Messages in this thread | | | Date | Thu, 19 Mar 2015 15:17:10 +0300 | Subject | Re: + lib-vsprintfc-even-faster-decimal-conversion.patch added to -mm tree | From | Alexey Dobriyan <> |
| |
On Wed, Mar 18, 2015 at 1:04 AM, Andrew Morton <akpm@linux-foundation.org> wrote: > On Mon, 16 Mar 2015 18:19:41 +0300 Alexey Dobriyan <adobriyan@gmail.com> wrote: > >> Rasmus, I redid benchmarks: > > tl;dr ;) Is this an ack or a nack?
New code executes slower for some input on one CPU I've benchmarked, both with -O2 and -Os (Core 2 Duo E6550). The slowdown is in 2-20% range depending on exact number being converted and is reproducible. With -O2 "bad" range is roughly 100-70000, with -Os it is 100-1000 and around 100000.
On another CPU (more modern Core i5-something), new code is uniformly faster giving advertised 10-20-35-40% speedups (yay!)
The ideal situation is still being a) one system call to push PIDs into userspace in bulk (memcpy directly from pid_ns->pidmap[i]), b) one system call to fetch data in binary to userspace given PID or set of PIDs,
without all of this string and /proc overhead.
Alexey
| |