/** * \file libyasm/floatnum.h * \brief YASM floating point (IEEE) interface. * * \license * Copyright (C) 2001-2007 Peter Johnson * * Based on public-domain x86 assembly code by Randall Hyde (8/28/91). * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND OTHER CONTRIBUTORS ``AS IS'' * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR OTHER CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * \endlicense */ #ifndef YASM_FLOATNUM_H #define YASM_FLOATNUM_H #ifndef YASM_LIB_DECL #define YASM_LIB_DECL #endif /** Initialize floatnum internal data structures. */ YASM_LIB_DECL void yasm_floatnum_initialize(void); /** Clean up internal floatnum allocations. */ YASM_LIB_DECL void yasm_floatnum_cleanup(void); /** Create a new floatnum from a decimal string. The input string must be in * standard C representation ([+-]123.456e[-+]789). * \param str floating point decimal string * \return Newly allocated floatnum. */ YASM_LIB_DECL /*@only@*/ yasm_floatnum *yasm_floatnum_create(const char *str); /** Duplicate a floatnum. * \param flt floatnum * \return Newly allocated floatnum with the same value as flt. */ YASM_LIB_DECL /*@only@*/ yasm_floatnum *yasm_floatnum_copy(const yasm_floatnum *flt); /** Destroy (free allocated memory for) a floatnum. * \param flt floatnum */ YASM_LIB_DECL void yasm_floatnum_destroy(/*@only@*/ yasm_floatnum *flt); /** Floating point calculation function: acc = acc op operand. * \note Not all operations in yasm_expr_op may be supported; unsupported * operations will result in an error. * \param acc floatnum accumulator * \param op operation * \param operand floatnum operand * \return Nonzero on error. */ YASM_LIB_DECL int yasm_floatnum_calc(yasm_floatnum *acc, yasm_expr_op op, yasm_floatnum *operand); /** Convert a floatnum to single-precision and return as 32-bit value. * The 32-bit value is a "standard" C value (eg, of unknown endian). * \param flt floatnum * \param ret_val pointer to storage for 32-bit output * \return Nonzero if flt can't fit into single precision: -1 if underflow * occurred, 1 if overflow occurred. */ YASM_LIB_DECL int yasm_floatnum_get_int(const yasm_floatnum *flt, /*@out@*/ unsigned long *ret_val); /** Output a #yasm_floatnum to buffer in little-endian or big-endian. Puts the * value into the least significant bits of the destination, or may be shifted * into more significant bits by the shift parameter. The destination bits are * cleared before being set. [0] should be the first byte output to the file. * \note Not all sizes are valid. Currently, only 32 (single-precision), 64 * (double-precision), and 80 (extended-precision) are valid sizes. * Use yasm_floatnum_check_size() to check for supported sizes. * \param flt floatnum * \param ptr pointer to storage for size bytes of output * \param destsize destination size (in bytes) * \param valsize size (in bits) * \param shift left shift (in bits) * \param bigendian endianness (nonzero=big, zero=little) * \param warn enables standard overflow/underflow warnings * \return Nonzero if flt can't fit into the specified precision: -1 if * underflow occurred, 1 if overflow occurred. */ YASM_LIB_DECL int yasm_floatnum_get_sized(const yasm_floatnum *flt, unsigned char *ptr, size_t destsize, size_t valsize, size_t shift, int bigendian, int warn); /** Basic check to see if size is valid for flt conversion (using * yasm_floatnum_get_sized()). Doesn't actually check for underflow/overflow * but rather checks for size=32,64,80 * (at present). * \param flt floatnum * \param size number of bits of output space * \return 1 if valid size, 0 if invalid size. */ YASM_LIB_DECL int yasm_floatnum_check_size(const yasm_floatnum *flt, size_t size); /** Print various representations of a floatnum. For debugging purposes only. * \param f file * \param flt floatnum */ YASM_LIB_DECL void yasm_floatnum_print(const yasm_floatnum *flt, FILE *f); #endif