Interesting line in the recent Dr Dobbs article about profiling?

[H. S. Teoh]

On Thu, Jul 25, 2013 at 08:15:42PM +0200, Gary Willoughby wrote:
> I've just read the article over at Dr Dobbs by Walter
> 
> http://www.drdobbs.com/cpp/increasing-compiler-speed-by-over-75/240158941
> 
> and this line caught my eye:
> 
> >Even if you know your code well, you're likely wrong about where
> >the performance bottlenecks are. Use a profiler. If you haven't
> >used one on your codebase in a while, it's highly likely there's
> >a bottleneck in there that's fixable with just a few lines of
> >code.

I can personally testify to the truth of this statement. All too often,
too much time is spent "optimizing" parts of code that are actually
irrelevant to the performance of the program, because the bottleneck
lies elsewhere, often in an unexpected place. (Not to mention that such
premature optimization makes your code harder to write, harder to read,
harder to maintain, and therefore more prone to bugs.)

I remember at least one instance in which a profiler immediately
revealed an fprintf (this was C/C++ code, back in the day) where it
shouldn't be, the removal of which instantly improved performance by at
least 30%, probably more.


> What profilers do you use with D especially for Linux and Mac? I've
> been compiling with -profile and trying to grok the output.

I use gdc with gprof, usually. I've only tried dmd -profile a few times.


T

-- 
I think Debian's doing something wrong, `apt-get install pesticide',
doesn't seem to remove the bugs on my system! -- Mike Dresser