I am new to CUDA programming. I am doing some experiments with the cuSOLVER Lapack APIs. One sample program calls cusolverDnSgetrf() and cusolverDnSgetrs() in an infinite loop. This code runs fine on A10. However, on A100 (A100-SXM4-40GB) it runs fine for a few iterations and then the cudaDeviceSync…

cusolverDnSgetrf() fails on A100 (but not on A10) when called in a tight loop

mnicely January 10, 2022, 3:04pm 5

Okay thanks for checking.
Can you file a bug and provide a complete reproducer?