MAGMA installation problem

[DUCOM]

Hey everyone,
I have been installing MAGMA 1.4.1 using intel mkl. I have already installed Intel mkl 10.3 and cuda 6 on ubuntu 14.04 LTS . After proper modification of the make.inc file I run make. But, i found this bug

magma_threadsetting.cpp:21: fatal error: mkl_service.h: No such file or directory
compilation terminated.
make[1]: *** [magma_threadsetting.o] Error 1


any body who can help me please?

[mgates3]

1) I recommend using the latest release, MAGMA 1.5.0.

2) Please include your make.inc file and any environment variables you set such as MKLROOT. Otherwise it is impossible to tell what is happening.

That file should be accessible in your Intel MKL installation, e.g.,

Code: Select all

prompt> ls $MKLROOT/include/mkl_service.h
/opt/intel/composer_xe_2013_sp1.2.144/mkl/include/mkl_service.h

-mark

[mgates3]

As I said before, please include your make.inc file and relevant environment variables. Otherwise, we cannot help debug your issue.
-mark

[DUCOM]

The following is the make.inc file i used


GPU_TARGET= Tesla

CC = gcc
MKLROOT=/home/aait/intel/composer_xe_2011_sp1.12.361/mkl/lib/intel64
CUDADIR=/usr/local/cuda-6.0
NVCC = $(CUDADIR)/bin/nvcc
FORT = gfortran

ARCH = ar
ARCHFLAGS = cr
RANLIB = ranlib

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

# see MKL Link Advisor at http://software.intel.com/sites/products/mkl/
# 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, GNU threads
#LIB = -lmkl_gf_lp64 -lmkl_gnu_thread -lmkl_core -lpthread -lcublas -lcudart -lstdc++ -lm -lgfortran

# gcc with MKL 10.3, Intel threads
#LIB = -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -lpthread -lcublas -lcudart -lstdc++ -lm -liomp5 -lgfortran

# define library directories preferably in your environment, or here.
# for MKL run, e.g.: source /opt/intel/composerxe/mkl/bin/mklvars.sh intel64

-include make.check-mkl
-include make.check-cuda

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

INC = -I$(CUDADIR)/include \
-I$(MKLROOT)/include

[DUCOM]

I have also tried Magma 1.5.0. But, still no change

[mgates3]

MKLROOT should be /home/aait/intel/composer_xe_2011_sp1.12.361/mkl.

It is best to set this in your environment using the scripts that Intel provides, instead of directly in the make.inc file, as is suggested in the sample make.inc files. E.g.,

source /opt/intel/composerxe/bin/compilervars.csh intel64

Also, are you really using a very old, CUDA capability 1.x Tesla card? If you have a CUDA capability 2.x Fermi card or CUDA capability 3.x Kepler card, you should set GPU_TARGET to Fermi or Kepler. Or don't set GPU_TARGET -- it will compile for all 3 and use the correct one at runtime. Unfortunately, NVIDIA uses "Tesla" for both their first architecture and for their high-end line of cards. Our sample make.inc files try to document exactly what we mean by Tesla.

-mark