GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-06-04 18:54:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

AltiVec signed integer pmadd removed, proved to be 2x slower than the scalar trait(!).

Browse Source
This commit is contained in:
Konstantinos Margaritis 2010-06-28 21:24:55 +03:00
parent 97889a7f46
commit cf3616b2c0
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
Eigen/src/Core/arch/AltiVec/PacketMath.h
View File

@ -174,6 +174,8 @@ template<> EIGEN_STRONG_INLINE Packet4f ei_pnegate(const Packet4f& a) { return e
template<> EIGEN_STRONG_INLINE Packet4i ei_pnegate(const Packet4i& a) { return ei_psub<Packet4i>(ei_p4i_ZERO, a); } template<> EIGEN_STRONG_INLINE Packet4i ei_pnegate(const Packet4i& a) { return ei_psub<Packet4i>(ei_p4i_ZERO, a); }
template<> EIGEN_STRONG_INLINE Packet4f ei_pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return vec_madd(a,b,ei_p4f_ZERO); } template<> EIGEN_STRONG_INLINE Packet4f ei_pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return vec_madd(a,b,ei_p4f_ZERO); }
/* Commented out: it's actually slower than processing it scalar
*
template<> EIGEN_STRONG_INLINE Packet4i ei_pmul<Packet4i>(const Packet4i& a, const Packet4i& b) template<> EIGEN_STRONG_INLINE Packet4i ei_pmul<Packet4i>(const Packet4i& a, const Packet4i& b)
{ {
// Detailed in: http://freevec.org/content/32bit_signed_integer_multiplication_altivec // Detailed in: http://freevec.org/content/32bit_signed_integer_multiplication_altivec
@ -207,6 +209,7 @@ template<> EIGEN_STRONG_INLINE Packet4i ei_pmul<Packet4i>(const Packet4i& a, con
return prod; return prod;
} }
*/
template<> EIGEN_STRONG_INLINE Packet4f ei_pdiv<Packet4f>(const Packet4f& a, const Packet4f& b) template<> EIGEN_STRONG_INLINE Packet4f ei_pdiv<Packet4f>(const Packet4f& a, const Packet4f& b)
{ {
Packet4f t, y_0, y_1, res; Packet4f t, y_0, y_1, res;