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 const char *shader_opcode_names[] =
37 /* WINED3DSIH_ABS */ "abs",
38 /* WINED3DSIH_ADD */ "add",
39 /* WINED3DSIH_BEM */ "bem",
40 /* WINED3DSIH_BREAK */ "break",
41 /* WINED3DSIH_BREAKC */ "breakc",
42 /* WINED3DSIH_BREAKP */ "breakp",
43 /* WINED3DSIH_CALL */ "call",
44 /* WINED3DSIH_CALLNZ */ "callnz",
45 /* WINED3DSIH_CMP */ "cmp",
46 /* WINED3DSIH_CND */ "cnd",
47 /* WINED3DSIH_CRS */ "crs",
48 /* WINED3DSIH_DCL */ "dcl",
49 /* WINED3DSIH_DEF */ "def",
50 /* WINED3DSIH_DEFB */ "defb",
51 /* WINED3DSIH_DEFI */ "defi",
52 /* WINED3DSIH_DP2ADD */ "dp2add",
53 /* WINED3DSIH_DP3 */ "dp3",
54 /* WINED3DSIH_DP4 */ "dp4",
55 /* WINED3DSIH_DST */ "dst",
56 /* WINED3DSIH_DSX */ "dsx",
57 /* WINED3DSIH_DSY */ "dsy",
58 /* WINED3DSIH_ELSE */ "else",
59 /* WINED3DSIH_ENDIF */ "endif",
60 /* WINED3DSIH_ENDLOOP */ "endloop",
61 /* WINED3DSIH_ENDREP */ "endrep",
62 /* WINED3DSIH_EXP */ "exp",
63 /* WINED3DSIH_EXPP */ "expp",
64 /* WINED3DSIH_FRC */ "frc",
65 /* WINED3DSIH_IF */ "if",
66 /* WINED3DSIH_IFC */ "ifc",
67 /* WINED3DSIH_LABEL */ "label",
68 /* WINED3DSIH_LIT */ "lit",
69 /* WINED3DSIH_LOG */ "log",
70 /* WINED3DSIH_LOGP */ "logp",
71 /* WINED3DSIH_LOOP */ "loop",
72 /* WINED3DSIH_LRP */ "lrp",
73 /* WINED3DSIH_M3x2 */ "m3x2",
74 /* WINED3DSIH_M3x3 */ "m3x3",
75 /* WINED3DSIH_M3x4 */ "m3x4",
76 /* WINED3DSIH_M4x3 */ "m4x3",
77 /* WINED3DSIH_M4x4 */ "m4x4",
78 /* WINED3DSIH_MAD */ "mad",
79 /* WINED3DSIH_MAX */ "max",
80 /* WINED3DSIH_MIN */ "min",
81 /* WINED3DSIH_MOV */ "mov",
82 /* WINED3DSIH_MOVA */ "mova",
83 /* WINED3DSIH_MUL */ "mul",
84 /* WINED3DSIH_NOP */ "nop",
85 /* WINED3DSIH_NRM */ "nrm",
86 /* WINED3DSIH_PHASE */ "phase",
87 /* WINED3DSIH_POW */ "pow",
88 /* WINED3DSIH_RCP */ "rcp",
89 /* WINED3DSIH_REP */ "rep",
90 /* WINED3DSIH_RET */ "ret",
91 /* WINED3DSIH_RSQ */ "rsq",
92 /* WINED3DSIH_SETP */ "setp",
93 /* WINED3DSIH_SGE */ "sge",
94 /* WINED3DSIH_SGN */ "sgn",
95 /* WINED3DSIH_SINCOS */ "sincos",
96 /* WINED3DSIH_SLT */ "slt",
97 /* WINED3DSIH_SUB */ "sub",
98 /* WINED3DSIH_TEX */ "texld",
99 /* WINED3DSIH_TEXBEM */ "texbem",
100 /* WINED3DSIH_TEXBEML */ "texbeml",
101 /* WINED3DSIH_TEXCOORD */ "texcrd",
102 /* WINED3DSIH_TEXDEPTH */ "texdepth",
103 /* WINED3DSIH_TEXDP3 */ "texdp3",
104 /* WINED3DSIH_TEXDP3TEX */ "texdp3tex",
105 /* WINED3DSIH_TEXKILL */ "texkill",
106 /* WINED3DSIH_TEXLDD */ "texldd",
107 /* WINED3DSIH_TEXLDL */ "texldl",
108 /* WINED3DSIH_TEXM3x2DEPTH */ "texm3x2depth",
109 /* WINED3DSIH_TEXM3x2PAD */ "texm3x2pad",
110 /* WINED3DSIH_TEXM3x2TEX */ "texm3x2tex",
111 /* WINED3DSIH_TEXM3x3 */ "texm3x3",
112 /* WINED3DSIH_TEXM3x3DIFF */ "texm3x3diff",
113 /* WINED3DSIH_TEXM3x3PAD */ "texm3x3pad",
114 /* WINED3DSIH_TEXM3x3SPEC */ "texm3x3spec",
115 /* WINED3DSIH_TEXM3x3TEX */ "texm3x3tex",
116 /* WINED3DSIH_TEXM3x3VSPEC */ "texm3x3vspec",
117 /* WINED3DSIH_TEXREG2AR */ "texreg2ar",
118 /* WINED3DSIH_TEXREG2GB */ "texreg2gb",
119 /* WINED3DSIH_TEXREG2RGB */ "texreg2rgb",
122 const struct wined3d_shader_frontend *shader_select_frontend(DWORD version_token)
124 switch (version_token >> 16)
128 return &sm1_shader_frontend;
133 return &sm4_shader_frontend;
136 FIXME("Unrecognised version token %#x\n", version_token);
141 static inline BOOL shader_is_version_token(DWORD token) {
142 return shader_is_pshader_version(token) ||
143 shader_is_vshader_version(token);
146 void shader_buffer_init(struct SHADER_BUFFER *buffer)
148 buffer->buffer = HeapAlloc(GetProcessHeap(), 0, SHADER_PGMSIZE);
149 buffer->buffer[0] = '\0';
152 buffer->newline = TRUE;
