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
9 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
10 * Copyright 2009 Henri Verbeet for CodeWeavers
12 * This library is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * This library is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with this library; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
30 #include "wined3d_private.h"
32 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
33 WINE_DECLARE_DEBUG_CHANNEL(d3d);
35 static void shader_dump_param(const DWORD param, const DWORD addr_token, int input, DWORD shader_version);
37 static inline BOOL shader_is_version_token(DWORD token) {
38 return shader_is_pshader_version(token) ||
39 shader_is_vshader_version(token);
42 void shader_buffer_init(struct SHADER_BUFFER *buffer)
44 buffer->buffer = HeapAlloc(GetProcessHeap(), 0, SHADER_PGMSIZE);
45 buffer->buffer[0] = '\0';
48 buffer->newline = TRUE;
51 void shader_buffer_free(struct SHADER_BUFFER *buffer)
53 HeapFree(GetProcessHeap(), 0, buffer->buffer);
56 int shader_vaddline(SHADER_BUFFER* buffer, const char *format, va_list args)
58 char* base = buffer->buffer + buffer->bsize;
61 rc = vsnprintf(base, SHADER_PGMSIZE - 1 - buffer->bsize, format, args);
63 if (rc < 0 || /* C89 */
64 rc > SHADER_PGMSIZE - 1 - buffer->bsize) { /* C99 */
66 ERR("The buffer allocated for the shader program string "
67 "is too small at %d bytes.\n", SHADER_PGMSIZE);
68 buffer->bsize = SHADER_PGMSIZE - 1;
72 if (buffer->newline) {
73 TRACE("GL HW (%u, %u) : %s", buffer->lineNo + 1, buffer->bsize, base);
74 buffer->newline = FALSE;
80 if (buffer->buffer[buffer->bsize-1] == '\n') {
82 buffer->newline = TRUE;
87 int shader_addline(SHADER_BUFFER* buffer, const char *format, ...)
92 va_start(args, format);
93 ret = shader_vaddline(buffer, format, args);
99 void shader_init(struct IWineD3DBaseShaderClass *shader,
100 IWineD3DDevice *device, const SHADER_OPCODE *instruction_table)
103 shader->device = device;
104 shader->shader_ins = instruction_table;
105 list_init(&shader->linked_programs);
108 const SHADER_OPCODE *shader_get_opcode(const SHADER_OPCODE *opcode_table, DWORD shader_version, DWORD code)
112 /** TODO: use dichotomic search */
113 while (opcode_table[i].name)
115 if ((code & WINED3DSI_OPCODE_MASK) == opcode_table[i].opcode
116 && shader_version >= opcode_table[i].min_version
117 && (!opcode_table[i].max_version || shader_version <= opcode_table[i].max_version))
119 return &opcode_table[i];
124 FIXME("Unsupported opcode %#x(%d) masked %#x, shader version %#x\n",
125 code, code, code & WINED3DSI_OPCODE_MASK, shader_version);
130 /* Read a parameter opcode from the input stream,
131 * and possibly a relative addressing token.
132 * Return the number of tokens read */
133 static int shader_get_param(const DWORD *pToken, DWORD shader_version, DWORD *param, DWORD *addr_token)
139 /* PS >= 3.0 have relative addressing (with token)
140 * VS >= 2.0 have relative addressing (with token)
141 * VS >= 1.0 < 2.0 have relative addressing (without token)
142 * The version check below should work in general */
143 if (*pToken & WINED3DSHADER_ADDRMODE_RELATIVE)
145 if (WINED3DSHADER_VERSION_MAJOR(shader_version) < 2)
147 *addr_token = (1 << 31)
148 | ((WINED3DSPR_ADDR << WINED3DSP_REGTYPE_SHIFT2) & WINED3DSP_REGTYPE_MASK2)
149 | ((WINED3DSPR_ADDR << WINED3DSP_REGTYPE_SHIFT) & WINED3DSP_REGTYPE_MASK)
150 | (WINED3DSP_NOSWIZZLE << WINED3DSP_SWIZZLE_SHIFT);
154 *addr_token = *(pToken + 1);
162 /* Return the number of parameters to skip for an opcode */
163 static inline int shader_skip_opcode(const SHADER_OPCODE *curOpcode, DWORD opcode_token, DWORD shader_version)
165 /* Shaders >= 2.0 may contain address tokens, but fortunately they
166 * have a useful length mask - use it here. Shaders 1.0 contain no such tokens */
167 return (WINED3DSHADER_VERSION_MAJOR(shader_version) >= 2)
168 ? ((opcode_token & WINED3DSI_INSTLENGTH_MASK) >> WINED3DSI_INSTLENGTH_SHIFT) : curOpcode->num_params;
171 /* Read the parameters of an unrecognized opcode from the input stream
172 * Return the number of tokens read.
174 * Note: This function assumes source or destination token format.
