laprade@ringworld.harris-atd.com (Ken Laprade) (08/26/89)
This is mentioned in the README file of the LUCAS distribution: It should be noted that when the 68020-68881 pair is installed the new IEEE math libraries which support the 68881 are used for floating-point transparently. Do the new IEEE math libraries also support the use of a 68000-68881 combination? If so, what address must the 68881 be located at (there is no coprocessor space in the 68000)? Does the Amiga check for a 68881 automatically on bootup? If not, what is required to tell it there is one there? Thanks for any help. -- Ken Laprade ARPA: laprade@trantor.harris-atd.com Harris Corporation - GISD UUCP: ...uunet!x102a!trantor!laprade Palm Bay, Florida voice: (407)727-4433
dale@boing.UUCP (Dale Luck) (08/27/89)
In article <2613@trantor.harris-atd.com> laprade@trantor.harris-atd.com (Ken Laprade) writes: > >Do the new IEEE math libraries also support the use of a 68000-68881 >combination? If so, what address must the 68881 be located at (there >is no coprocessor space in the 68000)? Does the Amiga check for a >68881 automatically on bootup? If not, what is required to tell it >there is one there? The V1.3 ieee libraries support the 68000/68881 peripheral style floating point accelerators such as from Microbotics or from CMI. However a bug was just found in the 1.3 ieee libraries that prevent them from working with 68882 chips. This has been fixed and then next release will support the 68000/68882 combination. The 6888x must be autoconfiged and an ieee math resource added to the system during binddrivers. Documentation for this is available. There is no predetermined address for the 6888x coprocessor chip. The amiga checks for 6888x only if there is a 68020 or better main cpu. The ieee libraries themselves check for the existance of various floating point options when they are first initialized. They initialize the library vector table to different routines depending on what is available. >Ken Laprade ARPA: laprade@trantor.harris-atd.com -- Dale Luck GfxBase/Boing, Inc. {uunet!cbmvax|pyramid}!amiga!boing!dale
billsey@agora.UUCP (Bill Seymour) (08/29/89)
From article <2613@trantor.harris-atd.com:, by laprade@ringworld.harris-atd.com (Ken Laprade):
: This is mentioned in the README file of the LUCAS distribution:
: It should be noted that when the 68020-68881 pair is
: installed the new IEEE math libraries which support the
: 68881 are used for floating-point transparently.
:
: Do the new IEEE math libraries also support the use of a 68000-68881
: combination? If so, what address must the 68881 be located at (there
: is no coprocessor space in the 68000)? Does the Amiga check for a
: 68881 automatically on bootup? If not, what is required to tell it
: there is one there?
For the 68000/68881 combo you need to have a math_resource set up
with the base address for the 68881. When the IEEE libraries are invoked,
and they don't see a 680(2|3)0/6888(1|2) combo, they look for this resource.
If they don't find the resource, then they go into software emulation mode.
Note that the 68000/68881 support via the 1.3 IEEE libraries is pretty non-
optimal. You don't get major speed improvements for many applications. (I
have seen a mandelbrot generator optimised for 68000/68881 that was running
close to 300 times faster than with the IEEE libraries... and 1200 times the
speed of 68000 alone... But then he'd really done a lot to optimise for his
combo! (68000 at 14.32 MHz, 68881 at 20MHz.))
: Thanks for any help.
:
: --
: Ken Laprade ARPA: laprade@trantor.harris-atd.com
: Harris Corporation - GISD UUCP: ...uunet!x102a!trantor!laprade
: Palm Bay, Florida voice: (407)727-4433
--
-Bill Seymour ...tektronix!reed!percival!agora!billsey
...tektronix!sequent.UUCP!calvin!billsey
Bejed, Inc. NES, Inc. Northwest Amiga Group At Home Sometimes
(503) 691-2552 (503) 246-9311 (503) 656-7393 BBS (503) 640-0842