[Info-vax] VSI OpenVMS Alpha 8.4-2L2

[Jan-Erik Soderholm]

Den 2017-07-25 kl. 15:05, skrev Arne Vajhøj:
> On 7/25/2017 12:22 AM, Camiel Vanderhoeven wrote:
>> Op maandag 24 juli 2017 16:46:02 UTC+2 schreef Arne Vajhøj:
>>> On 7/24/2017 7:00 AM, Jan-Erik Soderholm wrote:
>>>> Den 2017-07-13 kl. 00:43, skrev Jan-Erik Soderholm:
>>>>> First "look and feel" from 8.4-2L2 Alpha.
>>>>
>>>> Now, all three systems run 8.4-2L2. The two last (dev and prod)
>>>> was still at 8.2 and to get supported upgrade path I went:
>>>> - HP 8.2 => HP 8.4
>>>> - Add 8.4 UPDATE 10
>>>> - HP 8.4 => VSI 8.4-2L2
>>>>
>>>> A very simple DCL script shows slightly better performance.
>>>> A DCL loop with 10.000 f$user() calls runs approx 30% (!) faster.
>>>> Two other DCL loops with f$extract and f$trnlnm are 5% faster.
>>>> One test with f$getdvi is 5% slower. A test with open/write/delete
>>>> on a file is more or less unchanged, probaby expcted since it is
>>>> mainly I/O bound.
>>>>
>>>> How much of this that comes from 8.2 vs. 8.4 or from the
>>>> specific 8.4-2L2 changes, I don't know. But I seem to remember
>>>> that I got quite equal results from 8.2 and 8.4 anyway...
>>>
>>> Does any of these VMS versions have specific changes to
>>> DCL interpreter, SYS$GETJPI(W), SYS$TRNLNM and SYS$%GETDVI(W)?
>>>
>>> If not then I am a bit skeptic that general VMS changes
>>> would have that big an impact.
>>
>> No specific changes, but everything was compiled for EV6, enabling
>> the use of the byte and word extensions, which speeds up character string
>> handling (of which DCL does a lot).
> 
> Ah.
> 
> That could be an explanation.
> 
> Arne
> 
> 

Yes, of course. That was the explanation I expected since the recompile
with byte/word access support was discussed around the 9th and 10th July.
See thread "VSI OpenVMS Alpha V8.4-2L1 Layered Product Subset Available".

I just thought that it was interesting with some real world stats... :-)