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 void shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
52 unsigned int max_constants, float* constants, struct list *constant_list) {
53 constants_entry *constant;
54 local_constant* lconst;
58 if (TRACE_ON(d3d_shader)) {
59 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
64 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
65 constants[i * 4 + 0], constants[i * 4 + 1],
66 constants[i * 4 + 2], constants[i * 4 + 3]);
70 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
71 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
72 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
74 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
80 if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
81 else if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
82 else lcl_const[0] = constants[k + 0];
84 if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
85 else if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
86 else lcl_const[1] = constants[k + 1];
88 if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
89 else if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
90 else lcl_const[2] = constants[k + 2];
92 if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
93 else if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
94 else lcl_const[3] = constants[k + 3];
96 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
100 LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
105 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
109 checkGLcall("glProgramEnvParameter4fvARB()");
111 /* Load immediate constants */
112 if (TRACE_ON(d3d_shader)) {
113 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
114 GLfloat* values = (GLfloat*)lconst->value;
115 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
116 values[0], values[1], values[2], values[3]);
119 /* Immediate constants are clamped for 1.X shaders at loading times */
120 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
121 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
123 checkGLcall("glProgramEnvParameter4fvARB()");
127 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
129 * We only support float constants in ARB at the moment, so don't
130 * worry about the Integers or Booleans
132 void shader_arb_load_constants(
133 IWineD3DDevice* device,
135 char useVertexShader) {
137 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
138 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
139 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
141 if (useVertexShader) {
142 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
144 /* Load DirectX 9 float constants for vertex shader */
145 shader_arb_load_constantsF(vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
146 GL_LIMITS(vshader_constantsF),
147 stateBlock->vertexShaderConstantF,
148 &stateBlock->set_vconstantsF);
150 /* Upload the position fixup */
151 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
154 if (usePixelShader) {
156 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
157 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
159 /* Load DirectX 9 float constants for pixel shader */
160 shader_arb_load_constantsF(pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
161 GL_LIMITS(pshader_constantsF),
162 stateBlock->pixelShaderConstantF,
163 &stateBlock->set_pconstantsF);
164 if(((IWineD3DPixelShaderImpl *) pshader)->bumpenvmatconst) {
165 /* needsbumpmat stores the stage number from where to load the matrix. bumpenvmatconst stores the
166 * number of the constant to load the matrix into.
167 * The state manager takes care that this function is always called if the bump env matrix changes
169 float *data = (float *) &stateBlock->textureState[(int) psi->needsbumpmat][WINED3DTSS_BUMPENVMAT00];
170 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst, data));
172 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
173 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
177 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
178 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
179 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
181 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
182 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
184 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
185 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
187 mul_low[0] = 1.0; mul_low[1] = 1.0;
188 mul_low[2] = 1.0; mul_low[3] = 1.0;
190 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
191 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
192 checkGLcall("Load sRGB correction constants\n");
197 /* Generate the variable & register declarations for the ARB_vertex_program output target */
198 void shader_generate_arb_declarations(
199 IWineD3DBaseShader *iface,
200 shader_reg_maps* reg_maps,
201 SHADER_BUFFER* buffer,
202 WineD3D_GL_Info* gl_info) {
204 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
205 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
207 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
208 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
209 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
210 UINT extra_constants_needed = 0;
212 /* Temporary Output register */
