072 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3073 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3074 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3075 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3076 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3077 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3078 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3079 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3080 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3081 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3082 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3083 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3084 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3085 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3086 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3087 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3088 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3089 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3090 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3091 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3092 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3093 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3094 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3095 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3096 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3097 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3098 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3099 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3100 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3101 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3102 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3103 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3104 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3105 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3106 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3107 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3108 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3109 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3110 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3111 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3112 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3113 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3114 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3115 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3116 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3117 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3118 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3119 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3120 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3121 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3122 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3123 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3124 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3125 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3126 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3127 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3128 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3129 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3130 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3131 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3132 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3133 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3134 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3135 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3136 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3137 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3138 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3139 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3140 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3141 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3142 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3143 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3144 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3145 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3146 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3147 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3148 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3149 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3150 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3151 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3152 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3153 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3154 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3155 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3156 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3157 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3158 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3159 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3160 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3161 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3162 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3163 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3164 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3165 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3166 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3167 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3168 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3169 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3170 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3171 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3172 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3173 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3174 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3175 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3176 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3177 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3178 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3179 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3180 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3181 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3182 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3183 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3184 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3185 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3186 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3187 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3188 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3189 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3190 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3191 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3192 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3193 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3194 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3195 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3196 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3197 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3198 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3199 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3200 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3201 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3202 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3203 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3204 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3205 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3206 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3207 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3208 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3209 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3210 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3211 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3212 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3213 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3214 {0 /* Terminate */, { 0, 0 }, 0 },
3217 const struct fragment_pipeline arbfp_fragment_pipeline = {
3222 shader_arb_color_fixup_supported,
3223 arbfp_fragmentstate_template,
3224 TRUE /* We can disable projected textures */
3227 #define GLINFO_LOCATION device->adapter->gl_info
3229 struct arbfp_blit_priv {
3230 GLenum yuy2_rect_shader, yuy2_2d_shader;
3231 GLenum uyvy_rect_shader, uyvy_2d_shader;
3232 GLenum yv12_rect_shader, yv12_2d_shader;
3235 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3236 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3237 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3238 if(!device->blit_priv) {
3239 ERR("Out of memory\n");
3240 return E_OUTOFMEMORY;
3244 static void arbfp_blit_free(IWineD3DDevice *iface) {
3245 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3246 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3249 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3250 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3251 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3252 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3253 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3254 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3255 checkGLcall("Delete yuv programs\n");
3259 static BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, enum yuv_fixup yuv_fixup, GLenum textype, char *luminance)
3262 const char *tex, *texinstr;
3264 if (yuv_fixup == YUV_FIXUP_UYVY) {
3272 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3273 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3275 /* This is more tricky than just replacing the texture type - we have to navigate
3276 * properly in the texture to find the correct chroma values
3278 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3282 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3283 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3284 * filtering when we sample the texture.
3286 * These are the rules for reading the chroma:
3292 * So we have to get the sampling x position in non-normalized coordinates in integers
3294 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3295 shader_addline(buffer, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
3296 shader_addline(buffer, "MOV texcrd.w, size.x;\n");
3298 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3300 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3301 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3304 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3305 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3307 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3308 * even and odd pixels respectively
3310 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3311 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3313 /* Sample Pixel 1 */
3314 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3316 /* Put the value into either of the chroma values */
3317 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3318 shader_addline(buffer, "MUL chroma.x, luminance.%c, temp.x;\n", chroma);
3319 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3320 shader_addline(buffer, "MUL chroma.y, luminance.%c, temp.x;\n", chroma);
3322 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3323 * the pixel right to the current one. Otherwise, sample the left pixel.
