Tadej Ciglarič
Tadej Ciglarič
/verify with https://github.com/intel/llvm-test-suite/pull/1102
Running appropriate test suite branch resolves some of the test failures. Other tests fail for me locally even on sycl branch (and seem to be testing stuff unrelated to changes...
Could we implement this by just as reshaping A into a row vector and B into a vector and matrix-multiplying them (calculating outer product)?
I have implemented three versions of the algorithm. First uses CPU, second GPU, and the last is hybrid version that uses GPU only for larger matrix products. Here are some...
I will try.
I have used Eigen with MKL before, so I have everything already set up. Tests are already running.
I have updated the graphs with MKL results.
Actually sequantial MKL speeds up things by around 20%, but that is hard to see on the scale of the graphs. I don't think comparison is unfair, but speedups from...
Huh that is old. I found the branch here: https://github.com/bstatcomp/math/tree/gpu_qr