On 4/22/22 05:03, Giovanni Mascellani wrote:
Hi,
Il 21/04/22 20:57, Zebediah Figura ha scritto:
I dunno, I see the logic as not particularly complex...
Ok, given that Francisco reworked his patch in a rather different way, I did the same for mine. I am not really sure it's the ideal solution, but let's discuss it.
I'm also slightly concerned that we're generating terrible code too zealously. Simplicity of implementation is valuable, but it means that optimization passes have more busywork to do, and this is one of those cases where I'm not sure it's actually worth it...
The optimization step required this specific case is rather simple and already implemented. However, the new patch should please you better WRT point of view.
Eh, I mean extra CPU work for the pass itself, but even now it's probably not worth worrying about.