213 shader_addline(buffer, "TEMP TMP_OUT;\n");
215 for(i = 0; i < This->baseShader.limits.temporary; i++) {
216 if (reg_maps->temporary[i])
217 shader_addline(buffer, "TEMP R%u;\n", i);
220 for (i = 0; i < This->baseShader.limits.address; i++) {
221 if (reg_maps->address[i])
222 shader_addline(buffer, "ADDRESS A%d;\n", i);
225 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
226 if (reg_maps->texcoord[i])
227 shader_addline(buffer,"TEMP T%u;\n", i);
230 /* Texture coordinate registers must be pre-loaded */
231 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
232 if (reg_maps->texcoord[i])
233 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
236 if(reg_maps->bumpmat != -1 /* Only a pshader can use texbem */) {
237 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
238 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
241 if(max_constantsF < GL_LIMITS(pshader_constantsF)) {
242 ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst = max_constantsF;
243 shader_addline(buffer, "PARAM bumpenvmat = program.env[%d];\n", ((IWineD3DPixelShaderImpl *)This)->bumpenvmatconst);
245 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
247 extra_constants_needed += 1;
249 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
250 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
251 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
252 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
253 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
256 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
257 /* The idea is that if srgb is enabled, then disabled, the constant loading code
258 * can effectively disabling sRGB correction by passing 1.0 and INF as the multiplication
259 * and comparison constants. If it disables it that way, the shader won't be recompiled
260 * and the code will stay in, so sRGB writing can be turned on again by setting the
261 * constants from the spec
263 ps_impl->srgb_mode_hardcoded = 0;
264 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
265 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
266 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
267 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
269 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
270 srgb_mul_low, srgb_mul_low, srgb_mul_low);
271 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
272 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
273 ps_impl->srgb_mode_hardcoded = 1;
275 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
276 * path if the comparison value is set to INF
278 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
279 srgb_pow, srgb_pow, srgb_pow);
280 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
281 srgb_mul_high, srgb_mul_high, srgb_mul_high);
282 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
283 srgb_sub_high, srgb_sub_high, srgb_sub_high);
284 ps_impl->srgb_enabled = 1;
286 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
288 /* Do not write any srgb fixup into the shader to save shader size and processing time.
289 * As a consequence, we can't toggle srgb write on without recompilation
291 ps_impl->srgb_enabled = 0;
292 ps_impl->srgb_mode_hardcoded = 1;
295 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
296 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
297 max_constantsF, max_constantsF - 1);
300 static const char * const shift_tab[] = {
301 "dummy", /* 0 (none) */
302 "coefmul.x", /* 1 (x2) */
303 "coefmul.y", /* 2 (x4) */
304 "coefmul.z", /* 3 (x8) */
305 "coefmul.w", /* 4 (x16) */
306 "dummy", /* 5 (x32) */
307 "dummy", /* 6 (x64) */
308 "dummy", /* 7 (x128) */
309 "dummy", /* 8 (d256) */
310 "dummy", /* 9 (d128) */
311 "dummy", /* 10 (d64) */
312 "dummy", /* 11 (d32) */
313 "coefdiv.w", /* 12 (d16) */
314 "coefdiv.z", /* 13 (d8) */
315 "coefdiv.y", /* 14 (d4) */
316 "coefdiv.x" /* 15 (d2) */
319 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
320 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
321 char *ptr = write_mask;
322 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
324 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
327 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
329 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
330 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
331 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
332 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
338 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
339 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
340 * but addressed as "rgba". To fix this we need to swap the register's x
341 * and z components. */
342 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
343 char *ptr = swizzle_str;
345 /* swizzle bits fields: wwzzyyxx */
346 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
347 DWORD swizzle_x = swizzle & 0x03;
348 DWORD swizzle_y = (swizzle >> 2) & 0x03;
349 DWORD swizzle_z = (swizzle >> 4) & 0x03;
350 DWORD swizzle_w = (swizzle >> 6) & 0x03;
352 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
353 * generate a swizzle string. Unless we need to our own swizzling. */
354 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
356 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
