diff options
| author |  Gael Guennebaud <g.gael@free.fr> | 2016-05-24 21:54:03 +0200 |
|---|---|---|
| committer | Gael Guennebaud <g.gael@free.fr> | 2016-05-24 21:54:03 +0200 |
| commit | e68e165a2331e21004b61de689231a68e4b7e9d1 (patch) | |
| tree | 287afc84abac033eb169f09fe54861f21aeca191 /Eigen | |
| parent | 78390e418919c0b06a4dcf02200c1f03cf33e8c7 (diff) | |
bug #256: enable vectorization with unaligned loads/stores.
This concerns all architectures and all sizes.
This new behavior can be disabled by defining EIGEN_UNALIGNED_VECTORIZE=0
Diffstat (limited to 'Eigen')
| -rw-r--r-- | Eigen/src/Core/AssignEvaluator.h | 10 | ||||
| -rw-r--r-- | Eigen/src/Core/Matrix.h | 2 | ||||
| -rw-r--r-- | Eigen/src/Core/util/Macros.h | 5 |
3 files changed, 12 insertions, 5 deletions
diff --git a/Eigen/src/Core/AssignEvaluator.h b/Eigen/src/Core/AssignEvaluator.h index 3029a7758..766980e28 100644 --- a/Eigen/src/Core/AssignEvaluator.h +++ b/Eigen/src/Core/AssignEvaluator.h @@ -81,10 +81,10 @@ private: MayInnerVectorize = MightVectorize && int(InnerSize)!=Dynamic && int(InnerSize)%int(InnerPacketSize)==0 && int(OuterStride)!=Dynamic && int(OuterStride)%int(InnerPacketSize)==0 - && int(JointAlignment)>=int(InnerRequiredAlignment), + && (EIGEN_UNALIGNED_VECTORIZE || int(JointAlignment)>=int(InnerRequiredAlignment)), MayLinearize = bool(StorageOrdersAgree) && (int(DstFlags) & int(SrcFlags) & LinearAccessBit), MayLinearVectorize = bool(MightVectorize) && MayLinearize && DstHasDirectAccess - && ((int(DstAlignment)>=int(LinearRequiredAlignment)) || MaxSizeAtCompileTime == Dynamic), + && (EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment)) || MaxSizeAtCompileTime == Dynamic), /* If the destination isn't aligned, we have to do runtime checks and we don't unroll, so it's only good for large enough sizes. */ MaySliceVectorize = bool(MightVectorize) && bool(DstHasDirectAccess) @@ -130,8 +130,9 @@ public: : int(NoUnrolling) ) : int(Traversal) == int(LinearVectorizedTraversal) - ? ( bool(MayUnrollCompletely) && (int(DstAlignment)>=int(LinearRequiredAlignment)) ? int(CompleteUnrolling) - : int(NoUnrolling) ) + ? ( bool(MayUnrollCompletely) && ( EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment))) + ? int(CompleteUnrolling) + : int(NoUnrolling) ) : int(Traversal) == int(LinearTraversal) ? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling) : int(NoUnrolling) ) @@ -156,6 +157,7 @@ public: EIGEN_DEBUG_VAR(InnerMaxSize) EIGEN_DEBUG_VAR(LinearPacketSize) EIGEN_DEBUG_VAR(InnerPacketSize) + EIGEN_DEBUG_VAR(ActualPacketSize) EIGEN_DEBUG_VAR(StorageOrdersAgree) EIGEN_DEBUG_VAR(MightVectorize) EIGEN_DEBUG_VAR(MayLinearize) diff --git a/Eigen/src/Core/Matrix.h b/Eigen/src/Core/Matrix.h index ac40c82f0..86fce9571 100644 --- a/Eigen/src/Core/Matrix.h +++ b/Eigen/src/Core/Matrix.h @@ -27,7 +27,7 @@ private: default_alignment = compute_default_alignment<_Scalar,max_size>::value, actual_alignment = ((_Options&DontAlign)==0) ? default_alignment : 0, required_alignment = unpacket_traits<PacketScalar>::alignment, - packet_access_bit = packet_traits<_Scalar>::Vectorizable && (actual_alignment>=required_alignment) ? PacketAccessBit : 0 + packet_access_bit = (packet_traits<_Scalar>::Vectorizable && (EIGEN_UNALIGNED_VECTORIZE || (actual_alignment>=required_alignment))) ? PacketAccessBit : 0 }; public: diff --git a/Eigen/src/Core/util/Macros.h b/Eigen/src/Core/util/Macros.h index 466ca73ac..2d2d30cd2 100644 --- a/Eigen/src/Core/util/Macros.h +++ b/Eigen/src/Core/util/Macros.h @@ -756,6 +756,11 @@ namespace Eigen { #define EIGEN_DEFAULT_ALIGN_BYTES EIGEN_MAX_ALIGN_BYTES #endif + +#ifndef EIGEN_UNALIGNED_VECTORIZE +#define EIGEN_UNALIGNED_VECTORIZE 1 +#endif + //---------------------------------------------------------------------- |