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
13 * This library is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
18 * This library is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with this library; if not, write to the Free Software
25 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
33 #include "wined3d_private.h"
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 #define GLINFO_LOCATION (*gl_info)
40 /********************************************************
41 * ARB_[vertex/fragment]_program helper functions follow
42 ********************************************************/
45 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
46 * When constant_list == NULL, it will load all the constants.
48 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
49 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
51 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
52 unsigned int max_constants, float* constants, char *dirty_consts) {
53 local_constant* lconst;
57 if (TRACE_ON(d3d_shader)) {
58 for(i = 0; i < max_constants; i++) {
59 if(!dirty_consts[i]) continue;
60 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
61 constants[i * 4 + 0], constants[i * 4 + 1],
62 constants[i * 4 + 2], constants[i * 4 + 3]);
65 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
66 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
67 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
69 for(i = 0; i < max_constants; i++) {
70 if(!dirty_consts[i]) continue;
74 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
75 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
76 else lcl_const[0] = constants[j + 0];
78 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
79 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
80 else lcl_const[1] = constants[j + 1];
82 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
83 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
84 else lcl_const[2] = constants[j + 2];
86 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
87 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
88 else lcl_const[3] = constants[j + 3];
90 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
93 for(i = 0; i < max_constants; i++) {
96 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
100 checkGLcall("glProgramEnvParameter4fvARB()");
102 /* Load immediate constants */
103 if (TRACE_ON(d3d_shader)) {
104 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
105 GLfloat* values = (GLfloat*)lconst->value;
106 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
107 values[0], values[1], values[2], values[3]);
110 /* Immediate constants are clamped for 1.X shaders at loading times */
112 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
113 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
114 ret = max(ret, lconst->idx);
115 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
117 checkGLcall("glProgramEnvParameter4fvARB()");
122 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
124 * We only support float constants in ARB at the moment, so don't
125 * worry about the Integers or Booleans
127 void shader_arb_load_constants(
128 IWineD3DDevice* device,
130 char useVertexShader) {
132 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
133 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
134 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
136 if (useVertexShader) {
137 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
139 /* Load DirectX 9 float constants for vertex shader */
140 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
141 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
142 deviceImpl->highest_dirty_vs_const,
143 stateBlock->vertexShaderConstantF,
144 deviceImpl->activeContext->vshader_const_dirty);
146 /* Upload the position fixup */
147 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
150 if (usePixelShader) {
152 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
153 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
155 /* Load DirectX 9 float constants for pixel shader */
156 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
157 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
158 deviceImpl->highest_dirty_ps_const,
159 stateBlock->pixelShaderConstantF,
160 deviceImpl->activeContext->pshader_const_dirty);
162 if(((IWineD3DPixelShaderImpl *) pshader)->bumpenvmatconst != -1) {
163 /* needsbumpmat stores the stage number from where to load the matrix. bumpenvmatconst stores the
164 * number of the constant to load the matrix into.
165 * The state manager takes care that this function is always called if the bump env matrix changes
167 float *data = (float *) &stateBlock->textureState[(int) psi->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
168 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst, data));
170 if(((IWineD3DPixelShaderImpl *) pshader)->luminanceconst != -1) {
171 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
172 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
173 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
174 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
176 float *scale = (float *) &stateBlock->textureState[(int) psi->needsbumpmat][WINED3DTSS_BUMPENVLSCALE];
177 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst, scale));
178 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst] = 1;
181 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
182 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
186 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
187 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
188 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
190 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
191 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
193 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
194 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
196 mul_low[0] = 1.0; mul_low[1] = 1.0;
197 mul_low[2] = 1.0; mul_low[3] = 1.0;
