Edwin Dicker wrote: > I'd like to benchmark my own system with differently configured kernels. > Who knows some software to benchmark?? To be honest -- what are you trying to do? If there's just one program you'd like to get running faster, then use that. Think up something you can measure, and measure it. (If it's encoding Oggs, get a *standard* WAV file (your standard, but it shouldn't change), and encode that. If it's compiling kernels, get a *standard* kernel with a standard config, and compile it. If you have a series of programs that are "noticeably slow" and you'd like to accelerate, use them. But what you're going to end up doing is optimizing your kernel to run your benchmark. That's why it's important that your benchmark really reflects what you want to run fast. Kernels inherently involve a huge amount of tradeoffs. As you select one configuration over another, you're probably going to improve one aspect at the cost of others. You might, for example, choose a filesystem that is very fast at creating and deleting files (which you rarely do in bulk), but is a fair bit slower at file access times (which is usually more important). I do hope you understand the importance of turning off as many background services as possible, and making sure either that everything is in cache or nothing is in cache before you run a benchmark. Depending on what you're doing, you might even want to set up run-level 7 *just* to run your benchmark and write out a result to a file. Then you would boot directly to run-level 7 (append 7 to the kernel command line) to run the benchmark. But if you just want to "tune your system": forget it. That's *way* too underspecified an aim. James. -- E-mail address: james | John's Inverse Law of Physics: @westexe.demon.co.uk | You do Physics -- you get inverted.
