wined3d: Make "packed_output" a bitmap.
[wine/wine-gecko.git] / dlls / wined3d / baseshader.c
blobab3354facaff25c42b37607a7b8e0e029876f0dd
1 /*
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
27 #include "config.h"
28 #include <string.h>
29 #include <stdio.h>
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)
126 case WINED3D_SM1_VS:
127 case WINED3D_SM1_PS:
128 return &sm1_shader_frontend;
130 case WINED3D_SM4_PS:
131 case WINED3D_SM4_VS:
132 case WINED3D_SM4_GS:
133 return &sm4_shader_frontend;
135 default:
136 FIXME("Unrecognised version token %#x\n", version_token);
137 return NULL;
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';
150 buffer->bsize = 0;
151 buffer->lineNo = 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;
163 int rc;
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;
173 return -1;
176 if (buffer->newline) {
177 TRACE("GL HW (%u, %u) : %s", buffer->lineNo + 1, buffer->bsize, base);
178 buffer->newline = FALSE;
179 } else {
180 TRACE("%s", base);
183 buffer->bsize += rc;
184 if (buffer->buffer[buffer->bsize-1] == '\n') {
185 buffer->lineNo++;
186 buffer->newline = TRUE;
188 return 0;
191 int shader_addline(SHADER_BUFFER* buffer, const char *format, ...)
193 int ret;
194 va_list args;
196 va_start(args, format);
197 ret = shader_vaddline(buffer, format, args);
198 va_end(args);
200 return ret;
203 void shader_init(struct IWineD3DBaseShaderClass *shader, IWineD3DDevice *device)
205 shader->ref = 1;
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;
220 default:
221 FIXME("Unsupported register type: %d\n", register_type);
222 return register_idx;
226 static void shader_delete_constant_list(struct list* clist) {
228 struct list *ptr;
229 struct local_constant* constant;
231 ptr = list_head(clist);
232 while (ptr) {
233 constant = LIST_ENTRY(ptr, struct local_constant, entry);
234 ptr = list_next(clist, ptr);
235 HeapFree(GetProcessHeap(), 0, constant);
237 list_init(clist);
240 static inline void set_bitmap_bit(DWORD *bitmap, DWORD bit)
242 DWORD idx, shift;
243 idx = bit >> 5;
244 shift = bit & 0x1f;
245 bitmap[idx] |= (1 << shift);
248 static void shader_record_register_usage(IWineD3DBaseShaderImpl *This, struct shader_reg_maps *reg_maps,
249 const struct wined3d_shader_register *reg, BOOL pshader)
251 switch (reg->type)
253 case WINED3DSPR_TEXTURE: /* WINED3DSPR_ADDR */
254 if (pshader) reg_maps->texcoord[reg->idx] = 1;
255 else reg_maps->address[reg->idx] = 1;
256 break;
258 case WINED3DSPR_TEMP:
259 reg_maps->temporary[reg->idx] = 1;
260 break;
262 case WINED3DSPR_INPUT:
263 if (!pshader) reg_maps->input_registers |= 1 << reg->idx;
264 else
266 if (reg->rel_addr)
268 /* If relative addressing is used, we must assume that all registers
269 * are used. Even if it is a construct like v3[aL], we can't assume
270 * that v0, v1 and v2 aren't read because aL can be negative */
271 unsigned int i;
272 for (i = 0; i < MAX_REG_INPUT; ++i)
274 ((IWineD3DPixelShaderImpl *)This)->input_reg_used[i] = TRUE;
277 else
279 ((IWineD3DPixelShaderImpl *)This)->input_reg_used[reg->idx] = TRUE;
282 break;
284 case WINED3DSPR_RASTOUT:
285 if (reg->idx == 1) reg_maps->fog = 1;
286 break;
288 case WINED3DSPR_MISCTYPE:
289 if (pshader && reg->idx == 0) reg_maps->vpos = 1;
290 break;
292 case WINED3DSPR_CONST:
293 if (reg->rel_addr)
295 if (!pshader)
297 if (reg->idx <= ((IWineD3DVertexShaderImpl *)This)->min_rel_offset)
298 ((IWineD3DVertexShaderImpl *)This)->min_rel_offset = reg->idx;
299 else if (reg->idx >= ((IWineD3DVertexShaderImpl *)This)->max_rel_offset)
300 ((IWineD3DVertexShaderImpl *)This)->max_rel_offset = reg->idx;
302 reg_maps->usesrelconstF = TRUE;
304 else
306 set_bitmap_bit(reg_maps->constf, reg->idx);
308 break;
310 case WINED3DSPR_CONSTINT:
311 reg_maps->integer_constants |= (1 << reg->idx);
312 break;
314 case WINED3DSPR_CONSTBOOL:
315 reg_maps->boolean_constants |= (1 << reg->idx);
316 break;
318 default:
319 TRACE("Not recording register of type %#x and idx %u\n", reg->type, reg->idx);
320 break;
324 static unsigned int get_instr_extra_regcount(enum WINED3D_SHADER_INSTRUCTION_HANDLER instr, unsigned int param)
326 switch(instr)
328 case WINED3DSIH_M4x4:
