Performance Regressions; Previously: Death by error checks.

cdg@nullstone.com (Christopher Glaeser)
28 Dec 1995 11:34:19 -0500

          From comp.compilers

Related articles
Grammars for future languages schinz@guano.alphanet.ch (1995-10-22)
Re: Death by error checks. hbaker@netcom.com (1995-12-19)
Performance Regressions; Previously: Death by error checks. cdg@nullstone.com (1995-12-28)
Re: Performance Regressions; Previously: Death by error checks. hbaker@netcom.com (1995-12-30)
| List of all articles for this month |

From: cdg@nullstone.com (Christopher Glaeser)
Newsgroups: comp.compilers
Date: 28 Dec 1995 11:34:19 -0500
Organization: Compilers Central
References: 95-10-103 95-12-113
Keywords: performance

Henry Baker <hbaker@netcom.com> writes:


> Unfortunately, whereas before, he could rely upon the compiler staying
> just about as stupid from release to release, he now has to contend
> with compilers that get 'smarter' from release to release.


While it is true that many compilers continue to improve with each
release, the *implication* that the performance of a new release is
monotonically non-decreasing is false. It is not uncommon for new
compiler releases to have a ratio of performance improvements versus
performance regressions of about three to one.


Of course, SPEC performance is always monotonically non-decreasing,
even if the release must be postponed. It is possible that a code
fragment of a SPEC benchmark runs slower, but it must be offset by a
code fragment that improves by equal or greater amount, since the
emphasis is on the total sum gain.


Best regards,
Christopher Glaeser cdg@nullstone.com
Nullstone Corporation http://www.nullstone.com


--


Post a followup to this message

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