kernel32: Add support for completion routine invocation in ReadDirectoryChangesW.
[wine] / dlls / wined3d / shader_sm4.c
1 /*
2  * Copyright 2009 Henri Verbeet for CodeWeavers
3  *
4  * This library is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU Lesser General Public
6  * License as published by the Free Software Foundation; either
7  * version 2.1 of the License, or (at your option) any later version.
8  *
9  * This library is distributed in the hope that it will be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12  * Lesser General Public License for more details.
13  *
14  * You should have received a copy of the GNU Lesser General Public
15  * License along with this library; if not, write to the Free Software
16  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
17  */
18
19 #include "config.h"
20 #include "wine/port.h"
21
22 #include "wined3d_private.h"
23
24 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
25
26 #define WINED3D_SM4_INSTRUCTION_MODIFIER        (1 << 31)
27
28 #define WINED3D_SM4_INSTRUCTION_LENGTH_SHIFT    24
29 #define WINED3D_SM4_INSTRUCTION_LENGTH_MASK     (0xf << WINED3D_SM4_INSTRUCTION_LENGTH_SHIFT)
30
31 #define WINED3D_SM4_OPCODE_MASK                 0xff
32
33 #define WINED3D_SM4_REGISTER_MODIFIER           (1 << 31)
34
35 #define WINED3D_SM4_REGISTER_ORDER_SHIFT        20
36 #define WINED3D_SM4_REGISTER_ORDER_MASK         (0x3 << WINED3D_SM4_REGISTER_ORDER_SHIFT)
37
38 #define WINED3D_SM4_REGISTER_TYPE_SHIFT         12
39 #define WINED3D_SM4_REGISTER_TYPE_MASK          (0xf << WINED3D_SM4_REGISTER_TYPE_SHIFT)
40
41 #define WINED3D_SM4_IMMCONST_TYPE_SHIFT         0
42 #define WINED3D_SM4_IMMCONST_TYPE_MASK          (0x3 << WINED3D_SM4_IMMCONST_TYPE_SHIFT)
43
44 #define WINED3D_SM4_WRITEMASK_SHIFT             4
45 #define WINED3D_SM4_WRITEMASK_MASK              (0xf << WINED3D_SM4_WRITEMASK_SHIFT)
46
47 #define WINED3D_SM4_SWIZZLE_SHIFT               4
48 #define WINED3D_SM4_SWIZZLE_MASK                (0xff << WINED3D_SM4_SWIZZLE_SHIFT)
49
50 #define WINED3D_SM4_VERSION_MAJOR(version)      (((version) >> 4) & 0xf)
51 #define WINED3D_SM4_VERSION_MINOR(version)      (((version) >> 0) & 0xf)
52
53 enum wined3d_sm4_opcode
54 {
55     WINED3D_SM4_OP_ADD          = 0x00,
56     WINED3D_SM4_OP_AND          = 0x01,
57     WINED3D_SM4_OP_BREAK        = 0x02,
58     WINED3D_SM4_OP_BREAKC       = 0x03,
59     WINED3D_SM4_OP_CUT          = 0x09,
60     WINED3D_SM4_OP_DIV          = 0x0e,
61     WINED3D_SM4_OP_DP3          = 0x10,
62     WINED3D_SM4_OP_DP4          = 0x11,
63     WINED3D_SM4_OP_EMIT         = 0x13,
64     WINED3D_SM4_OP_ENDIF        = 0x15,
65     WINED3D_SM4_OP_ENDLOOP      = 0x16,
66     WINED3D_SM4_OP_EXP          = 0x19,
67     WINED3D_SM4_OP_FRC          = 0x1a,
68     WINED3D_SM4_OP_FTOI         = 0x1b,
69     WINED3D_SM4_OP_IADD         = 0x1e,
70     WINED3D_SM4_OP_IF           = 0x1f,
71     WINED3D_SM4_OP_IEQ          = 0x20,
72     WINED3D_SM4_OP_IGE          = 0x21,
73     WINED3D_SM4_OP_IMUL         = 0x26,
74     WINED3D_SM4_OP_ITOF         = 0x2b,
75     WINED3D_SM4_OP_LD           = 0x2d,
76     WINED3D_SM4_OP_LOG          = 0x2f,
77     WINED3D_SM4_OP_LOOP         = 0x30,
