On Wednesday, 11/26/2003 at 05:02 CET, Vladimir Fayzulaev <[EMAIL PROTECTED]> wrote: > Can anybody point me to the link (or to the code:) with benchmark to measure > CPU time for malloc(), calloc(), realloc() and free() calls mix. It must be > C code, multi-threaded is a plus. > It is required to test performance (only clear CPU time) in OMVS and compare > it later with Linux.
Yet another meaningless number since in the Real World more than one address space (MVS) or Linux server will be issuing those calls at the same time, invisibly consuming system resources that will affect the amount of time it takes to get through the call. The only thing you can ascertain is that *on this occasion*, OMVS malloc() took x ms and a Linux malloc() took y ms. This kind of benchmark on zSeries is, I feel, a real waste of time. Any attempt to make generalizations using such a benchmark will not give an accurate picture of the capabilities of the platform. How will you relate such a measurement to real world measurements of, say, the maxium arrival rate of HTTP transactions the platform can handle? Or bytes transferred per second per session? Or database queries? Or ....??? Alan Altmark Sr. Software Engineer IBM z/VM Development
