home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Stars of Shareware: Programmierung
/
SOURCE.mdf
/
programm
/
msdos
/
c
/
cephes22
/
qfloat
/
read.me
< prev
next >
Wrap
Text File
|
1992-11-17
|
1KB
|
47 lines
This archive contains extended precision floating point
arithmetic routines, a calculator program, and a few additional
functions.
Several definitions must be configured in the source code, as follows:
Define type of computer in mconf.h.
Set precision NQ to 12 words (144 bit precision)
or 24 words (336 bit precision) in qhead.h.
Set precision and iteration count in qhead.asm.
Set size of address LARGEMEM in qcalc.h.
For a version with strictly correct rounding rules:
define STICKY to be 1 in qflt.c
use qflta.c
use qsqrta.c
This gets a rating of "excellent!" from the PARANOIA test program.
For a faster version, all in C language, non strict rounding:
define STICKY to be 0 in qflt.c
use qfltb.c
use qsqrt.c
Fastest IBM PC version, assembly language inner loops, non strict rounding:
define STICKY to be 0 in qflt.c
use subm.asm, divn.asm, mulr.asm
use qsqrt.c
The .asm files work as written only with small memory model.
Fastest version, 68000 processor:
define STICKY to be 0 in qflt.c
use qf68k.a
use qsqrt.c
Fastest version, 68020 processor:
define STICKY to be 0 in qflt.c
use qfltd.asm
use qsqrt.c
There are many dialects of 68000 assembly language. Versions
for System V Unix and Definicon DASM exist but are not included
here.
This software is copyright (C) 1984 - 1988 by Stephen L. Moshier.
All rights are reserved. A library of over 100 transcendental
functions that use this arithmetic can be purchased from Oasys
Systems, Waltham, Massachusetts.