78     WINED3D_SM4_OP_LT           = 0x31,
79     WINED3D_SM4_OP_MAD          = 0x32,
80     WINED3D_SM4_OP_MIN          = 0x33,
81     WINED3D_SM4_OP_MAX          = 0x34,
82     WINED3D_SM4_OP_MOV          = 0x36,
83     WINED3D_SM4_OP_MOVC         = 0x37,
84     WINED3D_SM4_OP_MUL          = 0x38,
85     WINED3D_SM4_OP_RET          = 0x3e,
86     WINED3D_SM4_OP_RSQ          = 0x44,
87     WINED3D_SM4_OP_SAMPLE       = 0x45,
88     WINED3D_SM4_OP_SAMPLE_LOD   = 0x48,
89     WINED3D_SM4_OP_SAMPLE_GRAD  = 0x49,
90     WINED3D_SM4_OP_SQRT         = 0x4b,
91     WINED3D_SM4_OP_SINCOS       = 0x4d,
92     WINED3D_SM4_OP_UTOF         = 0x56,
93 };
94
95 enum wined3d_sm4_register_type
96 {
97     WINED3D_SM4_RT_TEMP         = 0x0,
98     WINED3D_SM4_RT_INPUT        = 0x1,
99     WINED3D_SM4_RT_OUTPUT       = 0x2,
100     WINED3D_SM4_RT_IMMCONST     = 0x4,
101     WINED3D_SM4_RT_SAMPLER      = 0x6,
102     WINED3D_SM4_RT_CONSTBUFFER  = 0x8,
103     WINED3D_SM4_RT_NULL         = 0xd,
104 };
105
106 enum wined3d_sm4_immconst_type
107 {
108     WINED3D_SM4_IMMCONST_SCALAR = 0x1,
109     WINED3D_SM4_IMMCONST_VEC4   = 0x2,
110 };
111
112 struct wined3d_sm4_data
113 {
114     struct wined3d_shader_version shader_version;
115     const DWORD *end;
116     const struct wined3d_shader_signature *output_signature;
117 };
118
119 struct wined3d_sm4_opcode_info
120 {
121     enum wined3d_sm4_opcode opcode;
122     enum WINED3D_SHADER_INSTRUCTION_HANDLER handler_idx;
123     UINT dst_count;
124     UINT src_count;
125 };
126
127 struct sysval_map
128 {
129     enum wined3d_sysval_semantic sysval;
130     WINED3DSHADER_PARAM_REGISTER_TYPE register_type;
131     UINT register_idx;
132 };
133
134 static const struct wined3d_sm4_opcode_info opcode_table[] =
135 {
136     {WINED3D_SM4_OP_ADD,        WINED3DSIH_ADD,         1,  2},
137     {WINED3D_SM4_OP_AND,        WINED3DSIH_AND,         1,  2},
138     {WINED3D_SM4_OP_BREAK,      WINED3DSIH_BREAK,       0,  0},
139     {WINED3D_SM4_OP_BREAKC,     WINED3DSIH_BREAKP,      0,  1},
140     {WINED3D_SM4_OP_CUT,        WINED3DSIH_CUT,         0,  0},
141     {WINED3D_SM4_OP_DIV,        WINED3DSIH_DIV,         1,  2},
142     {WINED3D_SM4_OP_DP3,        WINED3DSIH_DP3,         1,  2},
143     {WINED3D_SM4_OP_DP4,        WINED3DSIH_DP4,         1,  2},
144     {WINED3D_SM4_OP_EMIT,       WINED3DSIH_EMIT,        0,  0},
145     {WINED3D_SM4_OP_ENDIF,      WINED3DSIH_ENDIF,       0,  0},
146     {WINED3D_SM4_OP_ENDLOOP,    WINED3DSIH_ENDLOOP,     0,  0},
147     {WINED3D_SM4_OP_EXP,        WINED3DSIH_EXP,         1,  1},
148     {WINED3D_SM4_OP_FRC,        WINED3DSIH_FRC,         1,  1},
149     {WINED3D_SM4_OP_FTOI,       WINED3DSIH_FTOI,        1,  1},
150     {WINED3D_SM4_OP_IADD,       WINED3DSIH_IADD,        1,  2},
151     {WINED3D_SM4_OP_IF,         WINED3DSIH_IF,          0,  1},
152     {WINED3D_SM4_OP_IEQ,        WINED3DSIH_IEQ,         1,  2},
153     {WINED3D_SM4_OP_IGE,        WINED3DSIH_IGE,         1,  2},
154     {WINED3D_SM4_OP_IMUL,       WINED3DSIH_IMUL,        2,  2},
155     {WINED3D_SM4_OP_ITOF,       WINED3DSIH_ITOF,        1,  1},
156     {WINED3D_SM4_OP_LD,         WINED3DSIH_LD,          1,  2},
