diff options
| author |  Vamsi Sripathi <vamsi.sripathi@intel.com> | 2017-03-01 16:05:47 -0800 |
|---|---|---|
| committer |  Benoit Steiner <benoitsteiner@users.noreply.github.com> | 2017-03-01 16:05:47 -0800 |
| commit | 51331365b60dd042ebc0849ea4e1ab6a396b60fd (patch) | |
| tree | 3981cf3f8985b15dc8d88e27e56b318f88946e5f /third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint | |
| parent | b10f50ff15944badb7262a207f6628dfa52d6a9d (diff) | |
Fix TensorFlow compilation errors with KNL optimization flags (-mavx512f) (#7963)
* Fix compilation errors caused by intrinsic function and argument mis-match for AVX512 when -mavx512f is used
* More fixes for AVX512 compilation.
* Fix _mm512_ternarylogic_epi32() function to mimic masked blend functionality
Diffstat (limited to 'third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint')
| -rw-r--r-- | third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint b/third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint index b0a73aac79..eb604d38b1 100644 --- a/third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint +++ b/third_party/eigen3/unsupported/Eigen/CXX11/FixedPoint @@ -31,7 +31,7 @@ #include "src/FixedPoint/FixedPointTypes.h" // Use optimized implementations whenever available -#ifdef EIGEN_VECTORIZE_AVX512 +#if defined (EIGEN_VECTORIZE_AVX512DQ) || defined (EIGEN_VECTORIZE_AVX512BW) #include "src/FixedPoint/PacketMathAVX512.h" #include "src/FixedPoint/TypeCastingAVX512.h" |