Hi all, I've implemented some more matrix operation tests in PyViennaCL, and I've come across a curious bug when doing operations like the following:
A += B C = alpha * A + B where alpha is a scalar, A and C are dense row-major matrices, and B is a column-major dense matrix. One of these operations always fails, but it is not predictable which one, bizarrely. The only common factor across failures seems to be that the operations involve both a column-major and a row-major matrix. Most of the time, curiously, all the correct elements are along the main diagonal, and all the other elements are incorrect. Other times, the correct and incorrect elements seem randomly distributed (but I haven't investigated whether they are actually random); initially, the matrices are filled with random values. Don't count this as a bug report for ViennaCL yet, though -- I've not had time today to write an exactly equivalent test in C++; that's a task for tomorrow. (There is the unlikely possibility, of course, that the bug is in NumPy, which I am using as my benchmark..) Ciao, Toby ------------------------------------------------------------------------------ Introducing Performance Central, a new site from SourceForge and AppDynamics. Performance Central is your source for news, insights, analysis and resources for efficient Application Performance Management. Visit us today! http://pubads.g.doubleclick.net/gampad/clk?id=48897511&iu=/4140/ostg.clktrk _______________________________________________ ViennaCL-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/viennacl-devel
