questions about the CUDA kernel

[zjin-lcf]

In the warp_independent_search_kernel, new and delete are used to allocate memory. Could you please advise a workaround when dynamic memory allocation is not supported by a programming model ? Thank you.

[zhaoweijie12]

You could try to implement a global lightweight memory pool. Check Section V in https://arxiv.org/pdf/2210.07768.pdf for more details.