155 void shader_buffer_free(struct SHADER_BUFFER *buffer)
157 HeapFree(GetProcessHeap(), 0, buffer->buffer);
160 int shader_vaddline(SHADER_BUFFER* buffer, const char *format, va_list args)
162 char* base = buffer->buffer + buffer->bsize;
165 rc = vsnprintf(base, SHADER_PGMSIZE - 1 - buffer->bsize, format, args);
167 if (rc < 0 || /* C89 */
168 rc > SHADER_PGMSIZE - 1 - buffer->bsize) { /* C99 */
170 ERR("The buffer allocated for the shader program string "
171 "is too small at %d bytes.\n", SHADER_PGMSIZE);
172 buffer->bsize = SHADER_PGMSIZE - 1;
176 if (buffer->newline) {
177 TRACE("GL HW (%u, %u) : %s", buffer->lineNo + 1, buffer->bsize, base);
178 buffer->newline = FALSE;
184 if (buffer->buffer[buffer->bsize-1] == '\n') {
186 buffer->newline = TRUE;
191 int shader_addline(SHADER_BUFFER* buffer, const char *format, ...)
196 va_start(args, format);
197 ret = shader_vaddline(buffer, format, args);
203 void shader_init(struct IWineD3DBaseShaderClass *shader, IWineD3DDevice *device)
206 shader->device = device;
207 list_init(&shader->linked_programs);
210 /* Convert floating point offset relative
211 * to a register file to an absolute offset for float constants */
212 static unsigned int shader_get_float_offset(WINED3DSHADER_PARAM_REGISTER_TYPE register_type, UINT register_idx)
214 switch (register_type)
216 case WINED3DSPR_CONST: return register_idx;
217 case WINED3DSPR_CONST2: return 2048 + register_idx;
218 case WINED3DSPR_CONST3: return 4096 + register_idx;
219 case WINED3DSPR_CONST4: return 6144 + register_idx;
221 FIXME("Unsupported register type: %d\n", register_type);
226 static void shader_delete_constant_list(struct list* clist) {
229 struct local_constant* constant;
231 ptr = list_head(clist);
233 constant = LIST_ENTRY(ptr, struct local_constant, entry);
234 ptr = list_next(clist, ptr);
235 HeapFree(GetProcessHeap(), 0, constant);
240 static void shader_record_register_usage(IWineD3DBaseShaderImpl *This, struct shader_reg_maps *reg_maps,
241 DWORD register_type, UINT register_idx, BOOL has_rel_addr, BOOL pshader)
243 switch (register_type)
245 case WINED3DSPR_TEXTURE: /* WINED3DSPR_ADDR */
246 if (pshader) reg_maps->texcoord[register_idx] = 1;
247 else reg_maps->address[register_idx] = 1;
250 case WINED3DSPR_TEMP:
251 reg_maps->temporary[register_idx] = 1;
254 case WINED3DSPR_INPUT:
255 if (!pshader) reg_maps->attributes[register_idx] = 1;
260 /* If relative addressing is used, we must assume that all registers
261 * are used. Even if it is a construct like v3[aL], we can't assume
262 * that v0, v1 and v2 aren't read because aL can be negative */
264 for (i = 0; i < MAX_REG_INPUT; ++i)
266 ((IWineD3DPixelShaderImpl *)This)->input_reg_used[i] = TRUE;
271 ((IWineD3DPixelShaderImpl *)This)->input_reg_used[register_idx] = TRUE;
276 case WINED3DSPR_RASTOUT:
277 if (register_idx == 1) reg_maps->fog = 1;
280 case WINED3DSPR_MISCTYPE:
281 if (pshader && register_idx == 0) reg_maps->vpos = 1;
284 case WINED3DSPR_CONST:
289 if (register_idx <= ((IWineD3DVertexShaderImpl *)This)->min_rel_offset)
290 ((IWineD3DVertexShaderImpl *)This)->min_rel_offset = register_idx;
291 else if (register_idx >= ((IWineD3DVertexShaderImpl *)This)->max_rel_offset)
292 ((IWineD3DVertexShaderImpl *)This)->max_rel_offset = register_idx;
294 reg_maps->usesrelconstF = TRUE;
298 case WINED3DSPR_CONSTINT:
299 reg_maps->integer_constants |= (1 << register_idx);
302 case WINED3DSPR_CONSTBOOL:
303 reg_maps->boolean_constants |= (1 << register_idx);
307 TRACE("Not recording register of type %#x and idx %u\n", register_type, register_idx);
312 /* Note that this does not count the loop register
