1 // Copyright 2016 The SwiftShader Authors. All Rights Reserved.
2 //
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
6 //
7 //    http://www.apache.org/licenses/LICENSE-2.0
8 //
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
14 
15 #ifndef sw_Shader_hpp
16 #define sw_Shader_hpp
17 
18 #include "Common/Types.hpp"
19 
20 #include <string>
21 #include <vector>
22 
23 namespace sw
24 {
25 	class Shader
26 	{
27 	public:
28 		enum ShaderType
29 		{
30 			SHADER_PIXEL = 0xFFFF,
31 			SHADER_VERTEX = 0xFFFE,
32 			SHADER_GEOMETRY = 0xFFFD
33 		};
34 
35 		enum Opcode
36 		{
37 			// Matches order in d3d9types.h
38 			OPCODE_NOP = 0,
39 			OPCODE_MOV,
40 			OPCODE_ADD,
41 			OPCODE_SUB,
42 			OPCODE_MAD,
43 			OPCODE_MUL,
44 			OPCODE_RCPX,
45 			OPCODE_RSQX,
46 			OPCODE_DP3,
47 			OPCODE_DP4,
48 			OPCODE_MIN,
49 			OPCODE_MAX,
50 			OPCODE_SLT,
51 			OPCODE_SGE,
52 			OPCODE_EXP2X,   // D3DSIO_EXP
53 			OPCODE_LOG2X,   // D3DSIO_LOG
54 			OPCODE_LIT,
55 			OPCODE_ATT,   // D3DSIO_DST
56 			OPCODE_LRP,
57 			OPCODE_FRC,
58 			OPCODE_M4X4,
59 			OPCODE_M4X3,
60 			OPCODE_M3X4,
61 			OPCODE_M3X3,
62 			OPCODE_M3X2,
63 			OPCODE_CALL,
64 			OPCODE_CALLNZ,
65 			OPCODE_LOOP,
66 			OPCODE_RET,
67 			OPCODE_ENDLOOP,
68 			OPCODE_LABEL,
69 			OPCODE_DCL,
70 			OPCODE_POWX,
71 			OPCODE_CRS,
72 			OPCODE_SGN,
73 			OPCODE_ABS,
74 			OPCODE_NRM3,   // D3DSIO_NRM
75 			OPCODE_SINCOS,
76 			OPCODE_REP,
77 			OPCODE_ENDREP,
78 			OPCODE_IF,
79 			OPCODE_IFC,
80 			OPCODE_ELSE,
81 			OPCODE_ENDIF,
82 			OPCODE_BREAK,
83 			OPCODE_BREAKC,
84 			OPCODE_MOVA,
85 			OPCODE_DEFB,
86 			OPCODE_DEFI,
87 
88 			OPCODE_TEXCOORD = 64,
89 			OPCODE_TEXKILL,
90 			OPCODE_TEX,
91 			OPCODE_TEXBEM,
92 			OPCODE_TEXBEML,
93 			OPCODE_TEXREG2AR,
94 			OPCODE_TEXREG2GB,
95 			OPCODE_TEXM3X2PAD,
96 			OPCODE_TEXM3X2TEX,
97 			OPCODE_TEXM3X3PAD,
98 			OPCODE_TEXM3X3TEX,
99 			OPCODE_RESERVED0,
100 			OPCODE_TEXM3X3SPEC,
101 			OPCODE_TEXM3X3VSPEC,
102 			OPCODE_EXPP,
103 			OPCODE_LOGP,
104 			OPCODE_CND,
105 			OPCODE_DEF,
106 			OPCODE_TEXREG2RGB,
107 			OPCODE_TEXDP3TEX,
108 			OPCODE_TEXM3X2DEPTH,
109 			OPCODE_TEXDP3,
110 			OPCODE_TEXM3X3,
111 			OPCODE_TEXDEPTH,
112 			OPCODE_CMP0,   // D3DSIO_CMP
113 			OPCODE_BEM,
114 			OPCODE_DP2ADD,
115 			OPCODE_DFDX,   // D3DSIO_DSX
116 			OPCODE_DFDY,   // D3DSIO_DSY
117 			OPCODE_TEXLDD,
118 			OPCODE_CMP,   // D3DSIO_SETP
119 			OPCODE_TEXLDL,
120 			OPCODE_BREAKP,
121 
122 			OPCODE_PHASE = 0xFFFD,
123 			OPCODE_COMMENT = 0xFFFE,
124 			OPCODE_END = 0xFFFF,
125 
126 			OPCODE_PS_1_0 = 0xFFFF0100,
127 			OPCODE_PS_1_1 = 0xFFFF0101,
128 			OPCODE_PS_1_2 = 0xFFFF0102,
129 			OPCODE_PS_1_3 = 0xFFFF0103,
130 			OPCODE_PS_1_4 = 0xFFFF0104,
131 			OPCODE_PS_2_0 = 0xFFFF0200,
132 			OPCODE_PS_2_x = 0xFFFF0201,