199 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
200 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
201 checkGLcall("Load sRGB correction constants\n");
202 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
203 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
209 /* Generate the variable & register declarations for the ARB_vertex_program output target */
210 void shader_generate_arb_declarations(
211 IWineD3DBaseShader *iface,
212 shader_reg_maps* reg_maps,
213 SHADER_BUFFER* buffer,
214 WineD3D_GL_Info* gl_info) {
216 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
217 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
219 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
220 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
221 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
222 UINT extra_constants_needed = 0;
224 /* Temporary Output register */
225 shader_addline(buffer, "TEMP TMP_OUT;\n");
227 for(i = 0; i < This->baseShader.limits.temporary; i++) {
228 if (reg_maps->temporary[i])
229 shader_addline(buffer, "TEMP R%u;\n", i);
232 for (i = 0; i < This->baseShader.limits.address; i++) {
233 if (reg_maps->address[i])
234 shader_addline(buffer, "ADDRESS A%d;\n", i);
237 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
238 if (reg_maps->texcoord[i])
239 shader_addline(buffer,"TEMP T%u;\n", i);
242 /* Texture coordinate registers must be pre-loaded */
243 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
244 if (reg_maps->texcoord[i])
245 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
248 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = -1;
249 ((IWineD3DPixelShaderImpl *)This)->luminanceconst = -1;
250 if(reg_maps->bumpmat != -1 /* Only a pshader can use texbem */) {
251 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
252 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
255 if(max_constantsF < GL_LIMITS(pshader_constantsF)) {
256 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = max_constantsF;
257 shader_addline(buffer, "PARAM bumpenvmat = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst);
259 if(reg_maps->luminanceparams != -1 && max_constantsF +1 < GL_LIMITS(pshader_constantsF)) {
260 extra_constants_needed += 1;
261 ((IWineD3DPixelShaderImpl *)This)->luminanceconst = max_constantsF + 1;
262 shader_addline(buffer, "PARAM luminance = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->luminanceconst);
263 } else if(reg_maps->luminanceparams != -1) {
264 FIXME("No free constant to load the luminance parameters\n");
267 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
269 extra_constants_needed += 1;
271 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
272 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
273 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
274 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
275 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
278 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
279 /* The idea is that if srgb is enabled, then disabled, the constant loading code
280 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
281 * and comparison constants. If it disables it that way, the shader won't be recompiled
282 * and the code will stay in, so sRGB writing can be turned on again by setting the
283 * constants from the spec
285 ps_impl->srgb_mode_hardcoded = 0;
286 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
287 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
288 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
289 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
291 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
292 srgb_mul_low, srgb_mul_low, srgb_mul_low);
293 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
294 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
295 ps_impl->srgb_mode_hardcoded = 1;
297 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
298 * path if the comparison value is set to INF
300 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
301 srgb_pow, srgb_pow, srgb_pow);
302 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
303 srgb_mul_high, srgb_mul_high, srgb_mul_high);
304 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
305 srgb_sub_high, srgb_sub_high, srgb_sub_high);
306 ps_impl->srgb_enabled = 1;
308 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
310 /* Do not write any srgb fixup into the shader to save shader size and processing time.
311 * As a consequence, we can't toggle srgb write on without recompilation
313 ps_impl->srgb_enabled = 0;
314 ps_impl->srgb_mode_hardcoded = 1;
317 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
318 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
319 max_constantsF, max_constantsF - 1);
322 static const char * const shift_tab[] = {
323 "dummy", /* 0 (none) */
324 "coefmul.x", /* 1 (x2) */
325 "coefmul.y", /* 2 (x4) */
326 "coefmul.z", /* 3 (x8) */
327 "coefmul.w", /* 4 (x16) */
328 "dummy", /* 5 (x32) */
329 "dummy", /* 6 (x64) */
330 "dummy", /* 7 (x128) */
331 "dummy", /* 8 (d256) */
332 "dummy", /* 9 (d128) */
333 "dummy", /* 10 (d64) */
334 "dummy", /* 11 (d32) */
335 "coefdiv.w", /* 12 (d16) */
336 "coefdiv.z", /* 13 (d8) */
337 "coefdiv.y", /* 14 (d4) */
338 "coefdiv.x" /* 15 (d2) */
341 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
342 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
343 char *ptr = write_mask;
344 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
346 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
349 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
351 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
352 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
353 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
354 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
360 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
361 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
362 * but addressed as "rgba". To fix this we need to swap the register's x
363 * and z components. */
364 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
365 char *ptr = swizzle_str;
367 /* swizzle bits fields: wwzzyyxx */
368 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
369 DWORD swizzle_x = swizzle & 0x03;
370 DWORD swizzle_y = (swizzle >> 2) & 0x03;
371 DWORD swizzle_z = (swizzle >> 4) & 0x03;
372 DWORD swizzle_w = (swizzle >> 6) & 0x03;
374 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
375 * generate a swizzle string. Unless we need to our own swizzling. */
376 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
378 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
379 *ptr++ = swizzle_chars[swizzle_x];
381 *ptr++ = swizzle_chars[swizzle_x];
382 *ptr++ = swizzle_chars[swizzle_y];
383 *ptr++ = swizzle_chars[swizzle_z];
384 *ptr++ = swizzle_chars[swizzle_w];
391 static void pshader_get_register_name(
392 const DWORD param, char* regstr) {
394 DWORD reg = param & WINED3DSP_REGNUM_MASK;
395 DWORD regtype = shader_get_regtype(param);
398 case WINED3DSPR_TEMP:
399 sprintf(regstr, "R%u", reg);
401 case WINED3DSPR_INPUT:
403 strcpy(regstr, "fragment.color.primary");
405 strcpy(regstr, "fragment.color.secondary");
408 case WINED3DSPR_CONST:
409 sprintf(regstr, "C[%u]", reg);
411 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
412 sprintf(regstr,"T%u", reg);
414 case WINED3DSPR_COLOROUT:
416 sprintf(regstr, "TMP_COLOR");
418 /* TODO: See GL_ARB_draw_buffers */
419 FIXME("Unsupported write to render target %u\n", reg);
420 sprintf(regstr, "unsupported_register");
423 case WINED3DSPR_DEPTHOUT:
424 sprintf(regstr, "result.depth");
426 case WINED3DSPR_ATTROUT:
427 sprintf(regstr, "oD[%u]", reg);
429 case WINED3DSPR_TEXCRDOUT:
430 sprintf(regstr, "oT[%u]", reg);
433 FIXME("Unhandled register name Type(%d)\n", regtype);
434 sprintf(regstr, "unrecognized_register");
439 /* TODO: merge with pixel shader */
440 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
442 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
444 /* oPos, oFog and oPts in D3D */
445 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
447 DWORD reg = param & WINED3DSP_REGNUM_MASK;
448 DWORD regtype = shader_get_regtype(param);
450 BOOL is_color = FALSE;
452 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
453 strcat(hwLine, " -");
459 case WINED3DSPR_TEMP:
460 sprintf(tmpReg, "R%u", reg);
461 strcat(hwLine, tmpReg);
463 case WINED3DSPR_INPUT:
465 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
468 sprintf(tmpReg, "vertex.attrib[%u]", reg);
469 strcat(hwLine, tmpReg);
471 case WINED3DSPR_CONST:
472 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
473 if(reg >= This->rel_offset) {
474 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
476 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
479 sprintf(tmpReg, "C[%u]", reg);
481 strcat(hwLine, tmpReg);
483 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
484 sprintf(tmpReg, "A%u", reg);
485 strcat(hwLine, tmpReg);
487 case WINED3DSPR_RASTOUT:
488 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
489 strcat(hwLine, tmpReg);
491 case WINED3DSPR_ATTROUT:
493 strcat(hwLine, "result.color.primary");
495 strcat(hwLine, "result.color.secondary");
498 case WINED3DSPR_TEXCRDOUT:
499 sprintf(tmpReg, "result.texcoord[%u]", reg);
500 strcat(hwLine, tmpReg);
503 FIXME("Unknown reg type %d %d\n", regtype, reg);
504 strcat(hwLine, "unrecognized_register");
510 shader_arb_get_write_mask(arg, param, write_mask);
511 strcat(hwLine, write_mask);
514 shader_arb_get_swizzle(param, is_color, swizzle);
515 strcat(hwLine, swizzle);
519 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
520 SHADER_BUFFER* buffer = arg->buffer;
521 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
522 const char *tex_type;
524 switch(sampler_type) {
531 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
532 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
533 if(device->stateBlock->textures[sampler_idx] &&
534 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
542 case WINED3DSTT_VOLUME:
546 case WINED3DSTT_CUBE:
551 ERR("Unexpected texture type %d\n", sampler_type);
556 /* Shouldn't be possible, but let's check for it */
557 if(projected) FIXME("Biased and Projected texture sampling\n");
558 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
559 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
560 } else if (projected) {
561 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
563 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
567 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
568 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
569 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
570 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
572 WINED3DFORMAT conversion_group;
573 IWineD3DBaseTextureImpl *texture;
575 BOOL recorded = FALSE;
577 DWORD hex_version = shader->baseShader.hex_version;
581 switch(arg->opcode->opcode) {
583 if (hex_version < WINED3DPS_VERSION(2,0)) {
584 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
586 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
590 case WINED3DSIO_TEXLDL:
591 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
594 case WINED3DSIO_TEXDP3TEX:
595 case WINED3DSIO_TEXM3x3TEX:
596 case WINED3DSIO_TEXM3x3SPEC:
597 case WINED3DSIO_TEXM3x3VSPEC:
598 case WINED3DSIO_TEXBEM:
599 case WINED3DSIO_TEXREG2AR:
600 case WINED3DSIO_TEXREG2GB:
601 case WINED3DSIO_TEXREG2RGB:
602 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
606 /* Not a texture sampling instruction, nothing to do */
610 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
612 fmt = texture->resource.format;
613 conversion_group = texture->baseTexture.shader_conversion_group;
615 fmt = WINED3DFMT_UNKNOWN;
616 conversion_group = WINED3DFMT_UNKNOWN;
619 /* before doing anything, record the sampler with the format in the format conversion list,
620 * but check if it's not there already
622 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