357 *ptr++ = swizzle_chars[swizzle_x];
359 *ptr++ = swizzle_chars[swizzle_x];
360 *ptr++ = swizzle_chars[swizzle_y];
361 *ptr++ = swizzle_chars[swizzle_z];
362 *ptr++ = swizzle_chars[swizzle_w];
369 static void pshader_get_register_name(
370 const DWORD param, char* regstr) {
372 DWORD reg = param & WINED3DSP_REGNUM_MASK;
373 DWORD regtype = shader_get_regtype(param);
376 case WINED3DSPR_TEMP:
377 sprintf(regstr, "R%u", reg);
379 case WINED3DSPR_INPUT:
381 strcpy(regstr, "fragment.color.primary");
383 strcpy(regstr, "fragment.color.secondary");
386 case WINED3DSPR_CONST:
387 sprintf(regstr, "C[%u]", reg);
389 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
390 sprintf(regstr,"T%u", reg);
392 case WINED3DSPR_COLOROUT:
394 sprintf(regstr, "TMP_COLOR");
396 /* TODO: See GL_ARB_draw_buffers */
397 FIXME("Unsupported write to render target %u\n", reg);
398 sprintf(regstr, "unsupported_register");
401 case WINED3DSPR_DEPTHOUT:
402 sprintf(regstr, "result.depth");
404 case WINED3DSPR_ATTROUT:
405 sprintf(regstr, "oD[%u]", reg);
407 case WINED3DSPR_TEXCRDOUT:
408 sprintf(regstr, "oT[%u]", reg);
411 FIXME("Unhandled register name Type(%d)\n", regtype);
412 sprintf(regstr, "unrecognized_register");
417 /* TODO: merge with pixel shader */
418 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
420 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
422 /* oPos, oFog and oPts in D3D */
423 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
425 DWORD reg = param & WINED3DSP_REGNUM_MASK;
426 DWORD regtype = shader_get_regtype(param);
428 BOOL is_color = FALSE;
430 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
431 strcat(hwLine, " -");
437 case WINED3DSPR_TEMP:
438 sprintf(tmpReg, "R%u", reg);
439 strcat(hwLine, tmpReg);
441 case WINED3DSPR_INPUT:
443 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
446 sprintf(tmpReg, "vertex.attrib[%u]", reg);
447 strcat(hwLine, tmpReg);
449 case WINED3DSPR_CONST:
450 sprintf(tmpReg, "C[%s%u]", (param & WINED3DSHADER_ADDRMODE_RELATIVE) ? "A0.x + " : "", reg);
451 strcat(hwLine, tmpReg);
453 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
454 sprintf(tmpReg, "A%u", reg);
455 strcat(hwLine, tmpReg);
457 case WINED3DSPR_RASTOUT:
458 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
459 strcat(hwLine, tmpReg);
461 case WINED3DSPR_ATTROUT:
463 strcat(hwLine, "result.color.primary");
465 strcat(hwLine, "result.color.secondary");
468 case WINED3DSPR_TEXCRDOUT:
469 sprintf(tmpReg, "result.texcoord[%u]", reg);
470 strcat(hwLine, tmpReg);
473 FIXME("Unknown reg type %d %d\n", regtype, reg);
474 strcat(hwLine, "unrecognized_register");
480 shader_arb_get_write_mask(arg, param, write_mask);
481 strcat(hwLine, write_mask);
484 shader_arb_get_swizzle(param, is_color, swizzle);
485 strcat(hwLine, swizzle);
489 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
490 SHADER_BUFFER* buffer = arg->buffer;
491 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
492 const char *tex_type;
494 switch(sampler_type) {
503 case WINED3DSTT_VOLUME:
507 case WINED3DSTT_CUBE:
512 ERR("Unexpected texture type %d\n", sampler_type);
517 /* Shouldn't be possible, but let's check for it */
518 if(projected) FIXME("Biased and Projected texture sampling\n");
519 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
520 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
521 } else if (projected) {
522 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
524 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
528 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
529 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
530 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
531 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
533 WINED3DFORMAT conversion_group;
534 IWineD3DBaseTextureImpl *texture;
536 BOOL recorded = FALSE;
538 DWORD hex_version = shader->baseShader.hex_version;
542 switch(arg->opcode->opcode) {
544 if (hex_version < WINED3DPS_VERSION(2,0)) {
545 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
547 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
551 case WINED3DSIO_TEXLDL:
552 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
555 case WINED3DSIO_TEXDP3TEX:
556 case WINED3DSIO_TEXM3x3TEX:
557 case WINED3DSIO_TEXM3x3SPEC:
558 case WINED3DSIO_TEXM3x3VSPEC:
559 case WINED3DSIO_TEXBEM:
560 case WINED3DSIO_TEXREG2AR:
561 case WINED3DSIO_TEXREG2GB:
562 case WINED3DSIO_TEXREG2RGB:
563 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
567 /* Not a texture sampling instruction, nothing to do */
571 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
573 fmt = texture->resource.format;
574 conversion_group = texture->baseTexture.shader_conversion_group;
576 fmt = WINED3DFMT_UNKNOWN;
577 conversion_group = WINED3DFMT_UNKNOWN;
580 /* before doing anything, record the sampler with the format in the format conversion list,
581 * but check if it's not there already
583 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
584 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
589 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
590 shader->baseShader.num_sampled_samplers++;
591 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
594 pshader_get_register_name(arg->dst, reg);
595 shader_arb_get_write_mask(arg, arg->dst, writemask);
596 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
599 case WINED3DFMT_V8U8:
600 case WINED3DFMT_V16U16:
601 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
602 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
604 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
605 * disabled until an application that needs it is found because it causes unneeded
606 * shader recompilation in some game
608 if(strlen(writemask) >= 4) {
609 shader_addline(arg->buffer, "MOV %s.%c, one.z;\n", reg, writemask[3]);
613 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
614 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
615 * all registers, do so, this saves an instruction.
617 if(strlen(writemask) >= 5) {
618 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
619 } else if(strlen(writemask) >= 3) {
620 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
623 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
626 } else if(strlen(writemask) == 2) {
627 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n", reg, writemask[1],
633 case WINED3DFMT_X8L8V8U8:
634 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
635 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
636 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
638 if(strlen(writemask) >= 3) {
639 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
642 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
645 } else if(strlen(writemask) == 2) {
646 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
653 case WINED3DFMT_L6V5U5:
654 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
655 if(strlen(writemask) >= 4) {
656 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
657 shader_addline(arg->buffer, "MOV TMP.g, %s.%c;\n",
659 shader_addline(arg->buffer, "MAD %s.%c%c, %s.%c%c, coefmul.x, -one;\n",
660 reg, writemask[1], writemask[1],
661 reg, writemask[1], writemask[3]);
662 shader_addline(arg->buffer, "MOV %s.%c, TMP.g;\n", reg,
664 } else if(strlen(writemask) == 3) {
665 /* This is bad: We have VL, but we need VU */
666 FIXME("2 components sampled from a converted L6V5U5 texture\n");
668 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
675 case WINED3DFMT_Q8W8V8U8:
676 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
677 /* Correct the sign in all channels */
678 switch(strlen(writemask)) {
680 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
685 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
690 shader_addline(arg->buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
695 /* Should not occur, since it's at minimum '.' and a letter */
697 ERR("Unexpected writemask: \"%s\"\n", writemask);
702 shader_addline(arg->buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
707 /* stupid compiler */
714 static void pshader_gen_input_modifier_line (
715 SHADER_BUFFER* buffer,
720 /* Generate a line that does the input modifier computation and return the input register to use */
725 /* Assume a new line will be added */
728 /* Get register name */
729 pshader_get_register_name(instr, regstr);
730 shader_arb_get_swizzle(instr, FALSE, swzstr);
732 switch (instr & WINED3DSP_SRCMOD_MASK) {
733 case WINED3DSPSM_NONE:
734 sprintf(outregstr, "%s%s", regstr, swzstr);
737 case WINED3DSPSM_NEG:
738 sprintf(outregstr, "-%s%s", regstr, swzstr);
741 case WINED3DSPSM_BIAS:
742 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
744 case WINED3DSPSM_BIASNEG:
745 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
747 case WINED3DSPSM_SIGN:
748 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
750 case WINED3DSPSM_SIGNNEG:
751 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
753 case WINED3DSPSM_COMP:
754 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
757 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
759 case WINED3DSPSM_X2NEG:
760 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
763 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
764 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
767 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
768 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
771 sprintf(outregstr, "%s%s", regstr, swzstr);
775 /* Return modified or original register, with swizzle */
777 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
780 static inline void pshader_gen_output_modifier_line(
781 SHADER_BUFFER* buffer,
787 /* Generate a line that does the output modifier computation */
788 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
789 regstr, write_mask, regstr, shift_tab[shift]);
792 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
793 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
795 SHADER_BUFFER* buffer = arg->buffer;
797 char src_name[2][50];
800 pshader_get_register_name(arg->dst, dst_name);
801 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
802 strcat(dst_name, dst_wmask);
804 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
805 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
807 if(This->bumpenvmatconst != -1) {
808 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
809 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
810 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
811 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
