some rounding modes missing

[PeterJacko]

https://docs.julialang.org/en/v1/base/math/#Base.Rounding.RoundingMode gives a list of rounding modes including:
RoundNearestTiesAway
RoundNearestTiesUp

but it seems that rounding.jl doesn't implement these two.

julia> Float16( 1.23456789 , RoundNearestTiesUp )
ERROR: MethodError: no method matching _convert_rounding(::Type{Float16}, ::Float64, ::RoundingMode{
:NearestTiesUp})

julia> Float16( 1.23456789 , RoundNearestTiesAway )
ERROR: MethodError: no method matching _convert_rounding(::Type{Float16}, ::Float64, ::RoundingMode{
:NearestTiesAway})

I wanted to try these ones as I have noticed in my calculations that the sum of many positive Float16 numbers is always lower than the exact value, so suspecting that the rounding at ties could play a role.

Also, the documentation doesn't mention the method Float16 (unlike Float32 here https://docs.julialang.org/en/v1/base/numbers/#Core.Float32-Tuple{Any}).

julia> versioninfo()
Julia Version 1.0.1
Commit 0d713926f8 (2018-09-29 19:05 UTC)
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.0 (ORCJIT, skylake)
Environment:
  JULIA_NUM_THREADS = 2
  JULIA_PKG_SERVER = https://pkg.juliacomputing.com/
  JULIA_PKG_TOKEN_PATH = C:\Users\Peter\.julia\token.toml

[kshyatt]

rounding.jl does implement these two in round(x::AbstractFloat, ::RoundingMode{NearestTiesUp}) etc. but maybe not yet for Float16()? This would be a good PR if you felt up to it.

[PeterJacko]

rounding.jl does implement these two in round(x::AbstractFloat, ::RoundingMode{NearestTiesUp}) etc. but maybe not yet for Float16()? This would be a good PR if you felt up to it.

You are right that it works with round:

julia> round( 1.23456789 , RoundNearestTiesUp )
1.0

julia> round( 1.23456789 , RoundNearestTiesUp , digits = 5)
1.23457

but I can't find where rounding.jl implements these two. See https://github.com/JuliaLang/julia/blob/master/base/rounding.jl : they appear in the documentation there, but not in functions to_fenv and from_fenv - not sure what these do.

The issue is not only with Float16:

julia> Float32( 1.23456789 , RoundNearestTiesUp )
ERROR: MethodError: no method matching _convert_rounding(::Type{Float32}, ::Float64, ::RoundingMode{:NearestTiesUp})

julia> Float64( 1.23456789 , RoundNearestTiesUp )
ERROR: MethodError: no method matching _convert_rounding(::Type{Float64}, ::Float64, ::RoundingMode{:NearestTiesUp})

I am affraid I am not as skilled with julia development, I don't even know what the PR stands for!

[kshyatt]

PR = pull request!

Check out https://github.com/JuliaLang/julia/blob/master/base/floatfuncs.jl#L211

[PeterJacko]

Also, help on round gives as one of the many methods

round([T,] x, [r::RoundingMode])

but

julia> round( Float64 , 1.23456789 , RoundNearestTiesUp )
ERROR: MethodError: no method matching round(::Type{Float64}, ::Float64, ::RoundingMode{:NearestTiesUp})