2 * shaders implementation
4 * Copyright 2002-2003 Jason Edmeades
5 * Copyright 2002-2003 Raphael Junqueira
6 * Copyright 2004 Christian Costa
7 * Copyright 2005 Oliver Stieber
8 * Copyright 2006 Ivan Gyurdiev
10 * This library is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * This library is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with this library; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
28 #include "wined3d_private.h"
30 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
32 #define GLNAME_REQUIRE_GLSL ((const char *)1)
34 static inline BOOL shader_is_version_token(DWORD token) {
35 return shader_is_pshader_version(token) ||
36 shader_is_vshader_version(token);
40 SHADER_BUFFER* buffer,
41 const char *format, ...) {
43 char* base = buffer->buffer + buffer->bsize;
47 va_start(args, format);
48 rc = vsnprintf(base, SHADER_PGMSIZE - 1 - buffer->bsize, format, args);
51 if (rc < 0 || /* C89 */
52 rc > SHADER_PGMSIZE - 1 - buffer->bsize) { /* C99 */
54 ERR("The buffer allocated for the shader program string "
55 "is too small at %d bytes.\n", SHADER_PGMSIZE);
56 buffer->bsize = SHADER_PGMSIZE - 1;
60 if (buffer->newline) {
61 TRACE("GL HW (%u, %u) : %s", buffer->lineNo + 1, buffer->bsize, base);
62 buffer->newline = FALSE;
68 if (buffer->buffer[buffer->bsize-1] == '\n') {
70 buffer->newline = TRUE;
75 const SHADER_OPCODE* shader_get_opcode(
76 IWineD3DBaseShader *iface, const DWORD code) {
78 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl*) iface;
81 DWORD hex_version = This->baseShader.hex_version;
82 const SHADER_OPCODE *shader_ins = This->baseShader.shader_ins;
84 /** TODO: use dichotomic search */
85 while (NULL != shader_ins[i].name) {
86 if (((code & WINED3DSI_OPCODE_MASK) == shader_ins[i].opcode) &&
87 (((hex_version >= shader_ins[i].min_version) && (hex_version <= shader_ins[i].max_version)) ||
88 ((shader_ins[i].min_version == 0) && (shader_ins[i].max_version == 0)))) {
89 return &shader_ins[i];
93 FIXME("Unsupported opcode %#x(%d) masked %#x, shader version %#x\n",
94 code, code, code & WINED3DSI_OPCODE_MASK, hex_version);
98 /* Read a parameter opcode from the input stream,
99 * and possibly a relative addressing token.
100 * Return the number of tokens read */
101 int shader_get_param(
102 IWineD3DBaseShader* iface,
107 /* PS >= 3.0 have relative addressing (with token)
108 * VS >= 2.0 have relative addressing (with token)
109 * VS >= 1.0 < 2.0 have relative addressing (without token)
110 * The version check below should work in general */
112 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
113 char rel_token = WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2 &&
114 ((*pToken & WINED3DSHADER_ADDRESSMODE_MASK) == WINED3DSHADER_ADDRMODE_RELATIVE);
117 *addr_token = rel_token? *(pToken + 1): 0;
118 return rel_token? 2:1;
121 /* Return the number of parameters to skip for an opcode */
122 static inline int shader_skip_opcode(
123 IWineD3DBaseShaderImpl* This,
124 const SHADER_OPCODE* curOpcode,
125 DWORD opcode_token) {
127 /* Shaders >= 2.0 may contain address tokens, but fortunately they
128 * have a useful length mask - use it here. Shaders 1.0 contain no such tokens */
130 return (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2)?
131 ((opcode_token & WINED3DSI_INSTLENGTH_MASK) >> WINED3DSI_INSTLENGTH_SHIFT):
132 curOpcode->num_params;
135 /* Read the parameters of an unrecognized opcode from the input stream
136 * Return the number of tokens read.
138 * Note: This function assumes source or destination token format.
