benchmark_subarray.jl

using BenchmarkTools

@inline function partial_sum_inline(x, idx)
    s = zero(eltype(x))
    y = view(x, idx)
    for i in 1:endof(y)
        @inbounds s += y[i]
    end
    return s
end

@noinline function partial_sum_noinline(x, idx)
    s = zero(eltype(x))
    y = view(x, idx)
    for i in 1:endof(y)
        @inbounds s += y[i]
    end
    return s
end

function chunked_sum(partial_sum, x, chunk)
    n = length(x) ÷ chunk
    s = zero(eltype(x))
    for i in 0:n-1
        s += partial_sum(x, i*chunk+1:(i+1)*chunk)
    end
    return s
end

srand(1)
x = rand([-1, 1], 1000000)
@assert chunked_sum(partial_sum_inline, x, 10) == sum(x)
@assert chunked_sum(partial_sum_noinline, x, 10) == sum(x)

versioninfo()
@show Base.JLOptions().check_bounds  # 0: unspecified; 1: yes; 2: no

println()
println("** inline **")
display(@benchmark chunked_sum(partial_sum_inline, x, 10))
println()

println()
println("** noinline **")
display(@benchmark chunked_sum(partial_sum_noinline, x, 10))
println()

jl06.output

Julia Version 0.6.2
Commit d386e40 (2017-12-13 18:08 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
  WORD_SIZE: 64
  BLAS: libopenblas (DYNAMIC_ARCH NO_AFFINITY Penryn)
  LAPACK: libopenblas
  LIBM: libm
  LLVM: libLLVM-3.9.1 (ORCJIT, penryn)
Base.JLOptions().check_bounds = 0

** inline **
BenchmarkTools.Trial:
  memory estimate:  16 bytes
  allocs estimate:  1
  --------------
  minimum time:     2.224 ms (0.00% GC)
  median time:      2.876 ms (0.00% GC)
  mean time:        2.672 ms (0.00% GC)
  maximum time:     2.921 ms (0.00% GC)
  --------------
  samples:          1859
  evals/sample:     1

** noinline **
BenchmarkTools.Trial:
  memory estimate:  4.58 MiB
  allocs estimate:  100001
  --------------
  minimum time:     4.002 ms (0.00% GC)
  median time:      5.235 ms (0.00% GC)
  mean time:        5.485 ms (4.50% GC)
  maximum time:     8.833 ms (13.41% GC)
  --------------
  samples:          909
  evals/sample:     1

jl07.output

Julia Version 0.7.0-DEV.3354
Commit 9b5eed2b6c (2018-01-09 08:03 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Penryn)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, penryn)
Environment:
(Base.JLOptions()).check_bounds = 0

** inline **
BenchmarkTools.Trial:
  memory estimate:  16 bytes
  allocs estimate:  1
  --------------
  minimum time:     2.040 ms (0.00% GC)
  median time:      2.647 ms (0.00% GC)
  mean time:        2.639 ms (0.00% GC)
  maximum time:     2.701 ms (0.00% GC)
  --------------
  samples:          1881
  evals/sample:     1

** noinline **
BenchmarkTools.Trial:
  memory estimate:  16 bytes
  allocs estimate:  1
  --------------
  minimum time:     2.072 ms (0.00% GC)
  median time:      2.687 ms (0.00% GC)
  mean time:        2.664 ms (0.00% GC)
  maximum time:     2.747 ms (0.00% GC)
  --------------
  samples:          1864
  evals/sample:     1