I've been testing a C routine I've written that uses zlib. The module is
compiled with xlc with -O5.

The thing I can't explain is this. Testing the module on two p615 boxes, one
is a 7029-6E3, single CPU 1.2GHz running AIX 5.1 and the other is a 7029-6C3
single CPU 1.2GHz running AIX 5.2, both machines have 1GB RAM (the test
program is small) and 2 x mirrored 36GB disks (no IO involved). The program
runs the same 8K of random data through zlib compression, followed by a base
64 encode, and it does this 50,000 times.

On the AIX 5.1 machine, the runs average real   0m22.75s, user    0m22.48s

And on the 5.2 machine, real    0m34.68s, user    0m34.33s

I'm at a loss to explain why this statically linked binary runs so much
slower on 5.2 than 5.1, given the two p615s are identically configured (save
for one being the rack version and one is the tower).

Any ideas on what could be happening or how I could identify where the extra
time is going?

[ Post a follow-up to this message ]