Ok, just because I've read this mistake over and over again, I feel a need to correct it. The "client" and "server" VMs have NOTHING to do with things being run on the "server" or "client". The names come from the fact that the "client" VM is designed for normal user applications where startup time is a major factor. The "server" VM is designed for applications (like servers) that run longer and thus can trade off startup time for more optimized execution. It's kind of like making space-speed tradeoffs in g++ since using a LOT of space may slow down startup while the OS swaps things to disk and whatnot but it runs faster once its loaded. That being said, I think the g++ benchmarks should have used -O3, but I guess the real point that I see is that you aren't necessarily shooting yourself in the foot performance-wise if you use java instead of c++.
To give a quick idea of how verbose X can be, I have a program that runs a hardware simulation which I ocassionally have to run remotely. The last time I was running it, I happened to take a look at MRTG, and found it taking an average bandwidth of 4 MBits/s.
To clarify, this program basically has a bunch of text boxes that change quickly. That's it. If I didn't have a direct Ethernet link to campus, it would take me a lot more than 45 mins to run a full simulation.
Slashdot Mirror
Ok, just because I've read this mistake over and over again, I feel a need to correct it. The "client" and "server" VMs have NOTHING to do with things being run on the "server" or "client". The names come from the fact that the "client" VM is designed for normal user applications where startup time is a major factor. The "server" VM is designed for applications (like servers) that run longer and thus can trade off startup time for more optimized execution. It's kind of like making space-speed tradeoffs in g++ since using a LOT of space may slow down startup while the OS swaps things to disk and whatnot but it runs faster once its loaded. That being said, I think the g++ benchmarks should have used -O3, but I guess the real point that I see is that you aren't necessarily shooting yourself in the foot performance-wise if you use java instead of c++.
To give a quick idea of how verbose X can be, I have a program that runs a hardware simulation which I ocassionally have to run remotely. The last time I was running it, I happened to take a look at MRTG, and found it taking an average bandwidth of 4 MBits/s.
To clarify, this program basically has a bunch of text boxes that change quickly. That's it. If I didn't have a direct Ethernet link to campus, it would take me a lot more than 45 mins to run a full simulation.
You mean you don't like such lovely FreeBSD games as: /usr/games/factor?
/usr/games/pom
/usr/games/wargames
and