329 case WINED3DSIH_M3x4:
330 return param == 1 ? 3 : 0;
332 case WINED3DSIH_M4x3:
333 case WINED3DSIH_M3x3:
334 return param == 1 ? 2 : 0;
336 case WINED3DSIH_M3x2:
337 return param == 1 ? 1 : 0;
339 default:
340 return 0;
344 /* Note that this does not count the loop register
345 * as an address register. */
347 HRESULT shader_get_registers_used(IWineD3DBaseShader *iface, const struct wined3d_shader_frontend *fe,
348 struct shader_reg_maps *reg_maps, struct wined3d_shader_semantic *semantics_in,
349 struct wined3d_shader_semantic *semantics_out, const DWORD *byte_code, DWORD constf_size)
351 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
352 void *fe_data = This->baseShader.frontend_data;
353 struct wined3d_shader_version shader_version;
354 unsigned int cur_loop_depth = 0, max_loop_depth = 0;
355 const DWORD* pToken = byte_code;
356 char pshader;
358 /* There are some minor differences between pixel and vertex shaders */
360 memset(reg_maps, 0, sizeof(*reg_maps));
362 /* get_registers_used is called on every compile on some 1.x shaders, which can result
363 * in stacking up a collection of local constants. Delete the old constants if existing
365 shader_delete_constant_list(&This->baseShader.constantsF);
366 shader_delete_constant_list(&This->baseShader.constantsB);
367 shader_delete_constant_list(&This->baseShader.constantsI);
369 fe->shader_read_header(fe_data, &pToken, &shader_version);
370 reg_maps->shader_version = shader_version;
371 pshader = shader_is_pshader_version(shader_version.type);
373 reg_maps->constf = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
374 sizeof(*reg_maps->constf) * ((constf_size + 31) / 32));
375 if(!reg_maps->constf) {
376 ERR("Out of memory\n");
377 return E_OUTOFMEMORY;
380 while (!fe->shader_is_end(fe_data, &pToken))
382 struct wined3d_shader_instruction ins;
383 const char *comment;
384 UINT param_size;
386 /* Skip comments */
387 fe->shader_read_comment(&pToken, &comment);
388 if (comment) continue;
390 /* Fetch opcode */
391 fe->shader_read_opcode(fe_data, &pToken, &ins, &param_size);
393 /* Unhandled opcode, and its parameters */
394 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
396 TRACE("Skipping unrecognized instruction.\n");
397 pToken += param_size;
398 continue;
401 /* Handle declarations */
402 if (ins.handler_idx == WINED3DSIH_DCL)
404 struct wined3d_shader_semantic semantic;
406 fe->shader_read_semantic(&pToken, &semantic);
408 switch (semantic.reg.reg.type)
410 /* Vshader: mark attributes used
411 * Pshader: mark 3.0 input registers used, save token */
412 case WINED3DSPR_INPUT:
413 reg_maps->input_registers |= 1 << semantic.reg.reg.idx;
414 semantics_in[semantic.reg.reg.idx] = semantic;
415 break;
417 /* Vshader: mark 3.0 output registers used, save token */
418 case WINED3DSPR_OUTPUT:
419 reg_maps->output_registers |= 1 << semantic.reg.reg.idx;
420 semantics_out[semantic.reg.reg.idx] = semantic;
421 if (semantic.usage == WINED3DDECLUSAGE_FOG) reg_maps->fog = 1;
422 break;
424 /* Save sampler usage token */
425 case WINED3DSPR_SAMPLER:
426 reg_maps->sampler_type[semantic.reg.reg.idx] = semantic.sampler_type;
427 break;
429 default:
430 TRACE("Not recording DCL register type %#x.\n", semantic.reg.reg.type);
431 break;
434 else if (ins.handler_idx == WINED3DSIH_DEF)
436 struct wined3d_shader_dst_param dst;
437 struct wined3d_shader_src_param rel_addr;
439 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
440 if (!lconst) return E_OUTOFMEMORY;
442 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
443 lconst->idx = dst.reg.idx;
445 memcpy(lconst->value, pToken, 4 * sizeof(DWORD));
446 pToken += 4;
448 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
449 if (shader_version.major == 1 && pshader)
451 float *value = (float *) lconst->value;
452 if(value[0] < -1.0) value[0] = -1.0;
453 else if(value[0] > 1.0) value[0] = 1.0;
454 if(value[1] < -1.0) value[1] = -1.0;
455 else if(value[1] > 1.0) value[1] = 1.0;
456 if(value[2] < -1.0) value[2] = -1.0;
457 else if(value[2] > 1.0) value[2] = 1.0;
458 if(value[3] < -1.0) value[3] = -1.0;
459 else if(value[3] > 1.0) value[3] = 1.0;
462 list_add_head(&This->baseShader.constantsF, &lconst->entry);
