Discussion:
VSI VMS V8.4-2L2 availability?
Add Reply
Slo
2021-11-09 17:10:10 UTC
Reply
Permalink
I have VSI VMS V8.4-2L1 installed on an XP1000. It looks like
this box can run the enhanced VSI VMS V8.4-2L2 as well, but
I can't find the kit anywhere.

Q1: Why is there no ISO available in the same location as the
VSI VMS V8.4-2L1 kit? How does one qualify to obtain it?
Q2: Any experience to share when upgrading from L1 to L2?
Q3: Is it worth it, and in what areas one should expect better performance?

Thanks.
Dave Froble
2021-11-09 18:43:56 UTC
Reply
Permalink
Post by Slo
I have VSI VMS V8.4-2L1 installed on an XP1000. It looks like
this box can run the enhanced VSI VMS V8.4-2L2 as well, but
I can't find the kit anywhere.
Q1: Why is there no ISO available in the same location as the
VSI VMS V8.4-2L1 kit? How does one qualify to obtain it?
Q2: Any experience to share when upgrading from L1 to L2?
Q3: Is it worth it, and in what areas one should expect better performance?
Thanks.
2L2 is exactly the same as 2L1. 2L2 was compiled to use EV6 instructions.
Performance in the OS was reported to be perhaps 15% better. Note, "in the OS".

Unless you need extra performance in the OS, it really isn't needed.

ISVs and paying customers can get 2L2. Not really worth the effort for hobbyists.
--
David Froble Tel: 724-529-0450
Dave Froble Enterprises, Inc. E-Mail: ***@tsoft-inc.com
DFE Ultralights, Inc.
170 Grimplin Road
Vanderbilt, PA 15486
Slo
2021-11-09 22:19:20 UTC
Reply
Permalink
2L2 is exactly the same as 2L1. 2L2 was compiled to use EV6 instructions.
Performance in the OS was reported to be perhaps 15% better. Note, "in the OS".
Unless you need extra performance in the OS, it really isn't needed.
ISVs and paying customers can get 2L2. Not really worth the effort for hobbyists.
Thanks.
The other/main reason is playing with FreeAXP, which is an EV6 implementation,
and where I'm desperate to squeeze out any bits of performance. Compiling C takes
about 20x more than on my XP1000 (didn't really measure yet). Just planing for
a future when the hardware will eventually die...
Stephen Hoffman
2021-11-09 22:57:27 UTC
Reply
Permalink
Post by Slo
2L2 is exactly the same as 2L1. 2L2 was compiled to use EV6
instructions. Performance in the OS was reported to be perhaps 15%
better. Note, "in the OS". Unless you need extra performance in the OS,
it really isn't needed. ISVs and paying customers can get 2L2. Not
really worth the effort for hobbyists.
Thanks.
The other/main reason is playing with FreeAXP, which is an EV6
implementation, and where I'm desperate to squeeze out any bits of
performance. Compiling C takes about 20x more than on my XP1000
(didn't really measure yet). Just planing for a future when the
hardware will eventually die...
V8.4-2L2 has a minor change or two beyond being rebuilt for EV6. It's
not exactly the same as V8.4-2L1. V8.4-2L2 also incurs the VSI new-user
tax.

As for Alpha emulator performance, try some other emulators. AXPbox is
one that's been discussed around here. An emulator with JIT support
will likely be faster.

Scrounge (buy) an AlphaStation DS15, if you're in this for the long
haul. DS15 is easier to work with than is the AlphaStation XP1000. And
less fragile. And newer. And faster.
--
Pure Personal Opinion | HoffmanLabs LLC
Slo
2021-11-12 17:12:00 UTC
Reply
Permalink
Compiling a C package (73 modules) takes 12 minutes on my FreeAXP,
but about 2 on the XP1000. I'm not complaining; it's a miracle these
things work, really. I rarely do full builds, at most 2 or 3 files change
during an editing session, so this is acceptable to me.
Hoping to install AXPbox and see if I get a better ration than 6:1.
By the way, FreeAXP = 42, XP100 = 318 VUPs.
Mark Daniel
2021-11-12 22:40:03 UTC
Reply
Permalink

Post by Slo
By the way, FreeAXP = 42, XP100 = 318 VUPs.
Coincidentally I found today that if I run SimH with VMS 5.5-2H4 on a
Raspberry Pi Zero 2 W, I get 3.7 VUPs, which is pretty identical to the
VAXstation 3100M38 that it is emulating. Hardware cost is $15 and size
tiny. A throwaway VAX! See
https://www.raspberrypi.com/products/raspberry-pi-zero-2-w/.
Actually, as it’s a quad-core cpu, I could probably run a cluster of three
or four instances on it for $15!
Not challenging the value-for-money of the above unit but recently I was
curious how my seven-year-old, dual core, 2.7GHz i5 (macOS 12.0.1)
compared to my 20+ year workhorse PWS500.

