Performance updates

[Bill Baxter]

On Fri, Nov 14, 2008 at 9:06 AM, Moritz Warning <moritzwarning at web.de> wrote:
> On Fri, 14 Nov 2008 08:48:53 +0900, Bill Baxter wrote:
>
>> On Fri, Nov 14, 2008 at 7:49 AM, Moritz Warning <moritzwarning at web.de>
>> wrote:
>>> On Thu, 13 Nov 2008 06:19:27 -0500, bearophile wrote:
>>>
>>>> In this post I show few things I have found/collected in the last
>>>> weeks related to the performance of the code compiled with DMD.
>>>>
>>> [..]
>>>>
>>>> The D1 docs strongly suggest to use foreach every time it's possible,
>>>> avoiding to use the less handy for(), so for almost a year I have
>>>> assumed foreach is as fast as the for() but this two versions shows
>>>> differences:
>>>>
>>>>
>>> I have noticed the same problem with foreach. A for loop is often
>>> faster. :/
>>
>> One thing to be clear about when talking about the speed of foreach is
>> whether you're using it on a built-in array or on a user type with an
>> opApply.  The former is expected to be as fast as for(), the latter is
>> definitely not because it has to call a delegate each time around the
>> loop.
>
> Nobody would use "for"-loops with opApply in real world code. :P

Er, right, which is why using a for() loop is faster that foreach() in
that case.  Sorry if I wasn't clear about that.

--bb