Does MAGMA1.0 & Above support Geforce 9series?

[Ian]

Does MAGMA 1.0 support Geforce 9 Series?

Thanks.

[mgates3]

See http://developer.nvidia.com/cuda-gpus for a list of CUDA-enabled GPUs. MAGMA supports both CUDA 1.x and CUDA 2.x (Fermi) GPUs. Set the appropriate CUDA version in the make.inc file (GPU_TARGET=0 for CUDA 1.x, GPU_TARGET=1 for CUDA 2.x). However, only GPUs with CUDA 1.3 and above support double precision.

-mark

[Ian]

Thanks Mark!

I am trying to install MAGMA 1.0 under ubuntu 10.10.
When compilation, I encountered such problem,

Code: Select all

/home/ian/magma_1.0.0/control'
gfortran -O3 -DADD_ -x f95-cpp-input -Dmagma_devptr_t="integer(kind=../control/sizeptr.c: In function ‘main’: ../control/sizeptr.c:6:3: warning: format ‘%lu’ expects type ‘long unsigned int’, but argument 2 has type ‘unsigned int’ 4)" -I/usr/local/cuda/include -I../include -I../quark/include -c magma_zfortran.f90 -o magma_zfortran.o
magma_zfortran.f90:279.20:

integer(kind=../control/sizeptr.c: In function \xE2\x80\x98main\xE2\x80\x99: ../control/si
1
Error: Expected initialization expression at (1)
magma_zfortran.f90:281.20:

Here is my configuration (atlas installed through binary package):

Code: Select all

GPU_TARGET = 0

CC        = gcc
NVCC      = nvcc
FORT      = gfortran

ARCH      = ar
ARCHFLAGS = cr
RANLIB    = ranlib

OPTS      = -O3 -DADD_
FOPTS     = -O3 -DADD_ -x f95-cpp-input
NVOPTS    = --compiler-options -fno-strict-aliasing -DUNIX -O3 -DADD_
LDOPTS    = -fPIC -Xlinker -zmuldefs

LIB       = /usr/lib/atlas-base/atlas/ \
            -lf77blas -latlas -lcblas -lf2c -lcublas -lm -lgfortran

CUDADIR   = /usr/local/cuda

LIBDIR    = -L/usr/lib/atlas-base/atlas/ \
            -L/usr/lib/atlas-base/ \
            -L$(CUDADIR)/lib
INC       = -I$(CUDADIR)/include

LIBMAGMA     = ../lib/libmagma.a
LIBMAGMABLAS = ../lib/libmagmablas.a

I have been lurking around the forum all day trying to find a solution. I don't know too much about compiler setting, so I am also reading literature about what the meaning of the flags. But I still feel kinda lost.
Thanks in advance for any input!

[mgates3]

Notice the Fortran compile command is gibberish:

gfortran -O3 -DADD_ -x f95-cpp-input -Dmagma_devptr_t="integer(kind=../control/sizeptr.c: In function ‘main’: ../control/sizeptr.c:6:3: warning: format ‘%lu’ expects type ‘long unsigned int’, but argument 2 has type ‘unsigned int’ 4)" ...

It should look like:

gfortran -O3 -DADD_ -x f95-cpp-input -Dmagma_devptr_t="integer(kind=4)" ...

What's happening is that it is compiling & running sizeptr to get the size of a pointer in C, and using the output of that in the Fortran compile command. The output should be just "4" or "8", but in this case there is a warning when compiling sizeptr, which is erroneously getting stuck into the Fortran compile line. The fix is simple: Replace control/sizeptr.c with the attached version, which should compile without errors.

[Ian]

Thanks for interpretation of problem. I learn a lot from it. The patch you wrote is straightforward and effective. I encountered several similar errors mentioned in post below.
viewtopic.php?f=2&t=278

Now all have been fixed. It compiles.
Thank you, Mark.