Lines Matching refs:VPERM
504 // We can custom expand all VECTOR_SHUFFLEs to VPERM, others we can handle
1023 case PPCISD::VPERM: return "PPCISD::VPERM";
7433 /// lowered into a vperm.
7544 // VPERM.
7591 vperm.
7602 // Determining when to avoid vperm is tricky. Many things affect the cost
7603 // of vperm, particularly how many times the perm mask needs to be computed.
7606 // mask?) the vperm has a cost of 1. OTOH, hoisting the permute mask out of
7611 // available, if this block is within a loop, we should avoid using vperm
7617 // Lower this to a VPERM(V1, V2, V3) expression, where V3 is a constant
7621 // The SHUFFLE_VECTOR mask is almost exactly what we want for vperm, except
7626 // necessary to produce proper semantics with the big-endian-biased vperm
7646 return DAG.getNode(PPCISD::VPERM, dl, V1.getValueType(),
7649 return DAG.getNode(PPCISD::VPERM, dl, V1.getValueType(),
10879 // Because vperm has a big-endian bias, we must reverse the order