139 * It will not work with specially-formatted tokens like DEF or DCL,
140 * but hopefully those would be recognized */
142 int shader_skip_unrecognized(
143 IWineD3DBaseShader* iface,
144 const DWORD* pToken) {
149 /* TODO: Think of a good name for 0x80000000 and replace it with a constant */
150 while (*pToken & 0x80000000) {
152 DWORD param, addr_token;
153 tokens_read += shader_get_param(iface, pToken, ¶m, &addr_token);
154 pToken += tokens_read;
156 FIXME("Unrecognized opcode param: token=%08x "
157 "addr_token=%08x name=", param, addr_token);
158 shader_dump_param(iface, param, addr_token, i);
165 /* Convert floating point offset relative
166 * to a register file to an absolute offset for float constants */
168 unsigned int shader_get_float_offset(const DWORD reg) {
170 unsigned int regnum = reg & WINED3DSP_REGNUM_MASK;
171 int regtype = shader_get_regtype(reg);
174 case WINED3DSPR_CONST: return regnum;
175 case WINED3DSPR_CONST2: return 2048 + regnum;
176 case WINED3DSPR_CONST3: return 4096 + regnum;
177 case WINED3DSPR_CONST4: return 6144 + regnum;
179 FIXME("Unsupported register type: %d\n", regtype);
184 static void shader_delete_constant_list(struct list* clist) {
187 struct local_constant* constant;
189 ptr = list_head(clist);
191 constant = LIST_ENTRY(ptr, struct local_constant, entry);
192 ptr = list_next(clist, ptr);
193 HeapFree(GetProcessHeap(), 0, constant);
198 /* Note that this does not count the loop register
199 * as an address register. */
201 HRESULT shader_get_registers_used(
202 IWineD3DBaseShader *iface,
203 shader_reg_maps* reg_maps,
204 semantic* semantics_in,
205 semantic* semantics_out,
207 IWineD3DStateBlockImpl *stateBlock) {
209 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
210 unsigned int cur_loop_depth = 0, max_loop_depth = 0;
212 /* There are some minor differences between pixel and vertex shaders */
213 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
215 reg_maps->bumpmat = -1;
216 reg_maps->luminanceparams = -1;
221 /* get_registers_used is called on every compile on some 1.x shaders, which can result
222 * in stacking up a collection of local constants. Delete the old constants if existing
224 shader_delete_constant_list(&This->baseShader.constantsF);
225 shader_delete_constant_list(&This->baseShader.constantsB);
226 shader_delete_constant_list(&This->baseShader.constantsI);
228 while (WINED3DVS_END() != *pToken) {
229 CONST SHADER_OPCODE* curOpcode;
233 if (shader_is_version_token(*pToken)) {
238 } else if (shader_is_comment(*pToken)) {
239 DWORD comment_len = (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
241 pToken += comment_len;
246 opcode_token = *pToken++;
247 curOpcode = shader_get_opcode(iface, opcode_token);
249 /* Unhandled opcode, and its parameters */
250 if (NULL == curOpcode) {
251 while (*pToken & 0x80000000)
254 /* Handle declarations */
255 } else if (WINED3DSIO_DCL == curOpcode->opcode) {
257 DWORD usage = *pToken++;
258 DWORD param = *pToken++;
259 DWORD regtype = shader_get_regtype(param);
260 unsigned int regnum = param & WINED3DSP_REGNUM_MASK;
262 /* Vshader: mark attributes used
263 Pshader: mark 3.0 input registers used, save token */
264 if (WINED3DSPR_INPUT == regtype) {
267 reg_maps->attributes[regnum] = 1;
269 reg_maps->packed_input[regnum] = 1;
271 semantics_in[regnum].usage = usage;
272 semantics_in[regnum].reg = param;
274 /* Vshader: mark 3.0 output registers used, save token */
275 } else if (WINED3DSPR_OUTPUT == regtype) {
276 reg_maps->packed_output[regnum] = 1;
277 semantics_out[regnum].usage = usage;
278 semantics_out[regnum].reg = param;
279 if (usage & (WINED3DDECLUSAGE_FOG << WINED3DSP_DCL_USAGE_SHIFT))
282 /* Save sampler usage token */
283 } else if (WINED3DSPR_SAMPLER == regtype)
284 reg_maps->samplers[regnum] = usage;
286 } else if (WINED3DSIO_DEF == curOpcode->opcode) {
288 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
289 if (!lconst) return E_OUTOFMEMORY;
290 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
291 memcpy(&lconst->value, pToken + 1, 4 * sizeof(DWORD));
293 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
294 if(WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 && pshader) {
295 float *value = (float *) lconst->value;
296 if(value[0] < -1.0) value[0] = -1.0;
297 else if(value[0] > 1.0) value[0] = 1.0;
298 if(value[1] < -1.0) value[1] = -1.0;
299 else if(value[1] > 1.0) value[1] = 1.0;
