[Info-vax] VMS on Raspberry Pi 5

[Jake Hamby (Solid State Jake)]

On Monday, November 13, 2023 at 4:50:06 AM UTC-8, Arne Vajhøj wrote:
> On 11/13/2023 5:26 AM, Pancho wrote: 
> > On 12/11/2023 21:47, Bob Eager wrote: 
> >> On Sun, 12 Nov 2023 21:43:35 +0000, Pancho wrote: 
> >>> On 12/11/2023 21:05, The Natural Philosopher wrote 
> >>> 
> >>>> My impression of a Pi Zero is that it would knock a PDP/11 into the 
> >>>> middle of next week, and then some. 
> >>>> 
> >>>> 
> >>> Yeah, I would have expected 100s or 1000s of VUPS. 
> >> 
> >> On an emulator? 
> > 
> > More than 30. 
> > 
> > I don't really know the performance penalty of emulators.
> The overhead of a non-JIT instruction set emulation 
> must be huge. 

Absolutely. Looking at the VUPS script, it's really only benchmarking DCL's ability to loop using an IF test and a GOTO. It's not using any floating-point, which would be interesting since VAX FP has to be emulated. So you're testing a language interpreter inside an emulator.

On my Xeon W-1290P @ 3.70GHz (base clock), running OpenVMS natively, I get a max VUPS of 907.0. I wouldn't be surprised if this improves when the OS is recompiled with native compilers with optimization enabled. It's an interesting data point, but a very limited one since the test is so tiny and unrepresentative of real workloads.