133 			OPCODE_PS_3_0 = 0xFFFF0300,
134 
135 			OPCODE_VS_1_0 = 0xFFFE0100,
136 			OPCODE_VS_1_1 = 0xFFFE0101,
137 			OPCODE_VS_2_0 = 0xFFFE0200,
138 			OPCODE_VS_2_x = 0xFFFE0201,
139 			OPCODE_VS_2_sw = 0xFFFE02FF,
140 			OPCODE_VS_3_0 = 0xFFFE0300,
141 			OPCODE_VS_3_sw = 0xFFFE03FF,
142 
143 			OPCODE_NULL = 0x10000000,   // Dead instruction, to be eliminated
144 			OPCODE_WHILE,
145 			OPCODE_ENDWHILE,
146 			OPCODE_COS,
147 			OPCODE_SIN,
148 			OPCODE_TAN,
149 			OPCODE_ACOS,
150 			OPCODE_ASIN,
151 			OPCODE_ATAN,
152 			OPCODE_ATAN2,
153 			OPCODE_COSH,
154 			OPCODE_SINH,
155 			OPCODE_TANH,
156 			OPCODE_ACOSH,
157 			OPCODE_ASINH,
158 			OPCODE_ATANH,
159 			OPCODE_DP1,
160 			OPCODE_DP2,
161 			OPCODE_TRUNC,
162 			OPCODE_FLOOR,
163 			OPCODE_ROUND,
164 			OPCODE_ROUNDEVEN,
165 			OPCODE_CEIL,
166 			OPCODE_SQRT,
167 			OPCODE_RSQ,
168 			OPCODE_LEN2,
169 			OPCODE_LEN3,
170 			OPCODE_LEN4,
171 			OPCODE_DIST1,
172 			OPCODE_DIST2,
173 			OPCODE_DIST3,
174 			OPCODE_DIST4,
175 			OPCODE_NRM2,
176 			OPCODE_NRM4,
177 			OPCODE_DIV,
178 			OPCODE_MOD,
179 			OPCODE_EXP2,
180 			OPCODE_LOG2,
181 			OPCODE_EXP,
182 			OPCODE_LOG,
183 			OPCODE_POW,
184 			OPCODE_F2B,   // Float to bool
185 			OPCODE_B2F,   // Bool to float
186 			OPCODE_F2I,   // Float to int
187 			OPCODE_I2F,   // Int to float
188 			OPCODE_F2U,   // Float to uint
189 			OPCODE_U2F,   // Uint to float
190 			OPCODE_I2B,   // Int to bool
191 			OPCODE_B2I,   // Bool to int
192 			OPCODE_DET2,
193 			OPCODE_DET3,
194 			OPCODE_DET4,
195 			OPCODE_ALL,
196 			OPCODE_ANY,
197 			OPCODE_NEG,
198 			OPCODE_NOT,
199 			OPCODE_OR,
200 			OPCODE_XOR,
201 			OPCODE_AND,
202 			OPCODE_EQ,
203 			OPCODE_NE,
204 			OPCODE_STEP,
205 			OPCODE_SMOOTH,
206 			OPCODE_ISNAN,
207 			OPCODE_ISINF,
208 			OPCODE_TEXOFFSET,
209 			OPCODE_TEXLODOFFSET,
210 			OPCODE_TEXELFETCH,
211 			OPCODE_TEXELFETCHOFFSET,
212 			OPCODE_TEXGRAD,
213 			OPCODE_TEXGRADOFFSET,
214 			OPCODE_TEXBIAS,
215 			OPCODE_TEXLOD,
216 			OPCODE_TEXOFFSETBIAS,
217 			OPCODE_TEXSIZE,
218 			OPCODE_FLOATBITSTOINT,
219 			OPCODE_FLOATBITSTOUINT,
220 			OPCODE_INTBITSTOFLOAT,
221 			OPCODE_UINTBITSTOFLOAT,
222 			OPCODE_PACKSNORM2x16,
223 			OPCODE_PACKUNORM2x16,
224 			OPCODE_PACKHALF2x16,
225 			OPCODE_UNPACKSNORM2x16,
226 			OPCODE_UNPACKUNORM2x16,
227 			OPCODE_UNPACKHALF2x16,
228 			OPCODE_FORWARD1,
229 			OPCODE_FORWARD2,
230 			OPCODE_FORWARD3,
231 			OPCODE_FORWARD4,
232 			OPCODE_REFLECT1,
233 			OPCODE_REFLECT2,
234 			OPCODE_REFLECT3,
235 			OPCODE_REFLECT4,
236 			OPCODE_REFRACT1,
237 			OPCODE_REFRACT2,
238 			OPCODE_REFRACT3,
239 			OPCODE_REFRACT4,
240 			OPCODE_ICMP,
241 			OPCODE_UCMP,
242 			OPCODE_SELECT,
243 			OPCODE_EXTRACT,
244 			OPCODE_INSERT,
245 			OPCODE_DISCARD,