3324 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3326 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3327 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3328 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3330 /* Put the value into the other chroma */
3331 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3332 shader_addline(buffer, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma);
3333 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3334 shader_addline(buffer, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma);
3336 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3337 * the current one and lerp the two U and V values
3340 /* This gives the correctly filtered luminance value */
3341 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3346 static BOOL gen_yv12_read(SHADER_BUFFER *buffer, GLenum textype, char *luminance)
3351 case GL_TEXTURE_2D: tex = "2D"; break;
3352 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3354 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3358 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3359 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3360 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3361 * pitch of the luminance plane, the packing into the gl texture is a bit
3362 * unfortunate. If the whole texture is interpreted as luminance data it looks
3363 * approximately like this:
3365 * +----------------------------------+----
3377 * +----------------+-----------------+----
3379 * | U even rows | U odd rows |
3381 * +----------------+------------------ -
3383 * | V even rows | V odd rows |
3385 * +----------------+-----------------+----
3389 * So it appears as if there are 4 chroma images, but in fact the odd rows
3390 * in the chroma images are in the same row as the even ones. So its is
3391 * kinda tricky to read
3393 * When reading from rectangle textures, keep in mind that the input y coordinates
3394 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3396 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3397 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3399 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3400 /* the chroma planes have only half the width */
3401 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3403 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3404 * the coordinate. Also read the right side of the image when reading odd lines
3406 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3409 if(textype == GL_TEXTURE_2D) {
3411 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3413 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3415 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3416 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3418 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3419 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3420 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3421 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3422 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3424 /* clamp, keep the half pixel origin in mind */
3425 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3426 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3427 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3428 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3430 /* Read from [size - size+size/4] */
3431 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3432 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3434 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3435 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3436 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3437 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3438 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3439 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3441 /* Make sure to read exactly from the pixel center */
3442 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3443 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3446 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3447 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3448 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3449 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3450 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3452 /* Read the texture, put the result into the output register */
3453 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3454 shader_addline(buffer, "MOV chroma.x, temp.w;\n");
3456 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3457 * No need to clamp because we're just reusing the already clamped value from above
3459 if(textype == GL_TEXTURE_2D) {
3460 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3462 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3464 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3465 shader_addline(buffer, "MOV chroma.y, temp.w;\n");
3467 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3468 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3469 * values due to filtering
3471 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3472 if(textype == GL_TEXTURE_2D) {
3473 /* Multiply the y coordinate by 2/3 and clamp it */
3474 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3475 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3476 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3477 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3479 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
3480 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3483 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3484 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3485 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3492 static GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, enum yuv_fixup yuv_fixup, GLenum textype)
3495 SHADER_BUFFER buffer;
3496 char luminance_component;
3497 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3500 shader_buffer_init(&buffer);
3503 GL_EXTCALL(glGenProgramsARB(1, &shader));
3504 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3505 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3506 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3509 shader_buffer_free(&buffer);
3513 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3514 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3515 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3516 * each single pixel it contains, and one U and one V value shared between both
3519 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3520 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3521 * take the format into account when generating the read swizzles
3523 * Reading the Y value is straightforward - just sample the texture. The hardware
3524 * takes care of filtering in the horizontal and vertical direction.
3526 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3527 * because that would mix the U and V values of one pixel or two adjacent pixels.
3528 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3529 * regardless of the filtering setting. Vertical filtering works automatically
3530 * though - the U and V values of two rows are mixed nicely.