300 if(value[2] < -1.0) value[2] = -1.0;
301 else if(value[2] > 1.0) value[2] = 1.0;
302 if(value[3] < -1.0) value[3] = -1.0;
303 else if(value[3] > 1.0) value[3] = 1.0;
306 list_add_head(&This->baseShader.constantsF, &lconst->entry);
307 pToken += curOpcode->num_params;
309 } else if (WINED3DSIO_DEFI == curOpcode->opcode) {
311 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
312 if (!lconst) return E_OUTOFMEMORY;
313 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
314 memcpy(&lconst->value, pToken + 1, 4 * sizeof(DWORD));
315 list_add_head(&This->baseShader.constantsI, &lconst->entry);
316 pToken += curOpcode->num_params;
318 } else if (WINED3DSIO_DEFB == curOpcode->opcode) {
320 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
321 if (!lconst) return E_OUTOFMEMORY;
322 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
323 memcpy(&lconst->value, pToken + 1, 1 * sizeof(DWORD));
324 list_add_head(&This->baseShader.constantsB, &lconst->entry);
325 pToken += curOpcode->num_params;
327 /* If there's a loop in the shader */
328 } else if (WINED3DSIO_LOOP == curOpcode->opcode ||
329 WINED3DSIO_REP == curOpcode->opcode) {
331 if(cur_loop_depth > max_loop_depth)
332 max_loop_depth = cur_loop_depth;
333 pToken += curOpcode->num_params;
335 } else if (WINED3DSIO_ENDLOOP == curOpcode->opcode ||
336 WINED3DSIO_ENDREP == curOpcode->opcode) {
339 /* For subroutine prototypes */
340 } else if (WINED3DSIO_LABEL == curOpcode->opcode) {
342 DWORD snum = *pToken & WINED3DSP_REGNUM_MASK;
343 reg_maps->labels[snum] = 1;
344 pToken += curOpcode->num_params;
346 /* Set texture, address, temporary registers */
350 /* Declare 1.X samplers implicitly, based on the destination reg. number */
351 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 &&
352 (WINED3DSIO_TEX == curOpcode->opcode ||
353 WINED3DSIO_TEXBEM == curOpcode->opcode ||
354 WINED3DSIO_TEXBEML == curOpcode->opcode ||
355 WINED3DSIO_TEXDP3TEX == curOpcode->opcode ||
356 WINED3DSIO_TEXM3x2TEX == curOpcode->opcode ||
357 WINED3DSIO_TEXM3x3SPEC == curOpcode->opcode ||
358 WINED3DSIO_TEXM3x3TEX == curOpcode->opcode ||
359 WINED3DSIO_TEXM3x3VSPEC == curOpcode->opcode ||
360 WINED3DSIO_TEXREG2AR == curOpcode->opcode ||
361 WINED3DSIO_TEXREG2GB == curOpcode->opcode ||
362 WINED3DSIO_TEXREG2RGB == curOpcode->opcode)) {
364 /* Fake sampler usage, only set reserved bit and ttype */
365 DWORD sampler_code = *pToken & WINED3DSP_REGNUM_MASK;
367 if(!stateBlock->textures[sampler_code]) {
368 ERR("No texture bound to sampler %d\n", sampler_code);
369 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_2D;
371 int texType = IWineD3DBaseTexture_GetTextureDimensions(stateBlock->textures[sampler_code]);
373 /* We have to select between texture rectangles and 2D textures later because 2.0 and
374 * 3.0 shaders only have WINED3DSTT_2D as well
376 case GL_TEXTURE_RECTANGLE_ARB:
378 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_2D;
382 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_VOLUME;
385 case GL_TEXTURE_CUBE_MAP_ARB:
386 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_CUBE;
390 ERR("Unexpected gl texture type found: %d\n", texType);
391 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_2D;
395 /* texbem is only valid with < 1.4 pixel shaders */
396 if(WINED3DSIO_TEXBEM == curOpcode->opcode ||
397 WINED3DSIO_TEXBEML == curOpcode->opcode) {
398 if(reg_maps->bumpmat != -1 && reg_maps->bumpmat != sampler_code) {
399 FIXME("Pixel shader uses texbem instruction on more than 1 sampler\n");
401 reg_maps->bumpmat = sampler_code;
402 if(WINED3DSIO_TEXBEML == curOpcode->opcode) {
403 reg_maps->luminanceparams = sampler_code;
408 if(WINED3DSIO_NRM == curOpcode->opcode) {
409 reg_maps->usesnrm = 1;
410 } else if(WINED3DSIO_BEM == curOpcode->opcode) {
411 DWORD regnum = *pToken & WINED3DSP_REGNUM_MASK;
412 if(reg_maps->bumpmat != -1 && reg_maps->bumpmat != regnum) {
413 FIXME("Pixel shader uses bem or texbem instruction on more than 1 sampler\n");
415 reg_maps->bumpmat = regnum;
417 } else if(WINED3DSIO_DSY == curOpcode->opcode) {
418 reg_maps->usesdsy = 1;
421 /* This will loop over all the registers and try to
422 * make a bitmask of the ones we're interested in.
424 * Relative addressing tokens are ignored, but that's
425 * okay, since we'll catch any address registers when
426 * they are initialized (required by spec) */
428 limit = (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED)?