246 			OPCODE_FWIDTH,
247 			OPCODE_LEAVE,   // Return before the end of the function
248 			OPCODE_CONTINUE,
249 			OPCODE_TEST,   // Marks the end of the code that can be skipped by 'continue'
250 			OPCODE_SWITCH,
251 			OPCODE_ENDSWITCH,
252 
253 			// Integer opcodes
254 			OPCODE_INEG,
255 			OPCODE_IABS,
256 			OPCODE_ISGN,
257 			OPCODE_IADD,
258 			OPCODE_ISUB,
259 			OPCODE_IMUL,
260 			OPCODE_IDIV,
261 			OPCODE_IMAD,
262 			OPCODE_IMOD,
263 			OPCODE_SHL,
264 			OPCODE_ISHR,
265 			OPCODE_IMIN,
266 			OPCODE_IMAX,
267 
268 			// Unsigned integer opcodes
269 			OPCODE_UDIV,
270 			OPCODE_UMOD,
271 			OPCODE_USHR,
272 			OPCODE_UMIN,
273 			OPCODE_UMAX,
274 		};
275 
276 		static Opcode OPCODE_DP(int);
277 		static Opcode OPCODE_LEN(int);
278 		static Opcode OPCODE_DIST(int);
279 		static Opcode OPCODE_NRM(int);
280 		static Opcode OPCODE_FORWARD(int);
281 		static Opcode OPCODE_REFLECT(int);
282 		static Opcode OPCODE_REFRACT(int);
283 
284 		enum Control
285 		{
286 			CONTROL_RESERVED0,
287 			CONTROL_GT,
288 			CONTROL_EQ,
289 			CONTROL_GE,
290 			CONTROL_LT,
291 			CONTROL_NE,
292 			CONTROL_LE,
293 			CONTROL_RESERVED1
294 		};
295 
296 		enum SamplerType
297 		{
298 			SAMPLER_UNKNOWN,
299 			SAMPLER_1D,
300 			SAMPLER_2D,
301 			SAMPLER_CUBE,
302 			SAMPLER_VOLUME
303 		};
304 
305 		enum Usage   // For vertex input/output declarations
306 		{
307 			USAGE_POSITION = 0,
308 			USAGE_BLENDWEIGHT = 1,
309 			USAGE_BLENDINDICES = 2,
310 			USAGE_NORMAL = 3,
311 			USAGE_PSIZE = 4,
312 			USAGE_TEXCOORD = 5,
313 			USAGE_TANGENT = 6,
314 			USAGE_BINORMAL = 7,
315 			USAGE_TESSFACTOR = 8,
316 			USAGE_POSITIONT = 9,
317 			USAGE_COLOR = 10,
318 			USAGE_FOG = 11,
319 			USAGE_DEPTH = 12,
320 			USAGE_SAMPLE = 13
321 		};
322 
323 		enum ParameterType
324 		{
325 			PARAMETER_TEMP = 0,
326 			PARAMETER_INPUT = 1,
327 			PARAMETER_CONST = 2,
328 			PARAMETER_TEXTURE = 3,
329 			PARAMETER_ADDR = 3,
330 			PARAMETER_RASTOUT = 4,
331 			PARAMETER_ATTROUT = 5,
332 			PARAMETER_TEXCRDOUT = 6,
333 			PARAMETER_OUTPUT = 6,
334 			PARAMETER_CONSTINT = 7,
335 			PARAMETER_COLOROUT = 8,
336 			PARAMETER_DEPTHOUT = 9,
337 			PARAMETER_SAMPLER = 10,
338 			PARAMETER_CONST2 = 11,
339 			PARAMETER_CONST3 = 12,
340 			PARAMETER_CONST4 = 13,
341 			PARAMETER_CONSTBOOL = 14,
342 			PARAMETER_LOOP = 15,
343 			PARAMETER_TEMPFLOAT16 = 16,
344 			PARAMETER_MISCTYPE = 17,
345 			PARAMETER_LABEL = 18,
346 			PARAMETER_PREDICATE = 19,
347 
348 		//	PARAMETER_FLOAT1LITERAL,
349 		//	PARAMETER_FLOAT2LITERAL,
350 		//	PARAMETER_FLOAT3LITERAL,
351 			PARAMETER_FLOAT4LITERAL,
352 			PARAMETER_BOOL1LITERAL,
353 		//	PARAMETER_BOOL2LITERAL,
354 		//	PARAMETER_BOOL3LITERAL,
355 		//	PARAMETER_BOOL4LITERAL,
356 		//	PARAMETER_INT1LITERAL,
