diff options
author | Andreas Sandberg <andreas@sandberg.pp.se> | 2013-09-30 09:40:26 +0200 |
---|---|---|
committer | Andreas Sandberg <andreas@sandberg.pp.se> | 2013-09-30 09:40:26 +0200 |
commit | d3937f3b373ee9b1af09176d8a7d86eb6fcf790f (patch) | |
tree | f12ad785822c855522a762d93f3212c2948da696 /ext/fputils/include | |
parent | 3af2d8eab0e5b1ba4e936291b01895cc82fb124e (diff) | |
download | gem5-d3937f3b373ee9b1af09176d8a7d86eb6fcf790f.tar.xz |
ext: Include libfputils
This changeset includes libfputils from revision bbf0d61d75. This
library can be used to convert to and from 80-bit floats and query the
type of an 80-bit float, which is needed to support the x87 FPU.
Diffstat (limited to 'ext/fputils/include')
-rw-r--r-- | ext/fputils/include/fputils/fp80.h | 253 |
1 files changed, 253 insertions, 0 deletions
diff --git a/ext/fputils/include/fputils/fp80.h b/ext/fputils/include/fputils/fp80.h new file mode 100644 index 000000000..e584baab4 --- /dev/null +++ b/ext/fputils/include/fputils/fp80.h @@ -0,0 +1,253 @@ +/* + * Copyright (c) 2013, Andreas Sandberg + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. 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 COPYRIGHT HOLDERS AND 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 + * COPYRIGHT OWNER OR 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. + */ + +#ifndef _FP80_H +#define _FP80_H 1 + +#include <math.h> +#include <stdint.h> +#include <stdio.h> + +#ifdef __cplusplus +extern "C" { +#endif + +/** + * @defgroup fp80 80-bit Floats + * Functions handling 80-bit floats. + * + * @{ + */ + +/** Internal representation of an 80-bit float. */ +typedef struct { + union { + char bits[10]; + struct { + uint64_t fi; + uint16_t se; + } repr; + } u; +} fp80_t; + +/** Constant representing +inf */ +extern const fp80_t fp80_pinf; +/** Constant representing -inf */ +extern const fp80_t fp80_ninf; +/** Constant representing a quiet NaN */ +extern const fp80_t fp80_qnan; +/** Constant representing a quiet indefinite NaN */ +extern const fp80_t fp80_qnani; +/** Constant representing a signaling NaN */ +extern const fp80_t fp80_snan; +/** Alias for fp80_qnan */ +extern const fp80_t fp80_nan; + +/** + * Is the value a special floating point value? + * + * Determine if a floating point value is one of the special values + * (i.e., one of the infinities or NaNs). In practice, this function + * only checks if the exponent is set to the maximum value supported + * by the binary representation, which is a reserved value used for + * such special numbers. + * + * @param fp80 value to analyze. + * @return 1 if the value is special, 0 otherwise. + */ +int fp80_isspecial(fp80_t fp80); +/** + * Is the value a quiet NaN? + * + * @param fp80 value to analyze. + * @return 1 if true, 0 otherwise. + */ +int fp80_isqnan(fp80_t fp80); +/** + * Is the value an indefinite quiet NaN? + * + * @param fp80 value to analyze. + * @return 1 if true, 0 otherwise. + */ +int fp80_isqnani(fp80_t fp80); +/** + * Is the value a signaling NaN? + * + * @param fp80 value to analyze. + * @return 1 if true, 0 otherwise. + */ +int fp80_issnan(fp80_t fp80); + +/** + * Classify a floating point number. + * + * This function implements the same classification as the standard + * fpclassify() function. It returns one of the following floating + * point classes: + * <ul> + * <li>FP_NAN - The value is NaN. + * <li>FP_INFINITE - The value is either +inf or -inf. + * <li>FP_ZERO - The value is either +0 or -0. + * <li>FP_SUBNORMAL - The value is to small to be represented as a + * normalized float. See fp80_issubnormal(). + * <li>FP_NORMAL - The value is neither of above. + * </ul> + * + * @param fp80 value to analyze. + * @return Floating point classification. + */ +int fp80_classify(fp80_t fp80); + +/** + * Is a value finite? + * + * Check