CUDA Accelerated Linpack and NVIDIA Tesla C2075


After some time and registrations I’ve finally found the CUDA Accelerated Linpack at However, it says it only works for Kepler and Fermi models. So my question is: is there any version for older GPUs like ours, Tesla C2075?

Thanks in advance,


C2075 has a Fermi GPU.

Thanks Massimiliano! Now I have problems building it. I’m using Intel MPI, MKL 11 and CUDA Toolkit 4.1.28. These are the errors I get:

mpiicc -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DCUDA -I/root/hpl-2.0_FERMI_v15/include -I/root/hpl-2.0_FERMI_v15/include/CUDA -I/usr/local/apps/intel/impi/ -I/usr/local/apps/cuda/include -O3 -axS -w -fomit-frame-pointer -funroll-loops -openmp -o /root/hpl-2.0_FERMI_v15/bin/CUDA/xhpl HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /root/hpl-2.0_FERMI_v15/lib/CUDA/libhpl.a -L/root/hpl-2.0_FERMI_v15/src/cuda -ldgemm -L/usr/local/apps/cuda/lib64 -lcudart -lcublas -L/usr/local/apps/intel/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_intel_thread -lpthread -lm
/root/hpl-2.0_FERMI_v15/src/cuda/ undefined reference to cuStreamSynchronize' /root/hpl-2.0_FERMI_v15/src/cuda/ undefined reference to cuCtxSynchronize’
/root/hpl-2.0_FERMI_v15/src/cuda/ undefined reference to cuStreamCreate' /root/hpl-2.0_FERMI_v15/src/cuda/ undefined reference to cuMemcpy2DAsync_v2’
/root/hpl-2.0_FERMI_v15/src/cuda/ undefined reference to `cuCtxGetCurrent’
make: *** [dexe.grd] Error 1

which is strange since I’m linking with libcudart. I had to change some Makefiles because there is not libcuda on version 4.1.28, and besides I’m building HPL in a machine with no GPU (and therefore, no NVIDIA drivers), but which has access to the CUDA Toolkit through a shared directory. This last point is not important, is it?

Any help is appreciated. Thanks again,


If you are using the fast DGEMM for Fermi, you need to link is part of the driver, not the toolkit, so you will need to copy it from a node where it is present.

The other option is to use the normal DGEMM in CUBLAS ( just comment out the FERMI variable), but the code will run
slightly slower.

I am using GTX1080 and got successfully install the accelerated linpack.
but when I tried to run ./xhpl the program couldn’t find
I also couldn’t find the file anywhere.

any suggestion?

thanks in advance

I found the file in …/src/cuda/ and got successfully run this benchmark.
Unfortunately it seems that this benchmark didn’t utilize the gpu