175 * It will not work with specially-formatted tokens like DEF or DCL,
176 * but hopefully those would be recognized */
177 static int shader_skip_unrecognized(const DWORD *pToken, DWORD shader_version)
182 /* TODO: Think of a good name for 0x80000000 and replace it with a constant */
183 while (*pToken & 0x80000000) {
185 DWORD param, addr_token = 0;
186 tokens_read += shader_get_param(pToken, shader_version, ¶m, &addr_token);
187 pToken += tokens_read;
189 FIXME("Unrecognized opcode param: token=0x%08x "
190 "addr_token=0x%08x name=", param, addr_token);
191 shader_dump_param(param, addr_token, i, shader_version);
198 /* Convert floating point offset relative
199 * to a register file to an absolute offset for float constants */
200 static unsigned int shader_get_float_offset(const DWORD reg)
202 unsigned int regnum = reg & WINED3DSP_REGNUM_MASK;
203 int regtype = shader_get_regtype(reg);
206 case WINED3DSPR_CONST: return regnum;
207 case WINED3DSPR_CONST2: return 2048 + regnum;
208 case WINED3DSPR_CONST3: return 4096 + regnum;
209 case WINED3DSPR_CONST4: return 6144 + regnum;
211 FIXME("Unsupported register type: %d\n", regtype);
216 static void shader_delete_constant_list(struct list* clist) {
219 struct local_constant* constant;
221 ptr = list_head(clist);
223 constant = LIST_ENTRY(ptr, struct local_constant, entry);
224 ptr = list_next(clist, ptr);
225 HeapFree(GetProcessHeap(), 0, constant);
230 static void shader_parse_dst_param(DWORD param, const struct wined3d_shader_src_param *rel_addr,
231 struct wined3d_shader_dst_param *dst)
233 dst->register_type = ((param & WINED3DSP_REGTYPE_MASK) >> WINED3DSP_REGTYPE_SHIFT)
234 | ((param & WINED3DSP_REGTYPE_MASK2) >> WINED3DSP_REGTYPE_SHIFT2);
235 dst->register_idx = param & WINED3DSP_REGNUM_MASK;
236 dst->write_mask = param & WINED3DSP_WRITEMASK_ALL;
237 dst->modifiers = param & WINED3DSP_DSTMOD_MASK;
238 dst->shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
239 dst->rel_addr = rel_addr;
242 static void shader_parse_src_param(DWORD param, const struct wined3d_shader_src_param *rel_addr,
243 struct wined3d_shader_src_param *src)
245 src->register_type = ((param & WINED3DSP_REGTYPE_MASK) >> WINED3DSP_REGTYPE_SHIFT)
246 | ((param & WINED3DSP_REGTYPE_MASK2) >> WINED3DSP_REGTYPE_SHIFT2);
247 src->register_idx = param & WINED3DSP_REGNUM_MASK;
248 src->swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
249 src->modifiers = (param & WINED3DSP_SRCMOD_MASK) >> WINED3DSP_SRCMOD_SHIFT;
250 src->rel_addr = rel_addr;
253 /* Note that this does not count the loop register
254 * as an address register. */
256 HRESULT shader_get_registers_used(IWineD3DBaseShader *iface, struct shader_reg_maps *reg_maps,
257 struct wined3d_shader_semantic *semantics_in, struct wined3d_shader_semantic *semantics_out,
258 const DWORD *byte_code)
260 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
261 const SHADER_OPCODE *shader_ins = This->baseShader.shader_ins;
262 DWORD shader_version;
263 unsigned int cur_loop_depth = 0, max_loop_depth = 0;
264 const DWORD* pToken = byte_code;
267 /* There are some minor differences between pixel and vertex shaders */
269 memset(reg_maps->bumpmat, 0, sizeof(reg_maps->bumpmat));
270 memset(reg_maps->luminanceparams, 0, sizeof(reg_maps->luminanceparams));
272 /* get_registers_used is called on every compile on some 1.x shaders, which can result
273 * in stacking up a collection of local constants. Delete the old constants if existing
275 shader_delete_constant_list(&This->baseShader.constantsF);
276 shader_delete_constant_list(&This->baseShader.constantsB);
277 shader_delete_constant_list(&This->baseShader.constantsI);
279 /* The version token is supposed to be the first token */
280 if (!shader_is_version_token(*pToken))
282 FIXME("First token is not a version token, invalid shader.\n");
283 return WINED3DERR_INVALIDCALL;
285 reg_maps->shader_version = shader_version = *pToken++;
286 pshader = shader_is_pshader_version(shader_version);
288 while (WINED3DVS_END() != *pToken) {
289 CONST SHADER_OPCODE* curOpcode;
293 if (shader_is_comment(*pToken))
295 DWORD comment_len = (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
297 pToken += comment_len;
302 opcode_token = *pToken++;
303 curOpcode = shader_get_opcode(shader_ins, shader_version, opcode_token);
305 /* Unhandled opcode, and its parameters */
306 if (NULL == curOpcode) {
307 while (*pToken & 0x80000000)
310 /* Handle declarations */
311 } else if (WINED3DSIO_DCL == curOpcode->opcode) {
313 DWORD usage = *pToken++;
314 DWORD param = *pToken++;
315 DWORD regtype = shader_get_regtype(param);
316 unsigned int regnum = param & WINED3DSP_REGNUM_MASK;
318 /* Vshader: mark attributes used
319 Pshader: mark 3.0 input registers used, save token */
320 if (WINED3DSPR_INPUT == regtype) {
323 reg_maps->attributes[regnum] = 1;
325 reg_maps->packed_input[regnum] = 1;
