18:16fdobridge: <gfxstrand> Ugh... Looks like shared atomics aren't executing per-channel
18:16fdobridge: <gfxstrand> That or I'm only getting on invocation for some reason
18:18fdobridge: <gfxstrand> That or I'm only getting one invocation for some reason (edited)
19:12fdobridge: <gfxstrand> *sigh*... Needed a WARPSYNC
19:12fdobridge: <gfxstrand> shared memory should be working now
19:14fdobridge: <pixelcluster> warpsync? aren't invocations in one warp executing in lockstep so should be implicitly synced? or is it more like "sync all warps in this workgroup"?
19:16fdobridge: <gfxstrand> I need to warpsync before syncing all warps
19:16fdobridge: <gfxstrand> That way all the invocations participate in the sync
19:17fdobridge: <pixelcluster> huh alright