[Info-vax] Poulson info from Dave Cantor
glen herrmannsfeldt
gah at ugcs.caltech.edu
Thu Nov 18 14:36:38 EST 2010
Michael Kraemer <m.kraemer at gsi.de> wrote:
> In article <ic3qtf$9af$1 at news.eternal-september.org>, glen herrmannsfeldt
> <gah at ugcs.caltech.edu> writes:
> .
>> That tends to be a problem with RISC, and even worse with VLIW.
> not necessarily.
> I just tried an ancient "edt" built in 1991 for POWER1,
> still runs on POWER7, unmodified.
(snip)
Well, for RISC, and especially for VLIW, the problem isn't
allowing it to run, but to be able to extend the architecture
and allow older programs to run reasonably fast.
Expecting the compiler to optimize for a specific processor tends
to cause it to be much less optimal for other similar processors.
Now, that is still likely true for the (snipped) example of S/360.
It was usual in the early S/360 days for programs to use static
data fields mixed in with code. The 256 byte cache, with separate
instruction and data caches, on later machines causes such code
to run very slow. Even so, the constraints on the later
implementations aren't so restrictive.
For VLIW, a new implementation might have completely different
timing requirements, but is still forced to allow for the old ones.
There is one solution, though rarely used, and that is to have
the compilers output intermediate code, before the final bundling.
Then, either at program load time or install time, run the bundling
program for the specific processor available. It will then generate
the optimal code for that specific version.
> well, we don't know whether POWER will still be around in another 20 years,
> but if so, chances are not negligible that the "edt" binary will still run.
-- glen
More information about the Info-vax
mailing list