327 semantics_in[regnum].usage = (usage & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
328 semantics_in[regnum].usage_idx =
329 (usage & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
330 shader_parse_dst_param(param, NULL, &semantics_in[regnum].reg);
332 /* Vshader: mark 3.0 output registers used, save token */
333 } else if (WINED3DSPR_OUTPUT == regtype) {
334 reg_maps->packed_output[regnum] = 1;
335 semantics_out[regnum].usage = (usage & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
336 semantics_out[regnum].usage_idx =
337 (usage & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
338 shader_parse_dst_param(param, NULL, &semantics_out[regnum].reg);
340 if (usage & (WINED3DDECLUSAGE_FOG << WINED3DSP_DCL_USAGE_SHIFT))
343 /* Save sampler usage token */
344 } else if (WINED3DSPR_SAMPLER == regtype)
345 reg_maps->samplers[regnum] = usage;
347 } else if (WINED3DSIO_DEF == curOpcode->opcode) {
349 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
350 if (!lconst) return E_OUTOFMEMORY;
351 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
352 memcpy(lconst->value, pToken + 1, 4 * sizeof(DWORD));
354 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
355 if (WINED3DSHADER_VERSION_MAJOR(shader_version) == 1 && pshader)
357 float *value = (float *) lconst->value;
358 if(value[0] < -1.0) value[0] = -1.0;
359 else if(value[0] > 1.0) value[0] = 1.0;
360 if(value[1] < -1.0) value[1] = -1.0;
361 else if(value[1] > 1.0) value[1] = 1.0;
362 if(value[2] < -1.0) value[2] = -1.0;
363 else if(value[2] > 1.0) value[2] = 1.0;
364 if(value[3] < -1.0) value[3] = -1.0;
365 else if(value[3] > 1.0) value[3] = 1.0;
368 list_add_head(&This->baseShader.constantsF, &lconst->entry);
369 pToken += curOpcode->num_params;
371 } else if (WINED3DSIO_DEFI == curOpcode->opcode) {
373 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
374 if (!lconst) return E_OUTOFMEMORY;
375 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
376 memcpy(lconst->value, pToken + 1, 4 * sizeof(DWORD));
377 list_add_head(&This->baseShader.constantsI, &lconst->entry);
378 pToken += curOpcode->num_params;
380 } else if (WINED3DSIO_DEFB == curOpcode->opcode) {
382 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
383 if (!lconst) return E_OUTOFMEMORY;
384 lconst->idx = *pToken & WINED3DSP_REGNUM_MASK;
385 memcpy(lconst->value, pToken + 1, 1 * sizeof(DWORD));
386 list_add_head(&This->baseShader.constantsB, &lconst->entry);
387 pToken += curOpcode->num_params;
389 /* If there's a loop in the shader */
390 } else if (WINED3DSIO_LOOP == curOpcode->opcode ||
391 WINED3DSIO_REP == curOpcode->opcode) {
393 if(cur_loop_depth > max_loop_depth)
394 max_loop_depth = cur_loop_depth;
395 pToken += curOpcode->num_params;
397 /* Rep and Loop always use an integer constant for the control parameters */
398 This->baseShader.uses_int_consts = TRUE;
399 } else if (WINED3DSIO_ENDLOOP == curOpcode->opcode ||
400 WINED3DSIO_ENDREP == curOpcode->opcode) {
403 /* For subroutine prototypes */
404 } else if (WINED3DSIO_LABEL == curOpcode->opcode) {
406 DWORD snum = *pToken & WINED3DSP_REGNUM_MASK;
407 reg_maps->labels[snum] = 1;
408 pToken += curOpcode->num_params;
410 /* Set texture, address, temporary registers */
414 /* Declare 1.X samplers implicitly, based on the destination reg. number */
415 if (WINED3DSHADER_VERSION_MAJOR(shader_version) == 1
416 && pshader /* Filter different instructions with the same enum values in VS */
417 && (WINED3DSIO_TEX == curOpcode->opcode
418 || WINED3DSIO_TEXBEM == curOpcode->opcode
419 || WINED3DSIO_TEXBEML == curOpcode->opcode
420 || WINED3DSIO_TEXDP3TEX == curOpcode->opcode
421 || WINED3DSIO_TEXM3x2TEX == curOpcode->opcode
422 || WINED3DSIO_TEXM3x3SPEC == curOpcode->opcode
423 || WINED3DSIO_TEXM3x3TEX == curOpcode->opcode
424 || WINED3DSIO_TEXM3x3VSPEC == curOpcode->opcode
425 || WINED3DSIO_TEXREG2AR == curOpcode->opcode
426 || WINED3DSIO_TEXREG2GB == curOpcode->opcode
427 || WINED3DSIO_TEXREG2RGB == curOpcode->opcode))
429 /* Fake sampler usage, only set reserved bit and ttype */
430 DWORD sampler_code = *pToken & WINED3DSP_REGNUM_MASK;
432 TRACE("Setting fake 2D sampler for 1.x pixelshader\n");
433 reg_maps->samplers[sampler_code] = (0x1 << 31) | WINED3DSTT_2D;
435 /* texbem is only valid with < 1.4 pixel shaders */
436 if(WINED3DSIO_TEXBEM == curOpcode->opcode ||
437 WINED3DSIO_TEXBEML == curOpcode->opcode) {
438 reg_maps->bumpmat[sampler_code] = TRUE;
439 if(WINED3DSIO_TEXBEML == curOpcode->opcode) {
440 reg_maps->luminanceparams[sampler_code] = TRUE;
444 if(WINED3DSIO_NRM == curOpcode->opcode) {
445 reg_maps->usesnrm = 1;
446 } else if(WINED3DSIO_BEM == curOpcode->opcode && pshader) {
447 DWORD regnum = *pToken & WINED3DSP_REGNUM_MASK;
448 reg_maps->bumpmat[regnum] = TRUE;
449 } else if(WINED3DSIO_DSY == curOpcode->opcode) {
450 reg_maps->usesdsy = 1;
453 /* This will loop over all the registers and try to
454 * make a bitmask of the ones we're interested in.