812 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
814 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
816 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
820 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
822 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
823 SHADER_BUFFER* buffer = arg->buffer;
826 char src_name[3][50];
828 /* FIXME: support output modifiers */
830 /* Handle output register */
831 pshader_get_register_name(arg->dst, dst_name);
832 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
833 strcat(dst_name, dst_wmask);
835 /* Generate input register names (with modifiers) */
836 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
837 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
838 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
840 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
841 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
842 arg->opcode_token & WINED3DSI_COISSUE) {
843 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[1]);
845 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
846 shader_addline(buffer, "CMP %s, TMP, %s, %s;\n", dst_name, src_name[1], src_name[2]);
850 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
852 SHADER_BUFFER* buffer = arg->buffer;
855 char src_name[3][50];
857 /* FIXME: support output modifiers */
859 /* Handle output register */
860 pshader_get_register_name(arg->dst, dst_name);
861 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
862 strcat(dst_name, dst_wmask);
864 /* Generate input register names (with modifiers) */
865 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src_name[0]);
866 pshader_gen_input_modifier_line(buffer, arg->src[1], 1, src_name[1]);
867 pshader_gen_input_modifier_line(buffer, arg->src[2], 2, src_name[2]);
869 shader_addline(buffer, "CMP %s, %s, %s, %s;\n", dst_name,
870 src_name[0], src_name[2], src_name[1]);
873 /* Map the opcode 1-to-1 to the GL code */
874 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
876 CONST SHADER_OPCODE* curOpcode = arg->opcode;
877 SHADER_BUFFER* buffer = arg->buffer;
878 DWORD dst = arg->dst;
879 DWORD* src = arg->src;
884 /* Output token related */
885 char output_rname[256];
886 char output_wmask[20];
887 BOOL saturate = FALSE;
888 BOOL centroid = FALSE;
889 BOOL partialprecision = FALSE;
892 strcpy(tmpLine, curOpcode->glname);
894 /* Process modifiers */
895 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
896 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
898 saturate = mask & WINED3DSPDM_SATURATE;
899 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
900 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
901 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
903 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
906 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
908 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
910 /* Generate input and output registers */
911 if (curOpcode->num_params > 0) {
912 char operands[4][100];
914 /* Generate input register names (with modifiers) */
915 for (i = 1; i < curOpcode->num_params; ++i)
916 pshader_gen_input_modifier_line(buffer, src[i-1], i-1, operands[i]);
918 /* Handle output register */
919 pshader_get_register_name(dst, output_rname);
920 strcpy(operands[0], output_rname);
921 shader_arb_get_write_mask(arg, dst, output_wmask);
922 strcat(operands[0], output_wmask);
924 if (saturate && (shift == 0))
925 strcat(tmpLine, "_SAT");
926 strcat(tmpLine, " ");
927 strcat(tmpLine, operands[0]);
928 for (i = 1; i < curOpcode->num_params; i++) {
929 strcat(tmpLine, ", ");
930 strcat(tmpLine, operands[i]);
932 strcat(tmpLine,";\n");
933 shader_addline(buffer, tmpLine);
935 /* A shift requires another line. */
937 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
941 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
942 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
943 DWORD hex_version = This->baseShader.hex_version;
944 SHADER_BUFFER* buffer = arg->buffer;
947 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
948 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
950 pshader_get_register_name(arg->dst, reg_dest);
952 if(hex_version >= WINED3DPS_VERSION(2,0)) {
953 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
954 shader_addline(buffer, "KIL %s;\n", reg_dest);
956 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
957 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
959 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
960 shader_addline(buffer, "MOV TMP.w, one.w;\n", reg_dest);
961 shader_addline(buffer, "KIL TMP;\n", reg_dest);
965 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
966 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
967 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
969 DWORD dst = arg->dst;
970 DWORD* src = arg->src;
971 SHADER_BUFFER* buffer = arg->buffer;
972 DWORD hex_version = This->baseShader.hex_version;
973 BOOL projected = FALSE, bias = FALSE;
978 DWORD reg_sampler_code;
980 /* All versions have a destination register */
981 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
982 pshader_get_register_name(dst, reg_dest);
984 /* 1.0-1.3: Use destination register as coordinate source.