623 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
628 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
629 shader->baseShader.num_sampled_samplers++;
630 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
633 pshader_get_register_name(arg->dst, reg);
634 shader_arb_get_write_mask(arg, arg->dst, writemask);
635 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
638 case WINED3DFMT_V8U8:
639 case WINED3DFMT_V16U16:
640 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
641 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
643 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
644 * disabled until an application that needs it is found because it causes unneeded
645 * shader recompilation in some game
647 if(strlen(writemask) >= 4) {
648 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
652 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
653 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
654 * all registers, do so, this saves an instruction.
656 if(strlen(writemask) >= 5) {
657 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
658 } else if(strlen(writemask) >= 3) {
659 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
662 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
665 } else if(strlen(writemask) == 2) {
666 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
672 case WINED3DFMT_X8L8V8U8:
673 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
674 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
675 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
677 if(strlen(writemask) >= 3) {
678 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
681 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
684 } else if(strlen(writemask) == 2) {
685 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
692 case WINED3DFMT_L6V5U5:
693 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
694 if(strlen(writemask) >= 4) {
695 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
696 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
698 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
699 reg, writemask[1], writemask[1],
700 reg, writemask[1], writemask[3]);
701 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
703 } else if(strlen(writemask) == 3) {
704 /* This is bad: We have VL, but we need VU */
705 FIXME("2 components sampled from a converted L6V5U5 texture\n");
707 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
714 case WINED3DFMT_Q8W8V8U8:
715 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
716 /* Correct the sign in all channels */
717 switch(strlen(writemask)) {
719 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
724 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
729 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
734 /* Should not occur, since it's at minimum '.' and a letter */
736 ERR("Unexpected writemask: \"%s\"\n", writemask);
741 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
746 /* stupid compiler */
753 static void pshader_gen_input_modifier_line (
754 SHADER_BUFFER* buffer,
759 /* Generate a line that does the input modifier computation and return the input register to use */
764 /* Assume a new line will be added */
767 /* Get register name */
768 pshader_get_register_name(instr, regstr);
769 shader_arb_get_swizzle(instr, FALSE, swzstr);
771 switch (instr & WINED3DSP_SRCMOD_MASK) {
772 case WINED3DSPSM_NONE:
773 sprintf(outregstr, "%s%s", regstr, swzstr);
776 case WINED3DSPSM_NEG:
777 sprintf(outregstr, "-%s%s", regstr, swzstr);
780 case WINED3DSPSM_BIAS:
781 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
783 case WINED3DSPSM_BIASNEG:
784 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
786 case WINED3DSPSM_SIGN:
787 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
789 case WINED3DSPSM_SIGNNEG:
790 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
792 case WINED3DSPSM_COMP:
793 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
796 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
798 case WINED3DSPSM_X2NEG:
799 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
802 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
803 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
806 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
807 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
810 sprintf(outregstr, "%s%s", regstr, swzstr);
814 /* Return modified or original register, with swizzle */
816 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
819 static inline void pshader_gen_output_modifier_line(
820 SHADER_BUFFER* buffer,
826 /* Generate a line that does the output modifier computation */
827 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
828 regstr, write_mask, regstr, shift_tab[shift]);
831 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
832 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
834 SHADER_BUFFER* buffer = arg->buffer;
836 char src_name[2][50];
839 pshader_get_register_name(arg->dst, dst_name);
840 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
841 strcat(dst_name, dst_wmask);
843 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
844 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
846 if(This->bumpenvmatconst != -1) {
847 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
848 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
849 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
850 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
851 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
853 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
855 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
859 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
861 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
862 SHADER_BUFFER* buffer = arg->buffer;
865 char src_name[3][50];
866 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
867 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
869 /* FIXME: support output modifiers */