429 curOpcode->num_params + 1: curOpcode->num_params;
431 for (i = 0; i < limit; ++i) {
433 DWORD param, addr_token, reg, regtype;
434 pToken += shader_get_param(iface, pToken, ¶m, &addr_token);
436 regtype = shader_get_regtype(param);
437 reg = param & WINED3DSP_REGNUM_MASK;
439 if (WINED3DSPR_TEXTURE == regtype) { /* vs: WINED3DSPR_ADDR */
442 reg_maps->texcoord[reg] = 1;
444 reg_maps->address[reg] = 1;
447 else if (WINED3DSPR_TEMP == regtype)
448 reg_maps->temporary[reg] = 1;
450 else if (WINED3DSPR_INPUT == regtype) {
452 reg_maps->attributes[reg] = 1;
454 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
455 /* If relative addressing is used, we must assume that all registers
456 * are used. Even if it is a construct like v3[aL], we can't assume
457 * that v0, v1 and v2 aren't read because aL can be negative
460 for(i = 0; i < MAX_REG_INPUT; i++) {
461 ((IWineD3DPixelShaderImpl *) This)->input_reg_used[i] = TRUE;
464 ((IWineD3DPixelShaderImpl *) This)->input_reg_used[reg] = TRUE;
469 else if (WINED3DSPR_RASTOUT == regtype && reg == 1)
472 else if (WINED3DSPR_MISCTYPE == regtype && reg == 0 && pshader)
475 else if(WINED3DSPR_CONST == regtype) {
476 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
478 if(reg <= ((IWineD3DVertexShaderImpl *) This)->min_rel_offset) {
479 ((IWineD3DVertexShaderImpl *) This)->min_rel_offset = reg;
480 } else if(reg >= ((IWineD3DVertexShaderImpl *) This)->max_rel_offset) {
481 ((IWineD3DVertexShaderImpl *) This)->max_rel_offset = reg;
484 reg_maps->usesrelconstF = TRUE;
488 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and is used
489 * in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel shaders because TECRDOUT
490 * isn't used in them, but future register types might cause issues
492 else if(WINED3DSPR_TEXCRDOUT == regtype && i == 0 /* Only look at writes */ &&
493 !pshader && WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) < 3) {
494 reg_maps->texcoord_mask[reg] |= shader_get_writemask(param);
499 reg_maps->loop_depth = max_loop_depth;
504 static void shader_dump_decl_usage(
505 IWineD3DBaseShaderImpl* This,
509 DWORD regtype = shader_get_regtype(param);
513 if (regtype == WINED3DSPR_SAMPLER) {
514 DWORD ttype = decl & WINED3DSP_TEXTURETYPE_MASK;
517 case WINED3DSTT_2D: TRACE("_2d"); break;
518 case WINED3DSTT_CUBE: TRACE("_cube"); break;
519 case WINED3DSTT_VOLUME: TRACE("_volume"); break;
520 default: TRACE("_unknown_ttype(%08x)", ttype);
525 DWORD usage = decl & WINED3DSP_DCL_USAGE_MASK;
526 DWORD idx = (decl & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
528 /* Pixel shaders 3.0 don't have usage semantics */
529 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
530 if (pshader && This->baseShader.hex_version < WINED3DPS_VERSION(3,0))
536 case WINED3DDECLUSAGE_POSITION:
537 TRACE("position%d", idx);
539 case WINED3DDECLUSAGE_BLENDINDICES:
542 case WINED3DDECLUSAGE_BLENDWEIGHT:
545 case WINED3DDECLUSAGE_NORMAL:
546 TRACE("normal%d", idx);
548 case WINED3DDECLUSAGE_PSIZE:
551 case WINED3DDECLUSAGE_COLOR:
555 TRACE("specular%d", (idx - 1));
558 case WINED3DDECLUSAGE_TEXCOORD:
559 TRACE("texture%d", idx);
561 case WINED3DDECLUSAGE_TANGENT:
564 case WINED3DDECLUSAGE_BINORMAL:
567 case WINED3DDECLUSAGE_TESSFACTOR:
570 case WINED3DDECLUSAGE_POSITIONT:
571 TRACE("positionT%d", idx);
573 case WINED3DDECLUSAGE_FOG:
576 case WINED3DDECLUSAGE_DEPTH:
579 case WINED3DDECLUSAGE_SAMPLE:
583 FIXME("unknown_semantics(%08x)", usage);
588 static void shader_dump_arr_entry(
589 IWineD3DBaseShader *iface,
591 const DWORD addr_token,
596 ((param & WINED3DSHADER_ADDRESSMODE_MASK) == WINED3DSHADER_ADDRMODE_RELATIVE);
601 shader_dump_param(iface, addr_token, 0, input);
611 void shader_dump_param(
612 IWineD3DBaseShader *iface,
614 const DWORD addr_token,
617 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
618 static const char * const rastout_reg_names[] = { "oPos", "oFog", "oPts" };
619 static const char * const misctype_reg_names[] = { "vPos", "vFace"};
620 char swizzle_reg_chars[4];
622 DWORD reg = param & WINED3DSP_REGNUM_MASK;
623 DWORD regtype = shader_get_regtype(param);
624 DWORD modifier = param & WINED3DSP_SRCMOD_MASK;
626 /* There are some minor differences between pixel and vertex shaders */
627 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
629 /* For one, we'd prefer color components to be shown for pshaders.
