[Libreoffice] [PATCH] some changes for better performance, writer.

[Norbert Thiebaud]

On Tue, Jan 10, 2006 at 10:25 PM, Kayo Hamid
<kayo.hamid at gekkolinux.com.br> wrote:
> On 08/01/2011 17:00, Norbert Thiebaud wrote:
>>
>> I' confused:
>> -    sal_Int32 nCount = xAcc->getAccessibleChildCount();
>> -
>> -    for( sal_Int32 i=0; i<  nCount; i++ )
>> +    const sal_Int32 nCount = xAcc->getAccessibleChildCount();
>> +    sal_Int32 i = 0;
>> +
>> +    while(i<  nCount)
>>      {
>> -        if( xAcc->getAccessibleChild( i ) == xThis )
>> -            return i;
>> +           if( xAcc->getAccessibleChild( i ) == xThis )
>> +                   return i;
>> +           i++;
>>      }
>> +
>>
>> Unless I'm really dense: this simply change a for() loop into a
>> while() loop doing exactly the same thing
>> Why is it supposed to improve performance ?
>>
>> as for the rest:
>> Did you really measure that turning a passed-by-value scalar (which
>> caller's value cannot be modified by the callee, by definition) into a
>> const passed by value
>>  scalar, has _actually_ any performance impact ?
>>
>> Norbert
>
> Hello Norbert, first of all i'm new at coding, I really do not know if we
> have any improve, but after looking a lot for c++ performance tips (and I
> continue looking and reading tips for that) I have in my mind some things
> like:
>
> for/while -> it's better to use while, it's not a huge improve, but you can
> find tips that say 'use while, it's better'. I just do it.

I'm afraid that is a bad advice.
I doubt very much that there is any such performance impact (and if
so, open a bug with the compiler). and even if there was a one or 2
instructions differences, this would have to be in a very-very tight
loop to be significant (i.e measurable).

>
> const -> my goal is to change the local variable to const is to put it in
> read-only, wich is better,
Why is it better ?

>
> It's because all this that I ask for one review, I want to learn with
> everyone here. I'm in the right way?
Sure, and I gave you my review. Unfortunately, for this specific patch
it is not favorable because the changes in the patch do not achieve
what is advertised in the description of the patch.

but do not be discouraged by this particular outcome... As you said,
this is an opportunity to learn and come up with better patches :-)

? My goal is to code things that are
> better, even with minimal improves.
Performance improvements are a tricky business, not the first thing
you want to dive in. and 'minimal' (or not) performance improvements
should not be postulated, but measured or possibly demonstrated (less
probant).

May I suggest that you pick something in
http://wiki.documentfoundation.org/Easy_Hacks

Norbert

>
> revol_
>
> _______________________________________________
> LibreOffice mailing list
> LibreOffice at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libreoffice
>