> -----Original Message-----
> From: IBM Mainframe Discussion List On Behalf Of Tom Grieve
> Sent: Wednesday, October 25, 2006 4:04 AM
> To: IBM-MAIN@BAMA.UA.EDU
> Subject: Re: Java/LE Tuning Advice sought
>
> I still think it's probably a downstack size problem. Try
> playing with the size a bit -
>
> STACK(128K,128K,ANY,KEEP,256K,526K)
>
> or -
>
> STACK(128K,128K,ANY,KEEP,1024K,1024K)
>
> If the last one doesn't make a difference, then I'm completely wrong!
I took a console dump of the a/s and found what looks like the "main
TCB" (it had a little less than 1.5GB of storage associated with it)
with VSMDATA 'NOG SUMM'. Examined the LEDATA for that TCB and saw only
a single stack segment (above the line), but 158 heap segments (also
above the line). According to the LE doc, the "trailing" size parms are
for 24-bit (below the line) allocations.
Oh, I had previously tried HEAP(8M,2M,ANY,KEEP) but the "benchmark" job
took 1.5 HOURS with that setting.
It also seemed odd that the a/s startup was 2 minutes faster with 128K
specified in all four parms, but the "benchmark" job ran 2 minutes
slower. I guess "TANSTAAFL" applies here, too. :-|
Now, if I could just get something readable from -verbosegc.... I don't
know what charset it's using or how/where to change it, but it's neither
ASCII nor EBCDIC. Output looks compressed and/or encrypted.
-jc-
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
