www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - std.benchmarking and getting rid of TickDuration

reply Jonathan M Davis <jmdavisProg gmx.com> writes:
I just created a PR aimed at replacing the few pieces of Phobos 
that use TickDuration in their API, and it involves creating the 
std.benchmark module to hold the updated benchmarking functions, 
since they can't be overloads of the existing ones in 
std.datetime given that return types are involved. So, the new 
functions need to go in a different module. Putting them in 
std.benchmark made perfect sense given that that's where we were 
looking at putting them before when Andrei's std.benchmark was 
reviewed (and ultimately dropped). Normally, adding a module to 
Phobos would involve an extended review in the newsgroup, but in 
this case, it's just updated versions of existing functions moved 
to a new module. So, I don't think that a review of that level is 
required, but I figured that I'd bring it up in the newsgroup 
just in case.

https://github.com/D-Programming-Language/phobos/pull/3695

- Jonathan M Davis
Oct 04 2015
next sibling parent Jonathan M Davis <jmdavisProg gmx.com> writes:
LOL. I got the title wrong - std.benchmarking instead of 
std.benchmark. Oh well...

- Jonathan M Davis
Oct 04 2015
prev sibling parent reply Marco Leise <Marco.Leise gmx.de> writes:
Should the examples have `pragma(inline, false)` on the
benchmarked functions? I'm not so worried about inlining as I
am about const folding the benchmarked expressions away.

Concerning the module review I tend to agree with you. It
would only be ... well if I said in std.datetime the functions
were not so prominent, and went under the radar, but as their
own module they have to add enough value to warrant that. But
I am ok with that, too. std.benchmark is not something you'd
use in a production environment on a daily basis like you would
std.json, std.logger or std.allocator and you didn't add new
API.

If there _was_ a review for a new module I would swamp your
progress with feature requests like "validation of return
value of benchmarked functions" or "benchmark for x seconds
with y iterations of the inner loop for comparing benchmarks
for cases where one function is orders of magnitudes faster
than the other". And I'm sure others have lots of ideas, too
and it will turn into an endless bike-shedding discussion
and take another 6 months to get this merged and no one wants
that, right?

-- 
Marco
Oct 05 2015
parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On Monday, 5 October 2015 at 15:51:58 UTC, Marco Leise wrote:

 Should the examples have `pragma(inline, false)` on the 
 benchmarked functions? I'm not so worried about inlining as I 
 am about const folding the benchmarked expressions away.


I'm not sure that it's a good idea to start pasting pragmas for 
inlining all over the place rather than letting the inliner do 
its job, but regardless, there have been enough complaints about 
how pragma(inline...) is designed that I don't know if we can 
even rely on it staying the way it is right now.


 If there _was_ a review for a new module I would swamp your 
 progress with feature requests like "validation of return value 
 of benchmarked functions" or "benchmark for x seconds with y 
 iterations of the inner loop for comparing benchmarks for cases 
 where one function is orders of magnitudes faster than the 
 other". And I'm sure others have lots of ideas, too and it will 
 turn into an endless bike-shedding discussion and take another 
 6 months to get this merged and no one wants that, right?


Andrei previously had a proposal for std.benchmark which added a 
lot more, but it died in the review process (primarily due to it 
relying on static constructors to do some of what it was doing 
IIRC). We may very well get another proposal from him or someone 
else in the future. But this PR is entirely about replacing 
existing benchmarking functions with versions that use Duration 
and MonoTime instead of TickDuration and not about adding 
functionality. Major additions like Andrei worked on before can 
go through the review process as normal, and it's certainly not 
on my todo list. I'm behind enough on that as it is. As it 
stands, the main reason that the PR creates std.benchmark is 
because the new functions cannot got in std.datetime without 
conflicting with the functions that they're replacing (though I 
do think that the benchmarking functions are better placed in 
std.benchmark anyway, and Andrei's proposal did include moving 
them there).

- Jonathan M Davis
Oct 05 2015