985 1.4+: Use provided coordinate source register. */
986 if (hex_version < WINED3DPS_VERSION(1,4))
987 strcpy(reg_coord, reg_dest);
989 pshader_gen_input_modifier_line(buffer, src[0], 0, reg_coord);
991 /* 1.0-1.4: Use destination register number as texture code.
992 2.0+: Use provided sampler number as texure code. */
993 if (hex_version < WINED3DPS_VERSION(2,0))
994 reg_sampler_code = reg_dest_code;
996 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
999 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1000 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1001 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1003 if(hex_version < WINED3DPS_VERSION(1,4)) {
1005 if(reg_sampler_code < MAX_TEXTURES) {
1006 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1008 if (flags & WINED3DTTFF_PROJECTED) {
1011 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1012 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1013 if (src_mod == WINED3DSPSM_DZ) {
1015 } else if(src_mod == WINED3DSPSM_DW) {
1019 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1022 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1026 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1029 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1031 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1032 DWORD dst = arg->dst;
1033 SHADER_BUFFER* buffer = arg->buffer;
1034 DWORD hex_version = This->baseShader.hex_version;
1037 shader_arb_get_write_mask(arg, dst, tmp);
1038 if (hex_version != WINED3DPS_VERSION(1,4)) {
1039 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1040 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1042 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1045 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, reg_src);
1046 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1050 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1052 SHADER_BUFFER* buffer = arg->buffer;
1053 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1054 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1057 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1058 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1061 sprintf(dst_str, "T%u", reg1);
1062 shader_addline(buffer, "MOV TMP.r, T%u.a;\n", reg2);
1063 shader_addline(buffer, "MOV TMP.g, T%u.r;\n", reg2);
1064 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1065 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1068 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1070 SHADER_BUFFER* buffer = arg->buffer;
1071 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1072 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1075 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1076 DWORD reg2 = arg->src[0] & WINED3DSP_REGNUM_MASK;
1079 sprintf(dst_str, "T%u", reg1);
1080 shader_addline(buffer, "MOV TMP.r, T%u.g;\n", reg2);
1081 shader_addline(buffer, "MOV TMP.g, T%u.b;\n", reg2);
1082 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1083 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1086 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1087 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1089 DWORD dst = arg->dst;
1090 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1091 SHADER_BUFFER* buffer = arg->buffer;
1094 DWORD reg_dest_code;
1096 /* All versions have a destination register */
1097 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1098 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1099 pshader_get_register_name(dst, reg_coord);
1101 if(This->bumpenvmatconst != -1) {
1102 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1104 shader_addline(buffer, "SWZ TMP2, bumpenvmat, x, z, 0, 0;\n");
1105 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1106 shader_addline(buffer, "SWZ TMP2, bumpenvmat, y, w, 0, 0;\n");
1107 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1109 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1110 * so we can't let the GL handle this.
