fpc(3) — mips UMIPS RISC/os 5.01



FPC(3)              RISC/os Reference Manual               FPC(3)



NAME
     fpc - floating-point control registers

SYNOPSIS
     #include <sys/fpu.h>

     int getfpccsr()

     int setfpccsr(csr)
     int csr;

     int getfpcirr()

     int getfpceir()

     void setfpcled(value)
     int value;

     int swapRM(x)
     int x;

     int swapINX(x)
     int x;

DESCRIPTION
     These routines are to get and set the floating-point control
     registers of MIPS floating-point units.  All of these rou-
     tines take and or return their values as 32 bit integers.

     The file <sys/fpu.h> contains unions for each of the control
     registers. Each union contains a structure that breaks out
     the bit fields into the logical parts for each control
     register.  This file also contains constants for fields of
     the control registers.

     All implementations of MIPS floating-point have a control
     and status register and a implementation revsion register.
     The control and status register is returned by get_fpc_csr.
     The routine set_fpc_csr sets the control and status register
     and returns the old value.  The implementation revsion
     register is read-only and is returned by the routine
     get_fpc_irr.

     The R2360 floating-point units (floating-point boards) have
     two addtitonal control registers.  The exception instruction
     register is a read-only register and is returned by the rou-
     tine get_fpc_eir. The other floating-point control register
     on the R2360 is the leds register.  The low 8 bits
     corresponds to the leds where a one is off and a zero is on.
     The leds register is a write-only register and is set with
     the routine set_fpc_leds.




                        Printed 11/19/92                   Page 1



FPC(3)              RISC/os Reference Manual               FPC(3)



     The routine swapRN sets only the rounding mode and returns
     the old rounding mode.  The routine swapINX sets only the
     sticky inexact bit and returns the old one.  The bits in the
     arguments and return values to swapRN and swapINX are right
     justified.

ALSO SEE
     R2010 Floating Point Coprocessor Architecture
     R2360 Floating Point Board Product Description














































 Page 2                 Printed 11/19/92

Museum