• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    迪恩网络公众号
登陆 注册
描述文字

REX-Computing/unumjl: Julia Implementation of Unums

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称：

REX-Computing/unumjl

开源软件地址：

https://github.com/REX-Computing/unumjl

开源编程语言：

Julia 89.2%

开源软件介绍：

How to use the Julia Unum library (0.2):

  1. install julia.
  1. run julia.
  2. at the julia command line, run: Pkg.clone("git://github.com/REX-computing/unumjl.git", "Unums")
  3. at the prompt, type:
using Unums

You may now use unums. To convert floating points or integers to unums, use the convert function. Note that the conversion currently assumes that floating points represent exact values, even if the human-input value cannot be exact.

julia> Unum{3,4}(4.3)
Unum{3,4}(0x0000000000000003, 0x1333000000000000, 0x0001, 0x0001, 0x000F)
julia> Unum{4,6}(4.3)
Unum{4,6}(0x0000000000000401, 0x1333333333333000, 0x0000, 0x000A, 0x0033)

To retrieve the unum value, as a human-readable form, I recommend the describe() function which converts Unums and Ubounds to human-readable format. In the future, this input format will be natively parseable in julia to generate the appropriate unum.

julia> describe(Unum{3,4}(0x0000000000000003, 0x1333000000000000, 0x0001, 0x0001, 0x000F))
Unum{3,4}(4.29998779296875 op → 4.300048828125 op)

You can alternatively type in the specification directly into the Unum constructor.
The parts of a unum the unum constructor are as follows:

  1. exponent (64-bit unsigned float)
  2. fraction (64-bit unsigned float or array of 64-bit unsigned floats)
  3. flags (16-bit unsigned float, 2’s bit: sign, 1’s bit: ubit)
  4. esize (16-bit unsigned float)
  5. fsize (16-bit unsigned float)

Julia generally has support for special functions which construct important constants, and the Unum library implements these.

julia> describe(zero(Unum{3,4}))
Unum{3,4}(0.0 ex)
julia> describe(one(Unum{3,4}))
Unum{3,4}(1.0 ex)
julia> describe(inf(Unum{3,4}))
Unum{3,4}(∞ ex)
julia> describe(nan(Unum{3,4}))
Unum{3,4}(NaN)

The standard mathematical operators are overloaded to allow easy calculation with Unums.

julia> x = Unum{3,4}(4.3)
Unum{3,4}(0x0000000000000003, 0x1333000000000000, 0x0001, 0x0001, 0x000F)
julia> y = one(Unum{3,4})
Unum{3,4}(0x0000000000000001, 0x0000000000000000, 0x0000, 0x0001, 0x0000)
julia> x + y
Unum{3,4}(0x0000000000000003, 0x5333000000000000, 0x0001, 0x0001, 0x000F)
julia> describe(x + y)
Unum{3,4}(5.29998779296875 op → 5.300048828125 op)

julia> x = one(Unum{3,4})
Unum{3,4}(0x0000000000000001, 0x0000000000000000, 0x0000, 0x0001, 0x0000)
julia> y = Unum{3,4}(3)
Unum{3,4}(0x0000000000000001, 0x8000000000000000, 0x0000, 0x0000, 0x0000)
julia> x / y
Unum{3,4}(0x0000000000000001, 0x5555000000000000, 0x0001, 0x0002, 0x000F)
julia> describe(x / y)
Unum{3,4}(0.3333320617675781 op → 0.33333587646484375 op)

julia> x = one(Unum{4,7})
Unum{4,7}(0x0000000000000001, UInt64[0x0000000000000000,0x0000000000000000], 0x0000, 0x0001, 0x0000)
julia> y = Unum{4,7}(3)
Unum{4,7}(0x0000000000000001, UInt64[0x8000000000000000,0x0000000000000000], 0x0000, 0x0000, 0x0001)
julia> x / y
Unum{4,7}(0x0000000000000001, UInt64[0x5555555555555555,0x5555555555555555], 0x0001, 0x0002, 0x007F)
julia> describe(x / y)
Unum{4,7}(3.333333333333333333333333333333333333330884386769120234358398465547453654837878e-01 op → 3.333333333333333333333333333333333333338231226461759531283203068905092690324244e-01 op)

To observe the intended "bits" values as they are in "The End of Error", use the builtin bits(::Unum) function, bits(::Unum, ::String) will insert a spacer.

julia> bits(Unum{0,0}(1))
"0010"
julia> bits(Unum{4,7}(1) / Unum{4,7}(3), " ")
"0 001 01010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101 1 0010 1111111"

unumjl was created by Isaac Yonemoto on behalf of REX Computing Inc. and maintained on behalf of Etaphase, Inc.

Support for this work came in part from the following sources: DARPA Contract D15PC00135 awarded to REX Computing, Inc, DARPA Contract HR0011-17-9-0007 awarded to Etaphase, Inc.

鲜花
握手
雷人
路过
鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
More+
上一篇：
yunjhongwu/TensorDecompositions.jl: A Julia implementation of tensor decompositi ...发布时间：2022-07-09
下一篇：
HarrisonGrodin/Simplify.jl: Algebraic simplification in Julia发布时间：2022-07-09
热门推荐
More+
热门话题
More+
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服（服务时间 9:00～18:00）

在线QQ客服
地址：深圳市南山区西丽大学城创智工业园
电邮：jeky_zhao#qq.com
移动电话：139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap