Re: Software Pipelining

"Armel" <>
Wed, 24 Sep 2008 10:52:10 +0200

          From comp.compilers

Related articles
[6 earlier articles]
Re: Software Pipelining (kamal) (2008-09-10)
Re: Software Pipelining (johnhull2008) (2008-09-11)
Re: Software Pipelining (Tim Frink) (2008-09-16)
Re: Software Pipelining (Tim Frink) (2008-09-16)
Re: Software Pipelining (Pertti Kellomaki) (2008-09-17)
Re: Software Pipelining (Christopher Glaeser) (2008-09-21)
Re: Software Pipelining (Armel) (2008-09-24)
Software pipelining napi@rangkom.MY (1991-07-04)
Re: Software pipelining (1991-07-19)
| List of all articles for this month |

From: "Armel" <>
Newsgroups: comp.compilers
Date: Wed, 24 Sep 2008 10:52:10 +0200
Organization: les newsgroups par Orange
References: 08-08-072 08-08-086 08-08-092 08-09-060 08-09-079 08-09-111
Keywords: optimize
Posted-Date: 24 Sep 2008 12:12:52 EDT

"Christopher Glaeser" <> a ecrit
>> Even though I understand that this is an example, I still have the
>> feeling that with just a single functional unit, the benefits of
>> pipelining are marginal,
> Actually, the benefits of software pipelining can be significant, even
> for a single functional unit, if that functional unit has a multistage
> pipeline. Assuming the number of iterations is sufficiently large to
> mask the prologue and epiloque, the number of cycles required to
> execute any given iteration doesn't really matter. The key is to
> minimize the number of cycles needed to start each successive
> iteration, otherwise known as the minimum iteration interval (MII).

Yes, sure it (is/was) applied very often for unrolling various byte/small
data handling, when handling each item individually costs less than or is
similar to the prologue/epilogue of the loop!


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.