diff options
| author |  Gael Guennebaud <g.gael@free.fr> | 2018-12-06 15:58:06 +0100 |
|---|---|---|
| committer | Gael Guennebaud <g.gael@free.fr> | 2018-12-06 15:58:06 +0100 |
| commit | c53eececb0415834b961cb61cd466907261b4b2f (patch) | |
| tree | 2fda50e7549fda68501e3a33634dfc2ccfac743a /Eigen/src/Core/arch/SSE | |
| parent | 3fba59ea594eb26446352cb28813b38921439f23 (diff) | |
Implement AVX512 vectorization of std::complex<float/double>
Diffstat (limited to 'Eigen/src/Core/arch/SSE')
| -rwxr-xr-x | Eigen/src/Core/arch/SSE/PacketMath.h | 12 |
1 files changed, 7 insertions, 5 deletions
diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h index 0e8e0d2b3..b073fc8d4 100755 --- a/Eigen/src/Core/arch/SSE/PacketMath.h +++ b/Eigen/src/Core/arch/SSE/PacketMath.h @@ -61,20 +61,22 @@ template<> struct is_arithmetic<__m128> { enum { value = true }; }; template<> struct is_arithmetic<__m128i> { enum { value = true }; }; template<> struct is_arithmetic<__m128d> { enum { value = true }; }; +#define EIGEN_SSE_SHUFFLE_MASK(p,q,r,s) ((s)<<6|(r)<<4|(q)<<2|(p)) + #define vec4f_swizzle1(v,p,q,r,s) \ - (_mm_castsi128_ps(_mm_shuffle_epi32( _mm_castps_si128(v), ((s)<<6|(r)<<4|(q)<<2|(p))))) + (_mm_castsi128_ps(_mm_shuffle_epi32( _mm_castps_si128(v), EIGEN_SSE_SHUFFLE_MASK(p,q,r,s)))) #define vec4i_swizzle1(v,p,q,r,s) \ - (_mm_shuffle_epi32( v, ((s)<<6|(r)<<4|(q)<<2|(p)))) + (_mm_shuffle_epi32( v, EIGEN_SSE_SHUFFLE_MASK(p,q,r,s))) #define vec2d_swizzle1(v,p,q) \ - (_mm_castsi128_pd(_mm_shuffle_epi32( _mm_castpd_si128(v), ((q*2+1)<<6|(q*2)<<4|(p*2+1)<<2|(p*2))))) + (_mm_castsi128_pd(_mm_shuffle_epi32( _mm_castpd_si128(v), EIGEN_SSE_SHUFFLE_MASK(2*p,2*p+1,2*q,2*q+1)))) #define vec4f_swizzle2(a,b,p,q,r,s) \ - (_mm_shuffle_ps( (a), (b), ((s)<<6|(r)<<4|(q)<<2|(p)))) + (_mm_shuffle_ps( (a), (b), EIGEN_SSE_SHUFFLE_MASK(p,q,r,s))) #define vec4i_swizzle2(a,b,p,q,r,s) \ - (_mm_castps_si128( (_mm_shuffle_ps( _mm_castsi128_ps(a), _mm_castsi128_ps(b), ((s)<<6|(r)<<4|(q)<<2|(p)))))) + (_mm_castps_si128( (_mm_shuffle_ps( _mm_castsi128_ps(a), _mm_castsi128_ps(b), EIGEN_SSE_SHUFFLE_MASK(p,q,r,s))))) #define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \ const Packet4f p4f_##NAME = pset1<Packet4f>(X) |