630 * FIXME: use the swizzle function for this */
632 swizzle_reg_chars[0] = pshader? 'r': 'x';
633 swizzle_reg_chars[1] = pshader? 'g': 'y';
634 swizzle_reg_chars[2] = pshader? 'b': 'z';
635 swizzle_reg_chars[3] = pshader? 'a': 'w';
638 if ( (modifier == WINED3DSPSM_NEG) ||
639 (modifier == WINED3DSPSM_BIASNEG) ||
640 (modifier == WINED3DSPSM_SIGNNEG) ||
641 (modifier == WINED3DSPSM_X2NEG) ||
642 (modifier == WINED3DSPSM_ABSNEG) )
644 else if (modifier == WINED3DSPSM_COMP)
646 else if (modifier == WINED3DSPSM_NOT)
649 if (modifier == WINED3DSPSM_ABS || modifier == WINED3DSPSM_ABSNEG)
654 case WINED3DSPR_TEMP:
657 case WINED3DSPR_INPUT:
659 shader_dump_arr_entry(iface, param, addr_token, reg, input);
661 case WINED3DSPR_CONST:
662 case WINED3DSPR_CONST2:
663 case WINED3DSPR_CONST3:
664 case WINED3DSPR_CONST4:
666 shader_dump_arr_entry(iface, param, addr_token, shader_get_float_offset(param), input);
668 case WINED3DSPR_TEXTURE: /* vs: case D3DSPR_ADDR */
669 TRACE("%c%u", (pshader? 't':'a'), reg);
671 case WINED3DSPR_RASTOUT:
672 TRACE("%s", rastout_reg_names[reg]);
674 case WINED3DSPR_COLOROUT:
677 case WINED3DSPR_DEPTHOUT:
680 case WINED3DSPR_ATTROUT:
683 case WINED3DSPR_TEXCRDOUT:
685 /* Vertex shaders >= 3.0 use general purpose output registers
686 * (WINED3DSPR_OUTPUT), which can include an address token */
688 if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3) {
690 shader_dump_arr_entry(iface, param, addr_token, reg, input);
695 case WINED3DSPR_CONSTINT:
697 shader_dump_arr_entry(iface, param, addr_token, reg, input);
699 case WINED3DSPR_CONSTBOOL:
701 shader_dump_arr_entry(iface, param, addr_token, reg, input);
703 case WINED3DSPR_LABEL:
706 case WINED3DSPR_LOOP:
709 case WINED3DSPR_SAMPLER:
712 case WINED3DSPR_MISCTYPE:
714 FIXME("Unhandled misctype register %d\n", reg);
716 TRACE("%s", misctype_reg_names[reg]);
719 case WINED3DSPR_PREDICATE:
723 TRACE("unhandled_rtype(%#x)", regtype);
728 /* operand output (for modifiers and shift, see dump_ins_modifiers) */
730 if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
732 if (param & WINED3DSP_WRITEMASK_0) TRACE("%c", swizzle_reg_chars[0]);
733 if (param & WINED3DSP_WRITEMASK_1) TRACE("%c", swizzle_reg_chars[1]);
734 if (param & WINED3DSP_WRITEMASK_2) TRACE("%c", swizzle_reg_chars[2]);
735 if (param & WINED3DSP_WRITEMASK_3) TRACE("%c", swizzle_reg_chars[3]);
740 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
741 DWORD swizzle_r = swizzle & 0x03;
742 DWORD swizzle_g = (swizzle >> 2) & 0x03;
743 DWORD swizzle_b = (swizzle >> 4) & 0x03;
744 DWORD swizzle_a = (swizzle >> 6) & 0x03;
748 case WINED3DSPSM_NONE: break;
749 case WINED3DSPSM_NEG: break;
750 case WINED3DSPSM_NOT: break;
751 case WINED3DSPSM_BIAS: TRACE("_bias"); break;
752 case WINED3DSPSM_BIASNEG: TRACE("_bias"); break;
753 case WINED3DSPSM_SIGN: TRACE("_bx2"); break;
754 case WINED3DSPSM_SIGNNEG: TRACE("_bx2"); break;
755 case WINED3DSPSM_COMP: break;
756 case WINED3DSPSM_X2: TRACE("_x2"); break;
757 case WINED3DSPSM_X2NEG: TRACE("_x2"); break;
758 case WINED3DSPSM_DZ: TRACE("_dz"); break;
759 case WINED3DSPSM_DW: TRACE("_dw"); break;
760 case WINED3DSPSM_ABSNEG: TRACE(")"); break;
761 case WINED3DSPSM_ABS: TRACE(")"); break;
763 TRACE("_unknown_modifier(%#x)", modifier >> WINED3DSP_SRCMOD_SHIFT);
768 * swizzle bits fields:
771 if ((WINED3DVS_NOSWIZZLE >> WINED3DVS_SWIZZLE_SHIFT) != swizzle) {
772 if (swizzle_r == swizzle_g &&
773 swizzle_r == swizzle_b &&
774 swizzle_r == swizzle_a) {
775 TRACE(".%c", swizzle_reg_chars[swizzle_r]);
778 swizzle_reg_chars[swizzle_r],
779 swizzle_reg_chars[swizzle_g],
780 swizzle_reg_chars[swizzle_b],
781 swizzle_reg_chars[swizzle_a]);
787 /** Shared code in order to generate the bulk of the shader string.
