Currenlty all temporal loads are mapped to LDP or LDR. This patch will map all the non temporal 256-bit loads into LDNP. Future patches should address other non-temporal loads.
Could you add a comment similar to the one for the stores above?
nit: please add a newline before this comment.
Would be good to have a comment here to describe what's lowered here.
leftover commented-out code?
nit: newline here
I think this assertion doesn't really make sense any longer, because we now enable the custom lowering for additional types. This probably needs to became just a check (i.e add an early exit here if the cownciditon doesn't hold). At the moment this assertion triggers in multiple tests, causing them to fail.
I can't think of any reason why this should be needed now and also couldn't find anything when looking at the original review. @zjaffal it would be great if you could post a follow-up patch that removes isKnownEven from the checks for STNP or even better move the checks to a helper function used by both STNP and LDNP handling code.