search

intel / ARM_NEON_2_x86_SSE

The platform independent header allowing to compile any C/C++ code containing ARM NEON intrinsic functions for x86 target systems using SIMD up to AVX2 intrinsic functions
Other
430 stars 149 forks source link

how to convert vfmaq_f32 to SSE? #8

Closed adeagle closed 7 years ago

Zvictoria commented 7 years ago

Multiply accumulate functions are not a part of a standard arm_neon header, The are not a part of SSE instructions set either but implemented in FMA instruction set. I might add them in this porting tool as well if there is a need. Could you please explain what project do you need it for? Or you could do it yourself, it should be easy.

adeagle commented 7 years ago

I have change vfmaq_laneq_f32 to vmlaq_lane_f32+vget_low_f32; after convert,the speed is faster 2.5X than C code ,is that normal?

Zvictoria commented 7 years ago

yes, this level of speedup is expected in such kind of porting.

adeagle commented 7 years ago

will you support AVX?

Zvictoria commented 7 years ago

The short answer is "May be if there is a real need for that". The goal of this project has been to simplify porting from ARM based mobile devices to Intel Atom based mobile devices. And Atoms don't have AVX. If there appear some important projects for ARM -> x86 desktop porting we may add AVX there

caand commented 7 years ago

But AFAIK there is nothing equivalent to AVX on ARM: ARM only has 128bit SIMD.

As Victoria points out, this project targets mobile devices and does a very good job at it. If you need AVX, you are on another platform and you could (should) use directly AVX intrinsics.

Calin

From: Victoria [mailto:notifications@github.com] Sent: Friday, 08 September, 2017 14:25 To: intel/ARM_NEON_2_x86_SSE Cc: Subscribed Subject: Re: [intel/ARM_NEON_2_x86_SSE] how to convert vfmaq_f32 to SSE? (#8)

The short answer is "May be if there is a real need for that". The goal of this project has been to simplify porting from ARM based mobile devices to Intel Atom based mobile devices. And Atoms don't have AVX. If there appear some important projects for ARM -> x86 desktop porting we may add AVX there

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/intel/ARM_NEON_2_x86_SSE/issues/8#issuecomment-328089353 , or mute the thread https://github.com/notifications/unsubscribe-auth/ALa5uUgLRIswIgckUw_RgCJg1-6vyYACks5sgTIvgaJpZM4PLa45 .Image removed by sender.

adeagle commented 7 years ago

ok.thanks