313 * as an address register. */
315 HRESULT shader_get_registers_used(IWineD3DBaseShader *iface, const struct wined3d_shader_frontend *fe,
316 struct shader_reg_maps *reg_maps, struct wined3d_shader_semantic *semantics_in,
317 struct wined3d_shader_semantic *semantics_out, const DWORD *byte_code)
319 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
320 void *fe_data = This->baseShader.frontend_data;
321 DWORD shader_version;
322 unsigned int cur_loop_depth = 0, max_loop_depth = 0;
323 const DWORD* pToken = byte_code;
326 /* There are some minor differences between pixel and vertex shaders */
328 memset(reg_maps, 0, sizeof(*reg_maps));
330 /* get_registers_used is called on every compile on some 1.x shaders, which can result
331 * in stacking up a collection of local constants. Delete the old constants if existing
333 shader_delete_constant_list(&This->baseShader.constantsF);
334 shader_delete_constant_list(&This->baseShader.constantsB);
335 shader_delete_constant_list(&This->baseShader.constantsI);
337 fe->shader_read_header(fe_data, &pToken, &shader_version);
338 reg_maps->shader_version = shader_version;
339 pshader = shader_is_pshader_version(shader_version);
341 while (!fe->shader_is_end(fe_data, &pToken))
343 struct wined3d_shader_instruction ins;
348 fe->shader_read_comment(&pToken, &comment);
349 if (comment) continue;
352 fe->shader_read_opcode(fe_data, &pToken, &ins, ¶m_size, shader_version);
354 /* Unhandled opcode, and its parameters */
355 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
357 TRACE("Skipping unrecognized instruction.\n");
358 pToken += param_size;
362 /* Handle declarations */
363 if (ins.handler_idx == WINED3DSIH_DCL)
365 struct wined3d_shader_semantic semantic;
367 fe->shader_read_semantic(&pToken, &semantic);
369 switch (semantic.reg.register_type)
371 /* Vshader: mark attributes used
372 * Pshader: mark 3.0 input registers used, save token */
373 case WINED3DSPR_INPUT:
374 if (!pshader) reg_maps->attributes[semantic.reg.register_idx] = 1;
375 else reg_maps->packed_input[semantic.reg.register_idx] = 1;
376 semantics_in[semantic.reg.register_idx] = semantic;
379 /* Vshader: mark 3.0 output registers used, save token */
380 case WINED3DSPR_OUTPUT:
381 reg_maps->packed_output[semantic.reg.register_idx] = 1;
382 semantics_out[semantic.reg.register_idx] = semantic;
383 if (semantic.usage == WINED3DDECLUSAGE_FOG) reg_maps->fog = 1;
386 /* Save sampler usage token */
387 case WINED3DSPR_SAMPLER:
388 reg_maps->sampler_type[semantic.reg.register_idx] = semantic.sampler_type;
392 TRACE("Not recording DCL register type %#x.\n", semantic.reg.register_type);
396 else if (ins.handler_idx == WINED3DSIH_DEF)
398 struct wined3d_shader_dst_param dst;
399 struct wined3d_shader_src_param rel_addr;
401 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
402 if (!lconst) return E_OUTOFMEMORY;
404 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
405 lconst->idx = dst.register_idx;
407 memcpy(lconst->value, pToken, 4 * sizeof(DWORD));
410 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
411 if (WINED3DSHADER_VERSION_MAJOR(shader_version) == 1 && pshader)
413 float *value = (float *) lconst->value;
414 if(value[0] < -1.0) value[0] = -1.0;
415 else if(value[0] > 1.0) value[0] = 1.0;
416 if(value[1] < -1.0) value[1] = -1.0;
417 else if(value[1] > 1.0) value[1] = 1.0;
418 if(value[2] < -1.0) value[2] = -1.0;
419 else if(value[2] > 1.0) value[2] = 1.0;
420 if(value[3] < -1.0) value[3] = -1.0;
421 else if(value[3] > 1.0) value[3] = 1.0;
424 list_add_head(&This->baseShader.constantsF, &lconst->entry);
