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

[KUNLUN] update xccl lib & use native Reduce in dygraph #49941

Merged
merged 2 commits into from
Jan 20, 2023
Merged

[KUNLUN] update xccl lib & use native Reduce in dygraph #49941

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
1c119f1
update xccl lib & use native Reduce in dygraph
XiaociZhang Jan 19, 2023
62578d7
minor
XiaociZhang Jan 19, 2023
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
2 changes: 1 addition & 1 deletion cmake/external/xpu.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ else()
endif()

set(XPU_XCCL_BASE_URL
"https://klx-sdk-release-public.su.bcebos.com/xccl/release/1.0.6")
"https://klx-sdk-release-public.su.bcebos.com/xccl/release/1.0.7")

if(WITH_AARCH64)
set(XPU_XRE_DIR_NAME "xre-kylin_aarch64")
Expand Down
44 changes: 10 additions & 34 deletions paddle/fluid/distributed/collective/process_group_bkcl.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,41 +352,17 @@ std::shared_ptr<ProcessGroup::Task> ProcessGroupBKCL::Reduce(
const phi::DenseTensor& input,
BKCLContext_t comm,
const XPUStream& stream) {
phi::DenseTensor output_t;
paddle::framework::TensorCopy(*output, platform::XPUPlace(), &output_t);
const auto& place = input.place();
auto* calc_ctx = static_cast<phi::XPUContext*>(
platform::DeviceContextPool::Instance().Get(place));
switch (input.dtype()) {
case phi::DataType::FLOAT32:
calc_ctx->template Alloc<float>(&output_t);
break;
case phi::DataType::FLOAT16:
calc_ctx->template Alloc<float16>(&output_t);
break;
case phi::DataType::INT32:
calc_ctx->template Alloc<int>(&output_t);
break;
default:
VLOG(0) << "Error: type " << input.dtype() << " not supported for "
<< GetBackendName();
break;
}
int ret =
bkcl_all_reduce(comm,
input.data(),
output_t.data(),
input.numel(),
platform::ToBKCLDataType(
framework::TransToProtoVarType(input.type())),
ToBKCLRedType(opts.reduce_op),
stream);
if (rank_ == opts.root_rank) {
*output = output_t;
}
return ret;
return bkcl_reduce(comm,
input.data(),
output->data(),
input.numel(),
platform::ToBKCLDataType(
framework::TransToProtoVarType(input.type())),
ToBKCLRedType(opts.reduce_op),
opts.root_rank,
stream);
},
CommType::ALLREDUCE,
CommType::REDUCE,
sync_op,
use_calc_stream);
}
Expand Down