20:30sarbes: I've rewritten my scalar constant multiplier MR to go the NIR route, but I got worse numbers on Shader DB. It looks like the ffma op prevents some optimization here.
20:34anarsoul: sounds like your first approach to lower it in ppir is better :)
20:36anarsoul: But ffma support in ppir is a hack. It's just a consequence of lacking corresponding optimization pass in ppir
20:37sarbes: Yeah... I'll try to hack around the hack :)
20:38sarbes: I'll break down the ffma if I can sneak a multiplier in.
20:39anarsoul: or just handle constant multipliers in ppir?
20:40anarsoul: ffma allows to save a register when passing mul result to add
20:40anarsoul: so breaking down ffma may result in increased register pressure
20:44sarbes: ...or I'll add an ffma opcode which can accept two scalars. But this sounds even more hacky.
20:44sarbes: But it is doable.
21:20sarbes: On further thought, no.
21:23sarbes: But it shouldn't be too hard to rearrange ffmas by hand, I think.
21:59anarsoul: sarbes: there is nothing wrong in folding constant multipliers in ppir
21:59anarsoul: it is utgard-specific feature, so I don't think that it will be useful for nir in general