Enable vectorization of pack_rhs with a column-major RHS.

Rename and generalize Kernel<*> to PacketBlock<*,N>.
This commit is contained in:
Gael Guennebaud
2014-04-25 10:56:18 +02:00
parent b0e19db1cf
commit 3d8d0f6269
11 changed files with 64 additions and 43 deletions

View File

@@ -539,7 +539,7 @@ struct palign_impl<Offset,Packet4i>
};
template<> EIGEN_DEVICE_FUNC inline void
ptranspose(Kernel<Packet4f>& kernel) {
ptranspose(PacketBlock<Packet4f,4>& kernel) {
Packet4f t0, t1, t2, t3;
t0 = vec_mergeh(kernel.packet[0], kernel.packet[2]);
t1 = vec_mergel(kernel.packet[0], kernel.packet[2]);
@@ -552,7 +552,7 @@ ptranspose(Kernel<Packet4f>& kernel) {
}
template<> EIGEN_DEVICE_FUNC inline void
ptranspose(Kernel<Packet4i>& kernel) {
ptranspose(PacketBlock<Packet4i,4>& kernel) {
Packet4i t0, t1, t2, t3;
t0 = vec_mergeh(kernel.packet[0], kernel.packet[2]);
t1 = vec_mergel(kernel.packet[0], kernel.packet[2]);