diff options
| author |  Rasmus Munk Larsen <rmlarsen@google.com> | 2020-11-17 15:28:37 -0800 |
|---|---|---|
| committer | Rasmus Munk Larsen <rmlarsen@google.com> | 2020-11-17 15:28:37 -0800 |
| commit | 6bba58f10915266b56ec212ef406bc83b2efead0 (patch) | |
| tree | 910e844304d40fea4130ce425bc2f7ad42d73a00 | |
| parent | e9b55c4db80f292aa0a6dbe28ac84c0bcf604078 (diff) | |
Replace SSE_SHUFFLE_MASK macro with shuffle_mask.
| -rw-r--r-- | Eigen/src/Core/arch/AVX512/Complex.h | 20 |
1 files changed, 10 insertions, 10 deletions
diff --git a/Eigen/src/Core/arch/AVX512/Complex.h b/Eigen/src/Core/arch/AVX512/Complex.h index 747f7a5d5..25256ae1e 100644 --- a/Eigen/src/Core/arch/AVX512/Complex.h +++ b/Eigen/src/Core/arch/AVX512/Complex.h @@ -323,7 +323,7 @@ template<> EIGEN_STRONG_INLINE std::complex<double> pfirst<Packet4cd>(const Pack } template<> EIGEN_STRONG_INLINE Packet4cd preverse(const Packet4cd& a) { - return Packet4cd(_mm512_shuffle_f64x2(a.v, a.v, EIGEN_SSE_SHUFFLE_MASK(3,2,1,0))); + return Packet4cd(_mm512_shuffle_f64x2(a.v, a.v, shuffle_mask<3,2,1,0>::mask)); } template<> EIGEN_STRONG_INLINE std::complex<double> predux<Packet4cd>(const Packet4cd& a) @@ -426,15 +426,15 @@ ptranspose(PacketBlock<Packet8cf,8>& kernel) { EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock<Packet4cd,4>& kernel) { - __m512d T0 = _mm512_shuffle_f64x2(kernel.packet[0].v, kernel.packet[1].v, EIGEN_SSE_SHUFFLE_MASK(0,1,0,1)); // [a0 a1 b0 b1] - __m512d T1 = _mm512_shuffle_f64x2(kernel.packet[0].v, kernel.packet[1].v, EIGEN_SSE_SHUFFLE_MASK(2,3,2,3)); // [a2 a3 b2 b3] - __m512d T2 = _mm512_shuffle_f64x2(kernel.packet[2].v, kernel.packet[3].v, EIGEN_SSE_SHUFFLE_MASK(0,1,0,1)); // [c0 c1 d0 d1] - __m512d T3 = _mm512_shuffle_f64x2(kernel.packet[2].v, kernel.packet[3].v, EIGEN_SSE_SHUFFLE_MASK(2,3,2,3)); // [c2 c3 d2 d3] - - kernel.packet[3] = Packet4cd(_mm512_shuffle_f64x2(T1, T3, EIGEN_SSE_SHUFFLE_MASK(1,3,1,3))); // [a3 b3 c3 d3] - kernel.packet[2] = Packet4cd(_mm512_shuffle_f64x2(T1, T3, EIGEN_SSE_SHUFFLE_MASK(0,2,0,2))); // [a2 b2 c2 d2] - kernel.packet[1] = Packet4cd(_mm512_shuffle_f64x2(T0, T2, EIGEN_SSE_SHUFFLE_MASK(1,3,1,3))); // [a1 b1 c1 d1] - kernel.packet[0] = Packet4cd(_mm512_shuffle_f64x2(T0, T2, EIGEN_SSE_SHUFFLE_MASK(0,2,0,2))); // [a0 b0 c0 d0] + __m512d T0 = _mm512_shuffle_f64x2(kernel.packet[0].v, kernel.packet[1].v, shuffle_mask<0,1,0,1>::mask); // [a0 a1 b0 b1] + __m512d T1 = _mm512_shuffle_f64x2(kernel.packet[0].v, kernel.packet[1].v, shuffle_mask<2,3,2,3>::mask); // [a2 a3 b2 b3] + __m512d T2 = _mm512_shuffle_f64x2(kernel.packet[2].v, kernel.packet[3].v, shuffle_mask<0,1,0,1>::mask); // [c0 c1 d0 d1] + __m512d T3 = _mm512_shuffle_f64x2(kernel.packet[2].v, kernel.packet[3].v, shuffle_mask<2,3,2,3>::mask); // [c2 c3 d2 d3] + + kernel.packet[3] = Packet4cd(_mm512_shuffle_f64x2(T1, T3, shuffle_mask<1,3,1,3>::mask)); // [a3 b3 c3 d3] + kernel.packet[2] = Packet4cd(_mm512_shuffle_f64x2(T1, T3, shuffle_mask<0,2,0,2>::mask)); // [a2 b2 c2 d2] + kernel.packet[1] = Packet4cd(_mm512_shuffle_f64x2(T0, T2, shuffle_mask<1,3,1,3>::mask)); // [a1 b1 c1 d1] + kernel.packet[0] = Packet4cd(_mm512_shuffle_f64x2(T0, T2, shuffle_mask<0,2,0,2>::mask)); // [a0 b0 c0 d0] } } // end namespace internal |