1

/*

     2

 *  matiec - a compiler for the programming languages defined in IEC 61131-3

     3

 *

     4

 *  Copyright (C) 2003-2011  Mario de Sousa (msousa@fe.up.pt)

     5

 *  Copyright (C) 2012       Manuele Conti (conti.ma@alice.it)

     6

 *

     7

 *  This program is free software: you can redistribute it and/or modify

     8

 *  it under the terms of the GNU General Public License as published by

     9

 *  the Free Software Foundation, either version 3 of the License, or

    10

 *  (at your option) any later version.

    11

 *

    12

 *  This program is distributed in the hope that it will be useful,

    13

 *  but WITHOUT ANY WARRANTY; without even the implied warranty of

    14

 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

    15

 *  GNU General Public License for more details.

    16

 *

    17

 *  You should have received a copy of the GNU General Public License

    18

 *  along with this program.  If not, see <http://www.gnu.org/licenses/>.

    19

 *

    20

 *

    21

 * This code is made available on the understanding that it will not be

    22

 * used in safety-critical situations without a full and competent review.

    23

 */

    24

    25

/*

    26

 * An IEC 61131-3 compiler.

    27

 *

    28

 * Based on the

    29

 * FINAL DRAFT - IEC 61131-3, 2nd Ed. (2001-12-10)

    30

 *

    31

 */

    32

    33

    34

    35

    36

    37

/* Do constant folding...

    38

 *

    39

 * I.e., Determine the value of all expressions in which only constant values (i.e. literals) are used.

    40

 * The (constant) result of each operation is stored (annotated) in the respective operation symbol 

    41

 * (e.g.: add_expression_c) in the abstract syntax tree,

    42

 *

    43

 * For example:

    44

 *       2 + 3         -> the constant value '5'    is stored in the add_expression_c symbol.

    45

 *       22.2 - 5.0    -> the constant value '17.2' is stored in the add_expression_c symbol.

    46

 *       etc...

    47

 *

    48

 *

    49

 * NOTE 1 

    50

 *      Some operations and constants can have multiple data types. For example,

    51

 *        1 AND 0

    52

 *      may be either a BOOL, BYTE, WORD or LWORD.

    53

 *

    54

 *      The same happens with 

    55

 *        1 + 2

    56

 *      which may be signed (e.g. INT) or unsigned (UINT)

    57

 *

    58

 *      For the above reason, instead of storing a single constant value, we actually store 4:

    59

 *        - bool

    60

 *        - uint64

    61

 *        -  int64

    62

 *        - real64

    63

 *

    64

 *      Additionally, since the result of an operation may result in an overflow, we actually

    65

 *      store the result inside a struct (defined in absyntax.hh)

    66

 *

    67

 *             ** During stage 3 (semantic analysis/checking) we will be doing constant folding.

    68

 *              * That algorithm will anotate the abstract syntax tree with the result of operations

    69

 *              * on literals (i.e. 44 + 55 will store the result 99).

    70

 *              * Since the same source code (e.g. 1 + 0) may actually be a BOOL or an ANY_INT,

    71

 *              * or an ANY_BIT, we need to handle all possibilities, and determine the result of the

    72

 *              * operation assuming each type.

    73

 *              * For this reason, we have one entry for each possible type, with some expressions

    74

 *              * having more than one entry filled in!

    75

 *              **

    76

 *             typedef enum { cs_undefined,   // not defined --> const_value is not valid!

    77

 *                            cs_const_value, // const value is valid

    78

 *                            cs_overflow     // result produced overflow or underflow --> const_value is not valid!

    79

 *                          } const_status_t;

    80

 *    

    81

 *             typedef struct {

    82

 *                 const_status_t status;

    83

 *                 real64_t       value; 

    84

 *             } const_value_real64_t;

    85

 *             const_value_real64_t *const_value_real64; // when NULL --> UNDEFINED

    86

 *             

    87

 *             typedef struct {

    88

 *                 const_status_t status;

    89

 *                 int64_t        value; 

    90

 *             } const_value_int64_t;

    91

 *             const_value_int64_t *const_value_int64; // when NULL --> UNDEFINED

    92

 *             

    93

 *             typedef struct {

    94

 *                 const_status_t status;

    95

 *                 uint64_t       value; 

    96

 *             } const_value_uint64_t;

    97

 *             const_value_uint64_t *const_value_uint64; // when NULL --> UNDEFINED

    98

 *             

    99

 *             typedef struct {

   100

 *                 const_status_t status;

   101

 *                 bool           value; 

   102

 *             } const_value_bool_t;

   103

 *             const_value_bool_t *const_value_bool; // when NULL --> UNDEFINED

   104

 *

   105

 *

   106

 *

   107

 * NOTE 2 

   108

 *    This file does not print out any error messages!

   109

 *    We cannot really print out error messages when we find an overflow. Since each operation

   110

 *    (symbol in the absract syntax tree for that operation) will have up to 4 constant results,

   111

 *    it may happen that some of them overflow, while other do not.

   112

 *    We must wait for data type checking to determine the exact data type of each expression

   113

 *    before we can decide whether or not we should print out an overflow error message.

   114

 *

   115

 *    For this reason, this visitor merely annotates the abstract syntax tree, and leaves the

   116

 *    actuall printing of errors for the print_datatype_errors_c class!

   117

 */

   118

   119

#include "constant_folding.hh"

   120

#include <limits>

   121

#include <math.h> /* required for pow function */

   122

#include <stdlib.h> /* required for malloc() */

   123

   124

#if 1

   125

#define UINT64_MAX (std::numeric_limits< uint64_t >::max())

   126

#define  INT64_MAX (std::numeric_limits<  int64_t >::max())

   127

#define  INT64_MIN (std::numeric_limits<  int64_t >::min()) 

   128

#else

   129

#define __STDC_LIMIT_MACROS /* required for UINT64_MAX, INT64_MAX, INT64_MIN, ... */

   130

#include <stdint.h>         /* required for UINT64_MAX, INT64_MAX, INT64_MIN, ... */

   131

#endif

   132

   133

   134

#define FIRST_(symbol1, symbol2) (((symbol1)->first_order < (symbol2)->first_order)   ? (symbol1) : (symbol2))

   135

#define  LAST_(symbol1, symbol2) (((symbol1)->last_order  > (symbol2)->last_order)    ? (symbol1) : (symbol2))

   136

   137

#define STAGE3_ERROR(error_level, symbol1, symbol2, ...) {                                                                  \

   138

  if (current_display_error_level >= error_level) {                                                                         \

   139

    fprintf(stderr, "%s:%d-%d..%d-%d: error: ",                                                                             \

   140

            FIRST_(symbol1,symbol2)->first_file, FIRST_(symbol1,symbol2)->first_line, FIRST_(symbol1,symbol2)->first_column,\

   141

                                                 LAST_(symbol1,symbol2) ->last_line,  LAST_(symbol1,symbol2) ->last_column);\

   142

    fprintf(stderr, __VA_ARGS__);                                                                                           \

   143

    fprintf(stderr, "\n");                                                                                                  \

   144

    error_count++;                                                                                                     \

   145

  }                                                                                                                         \

   146

}

   147

   148

   149

#define STAGE3_WARNING(symbol1, symbol2, ...) {                                                                             \

   150

    fprintf(stderr, "%s:%d-%d..%d-%d: warning: ",                                                                           \

   151

            FIRST_(symbol1,symbol2)->first_file, FIRST_(symbol1,symbol2)->first_line, FIRST_(symbol1,symbol2)->first_column,\

   152

                                                 LAST_(symbol1,symbol2) ->last_line,  LAST_(symbol1,symbol2) ->last_column);\

   153

    fprintf(stderr, __VA_ARGS__);                                                                                           \

   154

    fprintf(stderr, "\n");                                                                                                  \

   155

    warning_found = true;                                                                                                   \

   156

}

   157

   158

   159

   160

   161

   162

   163

   164

   165

   166

   167

   168

   169

#define NEW_CVALUE(dtype, symbol) \

   170

 (symbol->const_value_##dtype) = new(symbol_c::const_value_##dtype##_t); \

   171

 if ((symbol->const_value_##dtype) == NULL) ERROR; \

   172

 (symbol->const_value_##dtype)->status = symbol_c::cs_undefined;

   173

   174

#define SET_CVALUE(dtype, symbol, new_value)  ((symbol)->const_value_##dtype->value) = new_value; ((symbol)->const_value_##dtype->status) = symbol_c::cs_const_value;  

   175

#define GET_CVALUE(dtype, symbol)             ((symbol)->const_value_##dtype->value) 

   176

#define SET_OVFLOW(dtype, symbol)             ((symbol)->const_value_##dtype->status) = symbol_c::cs_overflow

   177

    /* The following test is correct in the presence of a NULL pointer, as the logical evaluation will be suspended as soon as the first condition is false! */

   178

#define VALID_CVALUE(dtype, symbol)           ((NULL != (symbol)->const_value_##dtype) && (symbol_c::cs_const_value == (symbol)->const_value_##dtype->status))

   179

#define ISZERO_CVALUE(dtype, symbol)          ((VALID_CVALUE(dtype, symbol)) && (GET_CVALUE(dtype, symbol) == 0))

   181

#define DO_BINARY_OPER(dtype, oper)\

   182

	if (VALID_CVALUE(dtype, symbol->r_exp) && VALID_CVALUE(dtype, symbol->l_exp)) {                                \

   183

		NEW_CVALUE(dtype, symbol);                                                                             \

   184

		SET_CVALUE(dtype, symbol, GET_CVALUE(dtype, symbol->l_exp) oper GET_CVALUE(dtype, symbol->r_exp));     \

   185

	}

   186

   192

   193

   194

   195

   196

   197

   198

   199

/* NOTE:

   200

 *   Most of the conditions to detect overflows on signed and unsigned integer operations were adapted from

   201

 *   https://www.securecoding.cert.org/confluence/display/seccode/INT32-C.+Ensure+that+operations+on+signed+integers+do+not+result+in+overflow?showComments=false

   202

 *   https://www.securecoding.cert.org/confluence/display/seccode/INT30-C.+Ensure+that+unsigned+integer+operations+do+not+wrap

   203

 */

   204

   205

/* NOTE: If at all possible, all overflow tests are done by pre-condition tests, i.e. tests that 

   206

 *       can be run _before_ the operation is executed, and therefore without accessing the result!

   207

 *

   208

 *       The exception is for real/floating point values, that simply test if the result is NaN (not a number).

   209

 */

   210

   211

/* res = a + b */

   212

static void CHECK_OVERFLOW_uint64_SUM(symbol_c *res, symbol_c *a, symbol_c *b) {

   213

	if (!VALID_CVALUE(uint64, res))

   214

		return;

   215

	/* Test by post-condition: If sum is smaller than either operand => overflow! */

   216

	// if (GET_CVALUE(uint64, res) < GET_CVALUE(uint64, a))

   217

	/* Test by pre-condition: If (UINT64_MAX - a) < b => overflow! */

   218

	if ((UINT64_MAX - GET_CVALUE(uint64, a)) < GET_CVALUE(uint64, b))

   219

		SET_OVFLOW(uint64, res);

   220

}

   221

   222

   223

/* res = a - b */

   224

static void CHECK_OVERFLOW_uint64_SUB(symbol_c *res, symbol_c *a, symbol_c *b) {

   225

	if (!VALID_CVALUE(uint64, res))

   226

		return;

   227

	/* Test by post-condition: If diference is larger than a => overflow! */

   228

	// if (GET_CVALUE(uint64, res) > GET_CVALUE(uint64, a))

   229

	/* Test by pre-condition: if b > a => overflow! */

   230

	if (GET_CVALUE(uint64, b) > GET_CVALUE(uint64, a))

   231

		SET_OVFLOW(uint64, res);

   232

}

   233

   234

   235

/* res = a * b */

   236

static void CHECK_OVERFLOW_uint64_MUL(symbol_c *res, symbol_c *a, symbol_c *b) {

   237

	if (!VALID_CVALUE(uint64, res))

   238

		return;

   239

	/* Test by pre-condition: If (UINT64_MAX / a) < b => overflow! */

   240

	if ((UINT64_MAX / GET_CVALUE(uint64, a)) < GET_CVALUE(uint64, b))

   241

		SET_OVFLOW(uint64, res);

   242

}

   243

   244

   245

/* res = a / b */

   246

static void CHECK_OVERFLOW_uint64_DIV(symbol_c *res, symbol_c *a, symbol_c *b) {

   247

	if (!VALID_CVALUE(uint64, res))

   248

		return;

   249

	if (GET_CVALUE(uint64, b) == 0) /* division by zero! */

   250

		SET_OVFLOW(uint64, res);

   251

}

   252

   253

   254

/* res = a MOD b */

   255

static void CHECK_OVERFLOW_uint64_MOD(symbol_c *res, symbol_c *a, symbol_c *b) {

   256

	if (!VALID_CVALUE(uint64, res))

   257

		return;

   258

	/* no overflow condition exists, including division by zero, which IEC 61131-3 considers legal for MOD operation! */

   259

	if (false) 

   260

		SET_OVFLOW(uint64, res);

   261

}

   262

   263

   264

/* res = a + b */

   265

static void CHECK_OVERFLOW_int64_SUM(symbol_c *res, symbol_c *a_ptr, symbol_c *b_ptr) {

   266

	if (!VALID_CVALUE(int64, res))

   267

		return;

   268

	int64_t a = GET_CVALUE(int64, a_ptr);

   269

	int64_t b = GET_CVALUE(int64, b_ptr);

   270

	/* The following test is valid no matter what representation is being used (e.g. two's complement, etc...) */

   271

	if (((b > 0) && (a > (INT64_MAX - b)))

   272

	 || ((b < 0) && (a < (INT64_MIN - b))))

   273

		SET_OVFLOW(int64, res);

   274

}

   275

   276

   277

/* res = a - b */

   278

static void CHECK_OVERFLOW_int64_SUB(symbol_c *res, symbol_c *a_ptr, symbol_c *b_ptr) {

   279

	if (!VALID_CVALUE(int64, res))

   280

		return;

   281

	int64_t a = GET_CVALUE(int64, a_ptr);

   282

	int64_t b = GET_CVALUE(int64, b_ptr);

   283

	/* The following test is valid no matter what representation is being used (e.g. two's complement, etc...) */

   284

	if (((b > 0) && (a < (INT64_MIN + b)))

   285

	 || ((b < 0) && (a > (INT64_MAX + b))))

   286

		SET_OVFLOW(int64, res);

   287

}

   288

   289

   290

/* res = a * b */

   291

static void CHECK_OVERFLOW_int64_MUL(symbol_c *res, symbol_c *a_ptr, symbol_c *b_ptr) {

   292

	if (!VALID_CVALUE(int64, res))

   293

		return;

   294

	int64_t a = GET_CVALUE(int64, a_ptr);

   295

	int64_t b = GET_CVALUE(int64, b_ptr);

   296

	if (   ( (a > 0) &&  (b > 0) &&             (a > (INT64_MAX / b))) 

   297

	    || ( (a > 0) && !(b > 0) &&             (b < (INT64_MIN / a))) 

   298

	    || (!(a > 0) &&  (b > 0) &&             (a < (INT64_MIN / b))) 

   299

	    || (!(a > 0) && !(b > 0) && (a != 0) && (b < (INT64_MAX / a))))

   300

		SET_OVFLOW(int64, res);

   301

}

   302

   303

   304

/* res = a / b */

   305

static void CHECK_OVERFLOW_int64_DIV(symbol_c *res, symbol_c *a_ptr, symbol_c *b_ptr) {

   306

	if (!VALID_CVALUE(int64, res))

   307

		return;

   308

	int64_t a = GET_CVALUE(int64, a_ptr);

   309

	int64_t b = GET_CVALUE(int64, b_ptr);

   310

	if ((b == 0) || ((a == INT64_MIN) && (b == -1)))

   311

		SET_OVFLOW(int64, res);

   312

}

   313

   314

   315

/* res = a MOD b */

   316

static void CHECK_OVERFLOW_int64_MOD(symbol_c *res, symbol_c *a_ptr, symbol_c *b_ptr) {

   317

	if (!VALID_CVALUE(int64, res))

   318

		return;

   319

	int64_t a = GET_CVALUE(int64, a_ptr);

   320

	int64_t b = GET_CVALUE(int64, b_ptr);

   321

	/* IEC 61131-3 standard says IN1 MOD IN2 must be equivalent to

   322

	 *  IF (IN2 = 0) THEN OUT:=0 ; ELSE OUT:=IN1 - (IN1/IN2)*IN2 ; END_IF

   323

	 *

   324

	 * Note that, when IN1 = INT64_MIN, and IN2 = -1, an overflow occurs in the division,

   325

	 * so although the MOD operation should be OK, acording to the above definition, we actually have an overflow!!

   326

	 *

   327

	 * On the other hand, division by 0 is OK!!

   328

	 */

   329

	if ((a == INT64_MIN) && (b == -1))

   330

		SET_OVFLOW(int64, res);

   331

}

   332

   333

   334

/* res = - a */

   335

static void CHECK_OVERFLOW_int64_NEG(symbol_c *res, symbol_c *a_ptr) {

   336

	if (!VALID_CVALUE(int64, res))

   337

		return;

   338

	int64_t a = GET_CVALUE(int64, a_ptr);

   339

	if (a == INT64_MIN)

   340

		SET_OVFLOW(int64, res);

   341

}

   342

   343

   344

   345

static void CHECK_OVERFLOW_real64(symbol_c *res) {

   346

	if (!VALID_CVALUE(real64, res))

   347

		return;

   348

       	/* NaN => underflow, overflow, number is a higher precision format, is a complex number (IEEE standard) */

   349

	if (isnan(GET_CVALUE(real64, res)))

