/* * Copyright (c) 2010 Apple Inc. All rights reserved. * * @APPLE_LICENSE_HEADER_START@ * * 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. * 3. Neither the name of Apple Inc. ("Apple") nor the names of its * contributors may be used to endorse or promote products derived from * this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS 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 APPLE OR ITS 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. * * Portions of this software have been released under the following terms: * * (c) Copyright 1989-1993 OPEN SOFTWARE FOUNDATION, INC. * (c) Copyright 1989-1993 HEWLETT-PACKARD COMPANY * (c) Copyright 1989-1993 DIGITAL EQUIPMENT CORPORATION * * To anyone who acknowledges that this file is provided "AS IS" * without any express or implied warranty: * permission to use, copy, modify, and distribute this file for any * purpose is hereby granted without fee, provided that the above * copyright notices and this notice appears in all source code copies, * and that none of the names of Open Software Foundation, Inc., Hewlett- * Packard Company or Digital Equipment Corporation be used * in advertising or publicity pertaining to distribution of the software * without specific, written prior permission. Neither Open Software * Foundation, Inc., Hewlett-Packard Company nor Digital * Equipment Corporation makes any representations about the suitability * of this software for any purpose. * * Copyright (c) 2007, Novell, Inc. 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. * 3. Neither the name of Novell Inc. nor the names of its contributors * may be used to endorse or promote products derived from this * this software without specific prior written permission. * * 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 HOLDERS 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. * * @APPLE_LICENSE_HEADER_END@ */ /* ** NAME: ** ** ndrfloat.c ** ** FACILITY: ** ** IDL Stub Runtime Support ** ** ABSTRACT: ** ** NDR floating point conversion routines. ** ** Currently implemented: ** ** To || IEEE | VAX | Cray | IBM ** ==============+======+======+====== ** F IEEE || Yes | Yes | Yes | Yes ** ------||------+------+------+------ ** r VAX || Yes | Yes | Yes | Yes ** ------||------+------+------+------ ** o Cray || Yes | Yes | Yes | Yes ** ------||------+------+------+------ ** m IBM || Yes | Yes | Yes | Yes ** ** ** VERSION: DCE 1.0 ** */ #if HAVE_CONFIG_H #include <config.h> #endif /* * VAX Floating point * ------------------ * * VAX has four floating types: F, D, G, and H: * * F is 32 bits with a 1-bit sign, an 8-bit exponent field (excess * 128), and a 23-bit mantissa representing a fraction in the range * 0.5 (inclusive) to 1.0 (exclusive). (The redundant leading "1" * bit is omitted.) * * D is 64 bits with a 1 bit sign, an 8-bit exponent, and a 56-bit * mantissa. * * G is 64 bits with a 1 bit sign, an 11-bit exponent (excess 1024), * and a 52-bit mantissa. * * H is 128 bits. ... * * * IEEE Floating point * ------------------- * * IEEE has two floating types: single and double: * * Single is 32 bits with a 1-bit sign, an 8-bit exponent field * (excess 127 = 16#7F), and a 23-bit mantissa representing a fraction * in the range 1.0 (inclusive) to 2.0 (exclusive). (The redundant, * leading "1" bit is omitted.) * * -38 +38 * range = 10 .. 10 * * MaxReal = 3.402823E+38 * * Double is 64 bits with a 1-bit sign, an 11-bit exponent field * (excess 1023 = 16#3FF) and a 52-bit mantissa. * * -308 +308 * range = 10 .. 10 * * MaxReal = 1.797693134862316E+308 * * Note that (unlike all other allowed NDR floating point representations) * IEEE floating point numbers are generated on machines made by different * manufacturers (and with different architectures). In particular, some * of those machines are little-endian (e.g. 80x86) and some are big-endian * (e.g. M680x0). NDR defines that to interpret bytes in a stream tagged * as using IEEE floating point numbers, the integer representation tag * is used to determine the order of the bytes of the IEEE floating point * number. This means that the bytes are in reverse order if sent from * machine that uses little-endian integers as compared to one that uses * big-endian integers. * * * CRAY Floating Point * ------------------- * * Cray-1, and X-MP's have one hardware floating-point type: single. * (95-bit double-precision routines are available in software.) * * Single is 64 bits with a 1-bit sign, a 15-bit exponent field * (biased by 16384 = 16#4000), and a 48-bit mantissa representing * a fraction in the range of 0.0 (inclusive) to 1.0 (exclusive). * * -2466 +2466 * range = 10 .. 