Phil, thank you for the prompt answer an for clear explanations. I understand, that pipe() may be not very well optimized, since in most cases this is not a path for many small messages passed between processes, and STREAMS is ised. I just thought (for my naiveness) that this will really have small overhead over context switches.
I'll try to measure other things that generate context switches on a 1CPU/2CPU machine, and if I'll have question, I'd ask... Generally, getrusage() is my tool. I'll learn the whitepaper, and will take a [deep] look on your blog. > The other thing to remember is that Solaris is designed to scale well with > many threads and many CPUs. > Sometimes this stops us from being quite so > optimal on 1P/2P systems. As far as I understand Sun current strategy, Solaris 10 market tooks in account small [1- or 2-CPU] servers, too. Solaris 10 seems to be positioned as OS that covers the areas that Linux covers. So, the scaling issue is not accepted by me as final [potential] customer. And telling that uni- and dual-CPU machines will work slower using Solaris vs Linux, is... somewhat strange. One should not pay twice for the fact that his machine is slow. Anyway, I'll try do additional tests of our application with profiling on Solaris and on Linux, and shall ask questions (if any). This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list [EMAIL PROTECTED]