464 else if (ins.handler_idx == WINED3DSIH_DEFI)
466 struct wined3d_shader_dst_param dst;
467 struct wined3d_shader_src_param rel_addr;
469 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
470 if (!lconst) return E_OUTOFMEMORY;
472 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
473 lconst->idx = dst.reg.idx;
475 memcpy(lconst->value, pToken, 4 * sizeof(DWORD));
476 pToken += 4;
478 list_add_head(&This->baseShader.constantsI, &lconst->entry);
480 else if (ins.handler_idx == WINED3DSIH_DEFB)
482 struct wined3d_shader_dst_param dst;
483 struct wined3d_shader_src_param rel_addr;
485 local_constant* lconst = HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant));
486 if (!lconst) return E_OUTOFMEMORY;
488 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
489 lconst->idx = dst.reg.idx;
491 memcpy(lconst->value, pToken, sizeof(DWORD));
492 ++pToken;
494 list_add_head(&This->baseShader.constantsB, &lconst->entry);
496 /* If there's a loop in the shader */
497 else if (ins.handler_idx == WINED3DSIH_LOOP
498 || ins.handler_idx == WINED3DSIH_REP)
500 struct wined3d_shader_src_param src, rel_addr;
502 fe->shader_read_src_param(fe_data, &pToken, &src, &rel_addr);
504 /* Rep and Loop always use an integer constant for the control parameters */
505 if (ins.handler_idx == WINED3DSIH_REP)
507 reg_maps->integer_constants |= 1 << src.reg.idx;
509 else
511 fe->shader_read_src_param(fe_data, &pToken, &src, &rel_addr);
512 reg_maps->integer_constants |= 1 << src.reg.idx;
515 cur_loop_depth++;
516 if(cur_loop_depth > max_loop_depth)
517 max_loop_depth = cur_loop_depth;
519 else if (ins.handler_idx == WINED3DSIH_ENDLOOP
520 || ins.handler_idx == WINED3DSIH_ENDREP)
522 cur_loop_depth--;
524 /* For subroutine prototypes */
525 else if (ins.handler_idx == WINED3DSIH_LABEL)
527 struct wined3d_shader_src_param src, rel_addr;
529 fe->shader_read_src_param(fe_data, &pToken, &src, &rel_addr);
530 reg_maps->labels[src.reg.idx] = 1;
532 /* Set texture, address, temporary registers */
533 else
535 int i, limit;
537 /* This will loop over all the registers and try to
538 * make a bitmask of the ones we're interested in.
540 * Relative addressing tokens are ignored, but that's
541 * okay, since we'll catch any address registers when
542 * they are initialized (required by spec) */
544 if (ins.dst_count)
546 struct wined3d_shader_dst_param dst_param;
547 struct wined3d_shader_src_param dst_rel_addr;
549 fe->shader_read_dst_param(fe_data, &pToken, &dst_param, &dst_rel_addr);
551 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and
552 * is used in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel
553 * shaders because TECRDOUT isn't used in them, but future register types might cause issues */
554 if (!pshader && shader_version.major < 3 && dst_param.reg.type == WINED3DSPR_TEXCRDOUT)
556 reg_maps->texcoord_mask[dst_param.reg.type] |= dst_param.write_mask;
558 else
560 shader_record_register_usage(This, reg_maps, &dst_param.reg, pshader);
563 /* Declare 1.X samplers implicitly, based on the destination reg. number */
564 if (shader_version.major == 1
565 && pshader /* Filter different instructions with the same enum values in VS */
566 && (ins.handler_idx == WINED3DSIH_TEX
567 || ins.handler_idx == WINED3DSIH_TEXBEM
568 || ins.handler_idx == WINED3DSIH_TEXBEML
569 || ins.handler_idx == WINED3DSIH_TEXDP3TEX
570 || ins.handler_idx == WINED3DSIH_TEXM3x2TEX
571 || ins.handler_idx == WINED3DSIH_TEXM3x3SPEC
572 || ins.handler_idx == WINED3DSIH_TEXM3x3TEX
573 || ins.handler_idx == WINED3DSIH_TEXM3x3VSPEC
574 || ins.handler_idx == WINED3DSIH_TEXREG2AR
575 || ins.handler_idx == WINED3DSIH_TEXREG2GB
576 || ins.handler_idx == WINED3DSIH_TEXREG2RGB))
578 /* Fake sampler usage, only set reserved bit and ttype */
579 DWORD sampler_code = dst_param.reg.idx;
581 TRACE("Setting fake 2D sampler for 1.x pixelshader\n");
582 reg_maps->sampler_type[sampler_code] = WINED3DSTT_2D;
584 /* texbem is only valid with < 1.4 pixel shaders */
585 if (ins.handler_idx == WINED3DSIH_TEXBEM
586 || ins.handler_idx == WINED3DSIH_TEXBEML)
588 reg_maps->bumpmat[sampler_code] = TRUE;
589 if (ins.handler_idx == WINED3DSIH_TEXBEML)
591 reg_maps->luminanceparams[sampler_code] = TRUE;