10 * * MaxReal = 1.2650140831707E+2466 !! not correct !! * * A zero value or an underflow result is not biased and is represented * as 64 bits of zero. * * An overflow result is represented by an exponent >= 24576 (=16#6000). * * Conversions * ----------- * * NDR defines two floating types: single and double. NDR defines a set * of possible representations of these types: IEEE, VAX, Cray, IBM. The * correspondences between the NDR types and the representations is shown * below: * * NDR IEEE VAX CRAY IBM * +--------+--------+--------+--------+ * Single | single | F | single | | * +--------+--------+--------+--------+ * Double | double | G | single | | * +--------+--------+--------+--------+ * * Note that the original VAX architecture defined only F and D, not G or H. * Most VAX compilers seem by default to interpret "double precision float" * as meaning D format; however, some compilers have an option telling * them to interpret it as G. G seems to be the more "modern" and "popular" * flavor of double so that is why NDR uses it; also G corresponds better to * IEEE double. * * NOTE WELL: The conversion routines as they currently exist are sure to * have bugs in the boundary cases. Note also that the routines that * convert to VAX format are likely to work only on a VAX. * * The following diagrams show how the various representations appear in * an NDR byte stream: VAX F Floating ============== |<---- 8 bits ---->| byte offset field size +--+---------------+ |X2| FRAC1 | 0 1, 7 +--+---------------+ |SN| X1 | 1 1, 7 +--+---------------+ | FRAC3 | 2 8 +------------------+ | FRAC2 | 3 8 +------------------+ In terms of the VAX architecture handbook: X1 = 8:14 X2 = 7:7 SN = 15:15 FRAC1 = 0:6 FRAC2 = 24:31 FRAC3 = 16:23 FRACn are the segments of the mantissa. In increasing order of significance they are: FRAC3, FRAC2, FRAC1. Xn are the segments of the exponent. In increasing order of significance they are: X2, X1. SN is the sign bit. VAX D Floating (for illustration only) ============== |<---- 8 bits ---->| byte offset field size +--+---------------+ |X2| FRAC1 | 0 1, 7 +--+---------------+ |SN| X1 | 1 1, 7 +--+---------------+ | FRAC3 | 2 8 +------------------+ | FRAC2 | 3 8 +------------------+ | FRAC5 | 4 8 +------------------+ | FRAC4 | 5 8 +------------------+ | FRAC7 | 6 8 +------------------+ | FRAC6 | 7 8 +------------------+ VAX G Floating ============== |<---- 8 bits ---->| byte offset field size +--------+---------+ | X2 | FRAC1 | 0 4, 4 +--+-----+---------+ |SN| X1 | 1 1, 7 +--+---------------+ | FRAC3 | 2 8 +------------------+ | FRAC2 | 3 8 +------------------+ | FRAC5 | 4 8 +------------------+ | FRAC4 | 5 8 +------------------+ | FRAC7 | 6 8 +------------------+ | FRAC6 | 7 8 +------------------+ IEEE single =========== big-endian little-endian |<---- 8 bits ---->| |<---- 8 bits ---->| +--+---------------+ +------------------+ |SN| X1 | 0 | F3 | 0 +--+---------------+ +------------------+ |X2| F1 | 1 | F2 | 1 +--+---------------+ +--+---------------+ | F2 | 2 |X2| F1 | 2 +------------------+ +--+---------------+ | F3 | 3 |SN| X1 | 3 +------------------+ +--+---------------+ IEEE double =========== big-endian little-endian |<---- 8 bits ---->| |<---- 8 bits ---->| +--+---------------+ +------------------+ |SN| X1 | 0 | F7 | 0 +--+-----+---------+ +------------------+ | X2 | F1 | 1 | F6 | 1 +--------+---------+ +------------------+ | F2 | 2 | F5 | 2 +------------------+ +------------------+ | F3 | 3 | F4 | 3 +------------------+ +------------------+ | F4 | 4 | F3 | 4 +------------------+ +------------------+ | F5 | 5 | F2 | 5 +------------------+ +--+---------------+ | F6 | 6 |SN| X1 | 6 +------------------+ +--+-----+---------+ | F7 | 7 | X2 | F1 | 7 +------------------+ +--------+---------+ Cray single =========== |<---- 8 bits ---->| +--+---------------+ |SN| X1 | 0 +--+-----+---------+ | X2 | 1 +--------+---------+ | F1 | 2 +------------------+ | F2 | 3 +------------------+ | F3 | 4 +------------------+ | F4 | 5 +------------------+ | F5 | 6 +------------------+ | F6 | 7 +------------------+ */ #include <cvt.h> #include <cvt_pvt.h> /* The ordering of the following 3 includes should NOT be changed! */ #include <dce/rpc.h> #include <dce/stubbase.h> #include <lsysdep.h> # ifdef __STDC__ # include <stdlib.h> # endif int ndr_cray64_to_ieee32(unsigned32 *cray_flt, unsigned32 *ieee_flt); int ndr_ieee32_to_cray64(unsigned32 *ieee_flt, unsigned32 *cray_flt); #define INTERNAL static #define PUBLIC #define IEEE_FLOAT (NDR_LOCAL_FLOAT_REP == ndr_c_float_ieee) #define VAX_FLOAT (NDR_LOCAL_FLOAT_REP == ndr_c_float_vax) #define CRAY_FLOAT (NDR_LOCAL_FLOAT_REP == ndr_c_float_cray) #define IBM_FLOAT (NDR_LOCAL_FLOAT_REP == ndr_c_float_ibm) #define SWAB_64(dst, src) ( \ (dst)[0] = (src)[7], \ (dst)[1] = (src)[6], \ (dst)[2] = (src)[5], \ (dst)[3] = (src)[4], \ (dst)[4] = (src)[3], \ (dst)[5] = (src)[2], \ (dst)[6] = (src)[1], \ (dst)[7] = (src)[0] \ ) #define IBM_SWAB_64(dst, src) ( \ (dst)[0] = (src)[4], \ (dst)[1] = (src)[5], \ (dst)[2] = (src)[6], \ (dst)[3] = (src)[7], \ (dst)[4] = (src)[0], \ (dst)[5] = (src)[1], \ (dst)[6] = (src)[2], \ (dst)[7] = (src)[3] \ ) #ifndef cray #define SWAB_32(dst, src) ( \ (dst)[0] = (src)[3], \ (dst)[1] = (src)[2], \ (dst)[2] = (src)[1], \ (dst)[3] = (src)[0] \ ) #define VAXF_BL_SWAP(dst, src) ( \ (dst)[0] = (src)[1], \ (dst)[1] = (src)[0], \ (dst)[2] = (src)[3], \ (dst)[3] = (src)[2] \ ) #define VAXF_BL_SWAP1(dst) ( \ temp = (dst)[0], \ (dst)[0] = (dst)[1], \ (dst)[1] = temp, \ temp = (dst)[2], \ (dst)[2] = (dst)[3], \ (dst)[3] = temp \ ) #else #define SWAB_32(dst, src) ( \ (dst)[4] = (src)[7], \ (dst)[5] = (src)[6], \ (dst)[6] = (src)[5], \ (dst)[7] = (src)[4] \ ) #define VAXF_BL_SWAP(dst, src) ( \ (dst)[4] = (src)[5], \ (dst)[5] = (src)[4], \ (dst)[6] = (src)[7], \ (dst)[7] = (src)[6] \ ) #endif #define VAXG_BL_SWAP(dst, src) ( \ (dst)[0] = (src)[1], \ (dst)[1] = (src)[0], \ (dst)[2] = (src)[3], \ (dst)[3] = (src)[2], \ (dst)[4] = (src)[5], \ (dst)[5] = (src)[4], \ (dst)[6] = (src)[7], \ (dst)[7] = (src)[6] \ ) #define VAXG_BL_SWAP1(dst) ( \ temp = (dst)[0], \ (dst)[0] = (dst)[1], \ (dst)[1] = temp, \ temp = (dst)[2], \ (dst)[2] = (dst)[3], \ (dst)[3] = temp, \ temp = (dst)[4], \ (dst)[4] = (dst)[5], \ (dst)[5] = temp, \ temp = (dst)[6], \ (dst)[6] = (dst)[7], \ (dst)[7] = temp \ ) #define COPY_64(dst, src) ( \ *(long_float_p_t) (dst) = *(long_float_p_t) (src) \ ) #ifndef cray #define COPY_32(dst, src) ( \ *(short_float_p_t) (dst) = *(short_float_p_t) (src) \ ) #endif /* ** Floating point conversion routines: --> IEEE */ /* ** ** Routine: ** ** cvt_vax_f_to_ieee_single ** ** Functional Description: ** ** This routine converts a VAX F_Float floating point number ** into an IEEE single precision floating point number. ** ** Formal Parameters: ** ** input_value A VAX F_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The IEEE single precision representation of the VAX ** F_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_f_to_ieee_single ( CVT_VAX_F input_value, CVT_SIGNED_INT options, CVT_IEEE_SINGLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxf.c.h> #include <pkieees.c.h> } #endif /* ** ** Routine: ** ** cvt_vax_g_to_ieee_double ** ** Functional Description: ** ** This routine converts a VAX G_Float floating point number ** into an IEEE double precision floating point number. ** ** Formal Parameters: ** ** input_value A VAX G_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** output_value The IEEE double precision representation of the VAX ** G_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_g_to_ieee_double ( CVT_VAX_G input_value, CVT_SIGNED_INT options, CVT_IEEE_DOUBLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxg.c.h> #include <pkieeet.c.h> } #endif /* ** ** Routine: ** ** cvt_cray_to_ieee_single ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into an IEEE single precision floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** output_value The IEEE single precision representation of the CRAY ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_cray_to_ieee_single ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_IEEE_SINGLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkieees.c.h> } /* * Note 32 bit NDR floats for a cray are represented as IEEE 32 bit values. */ /* * This routine is not "internal" because cray stubbase.h macros need to invoke * it directly. */ int ndr_cray64_to_ieee32 ( unsigned32 *cray_flt, /* 64 bit entities */ unsigned32 *ieee_flt ) { /* * Cray does not have a 32 bit floating point format so ndr_short_floats * must be represented in 64 bits on Cray. The Cray floating point value * is converted to IEEE single for sending over the wire. Since the Cray * word size is 64 bits, make sure we use the least significant 4 bytes of * the 8 byte word to represent the IEEE single. Convert the IEEE single * in Cray's native big-endian format. */ cvt_cray_to_ieee_single((CVT_BYTE *)cray_flt, CVT_C_BIG_ENDIAN, (CVT_BYTE *)(&(((char *)ieee_flt)[4]))); return 0; } #endif /* ** ** Routine: ** ** cvt_cray_to_ieee_double ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into an IEEE double precision floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** output_value The IEEE double precision representation of the CRAY ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_cray_to_ieee_double ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_IEEE_DOUBLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkieeet.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_short_to_ieee_single ** ** Functional Description: ** ** This routine converts an IBM single precision floating point number ** into an IEEE single precision floating point number. ** ** Formal Parameters: ** ** input_value An IBM single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** output_value The IEEE single precision representation of the IBM ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_ibm_short_to_ieee_single ( CVT_IBM_SHORT input_value, CVT_SIGNED_INT options, CVT_IEEE_SINGLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibms.c.h> #include <pkieees.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_long_to_ieee_double ** ** Functional Description: ** ** This routine converts an IBM double precision floating point number ** into an IEEE double precision floating point number. ** ** Formal Parameters: ** ** input_value An IBM double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The IEEE double precision representation of the IBM ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IEEE_FLOAT || defined(DEBUG) INTERNAL void cvt_ibm_long_to_ieee_double ( CVT_IBM_LONG input_value, CVT_SIGNED_INT options, CVT_IEEE_DOUBLE output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibml.c.h> #include <pkieeet.c.h> } #endif /* ** Floating point conversion routines: --> VAX */ /* ** ** Routine: ** ** cvt_ieee_single_to_vax_f ** ** Functional Description: ** ** This routine converts an IEEE single precision floating point number ** into a VAX F_Float floating point number. ** ** Formal Parameters: ** ** input_value An IEEE single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding and ** underflow is not possible. Both options may be selected ** but will be effectively ignored (not signaled). ** ** output_value The VAX F_Float representation of the IEEE number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_single_to_vax_f ( CVT_IEEE_SINGLE input_value, CVT_SIGNED_INT options, CVT_VAX_F output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieees.c.h> #include <pkvaxf.c.h> } #endif /* ** ** Routine: ** ** cvt_ieee_double_to_vax_g ** ** Functional Description: ** ** This routine converts an IEEE double precision floating point number ** into a VAX G_Float floating point number. ** ** Formal Parameters: ** ** input_value An IEEE double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding. A ** rounding option may be selected but will be effectively ** ignored (not signaled). ** ** output_value The VAX G_Float representation of the IEEE number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_double_to_vax_g ( CVT_IEEE_DOUBLE input_value, CVT_SIGNED_INT options, CVT_VAX_G output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieeet.c.h> #include <pkvaxg.c.h> } #endif /* ** ** Routine: ** ** cvt_cray_to_vax_f ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into a VAX F_Float floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_ERR_UNDERFLOW - Raise underflows ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_VAX_ROUNDING. ** ** output_value The VAX F_Float representation of the CRAY number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_cray_to_vax_f ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_VAX_F output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkvaxf.c.h> } #endif /* ** ** Routine: ** ** cvt_cray_to_vax_g ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into a VAX G_Float floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding. A ** rounding option may be selected but will be effectively ** ignored (not signaled). ** ** output_value The VAX G_Float representation of the CRAY number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_cray_to_vax_g ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_VAX_G output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkvaxg.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_short_to_vax_f ** ** Functional Description: ** ** This routine converts an IBM single precision floating point number ** into a VAX F_Float floating point number. ** ** Formal Parameters: ** ** input_value An IBM single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding. A ** rounding option may be selected but will be effectively ** ignored (not signaled). ** ** output_value The VAX F_Float representation of the IBM number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_ibm_short_to_vax_f ( CVT_IBM_SHORT input_value, CVT_SIGNED_INT options, CVT_VAX_F output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibms.c.h> #include <pkvaxf.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_long_to_vax_g ** ** Functional Description: ** ** This routine converts an IBM double precision floating point number ** into a VAX G_Float floating point number. ** ** Formal Parameters: ** ** input_value An IBM double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_VAX_ROUNDING. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The VAX G_Float representation of the IBM number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if VAX_FLOAT || defined(DEBUG) INTERNAL void cvt_ibm_long_to_vax_g ( CVT_IBM_LONG input_value, CVT_SIGNED_INT options, CVT_VAX_G output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_VAX_ROUNDING; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibml.c.h> #include <pkvaxg.c.h> } #endif /* ** Floating point conversion routines: --> CRAY */ /* ** ** Routine: ** ** cvt_ieee_single_to_cray ** ** Functional Description: ** ** This routine converts an IEEE single precision floating point number ** into a CRAY floating point number. ** ** Formal Parameters: ** ** input_value An IEEE single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** ** NOTE: This routine does not require rounding and ** underflow is not possible. Both options may be selected ** but will be effectively ignored (not signaled). ** ** output_value The CRAY single precision representation of the IEEE ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_single_to_cray ( CVT_IEEE_SINGLE input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieees.c.h> #include <pkcray.c.h> } /* * This routine is not "internal" because cray idl_base.h macros need to invoke * it directly. */ int ndr_ieee32_to_cray64 ( unsigned32 *ieee_flt, /* 64 bit entities */ unsigned32 *cray_flt ) { /* * Cray does not have a 32 bit floating point format so ndr_short_floats * must be represented in 64 bits on Cray. The Cray floating point value * is converted to IEEE single for sending over the wire. If the callee * is also on a Cray, the IEEE single is converted back to a 64 bit Cray. * Assume the IEEE single is in Cray's native big-endian format and * resides in the least significant 32 bits of the 64 bit argument. */ cvt_ieee_single_to_cray((CVT_BYTE *)(&(((char *)ieee_flt)[4])), CVT_C_BIG_ENDIAN, (CVT_BYTE *)cray_flt); return 0; } #endif /* ** ** Routine: ** ** cvt_ieee_double_to_cray ** ** Functional Description: ** ** This routine converts an IEEE double precision floating point number ** into a CRAY single precision floating point number. ** ** Formal Parameters: ** ** input_value An IEEE double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The CRAY single precision representation of the IEEE ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_double_to_cray ( CVT_IEEE_DOUBLE input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieeet.c.h> #include <pkcray.c.h> } #endif /* ** ** Routine: ** ** cvt_vax_f_to_cray ** ** Functional Description: ** ** This routine converts a VAX F_Float floating point number ** into a CRAY floating point number. ** ** Formal Parameters: ** ** input_value A VAX F_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. There are no applicable ** options. ** ** NOTE: This routine does not require rounding and ** underflow is not possible. Both options may be selected ** but will be effectively ignored (not signaled). ** ** output_value The CRAY single precision representation of the VAX ** F_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_f_to_cray ( CVT_VAX_F input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxf.c.h> #include <pkcray.c.h> } #endif /* ** ** Routine: ** ** cvt_vax_g_to_cray ** ** Functional Description: ** ** This routine converts a VAX G_Float floating point number ** into a CRAY floating point number. ** ** Formal Parameters: ** ** input_value A VAX G_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The CRAY single precision representation of the VAX ** G_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_g_to_cray ( CVT_VAX_G input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxg.c.h> #include <pkcray.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_short_to_cray ** ** Functional Description: ** ** This routine converts an IBM single precision floating point number ** into a CRAY floating point number. ** ** Formal Parameters: ** ** input_value An IBM single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. ** ** output_value The CRAY single precision representation of the ** IBM number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT INTERNAL void cvt_ibm_short_to_cray ( CVT_IBM_SHORT input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibms.c.h> #include <pkcray.c.h> } #endif /* ** ** Routine: ** ** cvt_ibm_long_to_cray ** ** Functional Description: ** ** This routine converts an IBM double precision floating point number ** into a CRAY floating point number. ** ** Formal Parameters: ** ** input_value An IBM double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. ** ** output_value The CRAY single precision