housel@en.ecn.purdue.edu (Peter S. Housel) (07/29/89)
This package contains a floating-point arithmetic package for Minix-PC version 1.3, and a few ``mostly ANSI-compatible'' additions to the C library. Installation instructions are in the file INSTALL, and notes on implementation are in the file NOTES. This is more-or-less a medium quality implementation of floating point arithmetic for the Minix C compiler. No 80x87 math coprocessor is required (or even useful) to use this package. The binary format should be identical to that of most IEEE 754 implementations, and many operations give bit-for-bit identical results. It is by no means, however, fully standard-conforming. (There is no support for infinities, NaN's, or denormalized numbers, for example.) Whenever compiling a program which uses floating-point constants, and/or when linking any program which uses floating-point arithmetic, the "-f" option of the C compiler should be used. (This causes the floating point constant postprocessor to be run, and adds the floating point library to the list of libraries linked with the program.) Although the goal is to eventually provide all of the math functions mentioned in the math.h file (and in section B4 of K&R2), not all of these are available yet. (Due to portability problems, the BSD 4.3 freely-redistributable portions of libm are not yet usable.) Any bug reports, suggestions, questions, or improved code would be appreciated. -Peter S. Housel- housel@ecn.purdue.edu ...!pur-ee!housel