if a value is a finite value. That is, not one of the + * infinities or NaNs. + * + * @param fp80 value to analyze. + * @return -1 if negative finite, +1 if positive finite, 0 otherwise. + */ +int fp80_isfinite(fp80_t fp80); +/** + * Is the value a non-zero normal? + * + * This function checks if a floating point value is a normal (having + * an exponent larger or equal to 1) or not. See fp80_issubnormal() + * for a description of what a denormal value is. + * + * @see fp80_issubnormal() + * + * @param fp80 value to analyze. + * @return -1 if negative normal, +1 if positive normal, 0 otherwise. + */ +int fp80_isnormal(fp80_t fp80); +/** + * Is the value a NaN of any kind? + * + * @param fp80 value to analyze. + * @return -1 if negative NaN, +1 if positive NaN, 0 otherwise. + */ +int fp80_isnan(fp80_t fp80); +/** + * Is the value one of the infinities? + * + * @param fp80 value to analyze. + * @return -1 if -inf, +1 if +inf, 0 otherwise. + */ +int fp80_isinf(fp80_t fp80); +/** + * Determine value of the sign-bit of a floating point number. + * + * @note Floats can represent both positive and negative zeros. + * + * @param fp80 value to analyze. + * @return -1 if negative, +1 if positive. + */ +int fp80_sgn(fp80_t fp80); +/** + * Is the value zero? + * + * @param fp80 value to analyze. + * @return -1 if negative zero, +1 if positive zero, 0 otherwise. + */ +int fp80_iszero(fp80_t fp80); +/** + * Is the value a denormal? + * + * Numbers that are close to the minimum of what can be stored in a + * floating point number start loosing precision because bits in the + * fraction get used (implicitly) to store parts of the negative + * exponent (i.e., the exponent is saturated and the fraction is less + * than 1). Such numbers are known as denormals. This function checks + * whether a float is a denormal or not. + * + * @param fp80 value to analyze. + * @return -1 if negative denormal, +1 if positive denormal, 0 otherwise. + */ +int fp80_issubnormal(fp80_t fp80); + +/** + * Convert an 80-bit float to a 64-bit double. + * + * This function converts an 80-bit float into a standard 64-bit + * double. This conversion is inherently lossy since a double can only + * represent a subset of what an 80-bit float can represent. The + * fraction of the source value will always be truncated to fit the + * lower precision. If a value falls outside of the range that can be + * accurately represented by double by truncating the fraction, one of + * the following happens: + * <ul> + * <li>A denormal will be generated if that can approximate the + * value. + * <li>[-]0 will be generated if the magnitude of the value is too + * small to be represented at all. + * <li>+-Inf will be generated if the magnitude of the value is too + * large to be represented. + * </ul> + * + * NaN values will be preserved across the conversion. + * + * @param fp80 Source value to convert. + * @return 64-bit version of the float. + */ +double fp80_cvtd(fp80_t fp80); + +/** + * Convert an 64-bit double to an 80-bit float. + * + * This function converts a standard 64-bit double into an 80-bit + * float. This conversion is completely lossless since the 80-bit + * float represents a superset of what a 64-bit double can + * represent. + * + * @note Denormals will be converted to normalized values. + * + * @param fpd Source value to convert. + * @return 64-bit version of the float. + */ +fp80_t fp80_cvfd(double fpd); + +/** + * Dump the components of an 80-bit float to a file. + * + * @warning This function is intended for debugging and the format of + * the output is not guaranteed to be stable. + * + * @param fout Output stream (e.g., stdout) + * @param fp80 value to dump. + */ +void fp80_debug_dump(FILE *fout, fp80_t fp80); + +/** @} */ + +#ifdef __cplusplus +} /* extern "C" */ +#endif + +#endif |