> From
www.bitsavers.org, it seems that VAX/VMS 1.5 only supported
> D float, but by VMS 5.0 both D and G had Fortran support, with
> D float the default. I believe that some hardware included
> hardware support for G-float, while some did not.
Since we're taking a trip down memory lane...
G_float made its first appearance (along with quad-precision H_float) in
a microcode option (with some backplane rewiring) for the VAX-11/780
"Star" called the "Sparkle ECO". The first installation (I was there!)
was at the Smithsonian Astrophysical Observatory office in Cambridge,
MA. This was 1979. In 1980, the VAX-11/750 "Comet" came out and it had
a microcode G+H option that loaded from TU58 tapes (slow and
unreliable). The VAX-11/730 "Nebula" (1982?) was the first VAX to have
"hardware" support (FPGA) for the G and H formats and was, until the VAX
9000 many years later, the fastest H_float VAX.
VMS also supported software emulation of the G and H instructions - this
was added in VMS 2.0 or maybe a bit later. I think that the emulator
was originally written by Stan Rabinowitz and I enhanced and maintained
it for some years after. I did the integration of the emulator into the
VMS exception handler, with the help of VMS devo Kathy Morse.
Subsequent models tended to support G in hardware but not H. I think the
MicroVAX I was the only one to not have hardware support for G after that.
VAX FORTRAN support for these types was there as soon as the types
existed. You had to choose between D and G formats for your double
precision type. REAL*16 was always H. For Alpha, you could choose F/D,
F/G or IEEE S/T (as it was termed) formats for floating point. Quad
precision was always IEEE-style "X". As noted, the Alpha could load and
store D format but would round to 53-bit precision (D had 56) on load.
VAX F and D float were basically taken straight from the PDP-11/44
(different PDP-11 models had different FP formats!) DEC's Mary Payne
submitted a proposal styled after VAX G to the IEEE committee
standardizing floating point formats, but Kahan's proposal was accepted
instead.
--
Steve Lionel
Developer Products Division
Intel Corporation
Nashua, NH
For email address, replace "invalid" with "com"
User communities for Intel Software Development Products
http://softwareforums.intel.com/
Intel Fortran Support
http://support.intel.com/support/performancetools/fortran
My Fortran blog
http://www.intel.com/software/drfortran