595 else if (pshader && ins.handler_idx == WINED3DSIH_BEM)
597 reg_maps->bumpmat[dst_param.reg.idx] = TRUE;
601 if (ins.handler_idx == WINED3DSIH_NRM)
603 reg_maps->usesnrm = 1;
605 else if (ins.handler_idx == WINED3DSIH_DSY)
607 reg_maps->usesdsy = 1;
609 else if(ins.handler_idx == WINED3DSIH_TEXLDD)
611 reg_maps->usestexldd = 1;
613 else if(ins.handler_idx == WINED3DSIH_MOVA)
615 reg_maps->usesmova = 1;
618 limit = ins.src_count + (ins.predicate ? 1 : 0);
619 for (i = 0; i < limit; ++i)
621 struct wined3d_shader_src_param src_param, src_rel_addr;
622 unsigned int count;
624 fe->shader_read_src_param(fe_data, &pToken, &src_param, &src_rel_addr);
625 count = get_instr_extra_regcount(ins.handler_idx, i);
627 shader_record_register_usage(This, reg_maps, &src_param.reg, pshader);
628 while (count)
630 ++src_param.reg.idx;
631 shader_record_register_usage(This, reg_maps, &src_param.reg, pshader);
632 --count;
637 reg_maps->loop_depth = max_loop_depth;
639 This->baseShader.functionLength = ((char *)pToken - (char *)byte_code);
641 return WINED3D_OK;
644 static void shader_dump_decl_usage(const struct wined3d_shader_semantic *semantic,
645 const struct wined3d_shader_version *shader_version)
647 TRACE("dcl");
649 if (semantic->reg.reg.type == WINED3DSPR_SAMPLER)
651 switch (semantic->sampler_type)
653 case WINED3DSTT_2D: TRACE("_2d"); break;
654 case WINED3DSTT_CUBE: TRACE("_cube"); break;
655 case WINED3DSTT_VOLUME: TRACE("_volume"); break;
656 default: TRACE("_unknown_ttype(0x%08x)", semantic->sampler_type);
659 else
661 /* Pixel shaders 3.0 don't have usage semantics */
662 if (shader_is_pshader_version(shader_version->type) && shader_version->major < 3)
663 return;
664 else
665 TRACE("_");
667 switch (semantic->usage)
669 case WINED3DDECLUSAGE_POSITION:
670 TRACE("position%d", semantic->usage_idx);
671 break;
672 case WINED3DDECLUSAGE_BLENDINDICES:
673 TRACE("blend");
674 break;
675 case WINED3DDECLUSAGE_BLENDWEIGHT:
676 TRACE("weight");
677 break;
678 case WINED3DDECLUSAGE_NORMAL:
679 TRACE("normal%d", semantic->usage_idx);
680 break;
681 case WINED3DDECLUSAGE_PSIZE:
682 TRACE("psize");
683 break;
684 case WINED3DDECLUSAGE_COLOR:
685 if (semantic->usage_idx == 0) TRACE("color");
686 else TRACE("specular%d", (semantic->usage_idx - 1));
687 break;
688 case WINED3DDECLUSAGE_TEXCOORD:
689 TRACE("texture%d", semantic->usage_idx);
690 break;
691 case WINED3DDECLUSAGE_TANGENT:
692 TRACE("tangent");
693 break;
694 case WINED3DDECLUSAGE_BINORMAL:
695 TRACE("binormal");
696 break;
697 case WINED3DDECLUSAGE_TESSFACTOR:
698 TRACE("tessfactor");
699 break;
700 case WINED3DDECLUSAGE_POSITIONT:
701 TRACE("positionT%d", semantic->usage_idx);
702 break;
703 case WINED3DDECLUSAGE_FOG:
704 TRACE("fog");
705 break;
706 case WINED3DDECLUSAGE_DEPTH:
707 TRACE("depth");
708 break;
709 case WINED3DDECLUSAGE_SAMPLE:
710 TRACE("sample");
711 break;
712 default:
713 FIXME("unknown_semantics(0x%08x)", semantic->usage);
718 static void shader_dump_register(const struct wined3d_shader_register *reg,
719 const struct wined3d_shader_version *shader_version)
721 static const char * const rastout_reg_names[] = {"oPos", "oFog", "oPts"};
722 static const char * const misctype_reg_names[] = {"vPos", "vFace"};
723 UINT offset = reg->idx;
725 switch (reg->type)
727 case WINED3DSPR_TEMP:
728 TRACE("r");
729 break;
731 case WINED3DSPR_INPUT:
732 TRACE("v");
733 break;
735 case WINED3DSPR_CONST:
736 case WINED3DSPR_CONST2:
737 case WINED3DSPR_CONST3:
738 case WINED3DSPR_CONST4:
739 TRACE("c");
740 offset = shader_get_float_offset(reg->type, reg->idx);
741 break;
743 case WINED3DSPR_TEXTURE: /* vs: case WINED3DSPR_ADDR */
744 TRACE("%c", shader_is_pshader_version(shader_version->type) ? 't' : 'a');
745 break;
747 case WINED3DSPR_RASTOUT:
748 TRACE("%s", rastout_reg_names[reg->idx]);
749 break;
751 case WINED3DSPR_COLOROUT:
752 TRACE("oC");
753 break;
755 case WINED3DSPR_DEPTHOUT:
756 TRACE("oDepth");
757 break;
759 case WINED3DSPR_ATTROUT:
760 TRACE("oD");
761 break;
763 case WINED3DSPR_TEXCRDOUT:
764 /* Vertex shaders >= 3.0 use general purpose output registers
765 * (WINED3DSPR_OUTPUT), which can include an address token */