788 Use the shader_header_fct & shader_footer_fct to add strings
789 that are specific to pixel or vertex functions
790 NOTE: A description of how to parse tokens can be found at:
791 http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/usermodedisplaydriver_shader_cc8e4e05-f5c3-4ec0-8853-8ce07c1551b2.xml.asp */
792 void shader_generate_main(
793 IWineD3DBaseShader *iface,
794 SHADER_BUFFER* buffer,
795 shader_reg_maps* reg_maps,
796 CONST DWORD* pFunction) {
798 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
799 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device; /* To access shader backend callbacks */
800 const DWORD *pToken = pFunction;
801 const SHADER_OPCODE *curOpcode = NULL;
802 SHADER_HANDLER hw_fct = NULL;
804 SHADER_OPCODE_ARG hw_arg;
806 /* Initialize current parsing state */
807 hw_arg.shader = iface;
808 hw_arg.buffer = buffer;
809 hw_arg.reg_maps = reg_maps;
810 This->baseShader.parse_state.current_row = 0;
812 /* Second pass, process opcodes */
813 if (NULL != pToken) {
814 while (WINED3DPS_END() != *pToken) {
816 /* Skip version token */
817 if (shader_is_version_token(*pToken)) {
822 /* Skip comment tokens */
823 if (shader_is_comment(*pToken)) {
824 DWORD comment_len = (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
826 TRACE("#%s\n", (const char*)pToken);
827 pToken += comment_len;
832 hw_arg.opcode_token = *pToken++;
833 curOpcode = shader_get_opcode(iface, hw_arg.opcode_token);
836 if (curOpcode == NULL)
838 else if (This->baseShader.shader_mode == SHADER_GLSL)
839 hw_fct = curOpcode->hw_glsl_fct;
840 else if (This->baseShader.shader_mode == SHADER_ARB)
841 hw_fct = curOpcode->hw_fct;
843 /* Unknown opcode and its parameters */
844 if (NULL == curOpcode) {
845 FIXME("Unrecognized opcode: token=%08x\n", hw_arg.opcode_token);
846 pToken += shader_skip_unrecognized(iface, pToken);
849 } else if (WINED3DSIO_DCL == curOpcode->opcode ||
850 WINED3DSIO_NOP == curOpcode->opcode ||
851 WINED3DSIO_DEF == curOpcode->opcode ||
852 WINED3DSIO_DEFI == curOpcode->opcode ||
853 WINED3DSIO_DEFB == curOpcode->opcode ||
854 WINED3DSIO_PHASE == curOpcode->opcode ||
855 WINED3DSIO_RET == curOpcode->opcode) {
857 pToken += shader_skip_opcode(This, curOpcode, hw_arg.opcode_token);
859 /* If a generator function is set for current shader target, use it */
860 } else if (hw_fct != NULL) {
862 hw_arg.opcode = curOpcode;
864 /* Destination token */
865 if (curOpcode->dst_token) {
867 DWORD param, addr_token = 0;
868 pToken += shader_get_param(iface, pToken, ¶m, &addr_token);
870 hw_arg.dst_addr = addr_token;
873 /* Predication token */
874 if (hw_arg.opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED)
875 hw_arg.predicate = *pToken++;
877 /* Other source tokens */
878 for (i = 0; i < (curOpcode->num_params - curOpcode->dst_token); i++) {
880 DWORD param, addr_token = 0;
881 pToken += shader_get_param(iface, pToken, ¶m, &addr_token);
882 hw_arg.src[i] = param;
883 hw_arg.src_addr[i] = addr_token;
886 /* Call appropriate function for output target */
889 /* Add color correction if needed */
890 device->shader_backend->shader_color_correction(&hw_arg);
892 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
893 if (This->baseShader.shader_mode == SHADER_GLSL)
894 shader_glsl_add_instruction_modifiers(&hw_arg);
896 /* Unhandled opcode */
899 FIXME("Can't handle opcode %s in hwShader\n", curOpcode->name);
900 pToken += shader_skip_opcode(This, curOpcode, hw_arg.opcode_token);
903 /* TODO: What about result.depth? */
908 void shader_dump_ins_modifiers(const DWORD output) {
910 DWORD shift = (output & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
