Closed lucbv closed 4 months ago
I think this is the problem:
These are not freed. I'm also not sure why we're using new
at all here
@cgcgcg fyi, this is being worked on. @cwpearson actually that pattern appears twice in the test, my sense is simply passing a pointer to the spmv interface instead of storing pointers directly would be just fine?
Compiling Kokkos Core and Kernels with debug symbols and address sanitizer reveals a memory leak in the bsr_spmmv unit-test.
Kokkos Core was configured as follows (of course set variables to appropriate paths):
Kokkos Kernels was configure as follows (again set variable appropriately):
Note that the compiler used was
sems-gcc/13.2.0
available as a SEMS module.