representation of the ** IBM number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified, ** or the result is +infinity or -infinity ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if CRAY_FLOAT INTERNAL void cvt_ibm_long_to_cray ( CVT_IBM_LONG input_value, CVT_SIGNED_INT options, CVT_CRAY output_value ) { int i, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkibml.c.h> #include <pkcray.c.h> } #endif /* ** Floating point conversion routines: --> IBM */ /* ** ** Routine: ** ** cvt_ieee_single_to_ibm_short ** ** Functional Description: ** ** This routine converts an IEEE single precision floating point number ** into an IBM single precision floating point number. ** ** Formal Parameters: ** ** input_value An IEEE single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The IBM single precision representation of the IEEE ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_single_to_ibm_short ( CVT_IEEE_SINGLE input_value, CVT_SIGNED_INT options, CVT_IBM_SHORT output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieees.c.h> #include <pkibms.c.h> } #endif /* ** ** Routine: ** ** cvt_ieee_double_to_ibm_long ** ** Functional Description: ** ** This routine converts an IEEE double precision floating point number ** into a IBM double precision floating point number. ** ** Formal Parameters: ** ** input_value An IEEE double precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_BIG_ENDIAN - default is little endian ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding. A ** rounding option may be selected but will be effectively ** ignored (not signaled). ** ** output_value The IBM double precision representation of the IEEE ** number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT || defined(DEBUG) INTERNAL void cvt_ieee_double_to_ibm_long ( CVT_IEEE_DOUBLE input_value, CVT_SIGNED_INT options, CVT_IBM_LONG output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~(CVT_C_BIG_ENDIAN | CVT_C_ERR_UNDERFLOW) ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkieeet.c.h> #include <pkibml.c.h> } #endif /* ** ** Routine: ** ** cvt_vax_f_to_ibm_short ** ** Functional Description: ** ** This routine converts a VAX F_Float floating point number ** into an IBM single precision floating point number. ** ** Formal Parameters: ** ** input_value A VAX F_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_TRUNCATE - truncate ** CVT_C_ROUND_TO_POS - round to +infinity ** CVT_C_ROUND_TO_NEG - round to -infinity ** CVT_C_ROUND_TO_NEAREST - round to nearest ** CVT_C_VAX_ROUNDING - VAX rounding ** ** NOTE: If no rounding mode is selected the following ** default rounding mode is assumed: ** ** CVT_C_ROUND_TO_NEAREST. ** ** NOTE2: This routine can not underflow. The underflow ** option may be selected but will be effectively ignored ** (not signaled). ** ** output_value The IBM single precision representation of the VAX ** F_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_f_to_ibm_short ( CVT_VAX_F input_value, CVT_SIGNED_INT options, CVT_IBM_SHORT output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxf.c.h> #include <pkibms.c.h> } #endif /* ** ** Routine: ** ** cvt_vax_g_to_ibm_long ** ** Functional Description: ** ** This routine converts a VAX G_Float floating point number ** into an IBM double precision floating point number. ** ** Formal Parameters: ** ** input_value A VAX G_Float floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. Applicable options are: ** ** CVT_C_ERR_UNDERFLOW - Raise underflows ** ** NOTE: This routine does not require rounding. A ** rounding option may be selected but will be effectively ** ignored (not signaled). ** ** output_value The IBM double precision representation of the VAX ** G_Float number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT || defined(DEBUG) INTERNAL void cvt_vax_g_to_ibm_long ( CVT_VAX_G input_value, CVT_SIGNED_INT options, CVT_IBM_LONG output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkvaxg.c.h> #include <pkibml.c.h> } #endif /* ** ** Routine: ** ** cvt_cray_to_ibm_short ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into an IBM single precision floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. ** ** output_value The IBM single precision representation of the ** CRAY number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT INTERNAL void cvt_cray_to_ibm_short ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_IBM_SHORT output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkibms.c.h> } #endif /* ** ** Routine: ** ** cvt_cray_to_ibm_long ** ** Functional Description: ** ** This routine converts a CRAY single precision floating point number ** into an IBM double precision floating point number. ** ** Formal Parameters: ** ** input_value A CRAY single