766 if (shader_version->major >= 3) TRACE("o");
767 else TRACE("oT");
768 break;
770 case WINED3DSPR_CONSTINT:
771 TRACE("i");
772 break;
774 case WINED3DSPR_CONSTBOOL:
775 TRACE("b");
776 break;
778 case WINED3DSPR_LABEL:
779 TRACE("l");
780 break;
782 case WINED3DSPR_LOOP:
783 TRACE("aL");
784 break;
786 case WINED3DSPR_SAMPLER:
787 TRACE("s");
788 break;
790 case WINED3DSPR_MISCTYPE:
791 if (reg->idx > 1) FIXME("Unhandled misctype register %d\n", reg->idx);
792 else TRACE("%s", misctype_reg_names[reg->idx]);
793 break;
795 case WINED3DSPR_PREDICATE:
796 TRACE("p");
797 break;
799 case WINED3DSPR_IMMCONST:
800 TRACE("l");
801 break;
803 default:
804 TRACE("unhandled_rtype(%#x)", reg->type);
805 break;
808 if (reg->type == WINED3DSPR_IMMCONST)
810 TRACE("(");
811 switch (reg->immconst_type)
813 case WINED3D_IMMCONST_FLOAT:
814 TRACE("%.8e", *(float *)reg->immconst_data);
815 break;
817 case WINED3D_IMMCONST_FLOAT4:
818 TRACE("%.8e, %.8e, %.8e, %.8e",
819 *(float *)&reg->immconst_data[0], *(float *)&reg->immconst_data[1],
820 *(float *)&reg->immconst_data[2], *(float *)&reg->immconst_data[3]);
821 break;
823 default:
824 TRACE("<unhandled immconst_type %#x>", reg->immconst_type);
825 break;
827 TRACE(")");
829 else if (reg->type != WINED3DSPR_RASTOUT && reg->type != WINED3DSPR_MISCTYPE)
831 if (reg->rel_addr)
833 TRACE("[");
834 shader_dump_src_param(reg->rel_addr, shader_version);
835 TRACE(" + ");
837 TRACE("%u", offset);
838 if (reg->rel_addr) TRACE("]");
842 void shader_dump_dst_param(const struct wined3d_shader_dst_param *param,
843 const struct wined3d_shader_version *shader_version)
845 DWORD write_mask = param->write_mask;
847 shader_dump_register(&param->reg, shader_version);
849 if (write_mask != WINED3DSP_WRITEMASK_ALL)
851 static const char *write_mask_chars = "xyzw";
853 TRACE(".");
854 if (write_mask & WINED3DSP_WRITEMASK_0) TRACE("%c", write_mask_chars[0]);
855 if (write_mask & WINED3DSP_WRITEMASK_1) TRACE("%c", write_mask_chars[1]);
856 if (write_mask & WINED3DSP_WRITEMASK_2) TRACE("%c", write_mask_chars[2]);
857 if (write_mask & WINED3DSP_WRITEMASK_3) TRACE("%c", write_mask_chars[3]);
861 void shader_dump_src_param(const struct wined3d_shader_src_param *param,
862 const struct wined3d_shader_version *shader_version)
864 DWORD src_modifier = param->modifiers;
865 DWORD swizzle = param->swizzle;
867 if (src_modifier == WINED3DSPSM_NEG
868 || src_modifier == WINED3DSPSM_BIASNEG
869 || src_modifier == WINED3DSPSM_SIGNNEG
870 || src_modifier == WINED3DSPSM_X2NEG
871 || src_modifier == WINED3DSPSM_ABSNEG)
872 TRACE("-");
873 else if (src_modifier == WINED3DSPSM_COMP)
874 TRACE("1-");
875 else if (src_modifier == WINED3DSPSM_NOT)
876 TRACE("!");
878 if (src_modifier == WINED3DSPSM_ABS || src_modifier == WINED3DSPSM_ABSNEG)
879 TRACE("abs(");
881 shader_dump_register(&param->reg, shader_version);
883 if (src_modifier)
885 switch (src_modifier)
887 case WINED3DSPSM_NONE: break;
888 case WINED3DSPSM_NEG: break;
889 case WINED3DSPSM_NOT: break;
890 case WINED3DSPSM_BIAS: TRACE("_bias"); break;
891 case WINED3DSPSM_BIASNEG: TRACE("_bias"); break;
892 case WINED3DSPSM_SIGN: TRACE("_bx2"); break;
893 case WINED3DSPSM_SIGNNEG: TRACE("_bx2"); break;
894 case WINED3DSPSM_COMP: break;
895 case WINED3DSPSM_X2: TRACE("_x2"); break;
896 case WINED3DSPSM_X2NEG: TRACE("_x2"); break;
897 case WINED3DSPSM_DZ: TRACE("_dz"); break;
898 case WINED3DSPSM_DW: TRACE("_dw"); break;
899 case WINED3DSPSM_ABSNEG: TRACE(")"); break;
900 case WINED3DSPSM_ABS: TRACE(")"); break;
901 default:
902 TRACE("_unknown_modifier(%#x)", src_modifier);
906 if (swizzle != WINED3DSP_NOSWIZZLE)
908 static const char *swizzle_chars = "xyzw";
909 DWORD swizzle_x = swizzle & 0x03;
910 DWORD swizzle_y = (swizzle >> 2) & 0x03;
911 DWORD swizzle_z = (swizzle >> 4) & 0x03;
912 DWORD swizzle_w = (swizzle >> 6) & 0x03;
914 if (swizzle_x == swizzle_y
915 && swizzle_x == swizzle_z
916 && swizzle_x == swizzle_w)
918 TRACE(".%c", swizzle_chars[swizzle_x]);
920 else
922 TRACE(".%c%c%c%c", swizzle_chars[swizzle_x], swizzle_chars[swizzle_y],
923 swizzle_chars[swizzle_z], swizzle_chars[swizzle_w]);
928 /* Shared code in order to generate the bulk of the shader string.
