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

Cleanup faiss includes #1098

Merged
merged 2 commits into from
Dec 14, 2022
Merged

Cleanup faiss includes #1098

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cd78c60
Cleanup faiss includes
benfred Dec 14, 2022
fe245af
change to use raft::KeyValuePair
benfred Dec 14, 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
6 changes: 0 additions & 6 deletions cpp/include/raft/spatial/knn/detail/ann_quantized.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,16 +32,10 @@

#include <rmm/cuda_stream_view.hpp>

#include <faiss/gpu/GpuDistance.h>
#include <faiss/gpu/GpuIndexFlat.h>
#include <faiss/gpu/GpuIndexIVFFlat.h>
#include <faiss/gpu/GpuIndexIVFPQ.h>
#include <faiss/gpu/GpuIndexIVFScalarQuantizer.h>
#include <faiss/gpu/GpuResources.h>
#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/gpu/utils/Tensor.cuh>
#include <faiss/utils/Heap.h>

#include <thrust/iterator/transform_iterator.h>

Expand Down
2 changes: 0 additions & 2 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 @@ -38,9 +38,7 @@
#include <rmm/device_uvector.hpp>
#include <rmm/exec_policy.hpp>

#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/utils/Heap.h>

#include <thrust/fill.h>
#include <thrust/for_each.h>
Expand Down
1 change: 0 additions & 1 deletion cpp/include/raft/spatial/knn/detail/ball_cover/registers.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,6 @@

#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/utils/Heap.h>

#include <thrust/fill.h>

Expand Down
2 changes: 1 addition & 1 deletion cpp/include/raft/spatial/knn/detail/common_faiss.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
#include <raft/util/cuda_utils.cuh>
#include <raft/util/cudart_utils.hpp>

#include <faiss/gpu/GpuDistance.h>
#include <faiss/MetricType.h>
#include <raft/distance/distance_types.hpp>

namespace raft {
Expand Down
3 changes: 0 additions & 3 deletions cpp/include/raft/spatial/knn/detail/haversine_distance.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,11 +19,8 @@
#include <raft/util/cuda_utils.cuh>
#include <raft/util/cudart_utils.hpp>

#include <faiss/gpu/GpuDistance.h>
#include <faiss/gpu/GpuResources.h>
#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/utils/Heap.h>

#include <raft/core/handle.hpp>
#include <raft/distance/distance_types.hpp>
Expand Down
2 changes: 0 additions & 2 deletions cpp/include/raft/spatial/knn/detail/knn_brute_force_faiss.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,10 +23,8 @@
#include <rmm/device_uvector.hpp>

#include <faiss/gpu/GpuDistance.h>
#include <faiss/gpu/GpuResources.h>
#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/utils/Heap.h>

#include <cstdint>
#include <iostream>
Expand Down
5 changes: 0 additions & 5 deletions cpp/include/raft/spatial/knn/detail/selection_faiss.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,12 +20,7 @@
#include <raft/util/cudart_utils.hpp>
#include <raft/util/pow2_utils.cuh>

#include <faiss/gpu/GpuDistance.h>
#include <faiss/gpu/GpuIndexFlat.h>
#include <faiss/gpu/GpuResources.h>
#include <faiss/gpu/utils/Limits.cuh>
#include <faiss/gpu/utils/Select.cuh>
#include <faiss/utils/Heap.h>

namespace raft {
namespace spatial {
Expand Down
204 changes: 1 addition & 203 deletions cpp/include/raft/spatial/knn/detail/warp_select_faiss.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,41 +17,7 @@

namespace faiss {
namespace gpu {

template <typename _Key, typename _Value>
struct KeyValuePair {
typedef _Key Key; ///< Key data type
typedef _Value Value; ///< Value data type

Key key; ///< Item key
Value value; ///< Item value

/// Constructor
__host__ __device__ __forceinline__ KeyValuePair() {}

/// Copy Constructors
__host__ __device__ __forceinline__ KeyValuePair(cub::KeyValuePair<_Key, _Value>& kvp)
: key(kvp.key), value(kvp.value)
{
}

__host__ __device__ __forceinline__ KeyValuePair(faiss::gpu::KeyValuePair<_Key, _Value>& kvp)
: key(kvp.key), value(kvp.value)
{
}

/// Constructor
__host__ __device__ __forceinline__ KeyValuePair(Key const& key, Value const& value)
: key(key), value(value)
{
}

/// Inequality operator
__host__ __device__ __forceinline__ bool operator!=(const KeyValuePair& b)
{
return (value != b.value) || (key != b.key);
}
};
using cub::KeyValuePair;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Doesn't need to hold up this PR but it would be nice to use raft::KeyValuePair here for consistency. I've been converting the cub::keyvaluepair to the raft version everywhere for consistency except where the cub version is absolutely necessary.


