From: njn@cs.mu.oz.au (Nicholas James NETHERCOTE)
Subject: Re: Thoughts and Opinions or something like that
Date: 26 Apr 2001 02:12:35 GMT
Date: 2001-04-26T02:12:35+00:00 [thread overview]
Message-ID: <9c806j$kt4$1@mulga.cs.mu.OZ.AU> (raw)
In-Reply-To: 9c6evg$em1$1@marduk.litech.org
jmv16@cornell.edu (Jeffrey M. Vinocur) writes:
>Tarjei T. Jensen <tarjei.jensen@kvaerner.com> wrote:
>>
>>As far as instructions are concerned I always fancied deviding branch
>>instructions into two groups: Those who branch most of the time and those
>>who seldom branch. The idea was that one could skimp on branch prediction on
>>the CPU. For a while or for loop one would use the former type of
>>instructions. Since I have heard of nobody who have done this I assume that
>>the idea is not particularly good.
>Well, the goal isn't to lessen branch prediction -- the silicon
>is there either way -- but rather to reduce mispredicted
>branches. The most basic approach, I guess, is assume all
>backwards branches will be taken -- this gets loops. One
>not-very-complicated approach is to cache whether a branch was
>taken last time through and predict the same. This approach,
>however, does exactly what you suggest; it gets loop iterations
>2,3,4,...,n-1 all correct.
I remember seeing a suggestion somewhere for a pragma in a C-- implementation
that allowed you to estimate the probability that a condition succeeded. So it
sounds like it must be a worthwhile idea, assuming your estimates are accurate.
--
Nick Nethercote
njn[AT]cs.mu.oz.au
next prev parent reply other threads:[~2001-04-26 2:12 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-04-24 17:12 Thoughts and Opinions or something like that chris.danx
2001-04-24 21:35 ` Ted Dennison
2001-04-24 23:27 ` chris.danx
2001-04-25 7:40 ` Andrew Cooke
2001-04-25 10:42 ` chris.danx
2001-04-25 17:26 ` Warren W. Gay VE3WWG
2001-04-26 14:22 ` Marcin 'Qrczak' Kowalczyk
[not found] ` <3AE69665.63E362CD@info.unicaen.fr>
2001-04-25 10:51 ` chris.danx
2001-04-25 13:51 ` Jerzy Karczmarczuk
2001-04-25 13:52 ` chris.danx
[not found] ` <3AE6DAB3.899FF645@andrewcooke.free-online.co.uk>
2001-04-25 14:30 ` chris.danx
[not found] ` <3AE7E37B.F384DCDA@info.unicaen.fr>
2001-04-26 9:02 ` Play with virtual machines (Was: Thoughts and Opinions...) chris.danx
[not found] ` <3AE81234.267643AC@kfunigraz.ac.at>
2001-04-26 12:57 ` chris.danx
2001-04-27 2:14 ` Larry Elmore
2001-04-27 3:27 ` Goldhammer
2001-05-03 17:00 ` singlespeeder
2001-05-03 17:03 ` singlespeeder
2001-04-25 9:06 ` Thoughts and Opinions or something like that Tarjei T. Jensen
2001-04-25 12:12 ` Jeffrey M. Vinocur
2001-04-26 2:12 ` Nicholas James NETHERCOTE [this message]
2001-04-26 18:23 ` Keith Thompson
2001-04-25 12:09 ` Alain Fischer
2001-04-27 18:20 ` brian hiles
2001-04-28 1:27 ` Gregory Toomey
2001-04-28 21:34 ` chris.danx
2001-04-28 21:34 ` chris.danx
2001-04-30 8:31 ` Jon Beniston
2001-05-04 8:49 ` Biep @ http://www.biep.org
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox