Fix

paddle/fluid/framework/new_executor/interpreter/dependency_builder.cc

@@ -607,7 +607,7 @@ void PirDependencyBuilder::AddDependencyForCommunicationOp() {
   // c_allreduce_sum(b)
   // c_allreduce_sum(c)
   // c_sync_comm_stream(a)
-  const std::string kSyncComm = dialect::CSyncCommStreamOp::name();
+  const std::string kSyncComm = dialect::SyncCommStreamOp::name();
   dependence_op_idx = ULLONG_MAX;
   for (size_t op_idx = 0; op_idx < op_num_; ++op_idx) {
     if (instructions_.at(op_idx)->Name() == kSyncComm) {

paddle/fluid/ir_adaptor/translator/op_translator.cc

@@ -3902,6 +3902,20 @@ struct WithXShapeAndAxisGradOpTranscriber : public OpTranscriber {
   }
 };
 
+struct SyncCommStreamOpTranscriber : public OpTranscriber {
+  pir::OpInfo LookUpOpInfo(pir::IrContext* ctx,
+                           const OpDesc& op_desc) override {
+    std::string target_op_name = "pd_op.sync_comm_stream_";
+    const auto& op_info = ctx->GetRegisteredOpInfo(target_op_name);
+    if (!op_info) {
+      PADDLE_THROW(common::errors::InvalidArgument(
+          "Op c_sync_comm_stream should have corresponding "
+          "OpInfo pd_op.sync_comm_stream_."));
+    }
+    return op_info;
+  }
+};
+
 OpTranslator::OpTranslator() {
   pir::IrContext* ctx = pir::IrContext::Instance();
   ctx->GetOrRegisterDialect<paddle::dialect::OperatorDialect>();
@@ -4012,6 +4026,8 @@ OpTranslator::OpTranslator() {
       WithXShapeAndAxisGradOpTranscriber<dialect::SqueezeGradOp>();
   special_handlers["unsqueeze2_grad"] =
       WithXShapeAndAxisGradOpTranscriber<dialect::UnsqueezeGradOp>();
+
+  special_handlers["c_sync_comm_stream"] = SyncCommStreamOpTranscriber();
 }
 }  // namespace translator
 }  // namespace paddle

paddle/fluid/operators/collective/c_sync_comm_stream_op.cc

@@ -11,7 +11,7 @@ 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 "paddle/fluid/operators/collective/c_sync_comm_stream_op.h"
+#include "paddle/fluid/framework/op_registry.h"
 
 namespace paddle {
 namespace operators {

paddle/fluid/operators/ops_signature/c_sync_comm_stream_sig.cc

@@ -0,0 +1,27 @@
+// 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 "paddle/phi/core/compat/op_utils.h"
+
+namespace phi {
+
+KernelSignature CSyncCommStreamOpArgumentMapping(
+    const ArgumentMappingContext& ctx) {
+  return KernelSignature("sync_comm_stream", {"X"}, {"ring_id"}, {"Out"});
+}
+
+}  // namespace phi
+
+PD_REGISTER_ARG_MAPPING_FN(c_sync_comm_stream,
+                           phi::CSyncCommStreamOpArgumentMapping);

paddle/fluid/pir/dialect/op_generator/ops_api_gen.py

@@ -214,6 +214,8 @@
     'shadow_feed',
     'shadow_feed_tensors',
     'sparse_momentum',
+    'sync_comm_stream',
+    'sync_comm_stream_',
     'soft_relu',
     'match_matrix_tensor',
     'c_reduce_max',

paddle/fluid/pir/dialect/operator/utils/utils.cc

@@ -40,7 +40,6 @@ const std::unordered_set<std::string> LegacyOpList = {
     CConcatOp::name(),
     CBroadcast_Op::name(),
     CBroadcastOp::name(),
-    CSyncCommStream_Op::name(),
     DistributedPushSparseOp::name(),
     SendV2Op::name(),
     RecvV2Op::name(),

paddle/fluid/pir/transforms/pd_op_to_kernel_pass.cc

@@ -1351,9 +1351,9 @@ phi::KernelKey GetKernelKey(
     VLOG(8) << "LoadCombineOp's kernel data type must be FLOAT32";
   }
 
-  if (op->isa<CSyncCommStream_Op>() || op->isa<CSyncCommStreamOp>()) {
+  if (op->isa<SyncCommStream_Op>() || op->isa<SyncCommStreamOp>()) {
     res.set_dtype(phi::DataType::FLOAT32);
-    VLOG(8) << "CSyncCommStream_Op/CSyncCommStreamOp's kernel data type must "
+    VLOG(8) << "SyncCommStream_Op/SyncCommStreamOp's kernel data type must "
                "be FLOAT32";
   }
 

paddle/phi/kernels/gpu/sync_comm_stream_kernel.cu

@@ -0,0 +1,18 @@
+// 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 "paddle/phi/kernels/impl/sync_comm_stream_kernel.h"
+
+PD_REGISTER_KERNEL(
+    sync_comm_stream, GPU, ALL_LAYOUT, phi::SyncCommStreamKernel, float) {}

paddle/phi/kernels/impl/sync_comm_stream_kernel.h

@@ -0,0 +1,53 @@
+// 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.
+
+#pragma once
+
+#include <string>
+
+#include "paddle/phi/backends/xpu/xpu_info.h"
+#include "paddle/phi/core/dense_tensor.h"
+#include "paddle/phi/core/distributed/comm_context_manager.h"
+#include "paddle/phi/core/kernel_registry.h"
+
+#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
+#include "paddle/phi/core/distributed/nccl_comm_context.h"
+#elif defined(PADDLE_WITH_XPU_BKCL)
+#include "paddle/phi/core/distributed/bkcl_comm_context.h"
+#endif
+
+namespace phi {
+
+#if defined(PADDLE_WITH_XPU_BKCL)
+static void XPUStreamSync(XPUStream stream) {
+  PADDLE_ENFORCE_XDNN_SUCCESS(xpu_wait(stream), "xpu_wait");
+}
+#endif
+
+template <typename T, typename Context>
+void SyncCommStreamKernel(const Context &dev_ctx,
+                          const std::vector<const DenseTensor *> &x UNUSED,
+                          int ring_id UNUSED,
+                          std::vector<DenseTensor *> out UNUSED) {
+#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
+  phi::backends::gpu::GpuStreamSync(dev_ctx.stream());
+#elif defined(PADDLE_WITH_XPU_BKCL)
+  XPUStreamSync(dev_ctx.stream());
+#else
+  PADDLE_THROW(common::errors::PreconditionNotMet(
+      "PaddlePaddle should compile with GPU or XPU."));
+#endif
+}
+
+}  // namespace phi