Issue when run HPCG on DGX A100 by large size

Hi, I tried to test HPCG follow this link NVIDIA HPC-Benchmarks | NVIDIA NGC
When the test data size is small, not over 440-424-424, HPCG runs well and uses almost 40GB per GPU.
However if I increase the size, for example, to 512-512-512, the program crashes. After investigating the issue, I think it relates to overflow of integer, because in the current code, it is int(32 bit), not long long int(64 bit). I have succeeded in modifying the source code of HPCG to run it on CPU only by changing the lines below in Geometry.hpp file
Line 29: typedef int local_int_t; —> //typedef int local_int_t;
Line 30: //typedef long long local_int_t; —> typedef long long local_int_t;

I believe the HPCG binary from NVIDIA faces same issue. Could you please help me to fix this issue? I need to test HPCG on DGX A100 for our customer. Thanks.

Thanks,
Dac Liem