1112 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1113 & WINED3DTTFF_PROJECTED) {
1114 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1115 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1116 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1118 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1121 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1124 if(reg_dest_code < MAX_TEXTURES) {
1125 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1129 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1130 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1134 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1136 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1137 SHADER_BUFFER* buffer = arg->buffer;
1140 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1141 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1144 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1146 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1147 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1149 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1150 SHADER_BUFFER* buffer = arg->buffer;
1154 sprintf(dst_str, "T%u", reg);
1155 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1156 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1157 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1158 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1161 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1163 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1164 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1165 SHADER_BUFFER* buffer = arg->buffer;
1166 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1169 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1170 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1171 current_state->texcoord_w[current_state->current_row++] = reg;
1174 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1176 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1177 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1179 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1180 SHADER_BUFFER* buffer = arg->buffer;
1181 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1185 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1186 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1188 /* Sample the texture using the calculated coordinates */
1189 sprintf(dst_str, "T%u", reg);
1190 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1191 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1192 current_state->current_row = 0;
1195 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1197 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1198 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1200 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1201 SHADER_BUFFER* buffer = arg->buffer;
1202 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1206 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1207 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1209 /* Construct the eye-ray vector from w coordinates */
1210 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1211 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1212 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1214 /* Calculate reflection vector
1216 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1217 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1218 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1219 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1220 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1221 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1222 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1224 /* Sample the texture using the calculated coordinates */
1225 sprintf(dst_str, "T%u", reg);
1226 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1227 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1228 current_state->current_row = 0;
1231 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1233 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1234 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1236 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1237 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1238 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1239 SHADER_BUFFER* buffer = arg->buffer;
1243 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0_name);
1244 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1246 /* Calculate reflection vector.
1249 * TMP.xyz = 2 * --------- * N - E
1252 * Which normalizes the normal vector
1254 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1255 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1256 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1257 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1258 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1259 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1261 /* Sample the texture using the calculated coordinates */
1262 sprintf(dst_str, "T%u", reg);
1263 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1264 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1265 current_state->current_row = 0;
1268 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1269 SHADER_BUFFER* buffer = arg->buffer;
1272 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1273 * which is essentially an input, is the destiantion register because it is the first
1274 * param. According to the msdn, this must be register r5, but let's keep it more flexible
1277 pshader_get_register_name(arg->dst, dst_name);
1279 /* According to the msdn, the source register(must be r5) is unusable after
1280 * the texdepth instruction, so we're free to modify it
1282 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1284 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1285 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1286 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1288 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1289 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1290 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n", dst_name, dst_name);
1291 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n", dst_name, dst_name);
1294 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1295 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1296 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1297 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1298 SHADER_BUFFER* buffer = arg->buffer;
1299 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1303 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1304 shader_addline(buffer, "MOV TMP, 0.0;\n");
1305 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1307 sprintf(dst_str, "T%u", sampler_idx);
1308 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1311 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1312 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1313 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1317 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1318 SHADER_BUFFER* buffer = arg->buffer;
1320 /* Handle output register */
1321 pshader_get_register_name(arg->dst, dst_str);
1322 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1324 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1325 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1327 /* TODO: Handle output modifiers */
1330 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1331 * Perform the 3rd row of a 3x3 matrix multiply */
1332 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1333 SHADER_BUFFER* buffer = arg->buffer;
1337 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1339 pshader_get_register_name(arg->dst, dst_str);