871 /* Handle output register */
872 pshader_get_register_name(arg->dst, dst_name);
873 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
875 /* Generate input register names (with modifiers) */
876 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
877 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
878 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
880 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
881 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
882 arg->opcode_token & WINED3DSI_COISSUE) {
883 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
885 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
886 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
887 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
890 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
893 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
895 SHADER_BUFFER* buffer = arg->buffer;
898 char src_name[3][50];
899 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
900 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
902 /* FIXME: support output modifiers */
904 /* Handle output register */
905 pshader_get_register_name(arg->dst, dst_name);
906 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
908 /* Generate input register names (with modifiers) */
909 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
910 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
911 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
913 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
914 src_name[0], src_name[2], src_name[1]);
917 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
920 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
921 * dst = dot2(src0, src1) + src2 */
922 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
923 SHADER_BUFFER* buffer = arg->buffer;
926 char src_name[3][50];
927 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
928 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
930 pshader_get_register_name(arg->dst, dst_name);
931 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
933 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
934 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
935 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
937 /* Emulate a DP2 with a DP3 and 0.0 */
938 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
939 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
940 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
941 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
944 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
947 /* Map the opcode 1-to-1 to the GL code */
948 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
950 CONST SHADER_OPCODE* curOpcode = arg->opcode;
951 SHADER_BUFFER* buffer = arg->buffer;
952 DWORD dst = arg->dst;
953 DWORD* src = arg->src;
958 /* Output token related */
959 char output_rname[256];
960 char output_wmask[20];
961 BOOL saturate = FALSE;
962 BOOL centroid = FALSE;
963 BOOL partialprecision = FALSE;
966 strcpy(tmpLine, curOpcode->glname);
968 /* Process modifiers */
969 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
970 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
972 saturate = mask & WINED3DSPDM_SATURATE;
973 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
974 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
975 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
977 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
980 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
982 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
984 /* Generate input and output registers */
985 if (curOpcode->num_params > 0) {
986 char operands[4][100];
988 /* Generate input register names (with modifiers) */
989 for (i = 1; i < curOpcode->num_params; ++i)
990 pshader_gen_input_modifier_line(buffer, src[i-1], i-1, operands[i]);
992 /* Handle output register */
993 pshader_get_register_name(dst, output_rname);
994 strcpy(operands[0], output_rname);
995 shader_arb_get_write_mask(arg, dst, output_wmask);
996 strcat(operands[0], output_wmask);
998 if (saturate && (shift == 0))
999 strcat(tmpLine, "_SAT");
1000 strcat(tmpLine, " ");
1001 strcat(tmpLine, operands[0]);
1002 for (i = 1; i < curOpcode->num_params; i++) {
1003 strcat(tmpLine, ", ");
1004 strcat(tmpLine, operands[i]);
1006 strcat(tmpLine,";\n");
1007 shader_addline(buffer, tmpLine);
1009 /* A shift requires another line. */
1011 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1015 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1016 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1017 DWORD hex_version = This->baseShader.hex_version;
1018 SHADER_BUFFER* buffer = arg->buffer;
1021 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1022 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1024 pshader_get_register_name(arg->dst, reg_dest);
1026 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1027 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1028 shader_addline(buffer, "KIL %s;\n", reg_dest);
1030 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1031 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1033 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1034 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1035 shader_addline(buffer, "KIL TMP;\n");
1039 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1040 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1041 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1043 DWORD dst = arg->dst;
1044 DWORD* src = arg->src;
1045 SHADER_BUFFER* buffer = arg->buffer;
1046 DWORD hex_version = This->baseShader.hex_version;
1047 BOOL projected = FALSE, bias = FALSE;
1051 DWORD reg_dest_code;
1052 DWORD reg_sampler_code;
1054 /* All versions have a destination register */
1055 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1056 pshader_get_register_name(dst, reg_dest);
1058 /* 1.0-1.3: Use destination register as coordinate source.
