You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
graphengine/ge/graph/passes/buffer_pool_memory_pass.h

137 lines
5.4 KiB

/**
* Copyright 2020 Huawei Technologies Co., Ltd
*
* 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.
*/
#ifndef GE_GRAPH_PASSES_BUFFER_POOL_MEMORY_PASS_H_
#define GE_GRAPH_PASSES_BUFFER_POOL_MEMORY_PASS_H_
#include <queue>
#include "graph/graph.h"
#include "inc/graph_pass.h"
namespace ge {
class BufferPoolMemoryPass : public GraphPass {
public:
explicit BufferPoolMemoryPass() : logic_event_num_(0) {}
~BufferPoolMemoryPass() override = default;
struct BufferPool {
int64_t pool_id = 0;
int64_t pool_size = 0;
std::unordered_map<NodePtr, NodePtr> buffer_node_to_calc;
BufferPool(int64_t id, int64_t size, const std::unordered_map<NodePtr, NodePtr> &node_map)
: pool_id(id), pool_size(size), buffer_node_to_calc(node_map) {}
};
struct BufferPoolNodeItem {
NodePtr node = nullptr;
NodePtr out_calc_node = nullptr;
NodePtr pre_buffer_pool_node = nullptr;
int64_t total_size = 0;
int64_t offset_start = 0;
int64_t offset_end = 0;
bool is_last_input = true;
BufferPoolNodeItem(const NodePtr &buffer_n, const NodePtr &calc_n, const NodePtr &pre_buffer_n,
int64_t size, int64_t start, int64_t end, bool last)
: node(std::move(buffer_n)),
out_calc_node(std::move(calc_n)),
pre_buffer_pool_node(std::move(pre_buffer_n)),
total_size(size),
offset_start(start),
offset_end(end),
is_last_input(last) {}
BufferPoolNodeItem(const NodePtr &buffer_n, int64_t start, int64_t end)
: node(std::move(buffer_n)),
out_calc_node(nullptr),
pre_buffer_pool_node(nullptr),
total_size(0),
offset_start(start),
offset_end(end),
is_last_input(true) {}
};
Status Run(ComputeGraphPtr graph) override;
private:
static void ClearQueue(std::queue<std::pair<std::string, uint32_t>> &q);
static Status IsBufferPoolMemEnable(const ComputeGraphPtr &graph);
static Status CheckBufferPoolSize(int64_t total_size, int64_t pool_id, int64_t buffer_pool_size,
std::unordered_map<int64_t, int64_t> &calc_total_size);
static Status TryToFixNodeOrder(NodePtr &pre_node, NodePtr &curr_node, bool &not_change);
Status InsertMemCpyNodeAfter(ComputeGraphPtr &graph, NodePtr &node);
Status CopyOutForMultiUsedOutput(ComputeGraphPtr &graph);
Status GetBufferPoolAndPeerCalcNodes(const ComputeGraphPtr &graph);
Status SetBufferPoolSize(const std::string &batch_label, int64_t id, int64_t size);
Status AllocateAllBufferPoolSpace();
Status AllocateSpaceInBatch(const std::map<int64_t, std::vector<NodePtr>> &calc_nodes,
const std::unordered_map<int64_t, int64_t> &buffer_pool_size_map,
const std::unordered_map<NodePtr, NodePtr> &buffer_node_to_calc,
std::unordered_map<NodePtr, std::vector<BufferPoolNodeItem>> &buffer_pool_nodes_item);
Status AllocateSpaceInBufferPool(const BufferPool &buffer_pool,
const std::vector<NodePtr> &calc_nodes_in_pool,
std::unordered_map<NodePtr, std::vector<BufferPoolNodeItem>> &buffer_pool_nodes_item);
Status AllocateSpaceForBufferPoolNode(int64_t &next_start,
const BufferPool buffer_pool,
BufferPoolNodeItem &buffer_pool_node_item,
std::queue<BufferPoolNodeItem> &node_mem_range_in_pool);
NodePtr GetOffsetAndDependency(int64_t &next_start,
int64_t total_mem_size,
int64_t buffer_pool_size,
const std::unordered_map<NodePtr, NodePtr> &buffer_node_to_calc,
std::queue<BufferPoolNodeItem> &nodes_in_buffer);
Status FixTheTimingOfDependentNodes(NodePtr &dependent_calc_node, NodePtr &curr_pool_node);
uint32_t GenerateEventId(const std::string &node_name, std::queue<std::pair<std::string, uint32_t>> &event_queue);
Status SetResultOfMemoryAndEvent();
// Use map to ensure that each visit is in the order of batch label and pool id
std::map<std::string, std::map<int64_t, std::vector<NodePtr>>> calc_nodes_;
std::unordered_map<std::string, std::unordered_map<NodePtr, NodePtr>> buffer_node_to_calc_;
std::unordered_map<std::string, std::unordered_map<NodePtr, std::vector<BufferPoolNodeItem>>> peer_buffer_node_item_;
std::unordered_map<std::string, std::unordered_map<int64_t, int64_t>> buffer_pool_size_;
uint32_t logic_event_num_;
std::queue<std::pair<std::string, uint32_t>> mem_ctrl_event_;
std::queue<std::pair<std::string, uint32_t>> stream_ctrl_event_;
std::unordered_map<NodePtr, std::vector<std::string>> node_event_multiplexing_;
std::unordered_map<NodePtr, std::vector<int64_t>> buffer_node_logical_offset_;
};
} // namespace ge
#endif // GE_GRAPH_PASSES_BUFFER_POOL_MEMORY_PASS_H_