426 else if (ins.handler_idx == WINED3DSIH_DEFI)
428 struct wined3d_shader_dst_param dst;
429 struct wined3d_shader_src_param rel_addr;
431 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
432 if (!lconst) return E_OUTOFMEMORY;
434 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
435 lconst->idx = dst.register_idx;
437 memcpy(lconst->value, pToken, 4 * sizeof(DWORD));
440 list_add_head(&This->baseShader.constantsI, &lconst->entry);
442 else if (ins.handler_idx == WINED3DSIH_DEFB)
444 struct wined3d_shader_dst_param dst;
445 struct wined3d_shader_src_param rel_addr;
447 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
448 if (!lconst) return E_OUTOFMEMORY;
450 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
451 lconst->idx = dst.register_idx;
453 memcpy(lconst->value, pToken, sizeof(DWORD));
456 list_add_head(&This->baseShader.constantsB, &lconst->entry);
458 /* If there's a loop in the shader */
459 else if (ins.handler_idx == WINED3DSIH_LOOP
460 || ins.handler_idx == WINED3DSIH_REP)
462 struct wined3d_shader_src_param src, rel_addr;
464 fe->shader_read_src_param(&pToken, &src, &rel_addr, shader_version);
466 /* Rep and Loop always use an integer constant for the control parameters */
467 if (ins.handler_idx == WINED3DSIH_REP)
469 reg_maps->integer_constants |= 1 << src.register_idx;
473 fe->shader_read_src_param(&pToken, &src, &rel_addr, shader_version);
474 reg_maps->integer_constants |= 1 << src.register_idx;
478 if(cur_loop_depth > max_loop_depth)
479 max_loop_depth = cur_loop_depth;
481 else if (ins.handler_idx == WINED3DSIH_ENDLOOP
482 || ins.handler_idx == WINED3DSIH_ENDREP)
486 /* For subroutine prototypes */
487 else if (ins.handler_idx == WINED3DSIH_LABEL)
489 struct wined3d_shader_src_param src, rel_addr;
491 fe->shader_read_src_param(&pToken, &src, &rel_addr, shader_version);
492 reg_maps->labels[src.register_idx] = 1;
494 /* Set texture, address, temporary registers */
499 /* This will loop over all the registers and try to
500 * make a bitmask of the ones we're interested in.
502 * Relative addressing tokens are ignored, but that's
503 * okay, since we'll catch any address registers when
504 * they are initialized (required by spec) */
508 struct wined3d_shader_dst_param dst_param;
509 struct wined3d_shader_src_param dst_rel_addr;
511 fe->shader_read_dst_param(&pToken, &dst_param, &dst_rel_addr, shader_version);
513 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and
514 * is used in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel
515 * shaders because TECRDOUT isn't used in them, but future register types might cause issues */
516 if (!pshader && WINED3DSHADER_VERSION_MAJOR(shader_version) < 3
517 && dst_param.register_type == WINED3DSPR_TEXCRDOUT)
519 reg_maps->texcoord_mask[dst_param.register_type] |= dst_param.write_mask;
523 shader_record_register_usage(This, reg_maps, dst_param.register_type,
524 dst_param.register_idx, !!dst_param.rel_addr, pshader);
527 /* Declare 1.X samplers implicitly, based on the destination reg. number */
528 if (WINED3DSHADER_VERSION_MAJOR(shader_version) == 1
529 && pshader /* Filter different instructions with the same enum values in VS */
530 && (ins.handler_idx == WINED3DSIH_TEX
531 || ins.handler_idx == WINED3DSIH_TEXBEM
532 || ins.handler_idx == WINED3DSIH_TEXBEML
533 || ins.handler_idx == WINED3DSIH_TEXDP3TEX
534 || ins.handler_idx == WINED3DSIH_TEXM3x2TEX
535 || ins.handler_idx == WINED3DSIH_TEXM3x3SPEC
536 || ins.handler_idx == WINED3DSIH_TEXM3x3TEX
537 || ins.handler_idx == WINED3DSIH_TEXM3x3VSPEC
538 || ins.handler_idx == WINED3DSIH_TEXREG2AR