157     {WINED3D_SM4_OP_LOG,        WINED3DSIH_LOG,         1,  1},
158     {WINED3D_SM4_OP_LOOP,       WINED3DSIH_LOOP,        0,  0},
159     {WINED3D_SM4_OP_LT,         WINED3DSIH_LT,          1,  2},
160     {WINED3D_SM4_OP_MAD,        WINED3DSIH_MAD,         1,  3},
161     {WINED3D_SM4_OP_MIN,        WINED3DSIH_MIN,         1,  2},
162     {WINED3D_SM4_OP_MAX,        WINED3DSIH_MAX,         1,  2},
163     {WINED3D_SM4_OP_MOV,        WINED3DSIH_MOV,         1,  1},
164     {WINED3D_SM4_OP_MOVC,       WINED3DSIH_MOVC,        1,  3},
165     {WINED3D_SM4_OP_MUL,        WINED3DSIH_MUL,         1,  2},
166     {WINED3D_SM4_OP_RET,        WINED3DSIH_RET,         0,  0},
167     {WINED3D_SM4_OP_RSQ,        WINED3DSIH_RSQ,         1,  1},
168     {WINED3D_SM4_OP_SAMPLE,     WINED3DSIH_SAMPLE,      1,  3},
169     {WINED3D_SM4_OP_SAMPLE_LOD, WINED3DSIH_SAMPLE_LOD,  1,  4},
170     {WINED3D_SM4_OP_SAMPLE_GRAD,WINED3DSIH_SAMPLE_GRAD, 1,  5},
171     {WINED3D_SM4_OP_SQRT,       WINED3DSIH_SQRT,        1,  1},
172     {WINED3D_SM4_OP_SINCOS,     WINED3DSIH_SINCOS,      2,  1},
173     {WINED3D_SM4_OP_UTOF,       WINED3DSIH_UTOF,        1,  1},
174 };
175
176 static const WINED3DSHADER_PARAM_REGISTER_TYPE register_type_table[] =
177 {
178     /* WINED3D_SM4_RT_TEMP */           WINED3DSPR_TEMP,
179     /* WINED3D_SM4_RT_INPUT */          WINED3DSPR_INPUT,
180     /* WINED3D_SM4_RT_OUTPUT */         WINED3DSPR_OUTPUT,
181     /* UNKNOWN */                       0,
182     /* WINED3D_SM4_RT_IMMCONST */       WINED3DSPR_IMMCONST,
183     /* UNKNOWN */                       0,
184     /* WINED3D_SM4_RT_SAMPLER */        WINED3DSPR_SAMPLER,
185     /* WINED3D_SM4_RT_RESOURCE */       WINED3DSPR_RESOURCE,
186     /* WINED3D_SM4_RT_CONSTBUFFER */    WINED3DSPR_CONSTBUFFER,
187     /* UNKNOWN */                       0,
188     /* UNKNOWN */                       0,
189     /* UNKNOWN */                       0,
190     /* UNKNOWN */                       0,
191     /* WINED3D_SM4_RT_NULL */           WINED3DSPR_NULL,
192 };
193
194 static const struct sysval_map sysval_map[] =
195 {
196     {WINED3D_SV_DEPTH,      WINED3DSPR_DEPTHOUT,    0},
197     {WINED3D_SV_TARGET0,    WINED3DSPR_COLOROUT,    0},
198     {WINED3D_SV_TARGET1,    WINED3DSPR_COLOROUT,    1},
199     {WINED3D_SV_TARGET2,    WINED3DSPR_COLOROUT,    2},
200     {WINED3D_SV_TARGET3,    WINED3DSPR_COLOROUT,    3},
201     {WINED3D_SV_TARGET4,    WINED3DSPR_COLOROUT,    4},
202     {WINED3D_SV_TARGET5,    WINED3DSPR_COLOROUT,    5},
203     {WINED3D_SV_TARGET6,    WINED3DSPR_COLOROUT,    6},
204     {WINED3D_SV_TARGET7,    WINED3DSPR_COLOROUT,    7},
205 };
206
207 static const struct wined3d_sm4_opcode_info *get_opcode_info(enum wined3d_sm4_opcode opcode)
208 {
209     unsigned int i;
210
211     for (i = 0; i < sizeof(opcode_table) / sizeof(*opcode_table); ++i)
212     {
213         if (opcode == opcode_table[i].opcode) return &opcode_table[i];
214     }
215
216     return NULL;
217 }
218
219 static void map_sysval(enum wined3d_sysval_semantic sysval, struct wined3d_shader_register *reg)
220 {
221     unsigned int i;
222
223     for (i = 0; i < sizeof(sysval_map) / sizeof(*sysval_map); ++i)
224     {
225         if (sysval == sysval_map[i].sysval)
