Oppolzer - Informatik / Blog

From: "Bernd Oppolzer" <bernd.oppolzer@t-online.de>
Sent: Wednesday, 31 August 2011 5:34 AM

> This is IMHO the old RISC - CISC discussion.
>
> Should we have machine instructions to compute a polynome of grade n?
> I don't think so (but there were machines in the 60s which did just that,
> and - in that period - they were faster by using such instructions).

In a scientific center where such is done often, it would be useful.

As for the Intel product, some functions are provided in hardware- such as SQRT
and trig. functions.

> Should we have machine instructions to do a lookup in a binary tree?
> Maybe ...
>
> Should we have machine instructions for compression and decompression using the
> Lempel-Ziv algorithm? Of course, especially if DB2 with data compression is to be run
> on that hardware.

For many years, graphics functions have been provided by PC hardware.

> So I think the instruction set and the design decisions should not be driven by nostalgia
> (of course we have to take care of legacy workload), but by two things:
>
> - the kind of workload that will be run on the machine and performance goals to be met
>
> - and the needs of the compiler writers (they don't need MANY instructions)
>
> The comfort or discomfort of the ASSEMBLER programmers is not significant in this
> context, in my believe. Due to pipelining and cache issues, clever compilers will sooner
> or later outperform hand-written ASSEMBLER programs.