456 * Relative addressing tokens are ignored, but that's
457 * okay, since we'll catch any address registers when
458 * they are initialized (required by spec) */
460 limit = (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED)?
461 curOpcode->num_params + 1: curOpcode->num_params;
463 for (i = 0; i < limit; ++i) {
465 DWORD param, addr_token, reg, regtype;
466 pToken += shader_get_param(pToken, shader_version, ¶m, &addr_token);
468 regtype = shader_get_regtype(param);
469 reg = param & WINED3DSP_REGNUM_MASK;
471 if (WINED3DSPR_TEXTURE == regtype) { /* vs: WINED3DSPR_ADDR */
474 reg_maps->texcoord[reg] = 1;
476 reg_maps->address[reg] = 1;
479 else if (WINED3DSPR_TEMP == regtype)
480 reg_maps->temporary[reg] = 1;
482 else if (WINED3DSPR_INPUT == regtype) {
484 reg_maps->attributes[reg] = 1;
486 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
487 /* If relative addressing is used, we must assume that all registers
488 * are used. Even if it is a construct like v3[aL], we can't assume
489 * that v0, v1 and v2 aren't read because aL can be negative
492 for(i = 0; i < MAX_REG_INPUT; i++) {
493 ((IWineD3DPixelShaderImpl *) This)->input_reg_used[i] = TRUE;
496 ((IWineD3DPixelShaderImpl *) This)->input_reg_used[reg] = TRUE;
501 else if (WINED3DSPR_RASTOUT == regtype && reg == 1)
504 else if (WINED3DSPR_MISCTYPE == regtype && reg == 0 && pshader)
507 else if(WINED3DSPR_CONST == regtype) {
508 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
510 if(reg <= ((IWineD3DVertexShaderImpl *) This)->min_rel_offset) {
511 ((IWineD3DVertexShaderImpl *) This)->min_rel_offset = reg;
512 } else if(reg >= ((IWineD3DVertexShaderImpl *) This)->max_rel_offset) {
513 ((IWineD3DVertexShaderImpl *) This)->max_rel_offset = reg;
516 reg_maps->usesrelconstF = TRUE;
519 else if(WINED3DSPR_CONSTINT == regtype) {
520 This->baseShader.uses_int_consts = TRUE;
522 else if(WINED3DSPR_CONSTBOOL == regtype) {
523 This->baseShader.uses_bool_consts = TRUE;
526 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and is used
527 * in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel shaders because TECRDOUT
528 * isn't used in them, but future register types might cause issues
530 else if (WINED3DSPR_TEXCRDOUT == regtype && i == 0 /* Only look at writes */
531 && !pshader && WINED3DSHADER_VERSION_MAJOR(shader_version) < 3)
533 reg_maps->texcoord_mask[reg] |= shader_get_writemask(param);
539 reg_maps->loop_depth = max_loop_depth;
541 This->baseShader.functionLength = ((char *)pToken - (char *)byte_code);
546 static void shader_dump_decl_usage(DWORD decl, DWORD param, DWORD shader_version)
548 DWORD regtype = shader_get_regtype(param);
552 if (regtype == WINED3DSPR_SAMPLER) {
553 DWORD ttype = decl & WINED3DSP_TEXTURETYPE_MASK;
556 case WINED3DSTT_2D: TRACE("_2d"); break;
557 case WINED3DSTT_CUBE: TRACE("_cube"); break;
558 case WINED3DSTT_VOLUME: TRACE("_volume"); break;
559 default: TRACE("_unknown_ttype(0x%08x)", ttype);
564 DWORD usage = decl & WINED3DSP_DCL_USAGE_MASK;
565 DWORD idx = (decl & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
567 /* Pixel shaders 3.0 don't have usage semantics */
568 if (shader_is_pshader_version(shader_version) && shader_version < WINED3DPS_VERSION(3,0))
574 case WINED3DDECLUSAGE_POSITION:
575 TRACE("position%d", idx);
577 case WINED3DDECLUSAGE_BLENDINDICES:
580 case WINED3DDECLUSAGE_BLENDWEIGHT:
583 case WINED3DDECLUSAGE_NORMAL:
584 TRACE("normal%d", idx);
586 case WINED3DDECLUSAGE_PSIZE:
589 case WINED3DDECLUSAGE_COLOR:
593 TRACE("specular%d", (idx - 1));
596 case WINED3DDECLUSAGE_TEXCOORD:
597 TRACE("texture%d", idx);
599 case WINED3DDECLUSAGE_TANGENT:
602 case WINED3DDECLUSAGE_BINORMAL:
605 case WINED3DDECLUSAGE_TESSFACTOR:
608 case WINED3DDECLUSAGE_POSITIONT:
609 TRACE("positionT%d", idx);
611 case WINED3DDECLUSAGE_FOG:
614 case WINED3DDECLUSAGE_DEPTH:
617 case WINED3DDECLUSAGE_SAMPLE:
621 FIXME("unknown_semantics(0x%08x)", usage);
