-
Notifications
You must be signed in to change notification settings - Fork 2
Magma exercise
stomov edited this page Sep 13, 2013
·
16 revisions
Some useful links
- Homepage: http://icl.cs.utk.edu/magma/
- Forum : http://icl.cs.utk.edu/magma/forum/viewforum.php?f=2
- Download: http://icl.cs.utk.edu/magma/software/
Installing MAGMA
- Get MAGMA 1.4
wget http://icl.cs.utk.edu/projectsfiles/magma/downloads/magma-1.4.0.tar.gz - Unpack the library
tar zxvf magma-1.4.0.tar.gz - Modify variables in file make.inc to point to CUDA and LAPACK installations
- Load Intel MKL and CUDA toolkit
module load cudatoolkit intel - Run examples in testing
tomov@todi4:~/magma-1.4.0/testing> aprun -n1 -N1 -e OMP_NUM_THREADS=16 -d16 ./testing_dgemm -N 8000 -c -l
MAGMA 1.4.0 , capability 3.0
device 0: Tesla K20X, 732.0 MHz clock, 5759.6 MB memory, capability 3.5
Usage: ./testing_dgemm [options] [-h|--help]
If running lapack (option --lapack), MAGMA and CUBLAS error are both computed
relative to CPU BLAS result. Else, MAGMA error is computed relative to CUBLAS result.
transA = N, transB = N
M N K MAGMA Gflop/s (ms) CUBLAS Gflop/s (ms) CPU Gflop/s (ms) MAGMA error CUBLAS error
8000 8000 8000 682.51 (1500.35) 1166.05 ( 878.18) 12.87 (79575.56) 1.79e-14 1.79e-14
Application 2096250 resources: utime ~92s, stime ~3s, Rss ~2678724, inblocks ~366, outblocks ~414