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

fix issue in fusedL2knn which happens when rows are multiple of 256 #604

Merged
merged 2 commits into from
Mar 31, 2022
Merged

fix issue in fusedL2knn which happens when rows are multiple of 256 #604

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b852989
fix issue in fusedL2knn which happens when rows are multiple of 256, …
mdoijade Mar 31, 2022
f792ff6
remove redundant line changes
mdoijade Mar 31, 2022
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
10 changes: 5 additions & 5 deletions cpp/include/raft/spatial/knn/detail/fused_l2_knn.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,8 +105,8 @@ DI void storeWarpQGmem(myWarpSelect** heapArr,
for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
const auto idx = j * warpSize + lid;
if (idx < numOfNN) {
out_dists[gmemRowId * numOfNN + idx] = heapArr[i]->warpK[j];
out_inds[gmemRowId * numOfNN + idx] = (IdxT)heapArr[i]->warpV[j];
out_dists[std::size_t(gmemRowId) * numOfNN + idx] = heapArr[i]->warpK[j];
out_inds[std::size_t(gmemRowId) * numOfNN + idx] = (IdxT)heapArr[i]->warpV[j];
}
}
}
Expand All @@ -130,8 +130,8 @@ DI void loadPrevTopKsGmemWarpQ(myWarpSelect** heapArr,
for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
const auto idx = j * warpSize + lid;
if (idx < numOfNN) {
heapArr[i]->warpK[j] = out_dists[gmemRowId * numOfNN + idx];
heapArr[i]->warpV[j] = (uint32_t)out_inds[gmemRowId * numOfNN + idx];
heapArr[i]->warpK[j] = out_dists[std::size_t(gmemRowId) * numOfNN + idx];
heapArr[i]->warpV[j] = (uint32_t)out_inds[std::size_t(gmemRowId) * numOfNN + idx];
}
}
static constexpr auto kLaneWarpKTop = myWarpSelect::kNumWarpQRegisters - 1;
Expand Down Expand Up @@ -490,7 +490,7 @@ __global__ __launch_bounds__(Policy::Nthreads, 2) void fusedL2kNN(const DataT* x
}
}

if (((gridStrideX + Policy::Nblk * gridDim.x) > n) && gridDim.x == 1) {
if (((gridStrideX + Policy::Nblk * gridDim.x) >= n) && gridDim.x == 1) {
// This is last iteration of grid stride X
loadAllWarpQShmem<Policy, Pair>(heapArr, &shDumpKV[0], m, numOfNN);
storeWarpQGmem<Policy, Pair>(heapArr, out_dists, out_inds, m, numOfNN, starty);
Expand Down
5 changes: 4 additions & 1 deletion cpp/test/spatial/fused_l2_knn.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -212,20 +212,23 @@ class FusedL2KNNTest : public ::testing::TestWithParam<FusedL2KNNInputs> {

const std::vector<FusedL2KNNInputs> inputs = {
{100, 1000, 16, 10, raft::distance::DistanceType::L2Expanded},
{256, 256, 30, 10, raft::distance::DistanceType::L2Expanded},
{1000, 10000, 16, 10, raft::distance::DistanceType::L2Expanded},
{100, 1000, 16, 50, raft::distance::DistanceType::L2Expanded},
{20, 10000, 16, 10, raft::distance::DistanceType::L2Expanded},
{1000, 10000, 16, 50, raft::distance::DistanceType::L2Expanded},
{1000, 10000, 32, 50, raft::distance::DistanceType::L2Expanded},
{10000, 40000, 32, 30, raft::distance::DistanceType::L2Expanded},
{131072, 131072, 8, 60, raft::distance::DistanceType::L2Expanded},
// L2 unexpanded
{100, 1000, 16, 10, raft::distance::DistanceType::L2Unexpanded},
{1000, 10000, 16, 10, raft::distance::DistanceType::L2Unexpanded},
{100, 1000, 16, 50, raft::distance::DistanceType::L2Unexpanded},
{20, 10000, 16, 50, raft::distance::DistanceType::L2Unexpanded},
{1000, 10000, 16, 50, raft::distance::DistanceType::L2Unexpanded},
{1000, 10000, 32, 50, raft::distance::DistanceType::L2Unexpanded},
{10000, 40000, 32, 30, raft::distance::DistanceType::L2Unexpanded}};
{10000, 40000, 32, 30, raft::distance::DistanceType::L2Unexpanded},
{131072, 131072, 8, 60, raft::distance::DistanceType::L2Unexpanded}};

typedef FusedL2KNNTest<float> FusedL2KNNTestF;
TEST_P(FusedL2KNNTestF, FusedBruteForce) { this->testBruteForce(); }
Expand Down