On Thu, Jan 23, 2014 at 11:52 AM, Tom Gundersen <[email protected]> wrote: > On Thu, Jan 23, 2014 at 10:42 AM, Joonas Lahtinen > <[email protected]> wrote: >> On 17.01.2014 14:22, Tom Gundersen wrote: >>> We could then special case the particular bitmasks that >>> correspond to BGRX and skip the loop. Does that sound like something >>> you could work with? >> >> This sounds like a feature worth implementing to me. >> >> It'd require passing some option/information to instruct to skip blending >> even though the image is BGRA and just copy the alpha channel as is (it just >> gets discarded by UEFI BGRX format). Opinions on how to implement that? > > Maybe just set the alpha channel bitmask to 0?
With the latest numbers where the blending version only adds 10ms, I would like to see new data with an optimized compilation and a non-alpha blending case, before we introduce any further special-case magic. I'm not really convinced at that moment that all of this special casing is not a premature optimization. Kay _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
