W dniu sob, 21.10.2017 o godzinie 10∶01 +0200, użytkownik Paweł Hajdan,
Jr. napisał:
> On 20/10/2017 18:15, Michał Górny wrote:
> > W dniu pią, 20.10.2017 o godzinie 17∶42 +0200, użytkownik Paweł Hajdan,
> > Jr. napisał:
> > > Curious, do we have any measurements/estimates of the performance cost?
> > 
> > With a single thread serial processing of all hashes, it's just sum of
> > times involved in every hash, i.e. Th = T1 + T2 + T3 + ... You'd have to
> > get some numbers to get something smarter out of it.
> > 
> > If we assume we can do N threads, then cost of N algorithms is equal to
> > the slowest of them all. Which implies that having N algorithms is
> > fastest on systems capable of at least N threads.
> > 
> > Taking a random comparison [1], it seems that SHA3/512 is 3-5 times
> > slower than SHA2/512.
> 
> How large part of dependency calculation / other portage's operation is
> this though?
> 
> My point is, did profiling turn out hash computation as bottleneck, or
> is this more speculative?

Purely speculative.

> I'm still in favor of modernizing the hashes, just somewhat skeptical
> when performance is being mentioned.
> 

FWICS BLAKE2 can be even 2.5 times faster than SHA2, so we'll probably
go with that. In this case, the performance impact will be negligible --
in fact, it should be faster than the current set of three hashes.

-- 
Best regards,
Michał Górny


Reply via email to