Re: [02/10] gdiplus: If source and destination rectangles differ it's not a reason to fallback to software blitting.

[Vincent Povirk]

> Both StretchBlt and GdiAlphaBlend handle that case just fine.

Only if the graphics interpolation mode happens to be nearest
neighbor, which isn't the default.

[02/10] gdiplus: If source and destination rectangles differ it's not a reason to fallback to software blitting.

[Dmitry Timoshkov]

Vincent Povirk  wrote:
> > Both StretchBlt and GdiAlphaBlend handle that case just fine.
>
> Only if the graphics interpolation mode happens to be nearest
> neighbor, which isn't the default.

Feel free to add a check for it, there is no need to penalize a common
case without a reason.

[02/10] gdiplus: If source and destination rectangles differ it's not a reason to fallback to software blitting.

[Vincent Povirk]

> > > Both StretchBlt and GdiAlphaBlend handle that case just fine.
> >
> > Only if the graphics interpolation mode happens to be nearest
> > neighbor, which isn't the default.
>
> Feel free to add a check for it, there is no need to penalize a common
> case without a reason.

You also removed the test for rotation and skewing, which are
definitely NOT handled by gdi32.

[02/10] gdiplus: If source and destination rectangles differ it's not a reason to fallback to software blitting.

[Dmitry Timoshkov]

Vincent Povirk  wrote:
> You also removed the test for rotation and skewing, which are
> definitely NOT handled by gdi32.

I guess that this patch could be dropped then, other patches in the series
shouldn't be affected.