1059 1.4+: Use provided coordinate source register. */
1060 if (hex_version < WINED3DPS_VERSION(1,4))
1061 strcpy(reg_coord, reg_dest);
1063 pshader_gen_input_modifier_line(buffer, src[0], 0, reg_coord);
1065 /* 1.0-1.4: Use destination register number as texture code.
1066 2.0+: Use provided sampler number as texure code. */
1067 if (hex_version < WINED3DPS_VERSION(2,0))
1068 reg_sampler_code = reg_dest_code;
1070 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1073 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1074 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1075 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1077 if(hex_version < WINED3DPS_VERSION(1,4)) {
1079 if(reg_sampler_code < MAX_TEXTURES) {
1080 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1082 if (flags & WINED3DTTFF_PROJECTED) {
1085 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1086 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1087 if (src_mod == WINED3DSPSM_DZ) {
1089 } else if(src_mod == WINED3DSPSM_DW) {
1093 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1096 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1100 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1103 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1105 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1106 DWORD dst = arg->dst;
1107 SHADER_BUFFER* buffer = arg->buffer;
1108 DWORD hex_version = This->baseShader.hex_version;
1111 shader_arb_get_write_mask(arg, dst, tmp);
1112 if (hex_version != WINED3DPS_VERSION(1,4)) {
1113 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1114 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1116 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1119 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, reg_src);
1120 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1124 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1126 SHADER_BUFFER* buffer = arg->buffer;
1127 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1128 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1131 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1135 sprintf(dst_str, "T%u", reg1);
1136 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1137 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1138 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1139 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1140 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1143 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1145 SHADER_BUFFER* buffer = arg->buffer;
1146 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1147 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1150 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1154 sprintf(dst_str, "T%u", reg1);
1155 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1156 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1157 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1158 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1159 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1162 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1164 SHADER_BUFFER* buffer = arg->buffer;
1165 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1166 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1168 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1172 sprintf(dst_str, "T%u", reg1);
1173 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_str);
1174 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1175 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1178 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1179 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1181 DWORD dst = arg->dst;
1182 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1183 SHADER_BUFFER* buffer = arg->buffer;
1186 DWORD reg_dest_code;
1188 /* All versions have a destination register */
1189 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1190 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1191 pshader_get_register_name(dst, reg_coord);
1193 if(This->bumpenvmatconst != -1) {
1194 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1196 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
1197 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1198 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
1199 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1201 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1202 * so we can't let the GL handle this.
1204 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1205 & WINED3DTTFF_PROJECTED) {
1206 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1207 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1208 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1210 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1213 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1215 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && This->luminanceconst != -1) {
1216 shader_addline(buffer, "MAD TMP, T%u.z, luminance.x, luminance.y;\n", src);
1217 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1222 if(reg_dest_code < MAX_TEXTURES) {
1223 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1227 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1228 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1232 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1234 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1235 SHADER_BUFFER* buffer = arg->buffer;
1238 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1239 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1242 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1244 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1245 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1247 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1248 SHADER_BUFFER* buffer = arg->buffer;
1252 sprintf(dst_str, "T%u", reg);
1253 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1254 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1255 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1256 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1259 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1261 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1262 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1263 SHADER_BUFFER* buffer = arg->buffer;
1264 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1267 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1268 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1269 current_state->texcoord_w[current_state->current_row++] = reg;
1272 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1274 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1275 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1277 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1278 SHADER_BUFFER* buffer = arg->buffer;
1279 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1283 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1284 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1286 /* Sample the texture using the calculated coordinates */
1287 sprintf(dst_str, "T%u", reg);
1288 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1289 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1290 current_state->current_row = 0;
1293 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1295 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1296 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1298 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1299 SHADER_BUFFER* buffer = arg->buffer;
1300 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1304 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1305 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1307 /* Construct the eye-ray vector from w coordinates */
1308 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1309 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1310 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1312 /* Calculate reflection vector
1314 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1315 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1316 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1317 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1318 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1319 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1320 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1322 /* Sample the texture using the calculated coordinates */
1323 sprintf(dst_str, "T%u", reg);
1324 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1325 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1326 current_state->current_row = 0;
1329 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1331 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1332 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1334 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1335 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1336 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1337 SHADER_BUFFER* buffer = arg->buffer;
1341 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1342 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1344 /* Calculate reflection vector.