539 || ins.handler_idx == WINED3DSIH_TEXREG2GB
540 || ins.handler_idx == WINED3DSIH_TEXREG2RGB))
542 /* Fake sampler usage, only set reserved bit and ttype */
543 DWORD sampler_code = dst_param.register_idx;
545 TRACE("Setting fake 2D sampler for 1.x pixelshader\n");
546 reg_maps->sampler_type[sampler_code] = WINED3DSTT_2D;
548 /* texbem is only valid with < 1.4 pixel shaders */
549 if (ins.handler_idx == WINED3DSIH_TEXBEM
550 || ins.handler_idx == WINED3DSIH_TEXBEML)
552 reg_maps->bumpmat[sampler_code] = TRUE;
553 if (ins.handler_idx == WINED3DSIH_TEXBEML)
555 reg_maps->luminanceparams[sampler_code] = TRUE;
559 else if (pshader && ins.handler_idx == WINED3DSIH_BEM)
561 reg_maps->bumpmat[dst_param.register_idx] = TRUE;
565 if (ins.handler_idx == WINED3DSIH_NRM)
567 reg_maps->usesnrm = 1;
569 else if (ins.handler_idx == WINED3DSIH_DSY)
571 reg_maps->usesdsy = 1;
573 else if(ins.handler_idx == WINED3DSIH_TEXLDD)
575 reg_maps->usestexldd = 1;
578 limit = ins.src_count + (ins.predicate ? 1 : 0);
579 for (i = 0; i < limit; ++i)
581 struct wined3d_shader_src_param src_param, src_rel_addr;
583 fe->shader_read_src_param(&pToken, &src_param, &src_rel_addr, shader_version);
584 shader_record_register_usage(This, reg_maps, src_param.register_type,
585 src_param.register_idx, !!src_param.rel_addr, pshader);
589 reg_maps->loop_depth = max_loop_depth;
591 This->baseShader.functionLength = ((char *)pToken - (char *)byte_code);
596 static void shader_dump_decl_usage(const struct wined3d_shader_semantic *semantic, DWORD shader_version)
600 if (semantic->reg.register_type == WINED3DSPR_SAMPLER)
602 switch (semantic->sampler_type)
604 case WINED3DSTT_2D: TRACE("_2d"); break;
605 case WINED3DSTT_CUBE: TRACE("_cube"); break;
606 case WINED3DSTT_VOLUME: TRACE("_volume"); break;
607 default: TRACE("_unknown_ttype(0x%08x)", semantic->sampler_type);
612 /* Pixel shaders 3.0 don't have usage semantics */
613 if (shader_is_pshader_version(shader_version) && shader_version < WINED3DPS_VERSION(3,0))
618 switch (semantic->usage)
620 case WINED3DDECLUSAGE_POSITION:
621 TRACE("position%d", semantic->usage_idx);
623 case WINED3DDECLUSAGE_BLENDINDICES:
626 case WINED3DDECLUSAGE_BLENDWEIGHT:
629 case WINED3DDECLUSAGE_NORMAL:
630 TRACE("normal%d", semantic->usage_idx);
632 case WINED3DDECLUSAGE_PSIZE:
635 case WINED3DDECLUSAGE_COLOR:
636 if (semantic->usage_idx == 0) TRACE("color");
637 else TRACE("specular%d", (semantic->usage_idx - 1));
639 case WINED3DDECLUSAGE_TEXCOORD:
640 TRACE("texture%d", semantic->usage_idx);
642 case WINED3DDECLUSAGE_TANGENT:
645 case WINED3DDECLUSAGE_BINORMAL:
648 case WINED3DDECLUSAGE_TESSFACTOR:
651 case WINED3DDECLUSAGE_POSITIONT:
652 TRACE("positionT%d", semantic->usage_idx);
654 case WINED3DDECLUSAGE_FOG:
657 case WINED3DDECLUSAGE_DEPTH:
660 case WINED3DDECLUSAGE_SAMPLE:
664 FIXME("unknown_semantics(0x%08x)", semantic->usage);
669 static void shader_dump_register(WINED3DSHADER_PARAM_REGISTER_TYPE register_type, UINT register_idx,
670 const struct wined3d_shader_src_param *rel_addr, DWORD shader_version)
672 static const char * const rastout_reg_names[] = {"oPos", "oFog", "oPts"};
673 static const char * const misctype_reg_names[] = {"vPos", "vFace"};
674 UINT offset = register_idx;
676 switch (register_type)
678 case WINED3DSPR_TEMP:
682 case WINED3DSPR_INPUT:
686 case WINED3DSPR_CONST:
687 case WINED3DSPR_CONST2:
688 case WINED3DSPR_CONST3:
689 case WINED3DSPR_CONST4:
691 offset = shader_get_float_offset(register_type, register_idx);
694 case WINED3DSPR_TEXTURE: /* vs: case WINED3DSPR_ADDR */
695 TRACE("%c", shader_is_pshader_version(shader_version) ? 't' : 'a');
698 case WINED3DSPR_RASTOUT:
699 TRACE("%s", rastout_reg_names[register_idx]);
702 case WINED3DSPR_COLOROUT:
706 case WINED3DSPR_DEPTHOUT:
710 case WINED3DSPR_ATTROUT:
714 case WINED3DSPR_TEXCRDOUT:
715 /* Vertex shaders >= 3.0 use general purpose output registers
716 * (WINED3DSPR_OUTPUT), which can include an address token */
717 if (WINED3DSHADER_VERSION_MAJOR(shader_version) >= 3) TRACE("o");
721 case WINED3DSPR_CONSTINT:
725 case WINED3DSPR_CONSTBOOL:
729 case WINED3DSPR_LABEL:
733 case WINED3DSPR_LOOP:
737 case WINED3DSPR_SAMPLER:
741 case WINED3DSPR_MISCTYPE:
742 if (register_idx > 1) FIXME("Unhandled misctype register %d\n", register_idx);
743 else TRACE("%s", misctype_reg_names[register_idx]);
746 case WINED3DSPR_PREDICATE:
751 TRACE("unhandled_rtype(%#x)", register_type);
755 if (register_type != WINED3DSPR_RASTOUT && register_type != WINED3DSPR_MISCTYPE)
760 shader_dump_src_param(rel_addr, shader_version);
764 if (rel_addr) TRACE("]");
768 void shader_dump_dst_param(const struct wined3d_shader_dst_param *param, DWORD shader_version)
770 DWORD write_mask = param->write_mask;
772 shader_dump_register(param->register_type, param->register_idx, param->rel_addr, shader_version);
774 if (write_mask != WINED3DSP_WRITEMASK_ALL)
776 static const char *write_mask_chars = "xyzw";
779 if (write_mask & WINED3DSP_WRITEMASK_0) TRACE("%c", write_mask_chars[0]);
780 if (write_mask & WINED3DSP_WRITEMASK_1) TRACE("%c", write_mask_chars[1]);
781 if (write_mask & WINED3DSP_WRITEMASK_2) TRACE("%c", write_mask_chars[2]);
782 if (write_mask & WINED3DSP_WRITEMASK_3) TRACE("%c", write_mask_chars[3]);
786 void shader_dump_src_param(const struct wined3d_shader_src_param *param, DWORD shader_version)
788 DWORD src_modifier = param->modifiers;
789 DWORD swizzle = param->swizzle;
791 if (src_modifier == WINED3DSPSM_NEG
792 || src_modifier == WINED3DSPSM_BIASNEG
793 || src_modifier == WINED3DSPSM_SIGNNEG
794 || src_modifier == WINED3DSPSM_X2NEG
795 || src_modifier == WINED3DSPSM_ABSNEG)
797 else if (src_modifier == WINED3DSPSM_COMP)
799 else if (src_modifier == WINED3DSPSM_NOT)
802 if (src_modifier == WINED3DSPSM_ABS || src_modifier == WINED3DSPSM_ABSNEG)
805 shader_dump_register(param->register_type, param->register_idx, param->rel_addr, shader_version);
809 switch (src_modifier)
811 case WINED3DSPSM_NONE: break;
812 case WINED3DSPSM_NEG: break;
813 case WINED3DSPSM_NOT: break;
814 case WINED3DSPSM_BIAS: TRACE("_bias"); break;
815 case WINED3DSPSM_BIASNEG: TRACE("_bias"); break;
816 case WINED3DSPSM_SIGN: TRACE("_bx2"); break;
817 case WINED3DSPSM_SIGNNEG: TRACE("_bx2"); break;
818 case WINED3DSPSM_COMP: break;
819 case WINED3DSPSM_X2: TRACE("_x2"); break;
820 case WINED3DSPSM_X2NEG: TRACE("_x2"); break;
821 case WINED3DSPSM_DZ: TRACE("_dz"); break;
822 case WINED3DSPSM_DW: TRACE("_dw"); break;
823 case WINED3DSPSM_ABSNEG: TRACE(")"); break;
824 case WINED3DSPSM_ABS: TRACE(")"); break;
826 TRACE("_unknown_modifier(%#x)", src_modifier);
830 if (swizzle != WINED3DSP_NOSWIZZLE)
832 static const char *swizzle_chars = "xyzw";
833 DWORD swizzle_x = swizzle & 0x03;
834 DWORD swizzle_y = (swizzle >> 2) & 0x03;
835 DWORD swizzle_z = (swizzle >> 4) & 0x03;
836 DWORD swizzle_w = (swizzle >> 6) & 0x03;
838 if (swizzle_x == swizzle_y
839 && swizzle_x == swizzle_z
840 && swizzle_x == swizzle_w)
842 TRACE(".%c", swizzle_chars[swizzle_x]);
846 TRACE(".%c%c%c%c", swizzle_chars[swizzle_x], swizzle_chars[swizzle_y],
847 swizzle_chars[swizzle_z], swizzle_chars[swizzle_w]);
852 /* Shared code in order to generate the bulk of the shader string.
