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

Jump to bottom

add new field for event node [cherry-pick] #44245

Merged
merged 1 commit into from
Jul 12, 2022
Merged

add new field for event node [cherry-pick] #44245

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
327 changes: 245 additions & 82 deletions paddle/fluid/platform/profiler/chrometracing_logger.cc
View file
Open in desktop

Large diffs are not rendered by default.

1 change: 1 addition & 0 deletions paddle/fluid/platform/profiler/chrometracing_logger.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@ class ChromeTracingLogger : public BaseLogger {
void LogRuntimeTraceEventNode(const CudaRuntimeTraceEventNode&) override;
void LogNodeTrees(const NodeTrees&) override;
void LogMetaInfo(const std::unordered_map<std::string, std::string>);
void LogMemTraceEventNode(const MemTraceEventNode&) override;

private:
void OpenFile();
Expand Down
199 changes: 138 additions & 61 deletions paddle/fluid/platform/profiler/dump/test_serialization_logger.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,61 +19,102 @@
#include "paddle/fluid/platform/profiler/event_node.h"
#include "paddle/fluid/platform/profiler/event_python.h"

using paddle::platform::SerializationLogger;
using paddle::platform::DeserializationReader;
using paddle::platform::NodeTrees;
using paddle::platform::HostTraceEventNode;
using paddle::platform::CudaRuntimeTraceEventNode;
using paddle::platform::DeserializationReader;
using paddle::platform::DeviceTraceEvent;
using paddle::platform::DeviceTraceEventNode;
using paddle::platform::HostTraceEvent;
using paddle::platform::RuntimeTraceEvent;
using paddle::platform::DeviceTraceEvent;
using paddle::platform::TracerEventType;
using paddle::platform::HostTraceEventNode;
using paddle::platform::KernelEventInfo;
using paddle::platform::MemcpyEventInfo;
using paddle::platform::MemsetEventInfo;
using paddle::platform::MemTraceEvent;
using paddle::platform::NodeTrees;
using paddle::platform::OperatorSupplementEvent;
using paddle::platform::ProfilerResult;
using paddle::platform::RuntimeTraceEvent;
using paddle::platform::SerializationLogger;
using paddle::platform::TracerEventType;