626 static void shader_dump_arr_entry(const DWORD param, const DWORD addr_token,
627 unsigned int reg, int input, DWORD shader_version)
630 ((param & WINED3DSHADER_ADDRESSMODE_MASK) == WINED3DSHADER_ADDRMODE_RELATIVE);
635 shader_dump_param(addr_token, 0, input, shader_version);
645 static void shader_dump_param(const DWORD param, const DWORD addr_token, int input, DWORD shader_version)
647 static const char * const rastout_reg_names[] = { "oPos", "oFog", "oPts" };
648 static const char * const misctype_reg_names[] = { "vPos", "vFace"};
649 const char *swizzle_reg_chars = "xyzw";
651 DWORD reg = param & WINED3DSP_REGNUM_MASK;
652 DWORD regtype = shader_get_regtype(param);
653 DWORD modifier = param & WINED3DSP_SRCMOD_MASK;
655 /* There are some minor differences between pixel and vertex shaders */
656 char pshader = shader_is_pshader_version(shader_version);
659 if ( (modifier == WINED3DSPSM_NEG) ||
660 (modifier == WINED3DSPSM_BIASNEG) ||
661 (modifier == WINED3DSPSM_SIGNNEG) ||
662 (modifier == WINED3DSPSM_X2NEG) ||
663 (modifier == WINED3DSPSM_ABSNEG) )
665 else if (modifier == WINED3DSPSM_COMP)
667 else if (modifier == WINED3DSPSM_NOT)
670 if (modifier == WINED3DSPSM_ABS || modifier == WINED3DSPSM_ABSNEG)
675 case WINED3DSPR_TEMP:
678 case WINED3DSPR_INPUT:
680 shader_dump_arr_entry(param, addr_token, reg, input, shader_version);
682 case WINED3DSPR_CONST:
683 case WINED3DSPR_CONST2:
684 case WINED3DSPR_CONST3:
685 case WINED3DSPR_CONST4:
687 shader_dump_arr_entry(param, addr_token, shader_get_float_offset(param), input, shader_version);
689 case WINED3DSPR_TEXTURE: /* vs: case D3DSPR_ADDR */
690 TRACE("%c%u", (pshader? 't':'a'), reg);
692 case WINED3DSPR_RASTOUT:
693 TRACE("%s", rastout_reg_names[reg]);
695 case WINED3DSPR_COLOROUT:
698 case WINED3DSPR_DEPTHOUT:
701 case WINED3DSPR_ATTROUT:
704 case WINED3DSPR_TEXCRDOUT:
706 /* Vertex shaders >= 3.0 use general purpose output registers
707 * (WINED3DSPR_OUTPUT), which can include an address token */
709 if (WINED3DSHADER_VERSION_MAJOR(shader_version) >= 3) {
711 shader_dump_arr_entry(param, addr_token, reg, input, shader_version);
716 case WINED3DSPR_CONSTINT:
718 shader_dump_arr_entry(param, addr_token, reg, input, shader_version);
720 case WINED3DSPR_CONSTBOOL:
722 shader_dump_arr_entry(param, addr_token, reg, input, shader_version);
724 case WINED3DSPR_LABEL:
727 case WINED3DSPR_LOOP:
730 case WINED3DSPR_SAMPLER:
733 case WINED3DSPR_MISCTYPE:
735 FIXME("Unhandled misctype register %d\n", reg);
737 TRACE("%s", misctype_reg_names[reg]);
740 case WINED3DSPR_PREDICATE:
744 TRACE("unhandled_rtype(%#x)", regtype);
749 /* operand output (for modifiers and shift, see dump_ins_modifiers) */
751 if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
753 if (param & WINED3DSP_WRITEMASK_0) TRACE("%c", swizzle_reg_chars[0]);
754 if (param & WINED3DSP_WRITEMASK_1) TRACE("%c", swizzle_reg_chars[1]);
755 if (param & WINED3DSP_WRITEMASK_2) TRACE("%c", swizzle_reg_chars[2]);
756 if (param & WINED3DSP_WRITEMASK_3) TRACE("%c", swizzle_reg_chars[3]);
761 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
762 DWORD swizzle_x = swizzle & 0x03;
763 DWORD swizzle_y = (swizzle >> 2) & 0x03;
764 DWORD swizzle_z = (swizzle >> 4) & 0x03;
765 DWORD swizzle_w = (swizzle >> 6) & 0x03;
769 case WINED3DSPSM_NONE: break;
770 case WINED3DSPSM_NEG: break;
771 case WINED3DSPSM_NOT: break;
772 case WINED3DSPSM_BIAS: TRACE("_bias"); break;
773 case WINED3DSPSM_BIASNEG: TRACE("_bias"); break;
774 case WINED3DSPSM_SIGN: TRACE("_bx2"); break;
775 case WINED3DSPSM_SIGNNEG: TRACE("_bx2"); break;
776 case WINED3DSPSM_COMP: break;
777 case WINED3DSPSM_X2: TRACE("_x2"); break;
778 case WINED3DSPSM_X2NEG: TRACE("_x2"); break;
779 case WINED3DSPSM_DZ: TRACE("_dz"); break;
780 case WINED3DSPSM_DW: TRACE("_dw"); break;
781 case WINED3DSPSM_ABSNEG: TRACE(")"); break;
782 case WINED3DSPSM_ABS: TRACE(")"); break;
784 TRACE("_unknown_modifier(%#x)", modifier);
789 * swizzle bits fields:
792 if (swizzle != WINED3DSP_NOSWIZZLE)