//
// This file contains functions to:
Expand Down Expand Up @@ -591,173 +557,5 @@ inline __device__ void warpSortAnyRegistersKVP(K k[N], KeyValuePair<K, V> v[N])
{
BitonicSortStepKVP<K, V, N, Dir, Comp>::sort(k, v);
}

// `Dir` true, produce largest values.
// `Dir` false, produce smallest values.
template <typename K,
typename V,
bool Dir,
typename Comp,
int NumWarpQ,
int NumThreadQ,
int ThreadsPerBlock>
struct KeyValueWarpSelect {
static constexpr int kNumWarpQRegisters = NumWarpQ / faiss::gpu::kWarpSize;

__device__ inline KeyValueWarpSelect(K initKVal, faiss::gpu::KeyValuePair<K, V> initVVal, int k)
: initK(initKVal),
initV(initVVal),
numVals(0),
warpKTop(initKVal),
warpKTopRDist(initKVal),
kLane((k - 1) % faiss::gpu::kWarpSize)
{
static_assert(faiss::gpu::utils::isPowerOf2(ThreadsPerBlock), "threads must be a power-of-2");
static_assert(faiss::gpu::utils::isPowerOf2(NumWarpQ), "warp queue must be power-of-2");

// Fill the per-thread queue keys with the default value
#pragma unroll
for (int i = 0; i < NumThreadQ; ++i) {
threadK[i] = initK;
threadV[i].key = initV.key;
threadV[i].value = initV.value;
}

// Fill the warp queue with the default value
#pragma unroll
for (int i = 0; i < kNumWarpQRegisters; ++i) {
warpK[i] = initK;
warpV[i].key = initV.key;
warpV[i].value = initV.value;
}
}

__device__ inline void addThreadQ(K k, faiss::gpu::KeyValuePair<K, V>& v)
{
if (Dir ? Comp::gt(k, warpKTop) : Comp::lt(k, warpKTop)) {
// Rotate right
#pragma unroll
for (int i = NumThreadQ - 1; i > 0; --i) {
threadK[i] = threadK[i - 1];
threadV[i].key = threadV[i - 1].key;
threadV[i].value = threadV[i - 1].value;
}

threadK[0] = k;
threadV[0].key = v.key;
threadV[0].value = v.value;
++numVals;
}
}
/// This function handles sorting and merging together the
/// per-thread queues with the warp-wide queue, creating a sorted
/// list across both

// TODO
__device__ inline void mergeWarpQ()
{
// Sort all of the per-thread queues
faiss::gpu::warpSortAnyRegistersKVP<K, V, NumThreadQ, !Dir, Comp>(threadK, threadV);

// The warp queue is already sorted, and now that we've sorted the
// per-thread queue, merge both sorted lists together, producing
// one sorted list
faiss::gpu::warpMergeAnyRegistersKVP<K, V, kNumWarpQRegisters, NumThreadQ, !Dir, Comp, false>(
warpK, warpV, threadK, threadV);
}

/// WARNING: all threads in a warp must participate in this.
/// Otherwise, you must call the constituent parts separately.
__device__ inline void add(K k, faiss::gpu::KeyValuePair<K, V>& v)
{
addThreadQ(k, v);
checkThreadQ();
}

__device__ inline void reduce()
{
// Have all warps dump and merge their queues; this will produce
// the final per-warp results
mergeWarpQ();
}

__device__ inline void checkThreadQ()
{
bool needSort = (numVals == NumThreadQ);

#if CUDA_VERSION >= 9000
needSort = __any_sync(0xffffffff, needSort);
#else
needSort = __any(needSort);
#endif

if (!needSort) {
// no lanes have triggered a sort
return;
}

mergeWarpQ();

// Any top-k elements have been merged into the warp queue; we're
// free to reset the thread queues
numVals = 0;

#pragma unroll
for (int i = 0; i < NumThreadQ; ++i) {
threadK[i] = initK;
threadV[i].key = initV.key;
threadV[i].value = initV.value;
}

// We have to beat at least this element
warpKTopRDist = shfl(warpV[kNumWarpQRegisters - 1].key, kLane);
warpKTop = shfl(warpK[kNumWarpQRegisters - 1], kLane);
}

/// Dump final k selected values for this warp out
__device__ inline void writeOut(K* outK, V* outV, int k)
{
int laneId = faiss::gpu::getLaneId();

#pragma unroll
for (int i = 0; i < kNumWarpQRegisters; ++i) {
int idx = i * faiss::gpu::kWarpSize + laneId;

if (idx < k) {
outK[idx] = warpK[i];
outV[idx] = warpV[i].value;
}
}
}

// Default element key
const K initK;

// Default element value
const faiss::gpu::KeyValuePair<K, V> initV;

// Number of valid elements in our thread queue
int numVals;

// The k-th highest (Dir) or lowest (!Dir) element
K warpKTop;

// TopK's distance to closest landmark
K warpKTopRDist;

// Thread queue values
K threadK[NumThreadQ];
faiss::gpu::KeyValuePair<K, V> threadV[NumThreadQ];

// warpK[0] is highest (Dir) or lowest (!Dir)
K warpK[kNumWarpQRegisters];
faiss::gpu::KeyValuePair<K, V> warpV[kNumWarpQRegisters];

// This is what lane we should load an approximation (>=k) to the
// kth element from the last register in the warp queue (i.e.,
// warpK[kNumWarpQRegisters - 1]).
int kLane;
};

} // namespace gpu
} // namespace faiss