929 * NOTE: A description of how to parse tokens can be found on msdn */
930 void shader_generate_main(IWineD3DBaseShader *iface, SHADER_BUFFER *buffer,
931 const shader_reg_maps *reg_maps, const DWORD *pFunction, void *backend_ctx)
933 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
934 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device; /* To access shader backend callbacks */
935 const SHADER_HANDLER *handler_table = device->shader_backend->shader_instruction_handler_table;
936 const struct wined3d_shader_frontend *fe = This->baseShader.frontend;
937 void *fe_data = This->baseShader.frontend_data;
938 struct wined3d_shader_src_param src_rel_addr[4];
939 struct wined3d_shader_src_param src_param[4];
940 struct wined3d_shader_version shader_version;
941 struct wined3d_shader_src_param dst_rel_addr;
942 struct wined3d_shader_dst_param dst_param;
943 struct wined3d_shader_instruction ins;
944 struct wined3d_shader_context ctx;
945 const DWORD *pToken = pFunction;
946 SHADER_HANDLER hw_fct;
947 DWORD i;
949 /* Initialize current parsing state */
950 ctx.shader = iface;
951 ctx.reg_maps = reg_maps;
952 ctx.buffer = buffer;
953 ctx.backend_data = backend_ctx;
955 ins.ctx = &ctx;
956 ins.dst = &dst_param;
957 ins.src = src_param;
958 This->baseShader.parse_state.current_row = 0;
960 fe->shader_read_header(fe_data, &pToken, &shader_version);
962 while (!fe->shader_is_end(fe_data, &pToken))
964 const char *comment;
965 UINT param_size;
967 /* Skip comment tokens */
968 fe->shader_read_comment(&pToken, &comment);
969 if (comment) continue;
971 /* Read opcode */
972 fe->shader_read_opcode(fe_data, &pToken, &ins, &param_size);
974 /* Unknown opcode and its parameters */
975 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
977 TRACE("Skipping unrecognized instruction.\n");
978 pToken += param_size;
979 continue;
982 /* Nothing to do */
983 if (ins.handler_idx == WINED3DSIH_DCL
984 || ins.handler_idx == WINED3DSIH_NOP
985 || ins.handler_idx == WINED3DSIH_DEF
986 || ins.handler_idx == WINED3DSIH_DEFI
987 || ins.handler_idx == WINED3DSIH_DEFB
988 || ins.handler_idx == WINED3DSIH_PHASE
989 || ins.handler_idx == WINED3DSIH_RET)
991 pToken += param_size;
992 continue;
995 /* Select handler */
996 hw_fct = handler_table[ins.handler_idx];
998 /* Unhandled opcode */
999 if (!hw_fct)
1001 FIXME("Backend can't handle opcode %#x\n", ins.handler_idx);
1002 pToken += param_size;
1003 continue;
1006 /* Destination token */
1007 if (ins.dst_count) fe->shader_read_dst_param(fe_data, &pToken, &dst_param, &dst_rel_addr);
1009 /* Predication token */
1010 if (ins.predicate) ins.predicate = *pToken++;
1012 /* Other source tokens */
1013 for (i = 0; i < ins.src_count; ++i)
1015 fe->shader_read_src_param(fe_data, &pToken, &src_param[i], &src_rel_addr[i]);
1018 /* Call appropriate function for output target */
1019 hw_fct(&ins);
1021 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
1022 device->shader_backend->shader_add_instruction_modifiers(&ins);
1026 static void shader_dump_ins_modifiers(const struct wined3d_shader_dst_param *dst)
1028 DWORD mmask = dst->modifiers;
1030 switch (dst->shift)
1032 case 0: break;
1033 case 13: TRACE("_d8"); break;
1034 case 14: TRACE("_d4"); break;
1035 case 15: TRACE("_d2"); break;
1036 case 1: TRACE("_x2"); break;
1037 case 2: TRACE("_x4"); break;
1038 case 3: TRACE("_x8"); break;
1039 default: TRACE("_unhandled_shift(%d)", dst->shift); break;
1042 if (mmask & WINED3DSPDM_SATURATE) TRACE("_sat");
1043 if (mmask & WINED3DSPDM_PARTIALPRECISION) TRACE("_pp");
1044 if (mmask & WINED3DSPDM_MSAMPCENTROID) TRACE("_centroid");
1046 mmask &= ~(WINED3DSPDM_SATURATE | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_MSAMPCENTROID);
1047 if (mmask)
1048 FIXME("_unrecognized_modifier(%#x)", mmask);
1051 void shader_trace_init(const struct wined3d_shader_frontend *fe, void *fe_data, const DWORD *pFunction)
