[Info-vax] Some VMS/x86 perf test data from WASD maintainer.

[Mark Daniel]

On 28/4/2023 4:11 pm, Volker Halle wrote:
> Mark,
> 
> there are at least 2 variants of the 'VUPS' procedure. And some of these variants have my name in them.
> 
> Stromasys variant: I have been using this variant since about 19 years in CHARON-VAX and CHARON-AXP migration projects. This procedure was supposed to report the 'correct VUPS' values as documented for the older real VAX and Alpha systems:
> 
> VUPS numbers for 'real systems' were publicized by DEC, but most of these documents have disappeared from the WEB.
> 
> Here are some examples: http://www.vaxmacro.de/vvcc.html
> 
> I had to modify the Stromasys VUPS procedure (called SRI_VUPS.COM at that time, SRI = Software Resources International, which later became Stromasys), because it would enter an endless loop on the faster Alphas (> 1 GHz). And this modification found it's way into the other variants below.
> 
> FreeAXP variant:  http://www.migrationspecialties.com/Downloads/VUPS.txt
> EmuVM variant: https://emuvm.com/download/vups-com-benchmark/
> 
> The above 2 are strongly related (as can be seen by this line: $ cpu_multiplier = 10 ! VAX = 10 - Alpha/AXP = 40). And there is no code in these procedures to check the architecture they're running on and adapt the CPU_MULTIPLIER ! So the VUPS results produced can not be directly compared between the Stromasys variant and the other two - although there should be a linear relation between them. The DCL code in the 'inner loop' is the same for all these variants. Pure integer performance on one single CPU.
> 
> The idea of this procedure was the ability to compare 'VUPS' values between hardware systems and their emulated counterparts. Simple DCL procedure, could be mailed and run on the target system. No compiling required. When OpenVMS on I64 came around, I was interested in also calculating the 'VUPS' values for those systems. I used the runtime of the PRIME_SIEVE program to 'calibrate' the CPU_MODIFIER value in the SRI_VUPS.COM variant, so it would report VUPS values, which allowed to compare the VAX, Alpha, I64 and now x86-64 CPUs under OpenVMS.
> 
> So whenever someone provides VUPS values, they need to also document, which procedure has been used to collect them - as John has done by calling your procedure WASD_VUPS.COM
> 
> We can continue the VUPS discussion here, but should discuss the KERNEL mode question in the VSI forum.
>   
> Volker.

This is slightly off the topic of KERNEL mode consumption on X86 but ...

I have put all three into a ZIP archive at

    https://wasd.vsm.com.au/other/vups_230428a.zip
    https://wasd.vsm.com.au/other/vups_230428a.txt

for anyone who would like to experiment.

The comparative results across a number of platforms I have access to 
are as follows (extraneous white-space removed).

|$ @wasd_vups.com
|HP rx2660 (1.40GHz/6.0MB) with 4 CPU and 14335MB running VMS V8.4-2L3
|INFO: Preventing endless loop (10$) on fast CPUs
|Approximate System VUPs Rating : 469.0 ( min: 465.8 max: 470.6 )
|$ @migs_vups.com
|INFO: Preventing endless loop (10$) on fast CPUs
|Approximate System VUPs Rating :  456.5  ( min: 453  max: 458 )
|$ @emuvm_vups.com
|Approximate System VUPs Rating : 458.0 ( min: 458.0 max: 458.0 )

|$ @wasd_vups.com
|Digital Personal WorkStation with 1 CPU and 1536MB running VMS V8.4-2L1
|Approximate System VUPs Rating : 151.4 ( min: 151.0 max: 151.8 )
|$ @migs_vups.com
|Approximate System VUPs Rating :  144.5  ( min: 144  max: 144 )
|$ @emuvm_vups.com
|Approximate System VUPs Rating : 145.0 ( min: 144.8 max: 145.4 )

|! EISNER
|$ @WASD_VUPS.COM
|AlphaServer DS20 500 MHz with 2 CPU and 1536MB running VMS V8.4-2L2
|Approximate System VUPs Rating : 243.5 ( min: 238.6 max: 247.6 )
|$ @migs_VUPS.COM
|Approximate System VUPs Rating :  236.5  ( min: 234  max: 239 )
|$ @emuvm_VUPS.COM
|Approximate System VUPs Rating : 239.5 ( min: 238.2 max: 241.8 )

|! Dell Optiplex 9020 4 core i7 3.4Ghz 16GB Win 10 Pro 22H2
|$ @wasd_vups.com
|innotek GmbH VirtualBox with 2 CPU and 7574MB running VMS V9.2
|Approximate System VUPs Rating : 282.2 ( min: 281.4 max: 282.6 )
|$ @migs_vups.com
|INFO: Preventing endless loop (10$) on fast CPUs
|Approximate System VUPs Rating :  275.3  ( min: 273  max: 276 )
|$ @emuvm_vups.com
|Approximate System VUPs Rating : 275.8 ( min: 275.8 max: 275.8 )

|! BXNUC10i7FNH4 6 core i7 1.10GHz 32GB
|VSMX86$ @wasd_vups.com
|innotek GmbH VirtualBox with 2 CPU and 7936MB running VMS E9.2-1
|Approximate System VUPs Rating : 645.4 ( min: 642.4 max: 647.0 )
|$ @migs_vups.com
|INFO: Preventing endless loop (10$) on fast CPUs
|Approximate System VUPs Rating :  641.2  ( min: 641  max: 641 )
|$ @emuvm_vups.com
|Approximate System VUPs Rating : 636.3 ( min: 635.4 max: 638.2 )

Observations ... the VUPS.COM numbers all seem comparable between 
versions of the procedure.  The WASD_VUPS.COM numbers perhaps marginally 
higher than the other two.

When the adjustment re X86 per Volker's post are uncommented

|$!!! if f$getsyi("arch_name").eqs."x86_64" then $ cpu_multiplier=8

the numbers drop by some 20%.

|! Dell Optiplex 9020 4 core i7 3.4Ghz 16GB Win 10 Pro 22H2
|$ @wasd_vups.com
|innotek GmbH VirtualBox with 2 CPU and 7574MB running VMS V9.2
|Approximate System VUPs Rating : 225.6 ( min: 225.6 max: 225.6 )

|! Dell Optiplex 9020 4 core i7 3.4Ghz 16GB Win 10 Pro 22H2
|$ @wasd_vups.com
|innotek GmbH VirtualBox with 2 CPU and 7936MB running VMS E9.2-1
|INFO: Preventing endless loop (10$) on fast CPUs
|Approximate System VUPs Rating : 517.8 ( min: 515.8 max: 518.8 )

And now back to the main show, KERNEL mode numbers

   https://wasd.vsm.com.au/info-WASD/2023/0077

-- 
Anyone, who using social-media, forms an opinion regarding anything 
other than the relative cuteness of this or that puppy-dog, needs 
seriously to examine their critical thinking.