1347 * TMP.xyz = 2 * --------- * N - E
1350 * Which normalizes the normal vector
1352 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1353 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1354 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1355 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1356 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1357 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1359 /* Sample the texture using the calculated coordinates */
1360 sprintf(dst_str, "T%u", reg);
1361 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1362 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1363 current_state->current_row = 0;
1366 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1367 SHADER_BUFFER* buffer = arg->buffer;
1370 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1371 * which is essentially an input, is the destination register because it is the first
1372 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1375 pshader_get_register_name(arg->dst, dst_name);
1377 /* According to the msdn, the source register(must be r5) is unusable after
1378 * the texdepth instruction, so we're free to modify it
1380 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1382 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1383 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1384 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1386 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1387 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1388 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1389 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1392 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1393 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1394 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1395 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1396 SHADER_BUFFER* buffer = arg->buffer;
1397 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1401 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1402 shader_addline(buffer, "MOV TMP, 0.0;\n");
1403 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1405 sprintf(dst_str, "T%u", sampler_idx);
1406 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1409 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1410 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1411 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1415 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1416 SHADER_BUFFER* buffer = arg->buffer;
1418 /* Handle output register */
1419 pshader_get_register_name(arg->dst, dst_str);
1420 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1422 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1423 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1425 /* TODO: Handle output modifiers */
1428 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1429 * Perform the 3rd row of a 3x3 matrix multiply */
1430 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1431 SHADER_BUFFER* buffer = arg->buffer;
1435 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1437 pshader_get_register_name(arg->dst, dst_str);
1438 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1440 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1441 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1442 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1444 /* TODO: Handle output modifiers */
1447 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1448 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1449 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1450 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1452 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1453 SHADER_BUFFER* buffer = arg->buffer;
1454 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1457 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1458 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1460 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1461 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1462 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1464 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1465 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1466 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1467 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1470 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1471 Vertex/Pixel shaders to ARB_vertex_program codes */
1472 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1475 int nComponents = 0;
1476 SHADER_OPCODE_ARG tmpArg;
1478 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1480 /* Set constants for the temporary argument */
1481 tmpArg.shader = arg->shader;
1482 tmpArg.buffer = arg->buffer;
1483 tmpArg.src[0] = arg->src[0];
1484 tmpArg.src_addr[0] = arg->src_addr[0];
1485 tmpArg.src_addr[1] = arg->src_addr[1];
1486 tmpArg.reg_maps = arg->reg_maps;
1488 switch(arg->opcode->opcode) {
1489 case WINED3DSIO_M4x4:
1491 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1493 case WINED3DSIO_M4x3:
1495 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1497 case WINED3DSIO_M3x4:
1499 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1501 case WINED3DSIO_M3x3:
1503 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1505 case WINED3DSIO_M3x2:
1507 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1513 for (i = 0; i < nComponents; i++) {
1514 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1515 tmpArg.src[1] = arg->src[1]+i;
1516 vshader_hw_map2gl(&tmpArg);
1520 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1521 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1522 SHADER_BUFFER* buffer = arg->buffer;
1523 DWORD dst = arg->dst;
1524 DWORD src = arg->src[0];
1525 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1529 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1530 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1531 strcat(tmpLine, ",");
1532 vshader_program_add_param(arg, src, TRUE, tmpLine);
1533 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1534 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1537 strcat(tmpLine, ".w");
1540 shader_addline(buffer, "%s;\n", tmpLine);
1543 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1544 SHADER_BUFFER* buffer = arg->buffer;
1548 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1549 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1551 pshader_get_register_name(arg->dst, dst_name);
1552 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1554 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1555 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1556 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1557 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1558 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1562 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1565 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1566 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1567 * must contain fixed constants. So we need a separate function to filter those constants and