1053 struct wined3d_shader_version shader_version;
1054 const DWORD* pToken = pFunction;
1055 DWORD i;
1057 TRACE("Parsing %p\n", pFunction);
1059 fe->shader_read_header(fe_data, &pToken, &shader_version);
1061 TRACE("%s_%u_%u\n", shader_is_pshader_version(shader_version.type) ? "ps": "vs",
1062 shader_version.major, shader_version.minor);
1064 while (!fe->shader_is_end(fe_data, &pToken))
1066 struct wined3d_shader_instruction ins;
1067 const char *comment;
1068 UINT param_size;
1070 /* comment */
1071 fe->shader_read_comment(&pToken, &comment);
1072 if (comment)
1074 TRACE("//%s\n", comment);
1075 continue;
1078 fe->shader_read_opcode(fe_data, &pToken, &ins, &param_size);
1079 if (ins.handler_idx == WINED3DSIH_TABLE_SIZE)
1081 TRACE("Skipping unrecognized instruction.\n");
1082 pToken += param_size;
1083 continue;
1086 if (ins.handler_idx == WINED3DSIH_DCL)
1088 struct wined3d_shader_semantic semantic;
1090 fe->shader_read_semantic(&pToken, &semantic);
1092 shader_dump_decl_usage(&semantic, &shader_version);
1093 shader_dump_ins_modifiers(&semantic.reg);
1094 TRACE(" ");
1095 shader_dump_dst_param(&semantic.reg, &shader_version);
1097 else if (ins.handler_idx == WINED3DSIH_DEF)
1099 struct wined3d_shader_dst_param dst;
1100 struct wined3d_shader_src_param rel_addr;
1102 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
1104 TRACE("def c%u = %f, %f, %f, %f", shader_get_float_offset(dst.reg.type, dst.reg.idx),
1105 *(const float *)(pToken),
1106 *(const float *)(pToken + 1),
1107 *(const float *)(pToken + 2),
1108 *(const float *)(pToken + 3));
1109 pToken += 4;
1111 else if (ins.handler_idx == WINED3DSIH_DEFI)
1113 struct wined3d_shader_dst_param dst;
1114 struct wined3d_shader_src_param rel_addr;
1116 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
1118 TRACE("defi i%u = %d, %d, %d, %d", dst.reg.idx,
1119 *(pToken),
1120 *(pToken + 1),
1121 *(pToken + 2),
1122 *(pToken + 3));
1123 pToken += 4;
1125 else if (ins.handler_idx == WINED3DSIH_DEFB)
1127 struct wined3d_shader_dst_param dst;
1128 struct wined3d_shader_src_param rel_addr;
1130 fe->shader_read_dst_param(fe_data, &pToken, &dst, &rel_addr);
1132 TRACE("defb b%u = %s", dst.reg.idx, *pToken ? "true" : "false");
1133 ++pToken;
1135 else
1137 struct wined3d_shader_src_param dst_rel_addr, src_rel_addr;
1138 struct wined3d_shader_dst_param dst_param;
1139 struct wined3d_shader_src_param src_param;
1141 if (ins.dst_count)
1143 fe->shader_read_dst_param(fe_data, &pToken, &dst_param, &dst_rel_addr);
1146 /* Print out predication source token first - it follows
1147 * the destination token. */
1148 if (ins.predicate)
1150 fe->shader_read_src_param(fe_data, &pToken, &src_param, &src_rel_addr);
1151 TRACE("(");
1152 shader_dump_src_param(&src_param, &shader_version);
1153 TRACE(") ");
1156 /* PixWin marks instructions with the coissue flag with a '+' */
1157 if (ins.coissue) TRACE("+");
1159 TRACE("%s", shader_opcode_names[ins.handler_idx]);
1161 if (ins.handler_idx == WINED3DSIH_IFC
1162 || ins.handler_idx == WINED3DSIH_BREAKC)
1164 switch (ins.flags)
1166 case COMPARISON_GT: TRACE("_gt"); break;
1167 case COMPARISON_EQ: TRACE("_eq"); break;
1168 case COMPARISON_GE: TRACE("_ge"); break;
1169 case COMPARISON_LT: TRACE("_lt"); break;
1170 case COMPARISON_NE: TRACE("_ne"); break;
1171 case COMPARISON_LE: TRACE("_le"); break;
1172 default: TRACE("_(%u)", ins.flags);
1175 else if (ins.handler_idx == WINED3DSIH_TEX
1176 && shader_version.major >= 2
1177 && (ins.flags & WINED3DSI_TEXLD_PROJECT))
1179 TRACE("p");
1182 /* We already read the destination token, print it. */
1183 if (ins.dst_count)
1185 shader_dump_ins_modifiers(&dst_param);
1186 TRACE(" ");
1187 shader_dump_dst_param(&dst_param, &shader_version);
1190 /* Other source tokens */
1191 for (i = ins.dst_count; i < (ins.dst_count + ins.src_count); ++i)
