diff options
| author |  Gael Guennebaud <g.gael@free.fr> | 2014-04-22 16:03:19 +0200 |
|---|---|---|
| committer | Gael Guennebaud <g.gael@free.fr> | 2014-04-22 16:03:19 +0200 |
| commit | 5c5231ab71ba51432a24fdfd7d36a835b44805ea (patch) | |
| tree | 395149bfb0225a33f9db09b9c9766b8641acb404 /Eigen/src/Core/arch/SSE/MathFunctions.h | |
| parent | 2606abed535744fcaa41b923c71338a06b8ed3fa (diff) | |
Workaround gcc's default ABI not being able to distinghish between vector types of different sizes.
Diffstat (limited to 'Eigen/src/Core/arch/SSE/MathFunctions.h')
| -rw-r--r-- | Eigen/src/Core/arch/SSE/MathFunctions.h | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/Eigen/src/Core/arch/SSE/MathFunctions.h b/Eigen/src/Core/arch/SSE/MathFunctions.h index 09f74c651..8f78b3a6c 100644 --- a/Eigen/src/Core/arch/SSE/MathFunctions.h +++ b/Eigen/src/Core/arch/SSE/MathFunctions.h @@ -63,7 +63,7 @@ Packet4f plog<Packet4f>(const Packet4f& _x) x = _mm_or_ps(x, p4f_half); emm0 = _mm_sub_epi32(emm0, p4i_0x7f); - Packet4f e = padd(_mm_cvtepi32_ps(emm0), p4f_1); + Packet4f e = padd(Packet4f(_mm_cvtepi32_ps(emm0)), p4f_1); /* part2: if( x < SQRTHF ) { @@ -72,9 +72,9 @@ Packet4f plog<Packet4f>(const Packet4f& _x) } else { x = x - 1.0; } */ Packet4f mask = _mm_cmplt_ps(x, p4f_cephes_SQRTHF); - Packet4f tmp = _mm_and_ps(x, mask); + Packet4f tmp = pand(x, mask); x = psub(x, p4f_1); - e = psub(e, _mm_and_ps(p4f_1, mask)); + e = psub(e, pand(p4f_1, mask)); x = padd(x, tmp); Packet4f x2 = pmul(x,x); @@ -167,7 +167,7 @@ Packet4f pexp<Packet4f>(const Packet4f& _x) emm0 = _mm_cvttps_epi32(fx); emm0 = _mm_add_epi32(emm0, p4i_0x7f); emm0 = _mm_slli_epi32(emm0, 23); - return pmul(y, _mm_castsi128_ps(emm0)); + return pmul(y, Packet4f(_mm_castsi128_ps(emm0))); } template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet2d pexp<Packet2d>(const Packet2d& _x) @@ -241,7 +241,7 @@ Packet2d pexp<Packet2d>(const Packet2d& _x) emm0 = _mm_add_epi32(emm0, p4i_1023_0); emm0 = _mm_slli_epi32(emm0, 20); emm0 = _mm_shuffle_epi32(emm0, _MM_SHUFFLE(1,2,0,3)); - return pmul(x, _mm_castsi128_pd(emm0)); + return pmul(x, Packet2d(_mm_castsi128_pd(emm0))); } /* evaluation of 4 sines at onces, using SSE2 intrinsics. |