Hi Pietro,
I tried to compile the program with two different cuda version (10.2 and
11.1) and I got the following error ..
$ pgf90 -Mcuda=cc70,cuda11.0 test.f90
$ ./a.out
0: ALLOCATE: 4000 bytes requested; not enough memory: 35(CUDA driver
version is insufficient for CUDA runtime version)
$ pgf
Dear Rahul,
that seems indeed to be the case and I can only guess that the
environment seen by QE is different.
It's easy to check, can you compile and run this trivial program?
program ma
implicit none
integer, device :: a(1000)
integer :: i
a=1
!$cuf kernel do
do i=1,1000
a(i)=2
enddo
end
Hi Pietro,
Thank you but, as per this this article
https://docs.nvidia.com/deploy/cuda-compatibility/index.html
NVIDIA driver version 418 is compatible with cuda version 10.1 and 10.2.
Following the output from the command "nvidia-smi"
+--
Dear Rahul,
that is useful, you can lookup error codes of type 9xxx here:
https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html#group__CUDART__TYPES_1g3f51e3575c2178246db0a94a430e0038
and you'll find
cudaErrorInsufficientDriver = 35
This indicates that the installed NVIDI
Hi Pietro,
I have tried with configure option --disable-parallel but now, I am
getting message passing error..
*** error in Message Passing (mp) module ***
*** error code: 9035
I have also changed the cuda version to 10.2 from 11.1 but the error is same.
> Dear Rahul Verma,
>
> I would try
Dear Rahul Verma,
I would try the serial version first. Can you reconfigure and compile
with the additional option
--disable-parallel
and check if the misbehavior persists?
Best regards,
Pietro
On 12/31/20 7:50 AM, Rahul Verma wrote:
I have installed GPU enables QE package in a Volta V100
I have installed GPU enables QE package in a Volta V100 machine.
Unfortunately, while running pw.x it is not printing any output.
For reference I am attaching config.log and library dependency output with
this email.
Thank you
Regards
Rahul Verma
Department of Chemistry
IIT KanpurThis file conta