3532 * Appart of avoiding filtering issues, the code has to know which value it just
3533 * read, and where it can find the other one. To determine this, it checks if
3534 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3536 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3537 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3539 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3540 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3541 * in an unfiltered situation. Finding the luminance on the other hand requires
3542 * finding out if it is an odd or even pixel. The real drawback of this approach
3543 * is filtering. This would have to be emulated completely in the shader, reading
3544 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3545 * vertically. Beyond that it would require adjustments to the texture handling
3546 * code to deal with the width scaling
3548 shader_addline(&buffer, "!!ARBfp1.0\n");
3549 shader_addline(&buffer, "TEMP luminance;\n");
3550 shader_addline(&buffer, "TEMP temp;\n");
3551 shader_addline(&buffer, "TEMP chroma;\n");
3552 shader_addline(&buffer, "TEMP texcrd;\n");
3553 shader_addline(&buffer, "TEMP texcrd2;\n");
3554 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3555 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3556 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3560 case YUV_FIXUP_UYVY:
3561 case YUV_FIXUP_YUY2:
3562 if (!gen_planar_yuv_read(&buffer, yuv_fixup, textype, &luminance_component))
3564 shader_buffer_free(&buffer);
3569 case YUV_FIXUP_YV12:
3570 if (!gen_yv12_read(&buffer, textype, &luminance_component))
3572 shader_buffer_free(&buffer);
3578 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3579 shader_buffer_free(&buffer);
3583 /* Calculate the final result. Formula is taken from
3584 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3585 * ranges from -0.5 to 0.5
3587 shader_addline(&buffer, "SUB chroma.xy, chroma, coef.y;\n");
3589 shader_addline(&buffer, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component);
3590 shader_addline(&buffer, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component);
3591 shader_addline(&buffer, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
3592 shader_addline(&buffer, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component);
3593 shader_addline(&buffer, "END\n");
3596 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3598 if (glGetError() == GL_INVALID_OPERATION) {
3600 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3601 FIXME("Fragment program error at position %d: %s\n", pos,
3602 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3604 shader_buffer_free(&buffer);
3609 case YUV_FIXUP_YUY2:
3610 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yuy2_rect_shader = shader;
3611 else priv->yuy2_2d_shader = shader;
3614 case YUV_FIXUP_UYVY:
3615 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->uyvy_rect_shader = shader;
3616 else priv->uyvy_2d_shader = shader;
3619 case YUV_FIXUP_YV12:
3620 if (textype == GL_TEXTURE_RECTANGLE_ARB) priv->yv12_rect_shader = shader;
3621 else priv->yv12_2d_shader = shader;
3628 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3630 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3631 float size[4] = {width, height, 1, 1};
3632 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3633 const struct GlPixelFormatDesc *glDesc;
3634 enum yuv_fixup yuv_fixup;
3636 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3638 if (!is_yuv_fixup(glDesc->color_fixup))
3641 dump_color_fixup_desc(glDesc->color_fixup);
3642 /* Don't bother setting up a shader for unconverted formats */
3645 checkGLcall("glEnable(textype)");
3650 yuv_fixup = get_yuv_fixup(glDesc->color_fixup);
3654 case YUV_FIXUP_YUY2:
3655 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yuy2_rect_shader : priv->yuy2_2d_shader;
3658 case YUV_FIXUP_UYVY:
3659 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->uyvy_rect_shader : priv->uyvy_2d_shader;
3662 case YUV_FIXUP_YV12:
3663 shader = textype == GL_TEXTURE_RECTANGLE_ARB ? priv->yv12_rect_shader : priv->yv12_2d_shader;
3667 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup);
3670 checkGLcall("glEnable(textype)");
3675 if (!shader) shader = gen_yuv_shader(device, yuv_fixup, textype);
3678 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3679 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3680 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3681 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3682 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3683 checkGLcall("glProgramLocalParameter4fvARB");
3689 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3690 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3693 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3694 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3695 glDisable(GL_TEXTURE_2D);
3696 checkGLcall("glDisable(GL_TEXTURE_2D)");
3697 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3698 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3699 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3701 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3702 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3703 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3708 static BOOL arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup)
3710 enum yuv_fixup yuv_fixup;
3712 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
3714 TRACE("Checking support for fixup:\n");
3715 dump_color_fixup_desc(fixup);
3718 if (is_identity_fixup(fixup))
3724 /* We only support YUV conversions. */
3725 if (!is_yuv_fixup(fixup))
3727 TRACE("[FAILED]\n");
3731 yuv_fixup = get_yuv_fixup(fixup);
3734 case YUV_FIXUP_YUY2:
3735 case YUV_FIXUP_UYVY:
3736 case YUV_FIXUP_YV12:
3741 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup);
3742 TRACE("[FAILED]\n");
3747 const struct blit_shader arbfp_blit = {
3752 arbfp_blit_color_fixup_supported,
3755 #undef GLINFO_LOCATION