911 DWORD mmask = output & WINED3DSP_DSTMOD_MASK;
915 case 13: TRACE("_d8"); break;
916 case 14: TRACE("_d4"); break;
917 case 15: TRACE("_d2"); break;
918 case 1: TRACE("_x2"); break;
919 case 2: TRACE("_x4"); break;
920 case 3: TRACE("_x8"); break;
921 default: TRACE("_unhandled_shift(%d)", shift); break;
924 if (mmask & WINED3DSPDM_SATURATE) TRACE("_sat");
925 if (mmask & WINED3DSPDM_PARTIALPRECISION) TRACE("_pp");
926 if (mmask & WINED3DSPDM_MSAMPCENTROID) TRACE("_centroid");
928 mmask &= ~(WINED3DSPDM_SATURATE | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_MSAMPCENTROID);
930 FIXME("_unrecognized_modifier(%#x)", mmask >> WINED3DSP_DSTMOD_SHIFT);
933 /* First pass: trace shader, initialize length and version */
934 void shader_trace_init(
935 IWineD3DBaseShader *iface,
936 const DWORD* pFunction) {
938 IWineD3DBaseShaderImpl *This =(IWineD3DBaseShaderImpl *)iface;
940 const DWORD* pToken = pFunction;
941 const SHADER_OPCODE* curOpcode = NULL;
943 unsigned int len = 0;
946 TRACE("(%p) : Parsing programme\n", This);
948 if (NULL != pToken) {
949 while (WINED3DVS_END() != *pToken) {
950 if (shader_is_version_token(*pToken)) { /** version */
951 This->baseShader.hex_version = *pToken;
952 TRACE("%s_%u_%u\n", shader_is_pshader_version(This->baseShader.hex_version)? "ps": "vs",
953 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version),
954 WINED3DSHADER_VERSION_MINOR(This->baseShader.hex_version));
959 if (shader_is_comment(*pToken)) { /** comment */
960 DWORD comment_len = (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
962 TRACE("//%s\n", (const char*)pToken);
963 pToken += comment_len;
964 len += comment_len + 1;
967 opcode_token = *pToken++;
968 curOpcode = shader_get_opcode(iface, opcode_token);
971 if (NULL == curOpcode) {
973 FIXME("Unrecognized opcode: token=%08x\n", opcode_token);
974 tokens_read = shader_skip_unrecognized(iface, pToken);
975 pToken += tokens_read;
979 if (curOpcode->opcode == WINED3DSIO_DCL) {
981 DWORD usage = *pToken;
982 DWORD param = *(pToken + 1);
984 shader_dump_decl_usage(This, usage, param);
985 shader_dump_ins_modifiers(param);
987 shader_dump_param(iface, param, 0, 0);
991 } else if (curOpcode->opcode == WINED3DSIO_DEF) {
993 unsigned int offset = shader_get_float_offset(*pToken);
995 TRACE("def c%u = %f, %f, %f, %f", offset,
996 *(const float *)(pToken + 1),
997 *(const float *)(pToken + 2),
998 *(const float *)(pToken + 3),
999 *(const float *)(pToken + 4));
1003 } else if (curOpcode->opcode == WINED3DSIO_DEFI) {
1005 TRACE("defi i%u = %d, %d, %d, %d", *pToken & WINED3DSP_REGNUM_MASK,
1014 } else if (curOpcode->opcode == WINED3DSIO_DEFB) {
1016 TRACE("defb b%u = %s", *pToken & WINED3DSP_REGNUM_MASK,
1017 *(pToken + 1)? "true": "false");
1024 DWORD param, addr_token;
1027 /* Print out predication source token first - it follows
1028 * the destination token. */
1029 if (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED) {
1031 shader_dump_param(iface, *(pToken + 2), 0, 1);
1034 if (opcode_token & WINED3DSI_COISSUE) {
1035 /* PixWin marks instructions with the coissue flag with a '+' */
1039 TRACE("%s", curOpcode->name);
1041 if (curOpcode->opcode == WINED3DSIO_IFC ||
1042 curOpcode->opcode == WINED3DSIO_BREAKC) {
1044 DWORD op = (opcode_token & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
1046 case COMPARISON_GT: TRACE("_gt"); break;
1047 case COMPARISON_EQ: TRACE("_eq"); break;
1048 case COMPARISON_GE: TRACE("_ge"); break;
1049 case COMPARISON_LT: TRACE("_lt"); break;
1050 case COMPARISON_NE: TRACE("_ne"); break;
1051 case COMPARISON_LE: TRACE("_le"); break;
