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

Add gpuMemAlloc function #423

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
+344 −404
Open

Add gpuMemAlloc function #423

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
75a2ac5
Tackle build warnings
chhwang Dec 19, 2024
b8a1360
Add `gpuMemAlloc` function
chhwang Dec 19, 2024
1d1d07d
Merge branch 'main' into chhwang/malloc
chhwang Dec 20, 2024
78da4a4
Refine `gpu_utils.hpp`
chhwang Dec 20, 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
15 changes: 5 additions & 10 deletions apps/nccl/src/nccl.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -173,9 +173,9 @@ static std::shared_ptr<mscclpp::DeviceHandle<mscclpp::SmChannel>> setupSmChannel
std::transform(smChannels.begin(), smChannels.end(), std::back_inserter(smChannelDeviceHandles),
[](const mscclpp::SmChannel& smChannel) { return mscclpp::deviceHandle(smChannel); });
std::shared_ptr<mscclpp::DeviceHandle<mscclpp::SmChannel>> ptr =
mscclpp::allocSharedCuda<mscclpp::DeviceHandle<mscclpp::SmChannel>>(smChannelDeviceHandles.size());
mscclpp::memcpyCuda<mscclpp::DeviceHandle<mscclpp::SmChannel>>(ptr.get(), smChannelDeviceHandles.data(),
smChannelDeviceHandles.size(), cudaMemcpyHostToDevice);
mscclpp::detail::gpuCallocShared<mscclpp::DeviceHandle<mscclpp::SmChannel>>(smChannelDeviceHandles.size());
mscclpp::gpuMemcpy<mscclpp::DeviceHandle<mscclpp::SmChannel>>(ptr.get(), smChannelDeviceHandles.data(),
smChannelDeviceHandles.size(), cudaMemcpyHostToDevice);
return ptr;
}

Expand Down Expand Up @@ -341,7 +341,7 @@ static void ncclCommInitRankFallbackSingleNode(ncclComm* commPtr, std::shared_pt
commPtr->smSemaphores = std::move(smSemaphores);
commPtr->buffFlag = 0;
commPtr->numScratchBuff = 2;
commPtr->scratchBuff = mscclpp::allocExtSharedCuda<char>(SCRATCH_SIZE);
commPtr->scratchBuff = mscclpp::gpuMemAlloc(SCRATCH_SIZE);
commPtr->remoteScratchRegMemories =
setupRemoteMemories(commPtr->comm, rank, commPtr->scratchBuff.get(), SCRATCH_SIZE, mscclpp::Transport::CudaIpc);
}
Expand Down Expand Up @@ -726,17 +726,12 @@ NCCL_API ncclResult_t ncclCommDeregister(const ncclComm_t, void*) {
}

ncclResult_t ncclMemAlloc(void** ptr, size_t size) {
// Allocate memory using mscclpp::allocSharedPhysicalCuda
if (ptr == nullptr || size == 0) {
return ncclInvalidArgument;
}
std::shared_ptr<char> sharedPtr;
try {
if (mscclpp::isNvlsSupported()) {
sharedPtr = mscclpp::allocSharedPhysicalCuda<char>(size);
} else {
sharedPtr = mscclpp::allocExtSharedCuda<char>(size);
}
sharedPtr = mscclpp::gpuMemAlloc(size);
if (sharedPtr == nullptr) {
return ncclSystemError;
}
Expand Down
Loading
Loading