Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improve RBC eps-neighborhood query performance #2211

Merged
merged 13 commits into from
Mar 11, 2024
Merged

Improve RBC eps-neighborhood query performance #2211

Changes from 1 commit
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
2179ed6
optimize rbc eps-NN query
mfoerste4 Mar 4, 2024
1782558
remove L2 cache flush
mfoerste4 Mar 5, 2024
b3e998a
review suggestion constexpr
mfoerste4 Mar 5, 2024
0129748
utilize reordered index for rbc knn
mfoerste4 Mar 5, 2024
ab34c25
Merge branch 'branch-24.04' into rbc_eps_nn_performance
mfoerste4 Mar 5, 2024
fde845b
minor optimizations / suggestions
mfoerste4 Mar 7, 2024
663dabf
add restrict, free 1 register
mfoerste4 Mar 7, 2024
f861403
Merge branch 'rbc_eps_nn_performance' of github.com:mfoerste4/raft in…
mfoerste4 Mar 7, 2024
f417b1a
Merge branch 'branch-24.04' into rbc_eps_nn_performance
mfoerste4 Mar 7, 2024
2be4c17
apply modifications to all kernel variants
mfoerste4 Mar 7, 2024
768eb48
remove one more FLO per landmark visited
mfoerste4 Mar 7, 2024
2315e2f
change mask update
mfoerste4 Mar 8, 2024
2d06d0a
remove warp divergence
mfoerste4 Mar 8, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
remove L2 cache flush
  • Loading branch information
@mfoerste4
mfoerste4 committed Mar 5, 2024
commit 1782558dc61f321aeb722e6b5811b5f03b35747b
8 changes: 0 additions & 8 deletions cpp/include/raft/spatial/knn/detail/ball_cover.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -416,14 +416,6 @@ void rbc_build_index(raft::resources const& handle,
BallCoverIndex<value_idx, value_t, value_int, matrix_idx>& index,
distance_func dfunc)
{
{
/** flush the L2 cache - Hopper at 50MB */
size_t l2_cache_size = 50 * 1024 * 1024;
auto scratch_buf_ = rmm::device_buffer(l2_cache_size * 3, resource::get_cuda_stream(handle));
RAFT_CUDA_TRY(cudaMemsetAsync(
scratch_buf_.data(), 0, scratch_buf_.size(), resource::get_cuda_stream(handle)));
}

ASSERT(!index.is_index_trained(), "index cannot be previously trained");

rmm::device_uvector<value_idx> R_knn_inds(index.m, resource::get_cuda_stream(handle));
Expand Down