precision floating point number. ** ** options An integer bit mask. Set bits in the mask represent ** selected routine options. ** ** output_value The IBM double precision representation of the ** CRAY number. ** ** Side Effects/Signaled Errors: ** ** dcethread_aritherr_e an invalid input value was specified ** dcethread_fltovf_t floating point overflow ** dcethread_fltund_t floating point underflow ** */ #if IBM_FLOAT INTERNAL void cvt_cray_to_ibm_long ( CVT_CRAY input_value, CVT_SIGNED_INT options, CVT_IBM_LONG output_value ) { int i, j, round_bit_position; UNPACKED_REAL r; switch ( options & ~ CVT_C_ERR_UNDERFLOW ) { case 0 : options |= CVT_C_ROUND_TO_NEAREST; case CVT_C_ROUND_TO_NEAREST : case CVT_C_TRUNCATE : case CVT_C_ROUND_TO_POS : case CVT_C_ROUND_TO_NEG : case CVT_C_VAX_ROUNDING : break; default : DCETHREAD_RAISE(dcethread_aritherr_e); } #include <upkcray.c.h> #include <pkibml.c.h> } #endif /* * C V T _ F L O A T * * Internal common routine for "ndr_cvt_long/short_float". */ INTERNAL void cvt_float ( ndr_format_t sdrep, ndr_format_t ddrep, unsigned_char_t *srcp, unsigned_char_t *dstp, ndr_boolean longf ) { unsigned_char_t buff[8]; CVT_SIGNED_INT cvt_options = 0; /* * Conversion routines expect operands to be in their native format, * e.g. big-endian for CRAY floating point numbers. Conversion routines * involving IEEE numbers accept a flag indicating little- or big-endian. */ switch (ddrep.float_rep) { #if IEEE_FLOAT case ndr_c_float_ieee: if (ddrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; switch (sdrep.float_rep) { case ndr_c_float_ieee: /* IEEE -> IEEE */ if (ddrep.int_rep == sdrep.int_rep) if (longf) COPY_64(dstp, srcp); else COPY_32(dstp, srcp); else if (longf) SWAB_64(dstp, srcp); else SWAB_32(dstp, srcp); break; case ndr_c_float_vax: /* VAX -> IEEE */ /* * Since the conversion routines expect to operation on a * little-endian machine, byte swap before, and tell it * the result is little endian in order to get the correct * result. */ if (!((sdrep.int_rep == ndr_c_int_little_endian) && (ddrep.int_rep == ndr_c_int_little_endian))) { if (longf) { SWAB_32(buff, srcp); SWAB_32(buff+4, srcp+4); } else SWAB_32(buff, srcp); srcp = buff; } if (longf) cvt_vax_g_to_ieee_double( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_vax_f_to_ieee_single( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_cray: /* CRAY -> IEEE */ /* Cray short floats will be in IEEE 32 Bit format. */ if (longf) { /* Cray native format big-endian; swap if necessary */ #if !(defined(_IBMR2) && defined(_AIX)) && \ !(defined(__hppa) && defined(__hpux)) if (!((sdrep.int_rep == ndr_c_int_big_endian) && (ddrep.int_rep == ndr_c_int_little_endian))) { SWAB_64(buff, srcp); srcp = buff; } #endif cvt_cray_to_ieee_double( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); } else /* Already in IEEE format but might have to swap */ if (sdrep.int_rep != ddrep.int_rep) SWAB_32(dstp, srcp); break; case ndr_c_float_ibm: /* IBM -> IEEE */ /* IBM native format big-endian; swap if necessary */ if (sdrep.int_rep != ddrep.int_rep) { if (ddrep.int_rep != ndr_c_int_little_endian) { if (longf) SWAB_64(buff, srcp); else SWAB_32(buff, srcp); srcp = buff; } else { if (longf) { IBM_SWAB_64(buff, srcp); srcp = buff; } } } if (longf) cvt_ibm_long_to_ieee_double( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ibm_short_to_ieee_single( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; default: abort(); } break; #endif #if VAX_FLOAT case ndr_c_float_vax: switch (sdrep.float_rep) { case ndr_c_float_ieee: /* IEEE -> VAX */ if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; if (longf) cvt_ieee_double_to_vax_g( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ieee_single_to_vax_f( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_vax: /* VAX -> VAX */ if (ddrep.int_rep == sdrep.int_rep) if (longf) COPY_64(dstp, srcp); else COPY_32(dstp, srcp); else if (longf) VAXG_BL_SWAP(dstp, srcp); else VAXF_BL_SWAP(dstp, srcp); break; case ndr_c_float_cray: /* CRAY -> VAX */ /* Cray short floats will be in IEEE 32 Bit format. */ if (longf) { /* Cray native format big-endian; swap if necessary */ if (sdrep.int_rep != ndr_c_int_big_endian) { SWAB_64(buff, srcp); srcp = buff; } cvt_cray_to_vax_g( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); } else /* Source operand in IEEE format */ { if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; cvt_ieee_single_to_vax_f( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); } break; case ndr_c_float_ibm: /* IBM -> VAX */ /* IBM native format big-endian; swap if necessary */ if (sdrep.int_rep != ndr_c_int_big_endian) { if (longf) SWAB_64(buff, srcp); else SWAB_32(buff, srcp); srcp = buff; } if (longf) cvt_ibm_long_to_vax_g( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ibm_short_to_vax_f( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; default: abort(); } #if (NDR_LOCAL_INT_REP == ndr_c_int_big_endian) /* * Conversion routines produce native VAX floating point format. * If destination integer rep is big-endian VAX, convert from VAX * native floating point format to true big-endian byte ordering. */ if (sdrep.float_rep != ndr_c_float_vax && ddrep.int_rep == ndr_c_int_big_endian) if (longf) VAXG_BL_SWAP1(dstp); else VAXF_BL_SWAP1(dstp); #endif break; #endif #if CRAY_FLOAT case ndr_c_float_cray: switch (sdrep.float_rep) { case ndr_c_float_ieee: /* IEEE -> CRAY */ if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; if (longf) cvt_ieee_double_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ieee_single_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_vax: /* VAX -> CRAY */ /* * Assume that little-endian integer format implies VAX * native floating point format; big-endian integer format * implies true big-endian byte ordering, so swap bytes * within 16-bit words. */ if (sdrep.int_rep == ndr_c_int_big_endian) { if (longf) VAXG_BL_SWAP(buff, srcp); else VAXF_BL_SWAP(buff, srcp); srcp = buff; } if (longf) cvt_vax_g_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_vax_f_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_cray: /* CRAY -> CRAY */ /* Cray short floats will be in IEEE 32 Bit format. */ if (longf) if (ddrep.int_rep == sdrep.int_rep) COPY_64(dstp, srcp); else SWAB_64(dstp, srcp); else /* Source operand in IEEE format */ { if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; cvt_ieee_single_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); } break; case ndr_c_float_ibm: /* IBM -> CRAY */ if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; if (longf) cvt_ibm_long_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ibm_short_to_cray( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; default: abort(); } break; #endif #if IBM_FLOAT case ndr_c_float_ibm: switch (sdrep.float_rep) { case ndr_c_float_ieee: /* IEEE -> IBM */ if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; #if (NDR_LOCAL_INT_REP == ndr_c_int_big_endian) if (sdrep.int_rep != ddrep.int_rep) { /* convert to big endian format for SAA */ if (longf) SWAB_64(buff,srcp); else SWAB_32(buff,srcp); srcp = buff; } #endif if (longf) cvt_ieee_double_to_ibm_long( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_ieee_single_to_ibm_short( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_vax: /* VAX -> IBM */ /* * Assume that little-endian integer format implies VAX * native floating point format; big-endian integer format * implies true big-endian byte ordering, so swap bytes * within 16-bit words. */ if (sdrep.int_rep == ndr_c_int_big_endian) { if (longf) VAXG_BL_SWAP(buff, srcp); else VAXF_BL_SWAP(buff, srcp); srcp = buff; } if (longf) cvt_vax_g_to_ibm_long( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_vax_f_to_ibm_short( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_cray: /* CRAY -> IBM */ if (sdrep.int_rep == ndr_c_int_big_endian) cvt_options |= CVT_C_BIG_ENDIAN; if (longf) cvt_cray_to_ibm_long( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); else cvt_cray_to_ibm_short( (CVT_BYTE *)srcp, cvt_options, (CVT_BYTE *)dstp); break; case ndr_c_float_ibm: /* IBM -> IBM */ if (ddrep.int_rep == sdrep.int_rep) if (longf) COPY_64(dstp, srcp); else COPY_32(dstp, srcp); else if (longf) SWAB_64(dstp, srcp); else SWAB_32(dstp, srcp); break; default: abort(); } break; #endif default: abort(); } } /* * N D R _ C V T _ S H O R T _ F L O A T * * Convert between short (32 bit) floating point representations. */ PUBLIC void ndr_cvt_short_float ( ndr_format_t source_drep, ndr_format_t dst_drep, short_float_p_t srcp, short_float_p_t dstp ) { cvt_float(source_drep, dst_drep, (unsigned_char_t *)srcp, (unsigned_char_t *)dstp, ndr_false); } /* * N D R _ C V T _ L O N G _ F L O A T * * Convert between long (64 bit) floating point representations. */ PUBLIC void ndr_cvt_long_float ( ndr_format_t source_drep, ndr_format_t dst_drep, long_float_p_t srcp, long_float_p_t dstp ) { cvt_float(source_drep, dst_drep, (unsigned_char_t *)srcp, (unsigned_char_t *)dstp, ndr_true); } #ifdef DEBUG globaldef void (*cvt_vector[])() = { cvt_vax_f_to_ieee_single, cvt_vax_g_to_ieee_double, cvt_cray_to_ieee_single, cvt_cray_to_ieee_double, cvt_ibm_short_to_ieee_single, cvt_ibm_long_to_ieee_double, cvt_ieee_single_to_vax_f, cvt_ieee_double_to_vax_g, cvt_cray_to_vax_f, cvt_cray_to_vax_g, cvt_ibm_short_to_vax_f, cvt_ibm_long_to_vax_g, cvt_ieee_single_to_cray, cvt_ieee_double_to_cray, cvt_vax_f_to_cray, cvt_vax_g_to_cray, cvt_ieee_single_to_ibm_short, cvt_ieee_double_to_ibm_long, cvt_vax_f_to_ibm_short, cvt_vax_g_to_ibm_long }; #endif