<p>On datasets that are large enough, it looks like you get a .5x-.8x relative speedup per core. I tested at 4 and 16 cores. Because I parallelized CV iterations, a 10-fold CV doesn't actually fully utilize a 16-core machine.</p>

<p>The highlight in the data below is the 9.1x speedup on the birch3 dataset using a 16-core machine and 100-fold CV.</p>

<p>I didn't have the patience to run on covertype as fully as I had planned, but you can see a few tests that did complete. Using the other datasets, you could estimate the missing numbers. User time is handy for that.</p>

<p><a href="https://camo.githubusercontent.com/ba5dc732b4f5e5a4a361bc3a67af20b3a2c69591/687474703a2f2f692e696d6775722e636f6d2f33394e633170522e706e67" target="_blank"><img src="https://camo.githubusercontent.com/ba5dc732b4f5e5a4a361bc3a67af20b3a2c69591/687474703a2f2f692e696d6775722e636f6d2f33394e633170522e706e67" alt="Benchmarks" data-canonical-src="http://i.imgur.com/39Nc1pR.png" style="max-width:100%;"></a></p>

<p>Raw CSV:</p>

<pre><code>Dataset,Folds,Instance size,Commit level,Real time,User time,det_training time,,Rel speedup,Cores,Rel speedup per core
winequality,10,A3,4379c33,8.279,8.202,8.205,,,,
winequality,10,A3,a63ddcd,3.736,10.355,3.626,,2.2,4,0.55
winequality,10,A11,4379c33,3.618,3.612,3.584,,,,
winequality,10,A11,a63ddcd,0.821,4.272,0.787,,4.4,16,0.28
winequality,100,A3,4379c33,38.621,38.455,38.522,,,,
winequality,100,A3,a63ddcd,11.304,41.961,11.225,,3.4,4,0.85
winequality,100,A11,4379c33,16.521,16.488,16.488,,,,
winequality,100,A11,a63ddcd,1.867,17.68,1.832,,8.8,16,0.55
birch3,10,A3,4379c33,2005.799,2002.794,2005.571,,,,
birch3,10,A3,a63ddcd,1011.868,2707.466,1011.639,,2.0,4,0.50
birch3,10,A11,4379c33,860.472,860.356,860.364,,,,
birch3,10,A11,a63ddcd,209.309,1085.44,209.202,,4.1,16,0.26
birch3,100,A3,4379c33,9752.487,9739.609,9752.254,,,,
birch3,100,A3,a63ddcd,3403.719,12827.15,3403.468,,2.9,4,0.72
birch3,100,A11,4379c33,4057.705,4056.52,4057.59,,,,
birch3,100,A11,a63ddcd,448.225,5649.664,448.116,,9.1,16,0.57
corel,10,A3,4379c33,302.655,,301.33,,,,
corel,10,A3,a63ddcd,146.277,414.45,144.856,,2.1,4,0.52
corel,10,A11,4379c33,130.724,130.668,130.162,,,,
corel,10,A11,a63ddcd,25.647,167.328,25.087,,5.1,16,0.32
corel,100,A3,4379c33,1207.98,,1206.657,,,,
corel,100,A3,a63ddcd,417.857,1501.339,416.556,,2.9,4,0.72
corel,100,A11,4379c33,514.071,514.04,513.51,,,,
corel,100,A11,a63ddcd,59.166,722.54,58.602,,8.7,16,0.54
covertype,10,A3,4379c33,,,,,,,
covertype,10,A3,a63ddcd,48326.77,140865.588,48306.619,,,,
covertype,10,A11,4379c33,,,,,,,
covertype,10,A11,a63ddcd,10015.182,53387.82,10006.181,,,,
covertype,100,A3,4379c33,,,,,,,
covertype,100,A3,a63ddcd,,,,,,,
covertype,100,A11,4379c33,,,,,,,
covertype,100,A11,a63ddcd,14845.443,165220.52,14836.508,,,,
</code></pre>

<p style="font-size:small;-webkit-text-size-adjust:none;color:#666;">&mdash;<br>Reply to this email directly or <a href="https://github.com/mlpack/mlpack/pull/438#issuecomment-103117177">view it on GitHub</a>.<img alt="" height="1" src="https://github.com/notifications/beacon/AJ4bFDo2YLLJG25LKDagewCOCgOPbEizks5oKghDgaJpZM4Ea3Qy.gif" width="1" /></p>
<div itemscope itemtype="http://schema.org/EmailMessage">
  <div itemprop="action" itemscope itemtype="http://schema.org/ViewAction">
    <link itemprop="url" href="https://github.com/mlpack/mlpack/pull/438#issuecomment-103117177"></link>
    <meta itemprop="name" content="View Pull Request"></meta>
  </div>
  <meta itemprop="description" content="View this Pull Request on GitHub"></meta>
</div>