> I've found that server/client combinations using blcksock.pas cannot > communicate at higher than the kernel HZ setting -- quite a severe > restriction for me since the actual data packets in the client/server > protocol are small and rapidly processed. > > By commenting out the sleep call in blcksock.pas at: > > //not drain CP on large downloads.. > Sleep(0); > > the throughput grows enormously. I'm on Linux 2.6, by the way -- perhaps > Sleep(0) returns in less than a whole time slice on other systems. It seems > to me that the idle time triggered by the sleep call isn't usually a great > idea -- if a connection became a resource hog that would only be because it > had work to do, so isn't it better to use the kernel scheduler to arbitrate > such problems? Perhaps the repercussions of removing the sleep(0) are more > widespread than I understand, though?
This sleep is designed for windows platform. It not slowdown transfers on unload systems, and it made better responsibility on high-load systems. But task scheduling on Win32 and on Linux is totally different. Maybe this sleep is very ineffective on linux platform. Is not a problem to left this sleep for Windows platform only. -- Lukas Gebauer. E-mail: [EMAIL PROTECTED] http://www.ararat.cz/synapse/ - Ararat Synapse - TCP/IP Lib. ------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ synalist-public mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/synalist-public
