Also up the number of iterations the swift benchmark in the stdlib runs so that
it matches the number of iterations the c++ version does. This iteration number
is enough to cause the c++ version to run for ~1 second. With this same number
of iterations, the swift version takes ~10 seconds.
Swift SVN r15205
This reverts commit r15200. This readds prim with a smaller test case to get
around type checker slowness that causes the compiling with Benchmarks to be
prohibitively slow. Now we use a smaller dataset but run it for more times.
Swift SVN r15204
This reverts commit r15164. I need to reduce the size of the test case and run
it more times instead. I am reverting this for now to unblock Nadav. When it is
fixed I will recommit it.
Swift SVN r15200
We are actually doing a pretty good job optimizing this (even though I
am sure it could be better). The test case is a 5000 node prim run. IIRC
when I ran the cpp version of this on my local cpu it was instantaneous.
The swift version is ~1.5 seconds.
Swift SVN r15164
