juliamath / accuratearithmetic.jl Goto Github PK

I know nothing about Jekyll, less about its use with this repository.
Guidance requested ( or just go ahead and remedy the issue :) ).
[ note that the repository, master branch, has updated properly otherwise ]

message from github follows

JuliaMath/AccurateArithmetic.jl Page build failure

The page build failed for the master branch with the following error:
Unable to build page. Please try again later.

For information on troubleshooting Jekyll see:
https://docs.github.com/articles/troubleshooting-jekyll-builds
If you have any questions you can submit a request at https://support.github.com/contact?repo_id=119051030&page_build_id=253607617

Now that SIMDPirates.jl has been deprecated in favor of VectorizationBase.jl, we should start transitioning from the former to the latter.

CC: chriselrod/SIMDPirates.jl#15

The tag name "v.0.1.3" is not of the appropriate SemVer form (vX.Y.Z).
cc: @JeffreySarnoff

Thanks for the package! A new release would be much appreciated 👍

Many thanks in advance!

@JeffreySarnoff in one rust package https://docs.rs/safeeft/0.0.5/safeeft/ I found new safe EFT.
They said that M.Kashiwagi in his paper (in japanese) found special cases where twoSum and twoProducts not safe enough with IEEE 747-2008 floats. What do you think about that?

twosum ( 3 . 5 6 3 0 6 2 4 4 4 4 8 7 4 5 3 9 e +307 ,
−1 .7976931348623157 e +308)
( −1 .4413868904135704 e +308 , nan )
twoproduct ( 6 . 9 2 9 0 0 1 7 1 3 8 6 9 9 3 6 e +236 ,
2 . 5 9 4 4 4 7 5 2 5 1 9 5 2 0 0 3 e+71)
( 1 . 7 9 7 6 9 3 1 3 4 8 6 2 3 1 5 7 e +308 , i n f )

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

When I try to load this package I get: ERROR: LoadError: LoadError: UndefVarError: vload not defined

vload was removed from SIMDPirates.jl here and a new minor version that included this change was recently released: https://github.com/chriselrod/SIMDPirates.jl/releases/tag/v0.4.0

The compat section in Project.toml should probably be changed to include minor versions for pre-1.0 packages.

The naive dot product implementation currently uses "explicit" SIMD instructions:

This should prevent LLVM from transforming those into vfmadd instructions, which is a shame since those would probably be both more accurate and more efficient. We should

• either change those to non-explicit ("fuseable") SIMD instructions, relying on LLVM to turn those into vfmadds, or
• explicitly use an fma here.

I wonder if this is what explains the differences between OpenBLAS.ddot and AccurateArithmetic.dot_naive observed in the paper (fig. 4)

Somehow, I mis-derived four_sum.

oof, not quite right yet

it does nothing, a result of my mistyping

Probably related to #23:

julia> versioninfo()
Julia Version 1.6.0
Commit f9720dc2eb (2021-03-24 12:55 UTC)
Platform Info:
  OS: Linux (x86_64-redhat-linux)
  CPU: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.1 (ORCJIT, skylake)

(@v1.6) pkg> add AccurateArithmetic @0.3.5
   Resolving package versions...
ERROR: Unsatisfiable requirements detected for package SIMDPirates [21efa798]:
 SIMDPirates [21efa798] log:
 ├─possible versions are: 0.1.0-0.8.26 or uninstalled
 ├─restricted by julia compatibility requirements to versions: uninstalled
 └─restricted by compatibility requirements with AccurateArithmetic [22286c92] to versions: [0.2.0-0.3.16, 0.5.0, 0.6.3-0.8.26] — no versions left
   └─AccurateArithmetic [22286c92] log:
     ├─possible versions are: 0.2.1-0.3.5 or uninstalled
     └─restricted to versions 0.3.5 by an explicit requirement, leaving only versions 0.3.5

(@v1.6) pkg> add AccurateArithmetic @0.3.4
    Updating registry at `~/.julia/registries/General`
   Resolving package versions...
ERROR: Unsatisfiable requirements detected for package SIMDPirates [21efa798]:
 SIMDPirates [21efa798] log:
 ├─possible versions are: 0.1.0-0.8.26 or uninstalled
 ├─restricted by julia compatibility requirements to versions: uninstalled
 └─restricted by compatibility requirements with AccurateArithmetic [22286c92] to versions: [0.2.0-0.3.16, 0.5.0, 0.6.3-0.8.26] — no versions left
   └─AccurateArithmetic [22286c92] log:
     ├─possible versions are: 0.2.1-0.3.5 or uninstalled
     └─restricted to versions 0.3.4 by an explicit requirement, leaving only versions 0.3.4

...

(@v1.6) pkg> add AccurateArithmetic
   Resolving package versions...
    Updating `~/.julia/environments/v1.6/Project.toml`
  [22286c92] + AccurateArithmetic v0.3.0
    Updating `~/.julia/environments/v1.6/Manifest.toml`
  [22286c92] + AccurateArithmetic v0.3.0

cf. JuliaSIMD/VectorizationBase.jl#106 (comment). It looks like VectorizationBase is a dependency here, and from a cursory glance of the codebase it seems pretty ingrained into it. How much effort would it be to switch out, given that VectorizationBase has demonstrated segfaults on Julia v1.11?

We might end up using AccurateArithmetic in https://github.com/JuliaGeo/GeometryOps.jl, and would like to avoid as many dependency issues as possible 😄.

Ref: https://discourse.julialang.org/t/package-compatibility-caps/15301

Includes a revised sqrt.
Accurate Calculation of Euclidean Norms using Double Word Arithmetic (Nov 2022)