794 if (swizzle_x == swizzle_y &&
795 swizzle_x == swizzle_z &&
796 swizzle_x == swizzle_w) {
797 TRACE(".%c", swizzle_reg_chars[swizzle_x]);
800 swizzle_reg_chars[swizzle_x],
801 swizzle_reg_chars[swizzle_y],
802 swizzle_reg_chars[swizzle_z],
803 swizzle_reg_chars[swizzle_w]);
809 /* Shared code in order to generate the bulk of the shader string.
810 * NOTE: A description of how to parse tokens can be found on msdn */
811 void shader_generate_main(IWineD3DBaseShader *iface, SHADER_BUFFER* buffer,
812 const shader_reg_maps* reg_maps, CONST DWORD* pFunction)
814 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
815 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device; /* To access shader backend callbacks */
816 const SHADER_OPCODE *opcode_table = This->baseShader.shader_ins;
817 const SHADER_HANDLER *handler_table = device->shader_backend->shader_instruction_handler_table;
818 DWORD shader_version = reg_maps->shader_version;
819 struct wined3d_shader_src_param src_rel_addr[4];
820 struct wined3d_shader_src_param src_param[4];
821 struct wined3d_shader_src_param dst_rel_addr;
822 struct wined3d_shader_dst_param dst_param;
823 struct wined3d_shader_instruction ins;
824 struct wined3d_shader_context ctx;
825 const DWORD *pToken = pFunction;
826 const SHADER_OPCODE *curOpcode;
827 SHADER_HANDLER hw_fct;
830 /* Initialize current parsing state */
832 ctx.reg_maps = reg_maps;
836 ins.dst = &dst_param;
838 This->baseShader.parse_state.current_row = 0;
840 while (WINED3DPS_END() != *pToken)
844 /* Skip version token */
845 if (shader_is_version_token(*pToken))
851 /* Skip comment tokens */
852 if (shader_is_comment(*pToken))
854 pToken += (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
860 opcode_token = *pToken++;
861 curOpcode = shader_get_opcode(opcode_table, shader_version, opcode_token);
863 /* Unknown opcode and its parameters */
866 FIXME("Unrecognized opcode: token=0x%08x\n", opcode_token);
867 pToken += shader_skip_unrecognized(pToken, shader_version);
872 if (WINED3DSIO_DCL == curOpcode->opcode
873 || WINED3DSIO_NOP == curOpcode->opcode
874 || WINED3DSIO_DEF == curOpcode->opcode
875 || WINED3DSIO_DEFI == curOpcode->opcode
876 || WINED3DSIO_DEFB == curOpcode->opcode
877 || WINED3DSIO_PHASE == curOpcode->opcode
878 || WINED3DSIO_RET == curOpcode->opcode)
880 pToken += shader_skip_opcode(curOpcode, opcode_token, shader_version);
885 hw_fct = handler_table[curOpcode->handler_idx];
887 /* Unhandled opcode */
890 FIXME("Can't handle opcode %s in hwShader\n", curOpcode->name);
891 pToken += shader_skip_opcode(curOpcode, opcode_token, shader_version);
895 ins.handler_idx = curOpcode->handler_idx;
896 ins.flags = opcode_token & WINED3D_OPCODESPECIFICCONTROL_MASK;
897 ins.coissue = opcode_token & WINED3DSI_COISSUE;
899 /* Destination token */
900 ins.dst_count = curOpcode->dst_token ? 1 : 0;
903 DWORD param, addr_param;
904 pToken += shader_get_param(pToken, shader_version, ¶m, &addr_param);
906 if (param & WINED3DSHADER_ADDRMODE_RELATIVE)
908 shader_parse_src_param(addr_param, NULL, &dst_rel_addr);
909 shader_parse_dst_param(param, &dst_rel_addr, &dst_param);
913 shader_parse_dst_param(param, NULL, &dst_param);
917 /* Predication token */
918 if (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED) ins.predicate = *pToken++;
920 /* Other source tokens */
921 ins.src_count = curOpcode->num_params - curOpcode->dst_token;
922 for (i = 0; i < ins.src_count; ++i)
924 DWORD param, addr_param;
925 pToken += shader_get_param(pToken, shader_version, ¶m, &addr_param);
926 if (param & WINED3DSHADER_ADDRMODE_RELATIVE)
928 shader_parse_src_param(addr_param, NULL, &src_rel_addr[i]);
929 shader_parse_src_param(param, &src_rel_addr[i], &src_param[i]);
933 shader_parse_src_param(param, NULL, &src_param[i]);
937 /* Call appropriate function for output target */
940 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
941 /* FIXME: This should be internal to the shader backend.
