Hi List, As you know, I already have announced 3 beta versions of 2.2, and I think it is time to think in fixing already reported bugs and releasing 2.2 final. Besides the fixing bug task, I'd like to include some more optimizations in the way PyTables does internal computations (both for tables.Expr and the evaluation of conditions in table selections), as well as continue optimizing Blosc so that it can finally be faster than a plan memcpy() call, at least for decompressing purposes (we are almost there but not completely, see [1]).
For tackling these additional optimisations, and now that multi-core computers are pervasive, I'd like to experiment with threading so as to accelerate internal computations and compression. Also, I'd like to make PyTables to take advantage of Intel's MKL (Math Kernel Library) for improving the evaluation of complex functions (trigonometrical, exponential, logarithmic...) to the maximum possibilities of the underlying hardware. I'm even thinking of including MKL in the Pro version of PyTables Pro (so you don't have to pay Intel for a MKL license yourself), but this is not decided yet. In all of this work, the principles of memory-efficient computing (as explained in [2]) will be applied to get first-class performance in disk and memory-based computations. My goal is allow PyTables objects to act as a general, high-performance data container that can deal with compressed objects, allowing for an optimal use of your existing computer resources. Due to this, I expect the process to reach 2.2 final to be a bit long. I hope to release 2.2rc1, with some parallel code inside in a month or so, and 2.2rc2, with fully support for Intel's MKL (for Windows, Linux and Mac OS-X) in another month. In the end, my prevision is to release 2.2 final in the next month of June. Meanwhile, you are invited to test the current 2.2b3 release in order to make 2.2 final the better PyTables release ever ;-) Please tell me if you have suggestions about this process. Cheers! [1] http://blosc.pytables.org/trac/wiki/SyntheticBenchmarks [2] http://www.pytables.org/docs/CISE-12-2-ScientificPro.pdf -- Francesc Alted ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Pytables-users mailing list Pytables-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pytables-users
