devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

bug #973, improve AVX support by enabling vectorization of Vector4i-like types, and enforcing alignement of Vector4f/Vector2d-like types to preserve compatibility with SSE and future Eigen versions that will vectorize them with AVX enabled.

Browse Source
This commit is contained in:
Gael Guennebaud
2015-03-13 21:15:50 +01:00
parent d99ab35f9e
commit 1330f8bbd1
6 changed files with 108 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
test/unalignedassert.cpp
View File

@@ -81,7 +81,7 @@ void construct_at_boundary(int boundary)
void unalignedassert()
{
#if EIGEN_ALIGN_STATICALLY
#if EIGEN_ALIGN_STATICALLY
construct_at_boundary<Vector2f>(4);
construct_at_boundary<Vector3f>(4);
construct_at_boundary<Vector4f>(16);
@@ -100,7 +100,7 @@ void unalignedassert()
construct_at_boundary<Vector3cf>(4);
construct_at_boundary<Vector2cd>(EIGEN_ALIGN_BYTES);
construct_at_boundary<Vector3cd>(16);
#endif
#endif
check_unalignedassert_good<TestNew1>();
check_unalignedassert_good<TestNew2>();
@@ -112,11 +112,12 @@ void unalignedassert()
check_unalignedassert_good<Depends<true> >();
#if EIGEN_ALIGN_STATICALLY
if(EIGEN_ALIGN_BYTES==16)
if(EIGEN_ALIGN_BYTES>=16)
{
VERIFY_RAISES_ASSERT(construct_at_boundary<Vector4f>(8));
VERIFY_RAISES_ASSERT(construct_at_boundary<Vector2d>(8));
VERIFY_RAISES_ASSERT(construct_at_boundary<Vector2cf>(8));
VERIFY_RAISES_ASSERT(construct_at_boundary<Vector4i>(8));
}
for(int b=8; b<EIGEN_ALIGN_BYTES; b+=8)
{

2
test/vectorization_logic.cpp
View File

@@ -214,7 +214,7 @@ template<typename Scalar, bool Enable = internal::packet_traits<Scalar>::Vectori
>(DefaultTraversal,CompleteUnrolling)));
VERIFY((test_assign(Matrix11(), Matrix<Scalar,PacketSize,EIGEN_PLAIN_ENUM_MIN(2,PacketSize)>()*Matrix<Scalar,EIGEN_PLAIN_ENUM_MIN(2,PacketSize),PacketSize>(),
PacketSize>=EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD?DefaultTraversal:InnerVectorizedTraversal, CompleteUnrolling)));
InnerVectorizedTraversal, CompleteUnrolling)));
#endif
VERIFY(test_assign(MatrixXX(10,10),MatrixXX(20,20).block(10,10,2,3),