126 lines
4.0 KiB
Plaintext
126 lines
4.0 KiB
Plaintext
// Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
|
|
#include <algorithm>
|
|
#include <utility>
|
|
#include <vector>
|
|
|
|
#include "glog/logging.h"
|
|
#include "paddle/phi/core/distributed/utils.h"
|
|
#include "paddle/phi/core/kernel_registry.h"
|
|
#include "paddle/phi/kernels/gpu/partial_send_kernel.h"
|
|
#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
|
|
#include "paddle/phi/core/distributed/nccl_comm_context.h"
|
|
#endif
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void PartialSendKernel(const Context& dev_ctx,
|
|
const DenseTensor& x_in,
|
|
int peer,
|
|
int num,
|
|
int id) {
|
|
#if (defined(PADDLE_WITH_RCCL) || defined(PADDLE_WITH_NCCL)) && \
|
|
NCCL_VERSION_CODE >= 2703
|
|
auto x = &x_in;
|
|
int64_t numel = x->numel();
|
|
|
|
PADDLE_ENFORCE_GE(
|
|
peer,
|
|
0,
|
|
common::errors::InvalidArgument(
|
|
"The peer (%d) for partial_send op must be non-negative.", peer));
|
|
PADDLE_ENFORCE_GE(num,
|
|
1,
|
|
common::errors::InvalidArgument(
|
|
"The num (%d) for partial_send op must >=1", num));
|
|
PADDLE_ENFORCE_EQ(
|
|
(id >= 0 && id < num),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"The id (%d) for partial_send op must >=0 and <num (%d)", id, num));
|
|
PADDLE_ENFORCE_EQ(
|
|
(numel % num),
|
|
0,
|
|
common::errors::InvalidArgument(
|
|
"The input numel (%d) must be divisible by num(%d)", numel, num));
|
|
|
|
int64_t send_numel = numel / num;
|
|
int64_t offset = send_numel * id;
|
|
|
|
gpuStream_t stream = nullptr;
|
|
|
|
distributed::NCCLCommContext* comm_ctx = nullptr;
|
|
int nranks = 0;
|
|
int rank = 0;
|
|
|
|
comm_ctx =
|
|
static_cast<distributed::NCCLCommContext*>(dev_ctx.GetCommContext());
|
|
PADDLE_ENFORCE_NE(comm_ctx,
|
|
nullptr,
|
|
common::errors::Unavailable(
|
|
"NCCLCommContext is nullptr, collective op should "
|
|
"has ring_id attr."));
|
|
|
|
stream = dev_ctx.stream();
|
|
nranks = comm_ctx->GetSize();
|
|
rank = comm_ctx->GetRank();
|
|
|
|
PADDLE_ENFORCE_LT(
|
|
peer,
|
|
nranks,
|
|
common::errors::InvalidArgument("The value of peer (%d) you set must "
|
|
"be less than ranks (%d).",
|
|
peer,
|
|
nranks));
|
|
|
|
auto send_buf = distributed::GetPartialTensor(*x, offset, send_numel);
|
|
comm_ctx->Send(send_buf, send_numel, peer, stream);
|
|
|
|
VLOG(3) << "rank " << rank << " send " << send_numel << " from offset["
|
|
<< offset << "] to " << peer;
|
|
|
|
#else
|
|
PADDLE_THROW(
|
|
common::errors::Unavailable("PaddlePaddle should be compiled with NCCL "
|
|
"and NCCL version >= 2.7.3 is needed."));
|
|
#endif
|
|
}
|
|
} // namespace phi
|
|
|
|
#if (NCCL_VERSION_CODE >= 21000 && CUDA_VERSION >= 11000) || \
|
|
defined(PADDLE_WITH_HIP)
|
|
PD_REGISTER_KERNEL(partial_send,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::PartialSendKernel,
|
|
float,
|
|
double,
|
|
phi::bfloat16,
|
|
int,
|
|
int64_t,
|
|
phi::float16) {}
|
|
#else
|
|
PD_REGISTER_KERNEL(partial_send,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::PartialSendKernel,
|
|
float,
|
|
double,
|
|
int,
|
|
int64_t,
|
|
phi::float16) {}
|
|
#endif
|