942 * Also, right now this is the only reason "shader_mode" exists. */
943 if (This->baseShader.shader_mode == SHADER_GLSL) shader_glsl_add_instruction_modifiers(&ins);
947 static void shader_dump_ins_modifiers(const DWORD output)
949 DWORD shift = (output & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
950 DWORD mmask = output & WINED3DSP_DSTMOD_MASK;
954 case 13: TRACE("_d8"); break;
955 case 14: TRACE("_d4"); break;
956 case 15: TRACE("_d2"); break;
957 case 1: TRACE("_x2"); break;
958 case 2: TRACE("_x4"); break;
959 case 3: TRACE("_x8"); break;
960 default: TRACE("_unhandled_shift(%d)", shift); break;
963 if (mmask & WINED3DSPDM_SATURATE) TRACE("_sat");
964 if (mmask & WINED3DSPDM_PARTIALPRECISION) TRACE("_pp");
965 if (mmask & WINED3DSPDM_MSAMPCENTROID) TRACE("_centroid");
967 mmask &= ~(WINED3DSPDM_SATURATE | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_MSAMPCENTROID);
969 FIXME("_unrecognized_modifier(%#x)", mmask >> WINED3DSP_DSTMOD_SHIFT);
972 void shader_trace_init(const DWORD *pFunction, const SHADER_OPCODE *opcode_table)
974 const DWORD* pToken = pFunction;
975 const SHADER_OPCODE* curOpcode = NULL;
976 DWORD shader_version;
980 TRACE("Parsing %p\n", pFunction);
982 /* The version token is supposed to be the first token */
983 if (!shader_is_version_token(*pToken))
985 FIXME("First token is not a version token, invalid shader.\n");
988 shader_version = *pToken++;
989 TRACE("%s_%u_%u\n", shader_is_pshader_version(shader_version) ? "ps": "vs",
990 WINED3DSHADER_VERSION_MAJOR(shader_version), WINED3DSHADER_VERSION_MINOR(shader_version));
992 while (WINED3DVS_END() != *pToken)
994 if (shader_is_comment(*pToken)) /* comment */
996 DWORD comment_len = (*pToken & WINED3DSI_COMMENTSIZE_MASK) >> WINED3DSI_COMMENTSIZE_SHIFT;
998 TRACE("//%s\n", (const char*)pToken);
999 pToken += comment_len;
1002 opcode_token = *pToken++;
1003 curOpcode = shader_get_opcode(opcode_table, shader_version, opcode_token);
1008 FIXME("Unrecognized opcode: token=0x%08x\n", opcode_token);
1009 tokens_read = shader_skip_unrecognized(pToken, shader_version);
1010 pToken += tokens_read;
1014 if (curOpcode->opcode == WINED3DSIO_DCL)
1016 DWORD usage = *pToken;
1017 DWORD param = *(pToken + 1);
1019 shader_dump_decl_usage(usage, param, shader_version);
1020 shader_dump_ins_modifiers(param);
1022 shader_dump_param(param, 0, 0, shader_version);
1025 else if (curOpcode->opcode == WINED3DSIO_DEF)
1027 unsigned int offset = shader_get_float_offset(*pToken);
1029 TRACE("def c%u = %f, %f, %f, %f", offset,
1030 *(const float *)(pToken + 1),
1031 *(const float *)(pToken + 2),
1032 *(const float *)(pToken + 3),
1033 *(const float *)(pToken + 4));
1036 else if (curOpcode->opcode == WINED3DSIO_DEFI)
1038 TRACE("defi i%u = %d, %d, %d, %d", *pToken & WINED3DSP_REGNUM_MASK,
1045 else if (curOpcode->opcode == WINED3DSIO_DEFB)
1047 TRACE("defb b%u = %s", *pToken & WINED3DSP_REGNUM_MASK,
1048 *(pToken + 1)? "true": "false");
1053 DWORD param, addr_token = 0;
1056 /* Print out predication source token first - it follows
1057 * the destination token. */
1058 if (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED)
1061 shader_dump_param(*(pToken + 2), 0, 1, shader_version);
1064 if (opcode_token & WINED3DSI_COISSUE)
1066 /* PixWin marks instructions with the coissue flag with a '+' */
1070 TRACE("%s", curOpcode->name);
1072 if (curOpcode->opcode == WINED3DSIO_IFC
1073 || curOpcode->opcode == WINED3DSIO_BREAKC)
1075 DWORD op = (opcode_token & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
1079 case COMPARISON_GT: TRACE("_gt"); break;
1080 case COMPARISON_EQ: TRACE("_eq"); break;
1081 case COMPARISON_GE: TRACE("_ge"); break;
1082 case COMPARISON_LT: TRACE("_lt"); break;
1083 case COMPARISON_NE: TRACE("_ne"); break;
1084 case COMPARISON_LE: TRACE("_le"); break;
1085 default: TRACE("_(%u)", op);
1088 else if (curOpcode->opcode == WINED3DSIO_TEX
