Magma 1.4.0 and MKL

Open discussion for MAGMA

Magma 1.4.0 and MKL

Postby Linuxboy » Tue Sep 17, 2013 4:25 pm

Hi! I've compiled magma 1.4.0 with MKL. When i use sequential version all is working fine
MAGMA 1.4.0 , capability 3.0
device 0: GeForce GTX TITAN, 875.5 MHz clock, 6143.7 MB memory, capability 3.5
Usage: ./testing_dgetrf [options] [-h|--help]

ngpu 1
M N CPU GFlop/s (sec) GPU GFlop/s (sec) |PA-LU|/(N*|A|)
=========================================================================
1088 1088 21.72 ( 0.04) 75.03 ( 0.01) 3.18e-18
2112 2112 27.33 ( 0.23) 185.88 ( 0.03) 3.03e-18
3136 3136 28.36 ( 0.72) 223.99 ( 0.09) 2.37e-18
4160 4160 28.74 ( 1.67) 307.52 ( 0.16) 2.31e-18


But when i use Intel threads all compiles but i've got error
MAGMA 1.4.0 , capability 3.0
device 0: GeForce GTX TITAN, 875.5 MHz clock, 6143.7 MB memory, capability 3.5
Usage: ./testing_dgetrf [options] [-h|--help]

ngpu 1
M N CPU GFlop/s (sec) GPU GFlop/s (sec) |PA-LU|/(N*|A|)
=========================================================================
Ошибка сегментирования (segmentation error )


My make.inc
GPU_TARGET = Kepler

CC = gcc
NVCC = nvcc
FORT = gfortran

ARCH = ar
ARCHFLAGS = cr
RANLIB = ranlib

OPTS = -O3 -DADD_ -Wall -fno-strict-aliasing -fopenmp -DMAGMA_WITH_MKL -DMAGMA_SETAFFINITY -openmp -DMKL_ILP64
F77OPTS = -O3 -DADD_ -Wall
FOPTS = -O3 -DADD_ -Wall -x f95-cpp-input
NVOPTS = -O3 -DADD_ -Xcompiler -fno-strict-aliasing
LDOPTS = -fopenmp

# gcc with MKL 10.3, sequential version
#LIB = -lmkl_gf_lp64 -lmkl_sequential -lmkl_core -lcublas -lcudart -lstdc++ -lm -lgfortran
# gcc with MKL 10.3, Intel threads
LIB = -lmkl_intel_ilp64 -lmkl_intel_thread -lmkl_core -lpthread -lcublas -lcudart -lstdc++ -lm -liomp5 -lgfortran -fopenmp

MKLROOT = /opt/intel/mkl
CUDADIR = /usr/local/cuda


LIBDIR = -L$(MKLROOT)/lib/intel64 \
-L/opt/intel/composerxe/lib/intel64 \
-L$(CUDADIR)/lib64

INC = -I$(CUDADIR)/include -I$(MKLROOT)/include
Linuxboy
 
Posts: 14
Joined: Tue Nov 29, 2011 9:24 pm

Re: Magma 1.4.0 and MKL

Postby mgates3 » Mon Sep 23, 2013 1:00 pm

Don't use ilp64, use lp64.
Or if you really want to use ilp64, see make.inc.mkl-ilp64. In particular, add -DMAGMA_ILP64 and -DMKL_ILP64 to both OPTS and NVOPTS, and -fdefault-integer-8 to F77OPTS and FOPTS. (For future releases, I fixed it so magma recognizes -DMKL_ILP64.)
-mark
mgates3
 
Posts: 437
Joined: Fri Jan 06, 2012 2:13 pm


Return to User discussion

Who is online

Users browsing this forum: Yahoo [Bot] and 1 guest

cron