mirror of
https://git.planet-casio.com/Lephenixnoir/OpenLibm.git
synced 2025-01-01 06:23:39 +01:00
78 lines
2.9 KiB
Fortran
78 lines
2.9 KiB
Fortran
*DECK MPBLAS
|
|
SUBROUTINE MPBLAS (I1)
|
|
C***BEGIN PROLOGUE MPBLAS
|
|
C***SUBSIDIARY
|
|
C***PURPOSE Subsidiary to DQDOTA and DQDOTI
|
|
C***LIBRARY SLATEC
|
|
C***TYPE ALL (MPBLAS-A)
|
|
C***AUTHOR (UNKNOWN)
|
|
C***DESCRIPTION
|
|
C
|
|
C This subroutine is called to set up Brent's 'mp' package
|
|
C for use by the extended precision inner products from the BLAS.
|
|
C
|
|
C In the SLATEC library we require the Extended Precision MP number
|
|
C to have a mantissa twice as long as Double Precision numbers.
|
|
C The calculation of MPT (and MPMXR which is the actual array size)
|
|
C in this routine will give 2x (or slightly more) on the machine
|
|
C that we are running on. The INTEGER array size of 30 was chosen
|
|
C to be slightly longer than the longest INTEGER array needed on
|
|
C any machine that we are currently aware of.
|
|
C
|
|
C***SEE ALSO DQDOTA, DQDOTI
|
|
C***REFERENCES R. P. Brent, A Fortran multiple-precision arithmetic
|
|
C package, ACM Transactions on Mathematical Software 4,
|
|
C 1 (March 1978), pp. 57-70.
|
|
C R. P. Brent, MP, a Fortran multiple-precision arithmetic
|
|
C package, Algorithm 524, ACM Transactions on Mathema-
|
|
C tical Software 4, 1 (March 1978), pp. 71-81.
|
|
C***ROUTINES CALLED I1MACH, XERMSG
|
|
C***COMMON BLOCKS MPCOM
|
|
C***REVISION HISTORY (YYMMDD)
|
|
C 791001 DATE WRITTEN
|
|
C 890531 Changed all specific intrinsics to generic. (WRB)
|
|
C 890531 REVISION DATE from Version 3.2
|
|
C 891214 Prologue converted to Version 4.0 format. (BAB)
|
|
C 900402 Added TYPE section. (WRB)
|
|
C 920501 Reformatted the REFERENCES section. (WRB)
|
|
C 930124 Increased Array size in MPCON for SUN -r8, and calculate
|
|
C size for Quad Precision for 2x DP. (RWC)
|
|
C***END PROLOGUE MPBLAS
|
|
COMMON /MPCOM/ MPB, MPT, MPM, MPLUN, MPMXR, MPR(30)
|
|
C***FIRST EXECUTABLE STATEMENT MPBLAS
|
|
I1 = 1
|
|
C
|
|
C For full extended precision accuracy, MPB should be as large as
|
|
C possible, subject to the restrictions in Brent's paper.
|
|
C
|
|
C Statements below are for an integer wordlength of 48, 36, 32,
|
|
C 24, 18, and 16. Pick one, or generate a new one.
|
|
C 48 MPB = 4194304
|
|
C 36 MPB = 65536
|
|
C 32 MPB = 16384
|
|
C 24 MPB = 1024
|
|
C 18 MPB = 128
|
|
C 16 MPB = 64
|
|
C
|
|
MPBEXP = I1MACH(8)/2-2
|
|
MPB = 2**MPBEXP
|
|
C
|
|
C Set up remaining parameters
|
|
C UNIT FOR ERROR MESSAGES
|
|
MPLUN = I1MACH(4)
|
|
C NUMBER OF MP DIGITS
|
|
MPT = (2*I1MACH(14)+MPBEXP-1)/MPBEXP
|
|
C DIMENSION OF R
|
|
MPMXR = MPT+4
|
|
C
|
|
if (MPMXR.GT.30) THEN
|
|
CALL XERMSG('SLATEC', 'MPBLAS',
|
|
* 'Array space not sufficient for Quad Precision 2x ' //
|
|
* 'Double Precision, Proceeding.', 1, 1)
|
|
MPT = 26
|
|
MPMXR = 30
|
|
ENDIF
|
|
C EXPONENT RANGE
|
|
MPM = MIN(32767,I1MACH(9)/4-1)
|
|
RETURN
|
|
END
|