[PATCH] livelock in elevator scheduling

The current elevator_linus() doesn't obey the true elevator scheduling, and causes I/O livelock during frequent random write traffics. In such environment I/O (read/write) transactions may be delayed almost infinitely (more than 1 hour). Problem: Current elevator_linus() traverses the I/O reques

locks.c: removal of semaphores

Andrew, I got 5250 Req/s with your locks-sem.patch on normal Apache. It is good performance on normal Apache. Andrew Morton writes: > Kouichi, could you please test the performance of this on > your 8-way with Apache+fcntl serialisation? (the normal > Apache). Please use 2.4.0-test10-pre5, no

Re: Calling current() from interrupt context

Based of my measurement on i386 smp configuration, If a system has plenty of runnable tasks, schedule() produces noticable amount of cache misses at runqueue-head traversing and the goodness calculations. David S. Miller writes: > Some of us actually have instrumented it :-) I added a coloring

Re: zero-copy TCP

Ingo Molnar writes: > On Mon, 4 Sep 2000 [EMAIL PROTECTED] wrote: > > The experiment showed the following prefetching could reduce 20-30% of > > csum_partial_copy_generic() execution time. > > Please test it and post the numbers. csum_partial_copy_generic() already > does prefetching - the r

Re: zero-copy TCP

The experiment showed the following prefetching could reduce 20-30% of csum_partial_copy_generic() execution time. This suggests CPU cause heavy cache miss at checksum calclation and it is also supported by the PMC (performance monitoring counter) measurements. So I think the following statement i