Ah, sorry about that. I thought it was still only necessary for GEMM (i.e., my fault :p). In my experience, padding of 4 along the leading dimension can be pretty useful. But the 128x128 padding required by the GEMM kernel was a big design mistake of mine, hence my desire to see this restriction annihilated.
On Tue, Nov 29, 2016 at 5:03 PM, Karl Rupp <r...@iue.tuwien.ac.at> wrote: > Hi Philippe, > > I know I mentioned this to you already Karl, but ViennaCL could really >> benefit from using the GEMM code in Isaac, which has higher performance >> and doesn't require >> padding: https://github.com/ptillet/isaac/blob/master/lib/jit/generat >> ion/gemm.cpp >> (benchmarks on the main project page). >> > > Sure - yet the world does not only consist of GEMM ;-) But yes, the > padding was initially introduced because of GEMM. Yet it would not be wise > to give up on padding completely. There are a bunch of cases where this is > useful. > > Best regards, > Karli > > > On Tue, Nov 29, 2016 at 4:33 PM, Karl Rupp <r...@iue.tuwien.ac.at >> <mailto:r...@iue.tuwien.ac.at>> wrote: >> >> Hi, >> >> > I have been looking through the documentation and I can't find any >> > direction on how to create a matrix without the default padding. >> This >> > is providing to be a memory problem for me when working with very >> > 'narrow' matrices that are also quite long filling up the available >> > memory. If it is in the docs feel free to point it out but I have >> been >> > stumped at the moment. >> >> unfortunately there is no way to create a viennacl::matrix without >> padding, unless you provide your own host array, CUDA buffer, or >> OpenCL >> memory handle. There is not even a way of doing it through >> matrix_base. >> This is oversight and will be fixed asap. Allow for one more day :-) >> >> Best regards, >> Karli >> >> >> ------------------------------------------------------------ >> ------------------ >> _______________________________________________ >> ViennaCL-devel mailing list >> ViennaCL-devel@lists.sourceforge.net >> <mailto:ViennaCL-devel@lists.sourceforge.net> >> https://lists.sourceforge.net/lists/listinfo/viennacl-devel >> <https://lists.sourceforge.net/lists/listinfo/viennacl-devel> >> >> >> >
------------------------------------------------------------------------------
_______________________________________________ ViennaCL-devel mailing list ViennaCL-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/viennacl-devel
