Just for fun, I added pdqsort to the benchmark:

https://github.com/orlp/pdqsort

Here are some of the results on an Ivy Bridge hackintosh:

      size, qsort, inline, sort, stable, pdqsort, radix7
      1000,  88.0,   60.6, 31.2,   37.5,    24.6,   12.8
     10000, 109.3,   77.1, 45.6,   51.6,    29.1,   12.0
    100000, 137.1,   97.7, 57.7,   71.9,    33.3,   12.5
   1000000, 159.6,  114.2, 70.6,   88.6,    39.7,   20.4
  10000000, 185.2,  133.7, 82.0,   99.7,    43.6,   19.4

Edit: it's not quite as easy to plug radix7 into pdqsort's more varied benchmark program, which expects a template function with a type parameter of int.