We can get memory allocation information of instrumented programs using NVBit.
For example, we can set up call backs for cudaMalloc, using which we can get the memory allocated and the virtual address returned using the cuMemAlloc_v2_params struct.
Can we also get the source information of the allocation, i.e., the line number in the host code (enabled with debug symbols) from where the cudaMalloc (or any other memory allocation API) is called?
We can get memory allocation information of instrumented programs using NVBit. For example, we can set up call backs for cudaMalloc, using which we can get the memory allocated and the virtual address returned using the
cuMemAlloc_v2_params
struct. Can we also get the source information of the allocation, i.e., the line number in the host code (enabled with debug symbols) from where the cudaMalloc (or any other memory allocation API) is called?