add to_fixed() for F -> fixed rounding

I'm trying not to focus too much on performance yet, but I did happen to
look at some of the code generation and this does improve these stores
pretty dramatically, keeping them vectorized instead of breaking down
into scalar operations.

I think always converting float->int is the key,
and then you can convert int->whatever.

Change-Id: I4b986806a863b9f8f039f72fab40acf483196a0a
Reviewed-on: https://skia-review.googlesource.com/96800
Reviewed-by: Brian Osman <brianosman@google.com>
1 file changed