The original question in this thread was:
"Has anyone implemented COBOL 5.1 to the point that you have measured the
"promised" performance improvements in application code over the same code
compiled with earlier COBOL compilers?"
We have compiled a few of our programs that run as part of some "utility" jobs
that run daily and modified the JCL to run both the COBOL 5.1 version and the
COBOL 4.2 version. Mostly what these jobs do is read some input file, select
certain records or fields and create a new data set with what was selected.
So, none of them have a real large Procedure Division.
The output below is from a "Food Lion" SMF report (file 19 on the CBT) - it's
formatted in my email, I hope that doesn't get lost. The first entry for each
program on the listing the COBOL 5.1 run - in the case of COBSAM81 program,
the first one was compiled OPT(2) and the second OPT(1) (and we run on a
z10/BC, so we compiled ARCH(8)).
Mostly, the 5.1 programs show improvement in CPU time and in below the line
region. The elapsed time does not always improve, and sometimes gets worse,
but not significantly worse.
COBOL COMPARE JOBS RUN
PROGRAM CPU TIME ELAPSED TOTAL SERVICE REGION REGION
NAME MMMM:SS.TH TIME I/O UNITS BELOW ABOVE
COBSAM61 0:00.33 000:00:03 10,759 7K .3M 14.4M
COBSAM61 0:00.33 000:00:04 10,720 7K .6M 13.5M
COBSAM81 0:23.46 000:03:09 1,093,331 591K .3M 17.5M
COBSAM81 0:23.12 000:02:39 1,093,329 586K .3M 17.5M
COBSAM81 0:23.52 000:03:04 1,093,277 577K .6M 16.5M
MAKEEX 0:03.95 000:00:57 104,410 83K .3M 14.4M
MAKEEX 0:08.11 000:01:01 104,342 162K .4M 13.6M
SS090 0:02.97 000:01:02 97,874 64K .3M 14.1M
SS090 0:03.03 000:00:43 97,790 64K .3M 13.4M
SS092 0:12.87 000:01:34 110,679 265K .4M 14.9M
SS092 0:12.67 000:01:01 110,584 246K .3M 13.7M
SS604 0:04.18 000:00:50 104,250 88K .3M 14.1M
SS604 0:05.00 000:00:46 105,714 101K .3M 13.3M
So, this is not "application" code, but it is what we have measured so far.
HTH,
Greg Shirey
Ben E. Keith Company
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
