Benchmark result df

[chriselrod]

Example:

julia> benchmark_result_df(rb)
18×4 DataFrame
 Row │ Sizes  Library   Seconds      GFLOPS
     │ Int64  String    Float64      Float64
─────┼───────────────────────────────────────────
   1 │    20  BLIS      4.276e-5        0.374181
   2 │    77  BLIS      8.2822e-5      11.0244
   3 │   300  BLIS      0.000295945   182.466
   4 │    20  Gaius     2.63026e-7     60.8306
   5 │    77  Gaius     0.000269507     3.38791
   6 │   300  Gaius     0.000588027    91.8325
   7 │    20  MKL       4.41985e-7     36.2003
   8 │    77  MKL       4.38214e-6    208.361
   9 │   300  MKL       4.4458e-5    1214.63
  10 │    20  Octavian  2.4756e-7      64.6307
  11 │    77  Octavian  2.85825e-6    319.449
  12 │   300  Octavian  4.3652e-5    1237.06
  13 │    20  OpenBLAS  7.15642e-7     22.3576
  14 │    77  OpenBLAS  1.2527e-5      72.8878
  15 │   300  OpenBLAS  9.8837e-5     546.354
  16 │    20  Tullio    2.56611e-7     62.3512
  17 │    77  Tullio    9.1536e-5       9.97494
  18 │   300  Tullio    0.000415785   129.875

[DilumAluthge]

Could you add the following as a test:

import BLASBenchmarksCPU
import StatsPlots
benchmark_result = BLASBenchmarksCPU.runbench(Float64; sizes = [1, 2, 5, 10, 20, 50, 100, 200, 500, 1_000])
df = BLASBenchmarksCPU.benchmark_result_df(benchmark_result)
@test df isa DataFrame
df[!, :Size] = Float64.(df[!, :Size])
df[!, :GFLOPS] = Float64.(df[!, :GFLOPS])
df[!, :Time] = Float64.(df[!, :Time])
p = StatsPlots.@df df StatsPlots.plot(:Size, :GFLOPS; group = :Library, legend = :bottomright)

[DilumAluthge]

That test basically covers the main use case.

[chriselrod]

I named the time column Seconds instead of Time to be explicit about units.