Hi everybody,
During the last couples of days, I've explored a bit how Machine Learning
could help predictive autotuning, it seems that some techniques have led to
pretty good results in autotuning image processing kernels for GPUs.
Of course, it is out of question to rewrite complicated C++ Machine
Learning algorithm when some already exist in the public domain. Anyway, I
think that building a nice autotuning environment will introduce a lot of
new dependencies (Qt, some ML packages, etc...), and will result in a
binary rather than a library. It seems to me that it would require a
different build system, optional since we can ship reasonable precompiled
kernels. What do you think about having the autotuner as a separate
repository or a submodule?
Philippe
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk
_______________________________________________
ViennaCL-devel mailing list
ViennaCL-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/viennacl-devel
