Skip to content

opt(insert-and-evict): thrust prefix_sum introduce cudaMalloc/cudaFre… #182

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

opt(insert-and-evict): thrust prefix_sum introduce cudaMalloc/cudaFre… #182

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 5 additions & 9 deletions include/merlin/array_kernels.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@
#pragma once

#include <cooperative_groups.h>
#include "cub/cub.cuh"
#include "cuda_runtime.h"
#include "thrust/device_vector.h"
#include "thrust/execution_policy.h"
Expand Down Expand Up @@ -104,18 +105,13 @@ template <typename K, typename V, typename S, typename Tidx, int TILE_SIZE = 8>
void gpu_boolean_mask(size_t grid_size, size_t block_size, const bool* masks,
size_t n, size_t* n_evicted, Tidx* offsets,
K* __restrict keys, V* __restrict values,
S* __restrict scores, size_t dim, cudaStream_t stream) {
S* __restrict scores, Tidx* offset_ws,
size_t offset_ws_bytes, size_t dim, cudaStream_t stream) {
size_t n_offsets = (n + TILE_SIZE - 1) / TILE_SIZE;
gpu_cell_count<Tidx, TILE_SIZE>
<<<grid_size, block_size, 0, stream>>>(masks, offsets, n, n_evicted);
#if THRUST_VERSION >= 101600
auto policy = thrust::cuda::par_nosync.on(stream);
#else
auto policy = thrust::cuda::par.on(stream);
#endif
thrust::device_ptr<Tidx> d_src(offsets);
thrust::device_ptr<Tidx> d_dest(offsets);
thrust::exclusive_scan(policy, d_src, d_src + n_offsets, d_dest);
CUDA_CHECK(cub::DeviceScan::ExclusiveSum(offset_ws, offset_ws_bytes, offsets,
offsets, n_offsets, stream));
gpu_select_kvm_kernel<K, V, S, Tidx, TILE_SIZE>
<<<grid_size, block_size, 0, stream>>>(masks, n, offsets, keys, values,
scores, dim);
Expand Down
14 changes: 13 additions & 1 deletion include/merlin_hashtable.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@
#include <mutex>
#include <shared_mutex>
#include <type_traits>
#include "cub/cub.cuh"
#include "merlin/allocator.cuh"
#include "merlin/array_kernels.cuh"
#include "merlin/core_kernels.cuh"
Expand Down Expand Up @@ -598,9 +599,20 @@ class HashTable {

keys_not_empty<K>
<<<grid_size, block_size, 0, stream>>>(evicted_keys, d_masks, n);

void* d_temp_storage = nullptr;
size_t temp_storage_bytes = 0;
CUDA_CHECK(cub::DeviceScan::ExclusiveSum(d_temp_storage,
temp_storage_bytes, d_offsets,
d_offsets, n_offsets, stream));
auto helper_ws{
dev_mem_pool_->get_workspace<1>(temp_storage_bytes, stream)};
int64_t* d_temp_storage_i64 = helper_ws.get<int64_t*>(0);

gpu_boolean_mask<K, V, S, int64_t, TILE_SIZE>(
grid_size, block_size, d_masks, n, d_evicted_counter, d_offsets,
evicted_keys, evicted_values, evicted_scores, dim(), stream);
evicted_keys, evicted_values, evicted_scores, d_temp_storage_i64,
temp_storage_bytes, dim(), stream);
}
return;
}
Expand Down