1340 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1342 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1343 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1344 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1346 /* TODO: Handle output modifiers */
1349 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1350 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1351 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1352 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1354 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1355 SHADER_BUFFER* buffer = arg->buffer;
1356 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1359 pshader_gen_input_modifier_line(buffer, arg->src[0], 0, src0);
1360 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1362 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1363 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1364 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1366 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1367 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1368 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1369 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1372 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1373 Vertex shaders to ARB_vertex_program codes */
1374 void vshader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1377 int nComponents = 0;
1378 SHADER_OPCODE_ARG tmpArg;
1380 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1382 /* Set constants for the temporary argument */
1383 tmpArg.shader = arg->shader;
1384 tmpArg.buffer = arg->buffer;
1385 tmpArg.src[0] = arg->src[0];
1386 tmpArg.src_addr[0] = arg->src_addr[0];
1387 tmpArg.src_addr[1] = arg->src_addr[1];
1388 tmpArg.reg_maps = arg->reg_maps;
1390 switch(arg->opcode->opcode) {
1391 case WINED3DSIO_M4x4:
1393 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1395 case WINED3DSIO_M4x3:
1397 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1399 case WINED3DSIO_M3x4:
1401 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1403 case WINED3DSIO_M3x3:
1405 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1407 case WINED3DSIO_M3x2:
1409 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1415 for (i = 0; i < nComponents; i++) {
1416 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1417 tmpArg.src[1] = arg->src[1]+i;
1418 vshader_hw_map2gl(&tmpArg);
1422 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1423 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1424 SHADER_BUFFER* buffer = arg->buffer;
1425 DWORD dst = arg->dst;
1426 DWORD src = arg->src[0];
1427 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1431 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1432 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1433 strcat(tmpLine, ",");
1434 vshader_program_add_param(arg, src, TRUE, tmpLine);
1435 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1436 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1439 strcat(tmpLine, ".w");
1442 shader_addline(buffer, "%s;\n", tmpLine);
1445 /* TODO: merge with pixel shader */
1446 /* Map the opcode 1-to-1 to the GL code */
1447 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1449 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1450 SHADER_BUFFER* buffer = arg->buffer;
1451 DWORD dst = arg->dst;
1452 DWORD* src = arg->src;
1454 DWORD dst_regtype = shader_get_regtype(dst);
1458 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA)
1459 strcpy(tmpLine, "ARL");
1461 strcpy(tmpLine, curOpcode->glname);
1463 if (curOpcode->num_params > 0) {
1464 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1465 for (i = 1; i < curOpcode->num_params; ++i) {
1466 strcat(tmpLine, ",");
1467 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1470 shader_addline(buffer, "%s;\n", tmpLine);
1473 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1474 GLuint program_id = 0;
1475 const char *blt_vprogram =
1477 "PARAM c[1] = { { 1, 0.5 } };\n"
1478 "MOV result.position, vertex.position;\n"
1479 "MOV result.color, c[0].x;\n"
1480 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1481 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1484 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1485 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1486 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1488 if (glGetError() == GL_INVALID_OPERATION) {
1490 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1491 FIXME("Vertex program error at position %d: %s\n", pos,
1492 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1498 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1499 GLuint program_id = 0;
1500 const char *blt_fprogram =
1503 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1504 "MOV result.depth.z, R0.x;\n"
1507 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1508 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1509 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1511 if (glGetError() == GL_INVALID_OPERATION) {
1513 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1514 FIXME("Fragment program error at position %d: %s\n", pos,
1515 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1521 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1522 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1523 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1526 TRACE("Using vertex shader\n");
1528 /* Bind the vertex program */
1529 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB,
1530 ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId));
1531 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1533 /* Enable OpenGL vertex programs */
1534 glEnable(GL_VERTEX_PROGRAM_ARB);
1535 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1536 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n",
1537 This, ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId);
1538 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1539 glDisable(GL_VERTEX_PROGRAM_ARB);
1540 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1544 TRACE("Using pixel shader\n");
1546 /* Bind the fragment program */
1547 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB,
1548 ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId));
1549 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1551 /* Enable OpenGL fragment programs */
1552 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1553 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1554 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n",
1555 This, ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId);
1556 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1557 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1558 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1562 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1563 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1564 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1565 static GLuint vprogram_id = 0;
1566 static GLuint fprogram_id = 0;
1568 if (!vprogram_id) vprogram_id = create_arb_blt_vertex_program(gl_info);
1569 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vprogram_id));
1570 glEnable(GL_VERTEX_PROGRAM_ARB);
1572 if (!fprogram_id) fprogram_id = create_arb_blt_fragment_program(gl_info);
1573 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, fprogram_id));
1574 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1577 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1578 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1579 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1580 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1581 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1584 const shader_backend_t arb_program_shader_backend = {
1586 &shader_arb_select_depth_blt,
1587 &shader_arb_load_constants,
1588 &shader_arb_cleanup,
1589 &shader_arb_color_correction