From becd89df294a91ee3e6fefe7f5fc4e0c83dd7895 Mon Sep 17 00:00:00 2001
From: Gael Guennebaud <g.gael@free.fr>
Date: Thu, 6 Aug 2015 14:00:26 +0200
Subject: [PATCH] Enable runtime stack alignment in gemm_blocking_space.

---
 Eigen/src/Core/products/GeneralBlockPanelKernel.h |  2 +-
 Eigen/src/Core/products/GeneralMatrixMatrix.h     | 10 ++++++++++
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
index 1d62ccd93..b7c236bb4 100644
--- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h
+++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h
@@ -2066,7 +2066,7 @@ EIGEN_DONT_INLINE void gemm_pack_rhs<Scalar, Index, DataMapper, nr, ColMajor, Co
       const LinearMapper dm3 = rhs.getLinearMapper(0, j2 + 3);
 
       Index k=0;
-      if((PacketSize%4)==0) // TODO enbale vectorized transposition for PacketSize==2 ??
+      if((PacketSize%4)==0) // TODO enable vectorized transposition for PacketSize==2 ??
       {
         for(; k<peeled_k; k+=PacketSize) {
           PacketBlock<Packet,(PacketSize%4)==0?4:PacketSize> kernel;
diff --git a/Eigen/src/Core/products/GeneralMatrixMatrix.h b/Eigen/src/Core/products/GeneralMatrixMatrix.h
index be89181ee..3fba82ff8 100644
--- a/Eigen/src/Core/products/GeneralMatrixMatrix.h
+++ b/Eigen/src/Core/products/GeneralMatrixMatrix.h
@@ -293,8 +293,13 @@ class gemm_blocking_space<StorageOrder,_LhsScalar,_RhsScalar,MaxRows, MaxCols, M
       SizeB = ActualCols * MaxDepth
     };
 
+#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES
     EIGEN_ALIGN_MAX LhsScalar m_staticA[SizeA];
     EIGEN_ALIGN_MAX RhsScalar m_staticB[SizeB];
+#else
+    EIGEN_ALIGN_MAX char m_staticA[SizeA * sizeof(LhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1];
+    EIGEN_ALIGN_MAX char m_staticB[SizeB * sizeof(RhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1];
+#endif
 
   public:
 
@@ -303,8 +308,13 @@ class gemm_blocking_space<StorageOrder,_LhsScalar,_RhsScalar,MaxRows, MaxCols, M
       this->m_mc = ActualRows;
       this->m_nc = ActualCols;
       this->m_kc = MaxDepth;
+#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES
       this->m_blockA = m_staticA;
       this->m_blockB = m_staticB;
+#else
+      this->m_blockA = reinterpret_cast<LhsScalar*>((std::size_t(m_staticA) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1));
+      this->m_blockB = reinterpret_cast<RhsScalar*>((std::size_t(m_staticB) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1));
+#endif
     }
     
     void initParallel(Index, Index, Index, Index)