226         {
227             reg->type = sysval_map[i].register_type;
228             reg->idx = sysval_map[i].register_idx;
229         }
230     }
231 }
232
233 static void map_register(struct wined3d_sm4_data *priv, struct wined3d_shader_register *reg)
234 {
235     switch (priv->shader_version.type)
236     {
237         case WINED3D_SHADER_TYPE_PIXEL:
238             if (reg->type == WINED3DSPR_OUTPUT)
239             {
240                 unsigned int i;
241                 const struct wined3d_shader_signature *s = priv->output_signature;
242
243                 if (!s)
244                 {
245                     ERR("Shader has no output signature, unable to map register.\n");
246                     break;
247                 }
248
249                 for (i = 0; i < s->element_count; ++i)
250                 {
251                     if (s->elements[i].register_idx == reg->idx)
252                     {
253                         map_sysval(s->elements[i].sysval_semantic, reg);
254                         break;
255                     }
256                 }
257             }
258             break;
259
260         default:
261             break;
262     }
263 }
264
265 static void *shader_sm4_init(const DWORD *byte_code, const struct wined3d_shader_signature *output_signature)
266 {
267     struct wined3d_sm4_data *priv = HeapAlloc(GetProcessHeap(), 0, sizeof(*priv));
268     if (!priv)
269     {
270         ERR("Failed to allocate private data\n");
271         return NULL;
272     }
273
274     priv->output_signature = output_signature;
275
276     return priv;
277 }
278
279 static void shader_sm4_free(void *data)
280 {
281     HeapFree(GetProcessHeap(), 0, data);
282 }
283
284 static void shader_sm4_read_header(void *data, const DWORD **ptr, struct wined3d_shader_version *shader_version)
285 {
286     struct wined3d_sm4_data *priv = data;
287     DWORD version_token;
288
289     priv->end = *ptr;
290
291     version_token = *(*ptr)++;
292     TRACE("version: 0x%08x\n", version_token);
293
294     TRACE("token count: %u\n", **ptr);
295     priv->end += *(*ptr)++;
296
297     switch (version_token >> 16)
298     {
299         case WINED3D_SM4_PS:
300             priv->shader_version.type = WINED3D_SHADER_TYPE_PIXEL;
301             break;
302
303         case WINED3D_SM4_VS:
304             priv->shader_version.type = WINED3D_SHADER_TYPE_VERTEX;
305             break;
306
307         case WINED3D_SM4_GS:
308             priv->shader_version.type = WINED3D_SHADER_TYPE_GEOMETRY;
309             break;
310
311         default:
312             FIXME("Unrecognized shader type %#x\n", version_token >> 16);
313     }
314     priv->shader_version.major = WINED3D_SM4_VERSION_MAJOR(version_token);
315     priv->shader_version.minor = WINED3D_SM4_VERSION_MINOR(version_token);
316
317     *shader_version = priv->shader_version;
318 }
319
320 static void shader_sm4_read_opcode(void *data, const DWORD **ptr, struct wined3d_shader_instruction *ins,
321         UINT *param_size)
322 {
323     const struct wined3d_sm4_opcode_info *opcode_info;
324     DWORD token = *(*ptr)++;
325     DWORD opcode = token & WINED3D_SM4_OPCODE_MASK;
326