   350

		SET_OVFLOW(real64, res);

   351

}

   352

   353

   354

   355

   356

   357

   358

   359

   360

   361

   362

   363

   364

   365

   366

constant_folding_c::constant_folding_c(symbol_c *symbol) {

   367

    error_count = 0;

   368

    warning_found = false;

   369

    current_display_error_level = 0;

   370

   371

    /* check whether the platform on which the compiler is being run implements IEC 559 floating point data types. */

   372

    symbol_c null_symbol;

   373

    if (! (std::numeric_limits<real64_t>::is_iec559) )

   374

        STAGE3_WARNING(&null_symbol, &null_symbol, "The platform running the compiler does not implement IEC 559 floating point numbers. "

   375

                                                   "Any error and/or warning messages related to overflow/underflow of the result of operations on REAL/LREAL literals "

   376

                                                   " (i.e. constant folding) may themselves be erroneous, although are most probably correct.");

   377

}

   378

   379

   380

constant_folding_c::~constant_folding_c(void) {

   381

}

   382

   383

   384

int constant_folding_c::get_error_count() {

   385

	return error_count;

   386

}

   387

   388

   389

/*********************/

   390

/* B 1.2 - Constants */

   391

/*********************/

   392

/******************************/

   393

/* B 1.2.1 - Numeric Literals */

   394

/******************************/

   395

void *constant_folding_c::visit(real_c *symbol) {

   397

	return NULL;

   398

}

   399

   400

   401

void *constant_folding_c::visit(integer_c *symbol) {

   404

	return NULL;

   405

}

   406

   407

   408

void *constant_folding_c::visit(neg_real_c *symbol) {

   409

	symbol->exp->accept(*this);

   411

	CHECK_OVERFLOW_real64(symbol);

   412

	return NULL;

   413

}

   414

   415

/* | '-' integer	{$$ = new neg_integer_c($2, locloc(@$));} */

   416

void *constant_folding_c::visit(neg_integer_c *symbol) {

   417

	symbol->exp->accept(*this);

   419

	CHECK_OVERFLOW_int64_NEG(symbol, symbol->exp);

   420

	return NULL;

   421

}

   422

   423

   424

void *constant_folding_c::visit(binary_integer_c *symbol) {

   425

	return NULL;

   426

}

   427

   428

   429

void *constant_folding_c::visit(octal_integer_c *symbol) {

   430

	return NULL;

   431

}

   432

   433

   434

void *constant_folding_c::visit(hex_integer_c *symbol) {

   437

	return NULL;

   438

}

   439

   440

   441

/*

   442

integer_literal:

   443

  integer_type_name '#' signed_integer	{$$ = new integer_literal_c($1, $3, locloc(@$));}

   444

| integer_type_name '#' binary_integer	{$$ = new integer_literal_c($1, $3, locloc(@$));}

   445

| integer_type_name '#' octal_integer	{$$ = new integer_literal_c($1, $3, locloc(@$));}

   446

| integer_type_name '#' hex_integer	{$$ = new integer_literal_c($1, $3, locloc(@$));}

   447

*/

   448

// SYM_REF2(integer_literal_c, type, value)

   449

void *constant_folding_c::visit(integer_literal_c *symbol) {

   450

	symbol->value->accept(*this);

   453

	return NULL;

   454

}

   455

   456

   457

void *constant_folding_c::visit(real_literal_c *symbol) {

   458

	symbol->value->accept(*this);

   460

	return NULL;

   461

}

   462

   463

   464

void *constant_folding_c::visit(bit_string_literal_c *symbol) {

   465

	return NULL;

   466

}

   467

   468

   469

void *constant_folding_c::visit(boolean_literal_c *symbol) {

   470

	symbol->value->accept(*this);

   472

	return NULL;

   473

}

   474

   475

   476

void *constant_folding_c::visit(boolean_true_c *symbol) {

   478

	return NULL;

   479

}

   480

   481

   482

void *constant_folding_c::visit(boolean_false_c *symbol) {

   484

	return NULL;

   485

}

   486

   487

   488

/***************************************/

   489

/* B.3 - Language ST (Structured Text) */

   490

/***************************************/

   491

/***********************/

   492

/* B 3.1 - Expressions */

   493

/***********************/

   494

void *constant_folding_c::visit(or_expression_c *symbol) {

   495

	symbol->l_exp->accept(*this);

   496

	symbol->r_exp->accept(*this);

   497

	DO_BINARY_OPER(bool,   ||);

   498

	DO_BINARY_OPER(uint64, | );