图片。。。


【 在 sosei (fss.sosei) 的大作中提到: 】
: 见51楼
--
FROM 123.139.18.*

同一台机器上，
Python:
In [7]: a,b = 99999999999999999999999999999999999999999999999999999999999999
9999999999999999999999999999999999999999999999999999999999999999999999999999
99999999999999999999
   ...: 9999999999999999999999999999999999999999999999999999,999999
In [8]: from itertools import islice, chain, count
In [9]: %timeit list(islice(chain.from_iterable(zip(count(a, 4), count(-(a+2
), -4))), b//2))
82.8 ms ± 1.89 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
Julia:
julia> function f(b,l)
           a = Vector{BigInt}(undef, Int(ceil((l+1)/2)))
           p = 1
           @inbounds for i = b:4:b+l
               a[p] = i
               p += 2
           end
           p = 2
           @inbounds for i = -(b+2):-4:-(b+l)
               a[p] = i
               p += 2
           end
           a
       end
f (generic function with 1 method)
julia> a,b = 999999999999999999999999999999999999999999999999999999999999999
9999999999999999999999999999999999999999999999999999999999999999999999999999
99999999999999999999999999999999999999999999999999999999999999999999999,9999
99;
julia> f(2,2);
julia> BenchmarkTools.@btime f(a, b);
  50.668 ms (1500039 allocations: 64.85 MiB)
julia>

-----------------------
换了跟 %timeit 一样多次取样的 @btime 做 benchmark。
Julia 的 BigInt 用的是 GMP, vector里做不到内存连续，优势不像  primitive int
那么大。


【 在 sosei (fss.sosei) 的大作中提到: 】
: 呃呃呃，遇上打不开图的错误啦
: 破水木
: from itertools import islice, chain, count
: ...................
--
FROM 123.139.18.*

嗯嗯学到两个宏
在我这酷睿2 win7上跑，时间没啥变化。可能老型号CPU优化不行


【 在 KDr2 的大作中提到: 】
: 同一台机器上，
: Python:
: In [7]: a,b = 99999999999999999999999999999999999999999999999999999999999999
: ...................
--
FROM 60.1.11.*