853 * NOTE: A description of how to parse tokens can be found on msdn */
854 void shader_generate_main(IWineD3DBaseShader *iface, SHADER_BUFFER *buffer,
855 const shader_reg_maps *reg_maps, const DWORD *pFunction)
857 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
858 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device; /* To access shader backend callbacks */
859 const SHADER_HANDLER *handler_table = device->shader_backend->shader_instruction_handler_table;
860 const struct wined3d_shader_frontend *fe = This->baseShader.frontend;
861 void *fe_data = This->baseShader.frontend_data;
862 struct wined3d_shader_src_param src_rel_addr[4];
863 struct wined3d_shader_src_param src_param[4];
864 struct wined3d_shader_src_param dst_rel_addr;
865 struct wined3d_shader_dst_param dst_param;
866 struct wined3d_shader_instruction ins;
867 struct wined3d_shader_context ctx;
868 const DWORD *pToken = pFunction;
869 SHADER_HANDLER hw_fct;
870 DWORD shader_version;
873 /* Initialize current parsing state */
875 ctx.reg_maps = reg_maps;
879 ins.dst = &dst_param;
881 This->baseShader.parse_state.current_row = 0;
883 fe->shader_read_header(fe_data, &pToken, &shader_version);
885 while (!fe->shader_is_end(fe_data, &pToken))
890 /* Skip comment tokens */
891 fe->shader_read_comment(&pToken, &comment);
892 if (comment) continue;
895 fe->shader_read_opcode(fe_data, &pToken, &ins, ¶m_size, shader_version);
897 /* Unknown opcode and its parameters */
898 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
900 TRACE("Skipping unrecognized instruction.\n");
901 pToken += param_size;
906 if (ins.handler_idx == WINED3DSIH_DCL
907 || ins.handler_idx == WINED3DSIH_NOP
908 || ins.handler_idx == WINED3DSIH_DEF
909 || ins.handler_idx == WINED3DSIH_DEFI
910 || ins.handler_idx == WINED3DSIH_DEFB
911 || ins.handler_idx == WINED3DSIH_PHASE
912 || ins.handler_idx == WINED3DSIH_RET)
914 pToken += param_size;
919 hw_fct = handler_table[ins.handler_idx];
921 /* Unhandled opcode */
924 FIXME("Backend can't handle opcode %#x\n", ins.handler_idx);
925 pToken += param_size;
929 /* Destination token */
930 if (ins.dst_count) fe->shader_read_dst_param(&pToken, &dst_param, &dst_rel_addr, shader_version);
932 /* Predication token */
933 if (ins.predicate) ins.predicate = *pToken++;
935 /* Other source tokens */
936 for (i = 0; i < ins.src_count; ++i)
938 fe->shader_read_src_param(&pToken, &src_param[i], &src_rel_addr[i], shader_version);
941 /* Call appropriate function for output target */
944 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
945 /* FIXME: This should be internal to the shader backend.
946 * Also, right now this is the only reason "shader_mode" exists. */
947 if (This->baseShader.shader_mode == SHADER_GLSL) shader_glsl_add_instruction_modifiers(&ins);
951 static void shader_dump_ins_modifiers(const struct wined3d_shader_dst_param *dst)
953 DWORD mmask = dst->modifiers;
958 case 13: TRACE("_d8"); break;
959 case 14: TRACE("_d4"); break;
960 case 15: TRACE("_d2"); break;
961 case 1: TRACE("_x2"); break;
962 case 2: TRACE("_x4"); break;
963 case 3: TRACE("_x8"); break;
964 default: TRACE("_unhandled_shift(%d)", dst->shift); break;
967 if (mmask & WINED3DSPDM_SATURATE) TRACE("_sat");
968 if (mmask & WINED3DSPDM_PARTIALPRECISION) TRACE("_pp");
969 if (mmask & WINED3DSPDM_MSAMPCENTROID) TRACE("_centroid");
971 mmask &= ~(WINED3DSPDM_SATURATE | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_MSAMPCENTROID);
973 FIXME("_unrecognized_modifier(%#x)", mmask);
976 void shader_trace_init(const struct wined3d_shader_frontend *fe, void *fe_data, const DWORD *pFunction)
978 const DWORD* pToken = pFunction;
979 DWORD shader_version;
982 TRACE("Parsing %p\n", pFunction);
984 fe->shader_read_header(fe_data, &pToken, &shader_version);
986 TRACE("%s_%u_%u\n", shader_is_pshader_version(shader_version) ? "ps": "vs",
987 WINED3DSHADER_VERSION_MAJOR(shader_version), WINED3DSHADER_VERSION_MINOR(shader_version));
989 while (!fe->shader_is_end(fe_data, &pToken))
991 struct wined3d_shader_instruction ins;
996 fe->shader_read_comment(&pToken, &comment);
999 TRACE("//%s\n", comment);
1003 fe->shader_read_opcode(fe_data, &pToken, &ins, ¶m_size, shader_version);
1004 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
1006 TRACE("Skipping unrecognized instruction.\n");
1007 pToken += param_size;
1011 if (ins.handler_idx == WINED3DSIH_DCL)
1013 struct wined3d_shader_semantic semantic;
1015 fe->shader_read_semantic(&pToken, &semantic);
1017 shader_dump_decl_usage(&semantic, shader_version);
1018 shader_dump_ins_modifiers(&semantic.reg);
1020 shader_dump_dst_param(&semantic.reg, shader_version);
1022 else if (ins.handler_idx == WINED3DSIH_DEF)
1024 struct wined3d_shader_dst_param dst;
1025 struct wined3d_shader_src_param rel_addr;
1027 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
1029 TRACE("def c%u = %f, %f, %f, %f", shader_get_float_offset(dst.register_type, dst.register_idx),
1030 *(const float *)(pToken),
1031 *(const float *)(pToken + 1),
