MAGMA Forum

Hello,
The cmake building capabilities are not up to date. Please use the Makefiles as described in README. Please post your make.inc file if you still have problems compiling (and the errors that you get). Thanks.
Stan

Hello,
I looked at your code and was wondering if the problem comes from Maybe you should have

Hi Luis, I just ran the test and reproduced your result on a K20c. I added a check for the factorization though and for the orthogonality and everything looks fine: test> ./benchmark_cpu -N 2000 Reading input file Read succesful Calculating eigenvalues with MKL Quering space... Solving eigenvalues.....

Hello, It depends on your CPU/GPU configuration and the GPU BLAS used. If I use MAGMA BLAS's dsymv, the results on 16 Sandy Bridge CPU cores @2.6 GHz vs. a K20c GPU (13 x 192 CUDA cores @ 0.7 GHz) would look like this tomov:bunsen ~/trunk/testing> ./testing_dsyevd -l MAGMA 1.3.0 device 0: Tesla K20c...

Nick,
I am sure there would be many people interested. We will be interested to include a make.inc.arm file in the magma distribution and add CARMA as one more device supported. Some tuning may be needed, e.g., at first just to tune the block sizes in control/get_nb.cpp.
Stan

Good, so it looks like the libraries were created. You can ignore that last command for now. I'll double check with the other developers, but it looks like the last command in the Makefile for install should be cat $(MAGMA_DIR)/lib/pkgconfig/magma.pc.in | \ ... instead of cat $(MAGMA_DIR)/lib/pkgcon...

Hi Eric, Thanks for the report. Indeed we had a wrong assumption that certain out of bound values will be multiplied by 0 and therefore would be irrelevant for the result - which resulted in this bug when out of bound values are nan, and then nans propagated in the result. The other precisions have ...

You have to set -L/bummer/GotoBLAS to show the path where libgoto.a is on your system. So your make.inc should look something like this: ... OPTS = -O3 -DADD_ -DCUBLAS_GFORTRAN ... LIB = -lgoto -lpthread -lcublas -lcudart -llapack -lm CUDADIR = /usr/local/cuda-5.0 LIBDIR = -L/path-to-goto-blas -L$(C...

Hi Matt, This is what we currently do. The user is given different sources, e.g., MAGMA, clMAGMA, or MAGMA MIC, and in addition in make.inc a user may specify which accelerator is available. Then only a particular version gets installed. This mechanisms is not convenient though, e.g., for tools like...

Only the higher level (LAPACK) routines use both the CPUs and GPUs. The MAGMA BLAS is mostly for single GPU and some for multiple GPUs. These are used as building blocks in the LAPACK algorithms.