1570 SHADER_BUFFER* buffer = arg->buffer;
1574 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1575 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1577 pshader_get_register_name(arg->dst, dst_name);
1578 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1580 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name);
1581 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1585 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1589 /* TODO: merge with pixel shader */
1590 /* Map the opcode 1-to-1 to the GL code */
1591 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1593 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1594 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1595 SHADER_BUFFER* buffer = arg->buffer;
1596 DWORD dst = arg->dst;
1597 DWORD* src = arg->src;
1599 DWORD dst_regtype = shader_get_regtype(dst);
1603 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1604 if(shader->rel_offset) {
1605 memset(tmpLine, 0, sizeof(tmpLine));
1606 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1607 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1608 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1611 strcpy(tmpLine, "ARL");
1614 strcpy(tmpLine, curOpcode->glname);
1616 if (curOpcode->num_params > 0) {
1617 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1618 for (i = 1; i < curOpcode->num_params; ++i) {
1619 strcat(tmpLine, ",");
1620 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1623 shader_addline(buffer, "%s;\n", tmpLine);
1626 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1627 GLuint program_id = 0;
1628 const char *blt_vprogram =
1630 "PARAM c[1] = { { 1, 0.5 } };\n"
1631 "MOV result.position, vertex.position;\n"
1632 "MOV result.color, c[0].x;\n"
1633 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1634 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1637 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1638 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1639 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1641 if (glGetError() == GL_INVALID_OPERATION) {
1643 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1644 FIXME("Vertex program error at position %d: %s\n", pos,
1645 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1651 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1652 GLuint program_id = 0;
1653 const char *blt_fprogram =
1656 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1657 "MOV result.depth.z, R0.x;\n"
1660 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1661 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1662 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1664 if (glGetError() == GL_INVALID_OPERATION) {
1666 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1667 FIXME("Fragment program error at position %d: %s\n", pos,
1668 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1674 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1675 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1676 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1679 TRACE("Using vertex shader\n");
1681 /* Bind the vertex program */
1682 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1683 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1684 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1686 /* Enable OpenGL vertex programs */
1687 glEnable(GL_VERTEX_PROGRAM_ARB);
1688 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1689 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1690 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1691 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1692 glDisable(GL_VERTEX_PROGRAM_ARB);
1693 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1697 TRACE("Using pixel shader\n");
1699 /* Bind the fragment program */
1700 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1701 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1702 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1704 /* Enable OpenGL fragment programs */
1705 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1706 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1707 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1708 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1709 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1710 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1711 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1715 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1716 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1717 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1718 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1720 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1721 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1722 glEnable(GL_VERTEX_PROGRAM_ARB);
1724 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1725 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1726 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1729 static void shader_arb_destroy_depth_blt(IWineD3DDevice *iface) {
1730 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1731 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1732 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1734 if(priv->depth_blt_vprogram_id) {
1735 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1736 priv->depth_blt_vprogram_id = 0;
1738 if(priv->depth_blt_fprogram_id) {
1739 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1740 priv->depth_blt_fprogram_id = 0;
1744 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1745 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1746 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1747 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1748 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1751 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1752 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1753 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1756 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1757 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1759 This->baseShader.prgId = 0;
1760 This->baseShader.is_compiled = FALSE;
1763 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1764 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1765 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1769 static void shader_arb_free(IWineD3DDevice *iface) {
1770 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1771 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1774 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1778 const shader_backend_t arb_program_shader_backend = {
1780 &shader_arb_select_depth_blt,
1781 &shader_arb_destroy_depth_blt,
1782 &shader_arb_load_constants,
1783 &shader_arb_cleanup,
1784 &shader_arb_color_correction,
1785 &shader_arb_destroy,
1788 &shader_arb_dirty_const