tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Microbenchmarks

Could you be a bit more specific on this? Particulary what still needs
Whatever bugs you :-). The numbers for fork and exec are quite prominent. I
wonder if something bad is happening in the pmap/uvm.

Yet, some points merit enlightment. For example :

* Are we garanteed that Linux fork is our fork and not a vfork or something hybrid? Vfork is not being benchmarked, it is a pity.

* Even if compiled with -O0, the benchmark runs on a (I guess) optimized kernel so differences between gcc 4.1 and 4.3 might be significant;

* Did you run the benchmark once, or are the values computed as means over a few passes?

* I guess the system call performance is somehow linked to the command executed, and thus if that command is linked statically or dynamically;

* And so on.

As regarding fork, what is the exact policy of VM duplication? Do we make all data and stack pages RO in both father and child, do we copy stack and leave data RO only in child till a father write, do we invalidate every pages of the child and copy on demand, etc.?


Home | Main Index | Thread Index | Old Index