Yeppp.log([NaN]) = 710.1888

[floswald]

I found this to be a major problem in my application. anything one could do here? (I would be happy with anything but returning a number)

julia> Yeppp.log([NaN])
1-element Array{Float64,1}:
 710.188

[panlanfeng]

Unable to reproduce in my computer. What is your OS and Julia version?

[floswald]

That's Julia 0.5 on Mac OS El Capitan.

On Sat, 3 Dec 2016 at 04:20, Lanfeng Pan [email protected] wrote:

Unable to reproduce in my computer. What is your OS and Julia version?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/JuliaMath/Yeppp.jl/issues/31#issuecomment-264612678, or mute the thread https://github.com/notifications/unsubscribe-auth/AA-WdpcPw7T1GJh8s8cDESmQPRbNarEsks5rEN_2gaJpZM4LC_cZ .

[KristofferC]

On Ubuntu with 0.5 this gives NaN for me.

[floswald]

what can I say? I get 710. :-/

julia> using Yeppp

julia> Yeppp.log([NaN])
1-element Array{Float64,1}:
 710.188

julia> versioninfo()
Julia Version 0.5.0
Commit 3c9d753 (2016-09-19 18:14 UTC)
Platform Info:
  System: Darwin (x86_64-apple-darwin13.4.0)
  CPU: Intel(R) Core(TM) i5-5257U CPU @ 2.70GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.7.1 (ORCJIT, broadwell)

julia> Yeppp.
@checked_lib   cos             exp!            min!            subtract!
@yepppfunsAA_A cos!            libyeppp        multiply        sum
@yepppfunsA_A  dot             log             multiply!       sumabs
@yepppfunsA_S  eval            log!            release         sumabs2
__init__       evalpoly        max             sin             tan
add            evalpoly!       max!            sin!            tan!
add!           exp             min             subtract
julia> Yeppp.libyeppp
"/Users/florian.oswald/.julia/v0.5/Yeppp/deps/yeppp-1.0.0/binaries/macosx/x86_64/libyeppp.dylib"

[KristofferC]

I'm not doubting you, I was just giving another data point :)

[floswald]

Of course, I know. :-)

[ViralBShah]

cc @Maratyszcza

[Maratyszcza]

It might be a combination of three issues:

• Broadwell model is too new to be detected by Yeppp! 1.0.0 (which was released 3 years ago)
• Yeppp! chooses the non-optimized implementation (which is a bug)
• The non-optimized implementation has a bug in NaN handling

[floswald]

ok. is there any more recent release of Yeppp! ? let me know what to do to help trace this down.

[floswald]

still so on julia v1.1 FWIW.

[chriscoey]

The meaning of life is no longer 42. It is 710.

[ViralBShah]

This isn't a Julia issue. Just saying so for the benefit of any passers by.