Digital Personal WorkStation with 1 CPU and 1536MB running VMS V8.4-2L1
Approximate System VUPs Rating : 159.0 ( min: 159.0 max: 159.0 )

innotek GmbH VirtualBox with 2 CPU and 3584MB running VMS V9.1-A
Approximate System VUPs Rating : 199.1 ( min: 195.8 max: 204.2 )

So moving to a (virtual) instance of VMS won't cost me anything
performance-wise even if on this humble laptop. BTW: running two
concurrent VUPS.COM produces much the same result.

For comparison...

HP rx2660 (1.40GHz/6.0MB) with 4 CPU and 14335MB running VMS V8.4-2L1
INFO: Preventing endless loop (10$) on fast CPUs
Approximate System VUPs Rating : 495.1 ( min: 487.2 max: 499.6 )
--
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.
Dave Froble
2021-11-13 00:51:31 UTC
Reply
Permalink

Post by Slo
By the way, FreeAXP = 42, XP100 = 318 VUPs.
Coincidentally I found today that if I run SimH with VMS 5.5-2H4 on a
Raspberry Pi Zero 2 W, I get 3.7 VUPs, which is pretty identical to the
VAXstation 3100M38 that it is emulating. Hardware cost is $15 and size
tiny. A throwaway VAX! See
https://www.raspberrypi.com/products/raspberry-pi-zero-2-w/.
Actually, as it’s a quad-core cpu, I could probably run a cluster of three
or four instances on it for $15!
Not challenging the value-for-money of the above unit but recently I was curious
how my seven-year-old, dual core, 2.7GHz i5 (macOS 12.0.1) compared to my 20+
year workhorse PWS500.
Digital Personal WorkStation with 1 CPU and 1536MB running VMS V8.4-2L1
Approximate System VUPs Rating : 159.0 ( min: 159.0 max: 159.0 )
innotek GmbH VirtualBox with 2 CPU and 3584MB running VMS V9.1-A
Approximate System VUPs Rating : 199.1 ( min: 195.8 max: 204.2 )
So moving to a (virtual) instance of VMS won't cost me anything performance-wise
even if on this humble laptop. BTW: running two concurrent VUPS.COM produces
much the same result.
For comparison...
HP rx2660 (1.40GHz/6.0MB) with 4 CPU and 14335MB running VMS V8.4-2L1
INFO: Preventing endless loop (10$) on fast CPUs
Approximate System VUPs Rating : 495.1 ( min: 487.2 max: 499.6 )
This reliance on a DCL procedure to calculate VUPS isn't very reliable, other
than comparing the same command procedure on different systems to see the
difference.

I have two command procedures, CALCULATE_VUPS.COM and CALCVUPS.COM. They sure
don't agree with each other.

$ show lic/cha
VMS/LMF Charge Information for node DFE90A
This is a VAXstation 4000-90A, hardware model type 475
$ @calcvups
Approximate System VUPs Rating : 27.0
---------------------------------------------------------
AS800> show lic/cha
VMS/LMF Charge Information for node AS800
This is a AlphaServer 800 5/500, hardware model type 1585
AS800> @calcvups
Approximate System VUPs Rating : 141.8
---------------------------------------------------------

and

on the VAXstation 4000-90A
$ @calculate_vups
Approximate System VUPs Rating : 109.0 ( min: 109.0 max: 109.0 )

on the AlphaServer 800
AS800> @calculate_vups
Approximate System VUPs Rating : 516.0 ( min: 516.0 max: 516.0 )

So the gain of the AlphaServer over the VAXstation is about a factor of 4, using
the CALCVUPS.COM procedure, while the factor with CALCULATE_VUPS.COM is about
3.6. Ok, not too far off. However, consider the claims of performance sure is
a large difference between the two procedures. Don't know why.

Seems as if using either procedure can show differences, but not to be trusted
for showing total performance.

Now I think I'll see what the RX2660 has ...

Itanic> @calcvups
Approximate System VUPs Rating : 594.0
Itanic> @calculate_vups
INFO: Preventing endless loop (10$) on fast CPUs
Approximate System VUPs Rating : 2410.7 ( min: 2410.0 max: 2413.0 )

Again, the difference between the procedures is about a factor of 4.

Ok to use to see the Itanic is about 5 times the performance (using these
procedures) of the AlphaStation. Ain't process shrinks wonderful?

Just don't trust the totals ...
--
David Froble Tel: 724-529-0450
Dave Froble Enterprises, Inc. E-Mail: ***@tsoft-inc.com
DFE Ultralights, Inc.
170 Grimplin Road
Vanderbilt, PA 15486
Mark Daniel
2021-11-13 21:56:30 UTC
Reply
Permalink
8< snip 8<
Post by Dave Froble
Not challenging the value-for-money of the above unit but recently I was curious
how my seven-year-old, dual core, 2.7GHz i5 (macOS 12.0.1) compared to my 20+
year workhorse PWS500.
Digital Personal WorkStation with 1 CPU and 1536MB running VMS V8.4-2L1
Approximate System VUPs Rating : 159.0 ( min: 159.0 max: 159.0 )
innotek GmbH VirtualBox with 2 CPU and 3584MB running VMS V9.1-A
Approximate System VUPs Rating : 199.1 ( min: 195.8 max: 204.2 )
So moving to a (virtual) instance of VMS won't cost me anything performance-wise
even if on this humble laptop.  BTW: running two concurrent VUPS.COM produces
much the same result.
For comparison...
8< snip 8<
Post by Dave Froble
This reliance on a DCL procedure to calculate VUPS isn't very reliable,
other than comparing the same command procedure on different systems to
see the difference.8< snip 8<
...ba-boom.
--
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.
Dave Froble
2021-11-13 00:21:55 UTC
Reply
Permalink
Post by Slo
Compiling a C package (73 modules) takes 12 minutes on my FreeAXP,
but about 2 on the XP1000. I'm not complaining; it's a miracle these
things work, really. I rarely do full builds, at most 2 or 3 files change
during an editing session, so this is acceptable to me.
Hoping to install AXPbox and see if I get a better ration than 6:1.
By the way, FreeAXP = 42, XP100 = 318 VUPs.
I'm not sure why you're looking at emulators as long as the XP1000 is working.
If it is what you need, then use it. Sometime down the road you'll be able to
run VMS in a VM instance on x86. That will make things easier.
--
David Froble Tel: 724-529-0450
Dave Froble Enterprises, Inc. E-Mail: ***@tsoft-inc.com
DFE Ultralights, Inc.
170 Grimplin Road
Vanderbilt, PA 15486
Slo
2021-11-13 09:49:16 UTC
Reply
Permalink
Post by Dave Froble
I'm not sure why you're looking at emulators as long as the XP1000 is working.
If it is what you need, then use it. Sometime down the road you'll be able to
run VMS in a VM instance on x86. That will make things easier.
Because the hardware will die eventually, plus it's noisy and bulky.

What I do now is compile on the XP1000, then, for the .C or .H file I last
modified, I hit just one key from my editor, which does this:

$COPY/FTP/ASCII/PASSIVE/LOG {file} 192.168.0.84"username password"::"/dka0/myproj1/src/{file}"
192.168.0.84 is my FreeAXP OpenVMS on my desktop Windows 10.
(I don't use DECnet on either box, but the above syntax works beautifully.
File attributes are preserved, it just works!)

I then may run the (much slower) compile+link on FreeAXP, so my myproj1
sources on both boxes are 100% in sync.
But even this slow build is just fine, as just a few files are usually touched.
Of course, I can edit the sources on the FreeAXP, and then use the above
command to FTP over to the XP1000 -- it's perfectly symmetrical.

Next steps:
1. Try FreeAXP on my mediocre laptop, so I don't have to be at the desk at all.
2. Try AXPbox

