[WIP] Benchmark CAGRA select_k

[benfred]

No description provided.

[benfred]

Running ./cpp/build/MATRIX_BENCH --benchmark_filter=SelectKDataset/float/int64_t/kCagra/146/9594063/256/0/1/manual_time --benchmark_out_format=json --benchmark_out=select_k_times.json --select_k_dataset on this branch produces memory access errors like

=========     at 0x3ee0 in /home/ben/code/raft/cpp/include/raft/neighbors/detail/cagra/topk_for_cagra/topk_core.cuh:611:void raft::neighbors::experimental::cagra::detail::topk_cta_11_core<(int)1024, (int)8, (int)1, (int)256, (int)64, long>(unsigned int, unsigned int, const unsigned int *, const T6 *, unsigned int *, T6 *, unsigned char *, unsigned int *, bool, unsigned int *)
=========     by thread (155,0,0) in block (138,0,0)
=========     Address 0x7f66cc56f388 is out of bounds
=========     and is 694,750,328 bytes before the nearest allocation at 0x7f66f5c00000 of size 102,400 bytes
=========     Device Frame:/home/ben/code/raft/cpp/include/raft/neighbors/detail/cagra/topk_for_cagra/topk_core.cuh:784:void raft::neighbors::experimental::cagra::detail::kern_topk_cta_11<(int)1024, (int)8, (int)1, (int)256, (int)64, long>(unsigned int, unsigned int, unsigned int, const unsigned int *, unsigned int, const T6 *, unsigned int, unsigned int *, unsigned int, T6 *, unsigned int, unsigned char *, unsigned int *, bool) [0x13c0]

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB