Modernize test/perf/speed.jl

[IainNZ]

Output on Julia 1.0.0:

IAINMBP:JuMP idunning$ j1 --project=. test/perf/speed.jl
P-Median(100 facilities, 100 customers, 5000 locations) benchmark:
BenchmarkTools.Trial:
  memory estimate:  2.04 GiB
  allocs estimate:  28562192
  --------------
  minimum time:     2.877 s (40.82% GC)
  median time:      3.038 s (44.02% GC)
  mean time:        3.038 s (44.02% GC)
  maximum time:     3.200 s (46.89% GC)
  --------------
  samples:          2
  evals/sample:     1
Cont5(n=500) benchmark:
BenchmarkTools.Trial:
  memory estimate:  1.16 GiB
  allocs estimate:  15563463
  --------------
  minimum time:     1.519 s (36.65% GC)
  median time:      1.585 s (40.13% GC)
  mean time:        1.581 s (39.81% GC)
  maximum time:     1.637 s (42.14% GC)
  --------------
  samples:          4
  evals/sample:     1

Comparing with numbers from #1403:

PMedian

• release-0.18: ~1.0
• master, with name generation: ~4.6
• this PR: ~2.9

Cont5

• release-0.18: ~0.4
• master, with name generation: ~2.2
• this PR: ~1.6

[IainNZ]

I set up a Julia 1.0/JuMP 0.18 environment so I can actually properly run these in parallel:

IAINMBP:JuMP18 idunning$ j1 --project=. ../dev/JuMP/test/perf/speed.jl
P-Median(100 facilities, 100 customers, 5000 locations) benchmark:
BenchmarkTools.Trial:
  memory estimate:  268.90 MiB
  allocs estimate:  4507937
  --------------
  minimum time:     234.328 ms (0.00% GC)
  median time:      552.879 ms (56.86% GC)
  mean time:        597.856 ms (59.84% GC)
  maximum time:     1.427 s (82.95% GC)
  --------------
  samples:          9
  evals/sample:     1

[codecov]

Codecov Report

Merging #1601 into master will not change coverage.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           master   #1601   +/-   ##
======================================
  Coverage    90.8%   90.8%           
======================================
  Files          28      28           
  Lines        3708    3708           
======================================
  Hits         3367    3367           
  Misses        341     341

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4b04d34...b86460e. Read the comment docs.