[Simh] EXT :Re: some performance issues

[gérard Calliet]

Hello,

Thanks a lot. I'll try a new build.

I had the same test on VUPS (yes, basic DCL) with the same distribution 
which runs on a window server and I got 150 VUPS. Perhaps there is 
something specific about Itanium and OpenVMS there.

Another question is: are there ways of specific optimizations for CPU 
intensive applications. They are doing almost only there long Ada 
compilations.

I'll go back with new results.

Gérard Calliet

Le 06/03/2019 à 23:02, Mark Pizzolato a écrit :
> On Wednesday, March 6, 2019 at 11:02 AM, gérard Calliet wrote:
>> VAX hardware 89 VUPS
>>
>> SIMH 27 VUPS
>>
>> Gérard Calliet
> I suspect you may be using some generally available DCL procedure to
> report the simulated system's VUPS.  That procedure may not map very
> well to the actual total system performance doing things other than
> somewhat tight loops in DCL on the system. A simulator in such a tight
> loop may easily have almost all of its activity hitting the host system's
> CPU cache while a more natural instruction mix (as in your Ada compiler
> activity) may be very different.
>
> The simh version you are running (commit 8810571d) is 4 years and
> 3 months old.
>
> Maybe you should pick up the latest from https://github.com/simh/simh/archive/master.zip
>
> Once you've got a recent version, you may want to try building both with
> and without ASYNCH_IO.
>
> The default IA64 builds with asynch I/O enabled.  You can build without
> asynch I/O with the following command:
>
>      $ MMK/MACRO=(NOASYNCH=1) vax
>
> Let us know what you find.
>
> - Mark
>
>> Le 06/03/2019 à 19:46, Hittner, David T [US] (MS) a écrit :
>>> So from a performance analysis perspective, what are the VUPS ratings of
>>> the VS4000-90 and the SIMH VAX running on the RX2800?
>>> That's (more or less) the Integer performance rating.
>>>
>>> David