From b776458ccbd0b8dbde56d0d2dd0a683c6b4b0692 Mon Sep 17 00:00:00 2001
From: Benoit Steiner <benoit.steiner.goog@gmail.com>
Date: Thu, 27 Mar 2014 10:02:24 -0700
Subject: [PATCH] Vectorized the packing of a row-major matrix used as the left
 hand side argument in a matrix-matrix product.

---
 .../Core/products/GeneralBlockPanelKernel.h   | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
index d17752489..eeeb5290f 100644
--- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
@@ -954,9 +954,22 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, Pack1, Pack2, StorageOrder,
     }
     else
     {
-      for(Index k=0; k<depth; k++)
-      {
-        // TODO add a vectorized transpose here
+      const Index peeled_k = (depth/PacketSize)*PacketSize;
+      Index k=0;
+      for(; k<peeled_k; k+=PacketSize) {
+        for (Index m = 0; m < Pack1; m += PacketSize) {
+          Kernel<Packet> kernel;
+          for (int p = 0; p < PacketSize; ++p) {
+            kernel.packet[p] = ploadu<Packet>(&lhs(i+p+m, k));
+          }
+          ptranspose(kernel);
+          for (int p = 0; p < PacketSize; ++p) {
+            pstore(blockA+count+m+Pack1*p, cj.pconj(kernel.packet[p]));
+          }
+        }
+        count += PacketSize*Pack1;
+      }
+      for(; k<depth; k++) {
         Index w=0;
         for(; w<Pack1-3; w+=4)
         {