357 		//	PARAMETER_INT2LITERAL,
358 		//	PARAMETER_INT3LITERAL,
359 			PARAMETER_INT4LITERAL,
360 
361 			PARAMETER_VOID
362 		};
363 
364 		enum MiscParameterIndex
365 		{
366 			VPosIndex = 0,
367 			VFaceIndex = 1,
368 			InstanceIDIndex = 2,
369 			VertexIDIndex = 3,
370 		};
371 
372 		enum Modifier
373 		{
374 			MODIFIER_NONE,
375 			MODIFIER_NEGATE,
376 			MODIFIER_BIAS,
377 			MODIFIER_BIAS_NEGATE,
378 			MODIFIER_SIGN,
379 			MODIFIER_SIGN_NEGATE,
380 			MODIFIER_COMPLEMENT,
381 			MODIFIER_X2,
382 			MODIFIER_X2_NEGATE,
383 			MODIFIER_DZ,
384 			MODIFIER_DW,
385 			MODIFIER_ABS,
386 			MODIFIER_ABS_NEGATE,
387 			MODIFIER_NOT
388 		};
389 
390 		enum Analysis
391 		{
392 			// Flags indicating whether an instruction is affected by an execution enable mask
393 			ANALYSIS_BRANCH   = 0x00000001,
394 			ANALYSIS_BREAK    = 0x00000002,
395 			ANALYSIS_CONTINUE = 0x00000004,
396 			ANALYSIS_LEAVE    = 0x00000008,
397 		};
398 
399 		struct Relative
400 		{
401 			ParameterType type : 8;
402 			unsigned int index;
403 			unsigned int swizzle : 8;
404 			unsigned int scale;
405 			bool deterministic;   // Equal accross shader instances run in lockstep (e.g. unrollable loop couters)
406 		};
407 
408 		struct Parameter
409 		{
410 			union
411 			{
412 				struct
413 				{
414 					unsigned int index;   // For registers types
415 
416 					Relative rel;
417 				};
418 
419 				float value[4];       // For float constants
420 				int integer[4];       // For integer constants
421 				int boolean[4];       // For boolean constants
422 
423 				struct
424 				{
425 					unsigned int label;      // Label index
426 					unsigned int callSite;   // Call index (per label)
427 				};
428 			};
429 
Parametersw::Shader::Parameter430 			Parameter() : index(0), type(PARAMETER_VOID)
431 			{
432 				rel.type = PARAMETER_VOID;
433 				rel.index = 0;
434 				rel.swizzle = 0;
435 				rel.scale = 1;
436 				rel.deterministic = false;
437 			}
438 
439 			std::string string(ShaderType shaderType, unsigned short version) const;
440 			std::string typeString(ShaderType shaderType, unsigned short version) const;
441 			std::string relativeString() const;
442 
443 			ParameterType type : 8;
444 		};
445 
446 		struct DestinationParameter : Parameter
447 		{
448 			union
449 			{
450 				unsigned char mask;
451 
452 				struct
453 				{
454 					bool x : 1;
455 					bool y : 1;
456 					bool z : 1;
457 					bool w : 1;
458 				};
459 			};
460 
DestinationParametersw::Shader::DestinationParameter461 			DestinationParameter() : mask(0xF), saturate(false), partialPrecision(false), centroid(false), shift(0)
462 			{
463 			}
464 
465 			std::string modifierString() const;
466 			std::string shiftString() const;
467 			std::string maskString() const;
468 
