tbennun / mgbench Goto Github PK

Some multi-gpu servers behave differently according to the pci topology and the bus connection to the CPUs. I suggest adding a test that uses CPU affinity in order to check the speed between each CPU/GPU pair.

Hi,
I have a problem, I want to use MGbench at ARM TX2 platform.
But when I execute build.sh, it show:

-- The C compiler identification is GNU 8.3.1
-- The CXX compiler identification is GNU 8.3.1
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found CUDA: /usr/local/cuda (found suitable version "10.1", minimum required is "7")
-- Looking for C++ include unistd.h
-- Looking for C++ include unistd.h - found
-- Looking for C++ include stdint.h
-- Looking for C++ include stdint.h - found
-- Looking for C++ include inttypes.h
-- Looking for C++ include inttypes.h - found
-- Looking for C++ include sys/types.h
-- Looking for C++ include sys/types.h - found
-- Looking for C++ include sys/stat.h
-- Looking for C++ include sys/stat.h - found
-- Looking for C++ include fnmatch.h
-- Looking for C++ include fnmatch.h - found
-- Looking for C++ include stddef.h
-- Looking for C++ include stddef.h - found
-- Check size of uint32_t
-- Check size of uint32_t - done
-- Looking for strtoll
-- Looking for strtoll - found
-- Check size of pthread_rwlock_t
-- Check size of pthread_rwlock_t - done
CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_cublas_device_LIBRARY (ADVANCED)
linked by target "sgemm" in directory /root/mgbench

-- Configuring incomplete, errors occurred!
See also "/root/mgbench/build/CMakeFiles/CMakeOutput.log".
See also "/root/mgbench/build/CMakeFiles/CMakeError.log".
make: *** No targets specified and no makefile found. Stop.

Does ARM platform cause this problem? What can I do?
P.S. make log as below
CMakeError.log

I'm not sure that this is actually an issue rather than a lack of awareness on my part, but I've observed the following:

After enabling Intel IOMMU in GRUB, benchmarking throws lots of errors during uva such as:

DMAR: [DMA Write] Request Device [ : : . ] fault addr
DMAR: [fault reason 05] PTE Write access is not set

where [ : : . ] and fault addr contain a variety of addresses.

Log:

Running Matrix Multiplication with 4 GPUs, starting with 1 GPUs

Testing with 1 GPUs

SGEMM - MAPS (unmodified routine): 38.103477 ms
TEST passed

Testing with 2 GPUs

ERROR in deps/maps/include/maps/multi/scheduler.h:189: invalid argument (11)

I'd like to refer to this project in a paper I'm working on - is there a specific reference I should use?