1193 fe->shader_read_src_param(fe_data, &pToken, &src_param, &src_rel_addr);
1194 TRACE(!i ? " " : ", ");
1195 shader_dump_src_param(&src_param, &shader_version);
1198 TRACE("\n");
1202 void shader_cleanup(IWineD3DBaseShader *iface)
1204 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)iface;
1206 ((IWineD3DDeviceImpl *)This->baseShader.device)->shader_backend->shader_destroy(iface);
1207 HeapFree(GetProcessHeap(), 0, This->baseShader.reg_maps.constf);
1208 HeapFree(GetProcessHeap(), 0, This->baseShader.function);
1209 shader_delete_constant_list(&This->baseShader.constantsF);
1210 shader_delete_constant_list(&This->baseShader.constantsB);
1211 shader_delete_constant_list(&This->baseShader.constantsI);
1212 list_remove(&This->baseShader.shader_list_entry);
1214 if (This->baseShader.frontend && This->baseShader.frontend_data)
1216 This->baseShader.frontend->shader_free(This->baseShader.frontend_data);
1220 static const SHADER_HANDLER shader_none_instruction_handler_table[WINED3DSIH_TABLE_SIZE] = {0};
1221 static void shader_none_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {}
1222 static void shader_none_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {}
1223 static void shader_none_deselect_depth_blt(IWineD3DDevice *iface) {}
1224 static void shader_none_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1225 static void shader_none_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count) {}
1226 static void shader_none_load_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1227 static void shader_none_load_np2fixup_constants(IWineD3DDevice *iface, char usePS, char useVS) {}
1228 static void shader_none_destroy(IWineD3DBaseShader *iface) {}
1229 static HRESULT shader_none_alloc(IWineD3DDevice *iface) {return WINED3D_OK;}
1230 static void shader_none_free(IWineD3DDevice *iface) {}
1231 static BOOL shader_none_dirty_const(IWineD3DDevice *iface) {return FALSE;}
1232 static GLuint shader_none_generate_pshader(IWineD3DPixelShader *iface,
1233 SHADER_BUFFER *buffer, const struct ps_compile_args *args)
1235 FIXME("NONE shader backend asked to generate a pixel shader\n");
1236 return 0;
1238 static GLuint shader_none_generate_vshader(IWineD3DVertexShader *iface,
1239 SHADER_BUFFER *buffer, const struct vs_compile_args *args)
1241 FIXME("NONE shader backend asked to generate a vertex shader\n");
1242 return 0;
1244 static void shader_none_add_instruction_modifiers(const struct wined3d_shader_instruction *ins) {}
1246 #define GLINFO_LOCATION (*gl_info)
1247 static void shader_none_get_caps(WINED3DDEVTYPE devtype, const WineD3D_GL_Info *gl_info, struct shader_caps *pCaps)
1249 /* Set the shader caps to 0 for the none shader backend */
1250 pCaps->VertexShaderVersion = 0;
1251 pCaps->PixelShaderVersion = 0;
1252 pCaps->PixelShader1xMaxValue = 0.0;
1254 #undef GLINFO_LOCATION
1255 static BOOL shader_none_color_fixup_supported(struct color_fixup_desc fixup)
1257 if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
1259 TRACE("Checking support for fixup:\n");
1260 dump_color_fixup_desc(fixup);
1263 /* Faked to make some apps happy. */
1264 if (!is_yuv_fixup(fixup))
1266 TRACE("[OK]\n");
1267 return TRUE;
1270 TRACE("[FAILED]\n");
1271 return FALSE;
1274 const shader_backend_t none_shader_backend = {
1275 shader_none_instruction_handler_table,
1276 shader_none_select,
1277 shader_none_select_depth_blt,
1278 shader_none_deselect_depth_blt,
1279 shader_none_update_float_vertex_constants,
1280 shader_none_update_float_pixel_constants,
1281 shader_none_load_constants,
1282 shader_none_load_np2fixup_constants,
1283 shader_none_destroy,
1284 shader_none_alloc,
1285 shader_none_free,
1286 shader_none_dirty_const,
1287 shader_none_generate_pshader,
1288 shader_none_generate_vshader,
1289 shader_none_get_caps,
1290 shader_none_color_fixup_supported,
1291 shader_none_add_instruction_modifiers,