Curious. The "java" launcher for SDK 1.4.2 (old SR2) sets this to:
STACK(16384,4096,ANYWHERE,FREE,65536,16384)
Could it be that they got it wrong?
Maybe the combination of running an old version of the SDK with
newer z/OS and LE has caused a mismatch?
It might be interesting to see an LE options/storage report of
running your job without overriding the stack setting.
You would expect to see a ton of segment allocations/frees to the XPLINK
stack.
export _CEE_RUNOPTS="RPTSTG(ON),RPTOPTS(ON)"
Kirk
Chase, John wrote:
X-Posted to MVS-OE list.
-----Original Message-----
From: IBM Mainframe Discussion List On Behalf Of Tom Grieve
This looks like a downstack size problem. The JVM is built
using XPLink; if the downstack size is too small, almost
every new function will program check when it tries to save
registers on entry. LE then has to handle the
0C4 and extend the stack. Try -
STACK(128K,128K,ANY,KEEP,128K,128K)
You might be on to something.... Startup time for the application was
REDUCED by two minutes (from 5.5 min to 3.5). I've asked the
application tester to try the benchmarking job again; will post the
results when I get them.
-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