After about 15 years of not touching OpenVMS, I'm having fun again.
Bob Gezelter
2021-11-13 16:44:04 UTC
Reply
Permalink
Post by Slo
Post by Dave Froble
I'm not sure why you're looking at emulators as long as the XP1000 is working.
If it is what you need, then use it. Sometime down the road you'll be able to
run VMS in a VM instance on x86. That will make things easier.
Because the hardware will die eventually, plus it's noisy and bulky.
What I do now is compile on the XP1000, then, for the .C or .H file I last
$COPY/FTP/ASCII/PASSIVE/LOG {file} 192.168.0.84"username password"::"/dka0/myproj1/src/{file}"
192.168.0.84 is my FreeAXP OpenVMS on my desktop Windows 10.
(I don't use DECnet on either box, but the above syntax works beautifully.
File attributes are preserved, it just works!)
I then may run the (much slower) compile+link on FreeAXP, so my myproj1
sources on both boxes are 100% in sync.
But even this slow build is just fine, as just a few files are usually touched.
Of course, I can edit the sources on the FreeAXP, and then use the above
command to FTP over to the XP1000 -- it's perfectly symmetrical.
1. Try FreeAXP on my mediocre laptop, so I don't have to be at the desk at all.
2. Try AXPbox
After about 15 years of not touching OpenVMS, I'm having fun again.
Slo,

Not to flog a dead horse, but I have found that emulated instances and VM instances can require different tuning than real hardware. For one, the I/O penalties are different.

If the emulator is running slowly, it can be worthwhile exploring the nature of the bottleneck. Virtual devices, particularly virtual devices backed by the same physical spindle, can experience significant contention issues.

Your mileage (or, more to the point, performance ratio) can, and will often, vary.

- Bob Gezelter, http://www.rlgsc.com
^P
2021-11-14 10:26:29 UTC
Reply
Permalink
Post by Bob Gezelter
Post by Slo
Post by Dave Froble
I'm not sure why you're looking at emulators as long as the XP1000 is working.
If it is what you need, then use it. Sometime down the road you'll be able to
run VMS in a VM instance on x86. That will make things easier.
Because the hardware will die eventually, plus it's noisy and bulky.
What I do now is compile on the XP1000, then, for the .C or .H file I last
$COPY/FTP/ASCII/PASSIVE/LOG {file} 192.168.0.84"username password"::"/dka0/myproj1/src/{file}"
192.168.0.84 is my FreeAXP OpenVMS on my desktop Windows 10.
(I don't use DECnet on either box, but the above syntax works beautifully.
File attributes are preserved, it just works!)
I then may run the (much slower) compile+link on FreeAXP, so my myproj1
sources on both boxes are 100% in sync.
But even this slow build is just fine, as just a few files are usually touched.
Of course, I can edit the sources on the FreeAXP, and then use the above
command to FTP over to the XP1000 -- it's perfectly symmetrical.
1. Try FreeAXP on my mediocre laptop, so I don't have to be at the desk at all.
2. Try AXPbox
After about 15 years of not touching OpenVMS, I'm having fun again.
Slo,
Not to flog a dead horse, but I have found that emulated instances and VM instances can require different tuning than real hardware. For one, the I/O penalties are different.
If the emulator is running slowly, it can be worthwhile exploring the nature of the bottleneck. Virtual devices, particularly virtual devices backed by the same physical spindle, can experience significant contention issues.
Your mileage (or, more to the point, performance ratio) can, and will often, vary.
- Bob Gezelter, http://www.rlgsc.com
Indeed, this is the case, would it not therefore, be quite interesting to have a real bare metal solution for performance demanding emulated systems?

^P
Slo
2021-11-15 15:23:47 UTC
Reply
Permalink
Post by Bob Gezelter
Not to flog a dead horse, but I have found that emulated instances and VM instances can require different tuning than real hardware. For one, the I/O penalties are different.
If the emulator is running slowly, it can be worthwhile exploring the nature of the bottleneck. Virtual devices, particularly virtual devices backed by the same physical spindle, can experience significant contention issues.
Bob, not sure what you are referring to.
My FreeAXP stuff is all in C:\FreeAXP, *.cfg, .ISO, .vdisk.
Are saying that if I move, say, the latter two image files to another drive
(or SSD), I should expect improvements?

Note that I'm a home hobbyist with no production and performance worries.
I mostly play with C and DCL and try clean up the pile of decades of
code and junk that probably nobody will inherit...
Hunter Goatley
2021-11-16 05:31:38 UTC
Reply
Permalink
Actually, as it’s a quad-core cpu, I could probably run a cluster of three
or four instances on it for $15!
I run three SIMH VAXen 24/7 on a Pine A64 board (also a $15 quad-core
CPU). I keep meaning to try more just to see how many I could run before
the Pine is overloaded just trying to keep idle VAXen running. But for
occasional jobs, the three run just fine (also at 3.7 VUPs), and the
Pine can still be used for doing other things.
--
Hunter
------
Hunter Goatley, Process Software, http://www.process.com/
***@goatley.com http://hunter.goatley.com/
Loading...