469 			bool saturate         : 1;
470 			bool partialPrecision : 1;
471 			bool centroid         : 1;
472 			signed char shift     : 4;
473 		};
474 
475 		struct SourceParameter : Parameter
476 		{
SourceParametersw::Shader::SourceParameter477 			SourceParameter() : swizzle(0xE4), modifier(MODIFIER_NONE), bufferIndex(-1)
478 			{
479 			}
480 
481 			std::string string(ShaderType shaderType, unsigned short version) const;
482 			std::string swizzleString() const;
483 			std::string preModifierString() const;
484 			std::string postModifierString() const;
485 
486 			unsigned int swizzle : 8;
487 			Modifier modifier : 8;
488 			int bufferIndex : 8;
489 		};
490 
491 		struct Instruction
492 		{
493 			explicit Instruction(Opcode opcode);
494 			Instruction(const unsigned long *token, int size, unsigned char majorVersion);
495 
496 			virtual ~Instruction();
497 
498 			void parseOperationToken(unsigned long token, unsigned char majorVersion);
499 			void parseDeclarationToken(unsigned long token);
500 			void parseDestinationToken(const unsigned long *token, unsigned char majorVersion);
501 			void parseSourceToken(int i, const unsigned long *token, unsigned char majorVersion);
502 
503 			std::string string(ShaderType shaderType, unsigned short version) const;
504 			static std::string swizzleString(ParameterType type, unsigned char swizzle);
505 			std::string operationString(unsigned short version) const;
506 			std::string controlString() const;
507 
508 			bool isBranch() const;
509 			bool isCall() const;
510 			bool isBreak() const;
511 			bool isLoop() const;
512 			bool isEndLoop() const;
513 
514 			bool isPredicated() const;
515 
516 			Opcode opcode;
517 
518 			union
519 			{
520 				Control control;
521 
522 				struct
523 				{
524 					unsigned char project : 1;   // D3DSI_TEXLD_PROJECT
525 					unsigned char bias : 1;      // D3DSI_TEXLD_BIAS
526 				};
527 			};
528 
529 			bool predicate;
530 			bool predicateNot;   // Negative predicate
531 			unsigned char predicateSwizzle;
532 
533 			bool coissue;
534 			SamplerType samplerType;
535 			Usage usage;
536 			unsigned char usageIndex;
537 
538 			DestinationParameter dst;
539 			SourceParameter src[5];
540 
541 			union
542 			{
543 				unsigned int analysis;
544 
545 				struct
546 				{
547 					// Keep in sync with Shader::Analysis flags
548 					unsigned int analysisBranch : 1;
549 					unsigned int analysisBreak : 1;
550 					unsigned int analysisContinue : 1;
551 					unsigned int analysisLeave : 1;
552 				};
553 			};
554 		};
555 
556 		Shader();
557 
558 		virtual ~Shader();
559 
560 		int getSerialID() const;
561 		size_t getLength() const;
562 		ShaderType getShaderType() const;
563 		unsigned short getShaderModel() const;
564 
