Skip to content

enable RAFT under the hood of GPU FAISS #2840

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

Closed
alexanderguzhva wants to merge 1 commit into from
Closed

enable RAFT under the hood of GPU FAISS #2840

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
19 changes: 16 additions & 3 deletions faiss/gpu/StandardGpuResources.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -312,7 +312,12 @@ void StandardGpuResourcesImpl::initializeForDevice(int device) {
// If this is the first device that we're initializing, create our
// pinned memory allocation
if (defaultStreams_.empty() && pinnedMemSize_ > 0) {
pinnedMemAlloc_ = pmr->allocate(pinnedMemSize_);
try {
pinnedMemAlloc_ = pmr->allocate(pinnedMemSize_);
} catch (const std::bad_alloc& rmm_ex) {
FAISS_THROW_MSG("CUDA memory allocation error");
}

pinnedMemAllocSize_ = pinnedMemSize_;
}
#else
Expand Down Expand Up @@ -490,7 +495,11 @@ void* StandardGpuResourcesImpl::allocMemory(const AllocRequest& req) {

} else if (adjReq.space == MemorySpace::Device) {
#if defined USE_NVIDIA_RAFT
p = cmr->allocate(adjReq.size, adjReq.stream);
try {
p = cmr->allocate(adjReq.size, adjReq.stream);
} catch (const std::bad_alloc& rmm_ex) {
FAISS_THROW_MSG("CUDA memory allocation error");
}
#else
auto err = cudaMalloc(&p, adjReq.size);

Expand All @@ -516,7 +525,11 @@ void* StandardGpuResourcesImpl::allocMemory(const AllocRequest& req) {
#endif
} else if (adjReq.space == MemorySpace::Unified) {
#if defined USE_NVIDIA_RAFT
p = mmr->allocate(adjReq.size, adjReq.stream);
try {
p = mmr->allocate(adjReq.size, adjReq.stream);
} catch (const std::bad_alloc& rmm_ex) {
FAISS_THROW_MSG("CUDA memory allocation error");
}
#else
auto err = cudaMallocManaged(&p, adjReq.size);

Expand Down
2 changes: 1 addition & 1 deletion faiss/gpu/test/test_gpu_index.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,7 +189,7 @@ def test_ivfsq_cpu_coarse(self):

self.assertGreaterEqual(knn_intersection_measure(i_c, i_g), 0.9)

self.assertTrue(np.allclose(d_g, d_c, rtol=5e-5, atol=5e-5))
self.assertTrue(np.allclose(d_g, d_c, rtol=2e-4, atol=2e-4))

def test_ivfpq_cpu_coarse(self):
res = faiss.StandardGpuResources()
Expand Down