327     *param_size = ((token & WINED3D_SM4_INSTRUCTION_LENGTH_MASK) >> WINED3D_SM4_INSTRUCTION_LENGTH_SHIFT) - 1;
328
329     opcode_info = get_opcode_info(opcode);
330     if (!opcode_info)
331     {
332         FIXME("Unrecognized opcode %#x, token 0x%08x\n", opcode, token);
333         ins->handler_idx = WINED3DSIH_TABLE_SIZE;
334         return;
335     }
336
337     ins->handler_idx = opcode_info->handler_idx;
338     ins->flags = 0;
339     ins->coissue = 0;
340     ins->predicate = 0;
341     ins->dst_count = opcode_info->dst_count;
342     ins->src_count = opcode_info->src_count;
343
344     if (token & WINED3D_SM4_INSTRUCTION_MODIFIER)
345     {
346         DWORD modifier = *(*ptr)++;
347         FIXME("Skipping modifier 0x%08x.\n", modifier);
348     }
349 }
350
351 static void shader_sm4_read_src_param(void *data, const DWORD **ptr, struct wined3d_shader_src_param *src_param,
352         struct wined3d_shader_src_param *src_rel_addr)
353 {
354     struct wined3d_sm4_data *priv = data;
355     DWORD token = *(*ptr)++;
356     enum wined3d_sm4_register_type register_type;
357     DWORD order;
358
359     register_type = (token & WINED3D_SM4_REGISTER_TYPE_MASK) >> WINED3D_SM4_REGISTER_TYPE_SHIFT;
360     if (register_type >= sizeof(register_type_table) / sizeof(*register_type_table))
361     {
362         FIXME("Unhandled register type %#x\n", register_type);
363         src_param->reg.type = WINED3DSPR_TEMP;
364     }
365     else
366     {
367         src_param->reg.type = register_type_table[register_type];
368     }
369
370     if (token & WINED3D_SM4_REGISTER_MODIFIER)
371     {
372         DWORD modifier = *(*ptr)++;
373
374         /* FIXME: This will probably break down at some point. The SM4
375          * modifiers look like flags, while wined3d currently has an enum
376          * with possible combinations, e.g. WINED3DSPSM_ABSNEG. */
377         switch (modifier)
378         {
379             case 0x41:
380                 src_param->modifiers = WINED3DSPSM_NEG;
381                 break;
382
383             case 0x81:
384                 src_param->modifiers = WINED3DSPSM_ABS;
385                 break;
386
387             default:
388                 FIXME("Skipping modifier 0x%08x.\n", modifier);
389                 src_param->modifiers = WINED3DSPSM_NONE;
390                 break;
391         }
392     }
393     else
394     {
395         src_param->modifiers = WINED3DSPSM_NONE;
396     }
397
398     order = (token & WINED3D_SM4_REGISTER_ORDER_MASK) >> WINED3D_SM4_REGISTER_ORDER_SHIFT;
399
400     if (order < 1) src_param->reg.idx = ~0U;
401     else src_param->reg.idx = *(*ptr)++;
402
403     if (order < 2) src_param->reg.array_idx = ~0U;
404     else src_param->reg.array_idx = *(*ptr)++;
405
406     if (order > 2) FIXME("Unhandled order %u.\n", order);
407
408     if (register_type == WINED3D_SM4_RT_IMMCONST)
409     {
410         enum wined3d_sm4_immconst_type immconst_type =
411                 (token & WINED3D_SM4_IMMCONST_TYPE_MASK) >> WINED3D_SM4_IMMCONST_TYPE_SHIFT;
412         src_param->swizzle = WINED3DSP_NOSWIZZLE;
413
414         switch(immconst_type)
415         {
416             case WINED3D_SM4_IMMCONST_SCALAR:
