installation error: recipe for target 'magmablas/zaxpycp.o'

[railgun]

Hi,
I was installing MAGMA 2.2.0, and I got this error:

Code: Select all

/usr/local/cuda-8.0/include/surface_functions.h(134): error: expected a ";"

/usr/local/cuda-8.0/include/surface_functions.h(135): error: expected a ";"

/usr/local/cuda-8.0/include/surface_functions.h(136): error: expected a ";"
...

Error limit reached.
100 errors detected in the compilation of "/tmp/tmpxft_00002fea_00000000-13_zaxpycp.compute_35.cpp1.ii".
Compilation terminated.
Makefile:585: recipe for target 'magmablas/zaxpycp.o' failed
make: *** [magmablas/zaxpycp.o] Error 4

They are all same errors: expected a ";"．I build MAGMA with OpenBLAS, this is my make.inc file:

Code: Select all

# --------------------
# programs

CC        = gcc
CXX       = g++
NVCC      = nvcc
FORT      = gfortran

ARCH      = ar
ARCHFLAGS = cr
RANLIB    = ranlib


# --------------------
# flags

# Use -fPIC to make shared (.so) and static (.a) library;
# can be commented out if making only static library.
FPIC      = -fPIC

CFLAGS    = -O3 $(FPIC) -DADD_ -Wall -fopenmp 
FFLAGS    = -O3 $(FPIC) -DADD_ -Wall -Wno-unused-dummy-argument
F90FLAGS  = -O3 $(FPIC) -DADD_ -Wall -Wno-unused-dummy-argument -x f95-cpp-input
NVCCFLAGS = -O3         -DADD_       -Xcompiler "$(FPIC)"
LDFLAGS   =     $(FPIC)              -fopenmp

# C++11 (gcc >= 4.7) is not required, but has benefits like atomic operations
CXXFLAGS := $(CFLAGS) -std=c++11
CFLAGS   += -std=c99


# --------------------
# libraries

# gcc with OpenBLAS (includes LAPACK)
LIB       = -lopenblas

LIB      += -lcublas -lcusparse -lcudart


# --------------------
# directories

# define library directories preferably in your environment, or here.
OPENBLASDIR = /home/hlu10/OpenBLAS_install
CUDADIR = /usr/local/cuda-8.0
-include make.check-openblas
-include make.check-cuda

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

INC       = -I$(CUDADIR)/include

Actually I once successfully installed MAGMA on another GPU cluster, but now I need it work on another cluster, so I installed it again and got this error. On the old cluster the CUDA version is 7.5, on the new cluster the CUDA version is 8.0. I was wondering if it's because of MAGMA 2.2.0 is not compatible with CUDA 8.0 ? Does anyone know the reason and how to solve it ?
Thanks very much for reply.

[mgates3]

This seems odd. I can compile MAGMA 2.2 with CUDA 8.0. I'm using gcc 5.4.

Can you compile *any* CUDA code with CUDA 8.0? E.g., something as trivial as this includes surface_functions.h:

Code: Select all

// file foo.cu
// example compilation:
// nvcc -gencode arch=compute_30,code=sm_30 -o foo foo.cu
#include <stdio.h>

__global__
void kernel()
{
}

int main( int argc, char** argv )
{
    printf( "main\n" );
    dim3 blocks( 10 );
    dim3 threads( 20 );
    kernel<<< blocks, threads, 0, 0 >>>();
    return 0;
}

[railgun]

Hi,

CUDA 8.0 works well on other codes. I still don't know the reason, but the administrator of the cluster compiled MAGMA successfully, so I can use MAGMA now.

Thank you very much!

[shoopawax]

Hi!

I am having the exact problem. Is there anyone who has found the fix to this problem? My make.inc file is the same and the error is the same.