1089 && shader_version >= WINED3DPS_VERSION(2,0)
1090 && (opcode_token & WINED3DSI_TEXLD_PROJECT))
1095 /* Destination token */
1096 if (curOpcode->dst_token)
1098 tokens_read = shader_get_param(pToken, shader_version, ¶m, &addr_token);
1099 pToken += tokens_read;
1101 shader_dump_ins_modifiers(param);
1103 shader_dump_param(param, addr_token, 0, shader_version);
1106 /* Predication token - already printed out, just skip it */
1107 if (opcode_token & WINED3DSHADER_INSTRUCTION_PREDICATED)
1112 /* Other source tokens */
1113 for (i = curOpcode->dst_token; i < curOpcode->num_params; ++i)
1115 tokens_read = shader_get_param(pToken, shader_version, ¶m, &addr_token);
1116 pToken += tokens_read;
1118 TRACE((i == 0)? " " : ", ");
1119 shader_dump_param(param, addr_token, 1, shader_version);
1127 void shader_cleanup(IWineD3DBaseShader *iface)
1129 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)iface;
1131 ((IWineD3DDeviceImpl *)This->baseShader.device)->shader_backend->shader_destroy(iface);
1132 HeapFree(GetProcessHeap(), 0, This->baseShader.function);
1133 shader_delete_constant_list(&This->baseShader.constantsF);
1134 shader_delete_constant_list(&This->baseShader.constantsB);
1135 shader_delete_constant_list(&This->baseShader.constantsI);
1136 list_remove(&This->baseShader.shader_list_entry);
1139 static const SHADER_HANDLER shader_none_instruction_handler_table[WINED3DSIH_TABLE_SIZE] = {0};
1140 static void shader_none_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {}
1141 static void shader_none_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {}
1142 static void shader_none_deselect_depth_blt(IWineD3DDevice *iface) {}
1143 static void shader_none_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1144 static void shader_none_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1145 static void shader_none_load_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1146 static void shader_none_load_np2fixup_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1147 static void shader_none_destroy(IWineD3DBaseShader *iface) {}
1148 static HRESULT shader_none_alloc(IWineD3DDevice *iface) {return WINED3D_OK;}
1149 static void shader_none_free(IWineD3DDevice *iface) {}
1150 static BOOL shader_none_dirty_const(IWineD3DDevice *iface) {return FALSE;}
1151 static GLuint shader_none_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer, const struct ps_compile_args *args) {
1152 FIXME("NONE shader backend asked to generate a pixel shader\n");
1155 static GLuint shader_none_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer, const struct vs_compile_args *args) {
1156 FIXME("NONE shader backend asked to generate a vertex shader\n");
1160 #define GLINFO_LOCATION (*gl_info)
1161 static void shader_none_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
1163 /* Set the shader caps to 0 for the none shader backend */
1164 pCaps->VertexShaderVersion = 0;
1165 pCaps->PixelShaderVersion = 0;
1166 pCaps->PixelShader1xMaxValue = 0.0;
1168 #undef GLINFO_LOCATION
1169 static BOOL shader_none_color_fixup_supported(struct color_fixup_desc fixup)
1171 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
1173 TRACE("Checking support for fixup:\n");
1174 dump_color_fixup_desc(fixup);
1177 /* Faked to make some apps happy. */
1178 if (!is_yuv_fixup(fixup))
1184 TRACE("[FAILED]\n");
1188 const shader_backend_t none_shader_backend = {
1189 shader_none_instruction_handler_table,
1191 shader_none_select_depth_blt,
1192 shader_none_deselect_depth_blt,
1193 shader_none_update_float_vertex_constants,
1194 shader_none_update_float_pixel_constants,
1195 shader_none_load_constants,
1196 shader_none_load_np2fixup_constants,
1197 shader_none_destroy,
1200 shader_none_dirty_const,
1201 shader_none_generate_pshader,
1202 shader_none_generate_vshader,
1203 shader_none_get_caps,
1204 shader_none_color_fixup_supported,