1055 } else if (curOpcode->opcode == WINED3DSIO_TEX &&
1056 This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
1057 if(opcode_token & WINED3DSI_TEXLD_PROJECT) TRACE("p");
1060 /* Destination token */
1061 if (curOpcode->dst_token) {
1063 /* Destination token */
1064 tokens_read = shader_get_param(iface, pToken, ¶m, &addr_token);
1065 pToken += tokens_read;
1068 shader_dump_ins_modifiers(param);
1070 shader_dump_param(iface, param, addr_token, 0);
1073 /* Predication token - already printed out, just skip it */
1074 if (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED) {
1079 /* Other source tokens */
1080 for (i = curOpcode->dst_token; i < curOpcode->num_params; ++i) {
1082 tokens_read = shader_get_param(iface, pToken, ¶m, &addr_token);
1083 pToken += tokens_read;
1086 TRACE((i == 0)? " " : ", ");
1087 shader_dump_param(iface, param, addr_token, 1);
1093 This->baseShader.functionLength = (len + 1) * sizeof(DWORD);
1095 This->baseShader.functionLength = 1; /* no Function defined use fixed function vertex processing */
1099 static void shader_none_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {}
1100 static void shader_none_select_depth_blt(IWineD3DDevice *iface) {}
1101 static void shader_none_destroy_depth_blt(IWineD3DDevice *iface) {}
1102 static void shader_none_load_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1103 static void shader_none_cleanup(IWineD3DDevice *iface) {}
1104 static void shader_none_color_correction(SHADER_OPCODE_ARG* arg) {}
1105 static void shader_none_destroy(IWineD3DBaseShader *iface) {}
1106 static HRESULT shader_none_alloc(IWineD3DDevice *iface) {return WINED3D_OK;}
1107 static void shader_none_free(IWineD3DDevice *iface) {}
1109 const shader_backend_t none_shader_backend = {
1110 &shader_none_select,
1111 &shader_none_select_depth_blt,
1112 &shader_none_destroy_depth_blt,
1113 &shader_none_load_constants,
1114 &shader_none_cleanup,
1115 &shader_none_color_correction,
1116 &shader_none_destroy,
1121 /* *******************************************
1122 IWineD3DPixelShader IUnknown parts follow
1123 ******************************************* */
1124 HRESULT WINAPI IWineD3DBaseShaderImpl_QueryInterface(IWineD3DBaseShader *iface, REFIID riid, LPVOID *ppobj)
1126 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)iface;
1127 TRACE("(%p)->(%s,%p)\n",This,debugstr_guid(riid),ppobj);
1128 if (IsEqualGUID(riid, &IID_IUnknown)
1129 || IsEqualGUID(riid, &IID_IWineD3DBase)
1130 || IsEqualGUID(riid, &IID_IWineD3DBaseShader)
1131 || IsEqualGUID(riid, &IID_IWineD3DPixelShader)) {
1132 IUnknown_AddRef(iface);
1137 return E_NOINTERFACE;
1140 ULONG WINAPI IWineD3DBaseShaderImpl_AddRef(IWineD3DBaseShader *iface) {
1141 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1142 TRACE("(%p) : AddRef increasing from %d\n", This, This->baseShader.ref);
1143 return InterlockedIncrement(&This->baseShader.ref);
1146 ULONG WINAPI IWineD3DBaseShaderImpl_Release(IWineD3DBaseShader *iface) {
1147 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)iface;
1148 IWineD3DDeviceImpl *deviceImpl = (IWineD3DDeviceImpl *) This->baseShader.device;
1150 TRACE("(%p) : Releasing from %d\n", This, This->baseShader.ref);
1151 ref = InterlockedDecrement(&This->baseShader.ref);
1153 deviceImpl->shader_backend->shader_destroy(iface);
1154 HeapFree(GetProcessHeap(), 0, This->baseShader.function);
1155 shader_delete_constant_list(&This->baseShader.constantsF);
1156 shader_delete_constant_list(&This->baseShader.constantsB);
1157 shader_delete_constant_list(&This->baseShader.constantsI);
1158 list_remove(&This->baseShader.shader_list_entry);
1159 HeapFree(GetProcessHeap(), 0, This);