On Tue, May 26, 2015 at 11:15 AM, Claudio Jeker <[email protected]> wrote:
> On Tue, May 26, 2015 at 04:27:23PM +0200, Janne Johansson wrote: > > ..or, if it really is important for you, why not set up a test and > > benchmark it? > > > > I have no idea if or when amd64 would have surpassed i386 in performance, > > and if it is better how much better. > > But if this was an important topic worthy of the science, then by all > means > > do test it on your particular hardware and present the results. > > > > I do tend to agree with the comments already posted so I won't repeat > them, > > but if +/-5% is seriously interesting enough for people to keep > remembering > > and asking for, then it surely would be worth making a small effort too, > no? > > > > > > > > 2015-05-26 0:42 GMT+02:00 Bill Buhler <[email protected]>: > > > > > I'm preparing a new flash image for an Intel Atom dual core based > router > > > with 2gb of ram. I'm curious if there are current comments on the > current > > > performance of the two platforms? I know in the past the i386 was > actually > > > faster at things like PF, but that was several years ago. > > > > > > > > > > > Also ask yourself if a few % more speed is worth to have no proper W^X > support. At least unless you run -current and even then amd64 has probably > the most restrictive W^X policy for userland and kernel. Also more people > are running amd64 and so has better testing in general. > > -- > :wq Claudio > > I'll throw out an unsubstantiated guess, the change to 64 bit time makes amd64 perform better than i386 at packet filtering. Disclaimer, no idea how often time is interesting in a standard pure packet filtering environment.