417                 src_param->reg.immconst_type = WINED3D_IMMCONST_SCALAR;
418                 memcpy(src_param->reg.immconst_data, *ptr, 1 * sizeof(DWORD));
419                 *ptr += 1;
420                 break;
421
422             case WINED3D_SM4_IMMCONST_VEC4:
423                 src_param->reg.immconst_type = WINED3D_IMMCONST_VEC4;
424                 memcpy(src_param->reg.immconst_data, *ptr, 4 * sizeof(DWORD));
425                 *ptr += 4;
426                 break;
427
428             default:
429                 FIXME("Unhandled immediate constant type %#x\n", immconst_type);
430                 break;
431         }
432     }
433     else
434     {
435         src_param->swizzle = (token & WINED3D_SM4_SWIZZLE_MASK) >> WINED3D_SM4_SWIZZLE_SHIFT;
436     }
437
438     src_param->reg.rel_addr = NULL;
439
440     map_register(priv, &src_param->reg);
441 }
442
443 static void shader_sm4_read_dst_param(void *data, const DWORD **ptr, struct wined3d_shader_dst_param *dst_param,
444         struct wined3d_shader_src_param *dst_rel_addr)
445 {
446     struct wined3d_sm4_data *priv = data;
447     DWORD token = *(*ptr)++;
448     enum wined3d_sm4_register_type register_type;
449     DWORD order;
450
451     register_type = (token & WINED3D_SM4_REGISTER_TYPE_MASK) >> WINED3D_SM4_REGISTER_TYPE_SHIFT;
452     if (register_type >= sizeof(register_type_table) / sizeof(*register_type_table))
453     {
454         FIXME("Unhandled register type %#x\n", register_type);
455         dst_param->reg.type = WINED3DSPR_TEMP;
456     }
457     else
458     {
459         dst_param->reg.type = register_type_table[register_type];
460     }
461
462     order = (token & WINED3D_SM4_REGISTER_ORDER_MASK) >> WINED3D_SM4_REGISTER_ORDER_SHIFT;
463
464     if (order < 1) dst_param->reg.idx = ~0U;
465     else dst_param->reg.idx = *(*ptr)++;
466
467     if (order < 2) dst_param->reg.array_idx = ~0U;
468     else dst_param->reg.array_idx = *(*ptr)++;
469
470     if (order > 2) FIXME("Unhandled order %u.\n", order);
471
472     dst_param->write_mask = (token & WINED3D_SM4_WRITEMASK_MASK) >> WINED3D_SM4_WRITEMASK_SHIFT;
473     dst_param->modifiers = 0;
474     dst_param->shift = 0;
475     dst_param->reg.rel_addr = NULL;
476
477     map_register(priv, &dst_param->reg);
478 }
479
480 static void shader_sm4_read_semantic(const DWORD **ptr, struct wined3d_shader_semantic *semantic)
481 {
482     FIXME("ptr %p, semantic %p stub!\n", ptr, semantic);
483 }
484
485 static void shader_sm4_read_comment(const DWORD **ptr, const char **comment, UINT *comment_size)
486 {
487     FIXME("ptr %p, comment %p, comment_size %p stub!\n", ptr, comment, comment_size);
488     *comment = NULL;
489 }
490
491 static BOOL shader_sm4_is_end(void *data, const DWORD **ptr)
492 {
493     struct wined3d_sm4_data *priv = data;
494     return *ptr == priv->end;
495 }
496
497 const struct wined3d_shader_frontend sm4_shader_frontend =
498 {
499     shader_sm4_init,
500     shader_sm4_free,
501     shader_sm4_read_header,
502     shader_sm4_read_opcode,
503     shader_sm4_read_src_param,
504     shader_sm4_read_dst_param,
505     shader_sm4_read_semantic,
506     shader_sm4_read_comment,
507     shader_sm4_is_end,
508 };