Re: [PATCH] Optimised 1bit blitters

[Vladimir 'phcoder' Serbinenko]

On Tue, Aug 25, 2009 at 4:06 PM, Michal Suchanek<address@hidden> wrote:
> 2009/8/23 Robert Millan <address@hidden>:
>> On Fri, Aug 21, 2009 at 05:33:30PM +0200, Vladimir 'phcoder' Serbinenko 
>> wrote:
>>> +  for (j = 0; j < height; j++)
>>> +    {
>>> +      for (i = 0; i < width; i++)
>>> +        {
>>
>> It's a bit odd, but GCC doesn't seem to optimize those in a single loop.  
>> Could
>> you use "i = 0; i < height * width; i++" instead?  (for this and the other
>> similar instances)
>>
>> I can't comment much on the rest of this patch, as my understanding of
>> graphics is limited.  But please wait a few days before commit, hopefully
>> someone else will review.
>
> Well, this is not rocket science. You cache a function call which
> would be done on every iteration otherwise. This is not feasible with
> other bitmap types (except perhaps 8bit index->8bit index) because
> they use many more colours.
It's possible with RGB(A) because color transformation is formula-based.
>
> How well tested is this? There are quite a few blitters and some may
> not be ever used in current code.
I modified videotest to test every blitting function. This part isn't
included in the patch because it's dirty.
>
> The comment above the blend functions should probably not say they are
> replace blitters.
>
Thanks.
> Thanks
>
> Michal
>
>
> _______________________________________________
> Grub-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/grub-devel
>



-- 
Regards
Vladimir 'phcoder' Serbinenko

Personal git repository: http://repo.or.cz/w/grub2/phcoder.git

1bit.diff
Description: Text document