diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
index 5b0473598..e43529cc7 100644
--- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
@@ -1625,9 +1625,14 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
           prefetch(&blA[0]);
           const RhsScalar* blB = &blockB[j2*strideB+offsetB*nr];
 
-          if( (SwappedTraits::LhsProgress % 4)==0 )
+          // The following piece of code wont work for 512 bit registers
+          // Moreover it assumes that there is a half packet of the same size
+          // as nr (which is currently 4) for the return type.
+          typedef typename unpacket_traits<SResPacket>::half SResPacketHalf;
+          if ((SwappedTraits::LhsProgress % 4) == 0 &&
+              (SwappedTraits::LhsProgress <= 8) &&
+              unpacket_traits<SResPacketHalf>::size==4)
           {
-            // NOTE The following piece of code wont work for 512 bit registers
             SAccPacket C0, C1, C2, C3;
             straits.initAcc(C0);
             straits.initAcc(C1);