08:38 dolphin: ah, glad we've chosen "-d" as the dry run option in dim, "dim push -d" will gladly delete the branch in question
10:35 dolphin: airlied, sima: No drm-intel-next-fixes this week either
12:57 tzimmermann: airlied, sima, drm-misc-next-fixes is empty this week. hence no PR
22:54 karolherbst: so I found the "(('imul@64', a, '#b(is_bitcount2)'), ('iadd', ('ishl', a, ('ufind_msb', b)), ('ishl', a, ('find_lsb', b)))," rule today and I'm wondering if there is hardware where this is really helping? It sure doesn't help on NVK...
22:57 karolherbst: I know which shaders are hitting this patterns, so in case anybody wants to double check, I can share a list or so... I kinda plan to just remove this pattern for drivers that have has_imad32 set I think...
22:58 karolherbst: or has_imad32 and nir_lower_imul_2x32_64 not set...
22:58 karolherbst: the imul lowers to two instructions here
22:59 karolherbst: (high bits of the constant can be zero, so the imul lowering is dirt cheap if you got the imul_2x32)