This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Reorder some tree codes


Mark Mitchell <mark@codesourcery.com> writes:

| Gabriel Dos Reis wrote:
| > Mark Mitchell <mark@codesourcery.com> writes:
| > [...]
| > | This is clearly the kind of change that (if it's an otherwise good
| > | change) is acceptable in Stage 3; small, but with a non-trivial impact
| > | on compile-time performance.
| > I think 0.25% speed up is within the margin of statistical error.
| 
| You cannot know that unless you know how Nathan measured, what the
| variance on his measurements is, etc.

Probably true, but if Nathan keeps secret hot he did all thoses, I
believe it is legitimate to believe it is within the error margin.

| 
| I can state that the way in which we do these measurements typically
| has a variance in the range of 0.5 seconds over a total run, which
| takes about 500 seconds.  So, the normal variance is about 0.1%, which
| makes 0.25% significant.  This measurement involves building all of Qt
| (with the usual Qt makefiles) and an associated application, including
| linking.  Therefore, it underestimates the compiler speedup per se, as
| fixed costs (like linking) exist.
| 
| -- 
| Mark Mitchell
| CodeSourcery, LLC
| mark@codesourcery.com
| (916) 791-8304

-- 
                                                       Gabriel Dos Reis 
                                           gdr@integrable-solutions.net


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]