565 		void append(Instruction *instruction);
566 		void declareSampler(int i);
567 
568 		const Instruction *getInstruction(size_t i) const;
569 		int size(unsigned long opcode) const;
570 		static int size(unsigned long opcode, unsigned short shaderModel);
571 
572 		void print(const char *fileName, ...) const;
573 		void printInstruction(int index, const char *fileName) const;
574 
575 		static bool maskContainsComponent(int mask, int component);
576 		static bool swizzleContainsComponent(int swizzle, int component);
577 		static bool swizzleContainsComponentMasked(int swizzle, int component, int mask);
578 
579 		bool containsDynamicBranching() const;
580 		bool containsBreakInstruction() const;
581 		bool containsContinueInstruction() const;
582 		bool containsLeaveInstruction() const;
583 		bool containsDefineInstruction() const;
584 		bool usesSampler(int i) const;
585 
586 		struct Semantic
587 		{
Semanticsw::Shader::Semantic588 			Semantic(unsigned char usage = 0xFF, unsigned char index = 0xFF, bool flat = false) : usage(usage), index(index), centroid(false), flat(flat)
589 			{
590 			}
591 
operator ==sw::Shader::Semantic592 			bool operator==(const Semantic &semantic) const
593 			{
594 				return usage == semantic.usage && index == semantic.index;
595 			}
596 
activesw::Shader::Semantic597 			bool active() const
598 			{
599 				return usage != 0xFF;
600 			}
601 
602 			unsigned char usage;
603 			unsigned char index;
604 			bool centroid;
605 			bool flat;
606 		};
607 
608 		void optimize();
609 
610 		// FIXME: Private
611 		unsigned int dirtyConstantsF;
612 		unsigned int dirtyConstantsI;
613 		unsigned int dirtyConstantsB;
614 
615 		bool dynamicallyIndexedTemporaries;
616 		bool dynamicallyIndexedInput;
617 		bool dynamicallyIndexedOutput;
618 
619 	protected:
620 		void parse(const unsigned long *token);
621 
622 		void optimizeLeave();
623 		void optimizeCall();
624 		void removeNull();
625 
626 		void analyzeDirtyConstants();
627 		void analyzeDynamicBranching();
628 		void analyzeSamplers();
629 		void analyzeCallSites();
630 		void analyzeDynamicIndexing();
631 		void markFunctionAnalysis(unsigned int functionLabel, Analysis flag);
632 
633 		ShaderType shaderType;
634 
635 		union
636 		{
637 			unsigned short shaderModel;
638 
639 			struct
640 			{
641 				unsigned char minorVersion;
642 				unsigned char majorVersion;
643 			};
644 		};
645 
646 		std::vector<Instruction*> instruction;
647 
648 		unsigned short usedSamplers;   // Bit flags
649 
650 	private:
651 		const int serialID;
652 		static volatile int serialCounter;
653 
654 		bool dynamicBranching;
655 		bool containsBreak;
656 		bool containsContinue;
657 		bool containsLeave;
658 		bool containsDefine;
659 	};
660 }
661 
662 #endif   // sw_Shader_hpp
663