1032 *(const float *)(pToken + 2),
1033 *(const float *)(pToken + 3));
1036 else if (ins.handler_idx == WINED3DSIH_DEFI)
1038 struct wined3d_shader_dst_param dst;
1039 struct wined3d_shader_src_param rel_addr;
1041 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
1043 TRACE("defi i%u = %d, %d, %d, %d", dst.register_idx,
1050 else if (ins.handler_idx == WINED3DSIH_DEFB)
1052 struct wined3d_shader_dst_param dst;
1053 struct wined3d_shader_src_param rel_addr;
1055 fe->shader_read_dst_param(&pToken, &dst, &rel_addr, shader_version);
1057 TRACE("defb b%u = %s", dst.register_idx, *pToken ? "true" : "false");
1062 struct wined3d_shader_src_param dst_rel_addr, src_rel_addr;
1063 struct wined3d_shader_dst_param dst_param;
1064 struct wined3d_shader_src_param src_param;
1068 fe->shader_read_dst_param(&pToken, &dst_param, &dst_rel_addr, shader_version);
1071 /* Print out predication source token first - it follows
1072 * the destination token. */
1075 fe->shader_read_src_param(&pToken, &src_param, &src_rel_addr, shader_version);
1077 shader_dump_src_param(&src_param, shader_version);
1081 /* PixWin marks instructions with the coissue flag with a '+' */
1082 if (ins.coissue) TRACE("+");
1084 TRACE("%s", shader_opcode_names[ins.handler_idx]);
1086 if (ins.handler_idx == WINED3DSIH_IFC
1087 || ins.handler_idx == WINED3DSIH_BREAKC)
1091 case COMPARISON_GT: TRACE("_gt"); break;
1092 case COMPARISON_EQ: TRACE("_eq"); break;
1093 case COMPARISON_GE: TRACE("_ge"); break;
1094 case COMPARISON_LT: TRACE("_lt"); break;
1095 case COMPARISON_NE: TRACE("_ne"); break;
1096 case COMPARISON_LE: TRACE("_le"); break;
1097 default: TRACE("_(%u)", ins.flags);
1100 else if (ins.handler_idx == WINED3DSIH_TEX
1101 && shader_version >= WINED3DPS_VERSION(2,0)
1102 && (ins.flags & WINED3DSI_TEXLD_PROJECT))
1107 /* We already read the destination token, print it. */
1110 shader_dump_ins_modifiers(&dst_param);
1112 shader_dump_dst_param(&dst_param, shader_version);
1115 /* Other source tokens */
1116 for (i = ins.dst_count; i < (ins.dst_count + ins.src_count); ++i)
1118 fe->shader_read_src_param(&pToken, &src_param, &src_rel_addr, shader_version);
1119 TRACE(!i ? " " : ", ");
1120 shader_dump_src_param(&src_param, 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);
1138 if (This->baseShader.frontend && This->baseShader.frontend_data)
1140 This->baseShader.frontend->shader_free(This->baseShader.frontend_data);
1144 static const SHADER_HANDLER shader_none_instruction_handler_table[WINED3DSIH_TABLE_SIZE] = {0};
1145 static void shader_none_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {}
1146 static void shader_none_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {}
1147 static void shader_none_deselect_depth_blt(IWineD3DDevice *iface) {}
1148 static void shader_none_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1149 static void shader_none_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1150 static void shader_none_load_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1151 static void shader_none_load_np2fixup_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1152 static void shader_none_destroy(IWineD3DBaseShader *iface) {}
1153 static HRESULT shader_none_alloc(IWineD3DDevice *iface) {return WINED3D_OK;}
1154 static void shader_none_free(IWineD3DDevice *iface) {}
1155 static BOOL shader_none_dirty_const(IWineD3DDevice *iface) {return FALSE;}
1156 static GLuint shader_none_generate_pshader(IWineD3DPixelShader *iface,
1157 SHADER_BUFFER *buffer, const struct ps_compile_args *args)
1159 FIXME("NONE shader backend asked to generate a pixel shader\n");
1162 static GLuint shader_none_generate_vshader(IWineD3DVertexShader *iface,
1163 SHADER_BUFFER *buffer, const struct vs_compile_args *args)
1165 FIXME("NONE shader backend asked to generate a vertex shader\n");
1169 #define GLINFO_LOCATION (*gl_info)
1170 static void shader_none_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
1172 /* Set the shader caps to 0 for the none shader backend */
1173 pCaps->VertexShaderVersion = 0;
1174 pCaps->PixelShaderVersion = 0;
1175 pCaps->PixelShader1xMaxValue = 0.0;
1177 #undef GLINFO_LOCATION
1178 static BOOL shader_none_color_fixup_supported(struct color_fixup_desc fixup)
1180 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
1182 TRACE("Checking support for fixup:\n");
1183 dump_color_fixup_desc(fixup);
1186 /* Faked to make some apps happy. */
1187 if (!is_yuv_fixup(fixup))
1193 TRACE("[FAILED]\n");
1197 const shader_backend_t none_shader_backend = {
1198 shader_none_instruction_handler_table,
1200 shader_none_select_depth_blt,
1201 shader_none_deselect_depth_blt,
1202 shader_none_update_float_vertex_constants,
1203 shader_none_update_float_pixel_constants,
1204 shader_none_load_constants,
1205 shader_none_load_np2fixup_constants,
1206 shader_none_destroy,
1209 shader_none_dirty_const,
1210 shader_none_generate_pshader,
1211 shader_none_generate_vshader,
1212 shader_none_get_caps,
1213 shader_none_color_fixup_supported,