TEST(SerializationLoggerTest, dump_case0) {
std::list<HostTraceEvent> host_events;
std::list<RuntimeTraceEvent> runtime_events;
std::list<DeviceTraceEvent> device_events;
std::list<MemTraceEvent> mem_events;
std::list<OperatorSupplementEvent> op_supplement_events;
host_events.push_back(HostTraceEvent(std::string("dataloader#1"),
TracerEventType::Dataloader, 1000, 10000,
10, 10));
TracerEventType::Dataloader,
1000,
10000,
10,
10));
host_events.push_back(HostTraceEvent(
std::string("op1"), TracerEventType::Operator, 11000, 20000, 10, 10));
host_events.push_back(HostTraceEvent(
std::string("op2"), TracerEventType::Operator, 21000, 30000, 10, 10));
host_events.push_back(HostTraceEvent(
std::string("op3"), TracerEventType::Operator, 31000, 40000, 10, 11));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch1"), 15000,
17000, 10, 10, 1, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch2"), 25000,
35000, 10, 10, 2, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch3"), 33000,
37000, 10, 11, 3, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudaMemcpy1"), 18000,
19000, 10, 10, 4, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudaMemset1"), 38000,
39000, 10, 11, 5, 0));
device_events.push_back(
DeviceTraceEvent(std::string("kernel1"), TracerEventType::Kernel, 40000,
55000, 0, 10, 10, 1, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("kernel2"), TracerEventType::Kernel, 70000,
95000, 0, 10, 10, 2, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("kernel3"), TracerEventType::Kernel, 60000,
65000, 0, 10, 11, 3, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("memcpy1"), TracerEventType::Memcpy, 56000,
59000, 0, 10, 10, 4, MemcpyEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("memset1"), TracerEventType::Memset, 66000,
69000, 0, 10, 11, 5, MemsetEventInfo()));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch1"), 15000, 17000, 10, 10, 1, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch2"), 25000, 35000, 10, 10, 2, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch3"), 33000, 37000, 10, 11, 3, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudaMemcpy1"), 18000, 19000, 10, 10, 4, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudaMemset1"), 38000, 39000, 10, 11, 5, 0));
device_events.push_back(DeviceTraceEvent(std::string("kernel1"),
TracerEventType::Kernel,
40000,
55000,
0,
10,
10,
1,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("kernel2"),
TracerEventType::Kernel,
70000,
95000,
0,
10,
10,
2,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("kernel3"),
TracerEventType::Kernel,
60000,
65000,
0,
10,
11,
3,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("memcpy1"),
TracerEventType::Memcpy,
56000,
59000,
0,
10,
10,
4,
MemcpyEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("memset1"),
TracerEventType::Memset,
66000,
69000,
0,
10,
11,
5,
MemsetEventInfo()));
SerializationLogger logger("test_serialization_logger_case0.pb");
NodeTrees tree(host_events, runtime_events, device_events);
NodeTrees tree(host_events,
runtime_events,
device_events,
mem_events,
op_supplement_events);
std::map<uint64_t, std::vector<HostTraceEventNode*>> nodes =
tree.Traverse(true);
EXPECT_EQ(nodes[10].size(), 4u);
Expand Down Expand Up @@ -102,33 +143,69 @@ TEST(SerializationLoggerTest, dump_case1) {
std::list<HostTraceEvent> host_events;
std::list<RuntimeTraceEvent> runtime_events;
std::list<DeviceTraceEvent> device_events;
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch1"), 15000,
17000, 10, 10, 1, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch2"), 25000,
35000, 10, 10, 2, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudalaunch3"), 33000,
37000, 10, 11, 3, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudaMemcpy1"), 18000,
19000, 10, 10, 4, 0));
runtime_events.push_back(RuntimeTraceEvent(std::string("cudaMemset1"), 38000,
39000, 10, 11, 5, 0));
device_events.push_back(
DeviceTraceEvent(std::string("kernel1"), TracerEventType::Kernel, 40000,
55000, 0, 10, 10, 1, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("kernel2"), TracerEventType::Kernel, 70000,
95000, 0, 10, 10, 2, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("kernel3"), TracerEventType::Kernel, 60000,
65000, 0, 10, 11, 3, KernelEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("memcpy1"), TracerEventType::Memcpy, 56000,
59000, 0, 10, 10, 4, MemcpyEventInfo()));
device_events.push_back(
DeviceTraceEvent(std::string("memset1"), TracerEventType::Memset, 66000,
69000, 0, 10, 11, 5, MemsetEventInfo()));
std::list<MemTraceEvent> mem_events;
std::list<OperatorSupplementEvent> op_supplement_events;
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch1"), 15000, 17000, 10, 10, 1, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch2"), 25000, 35000, 10, 10, 2, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudalaunch3"), 33000, 37000, 10, 11, 3, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudaMemcpy1"), 18000, 19000, 10, 10, 4, 0));
runtime_events.push_back(RuntimeTraceEvent(
std::string("cudaMemset1"), 38000, 39000, 10, 11, 5, 0));
device_events.push_back(DeviceTraceEvent(std::string("kernel1"),
TracerEventType::Kernel,
40000,
55000,
0,
10,
10,
1,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("kernel2"),
TracerEventType::Kernel,
70000,
95000,
0,
10,
10,
2,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("kernel3"),
TracerEventType::Kernel,
60000,
65000,
0,
10,
11,
3,
KernelEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("memcpy1"),
TracerEventType::Memcpy,
56000,
59000,
0,
10,
10,
4,
MemcpyEventInfo()));
device_events.push_back(DeviceTraceEvent(std::string("memset1"),
TracerEventType::Memset,
66000,
69000,
0,
10,
11,
5,
MemsetEventInfo()));
SerializationLogger logger("test_serialization_logger_case1.pb");
NodeTrees tree(host_events, runtime_events, device_events);
NodeTrees tree(host_events,
runtime_events,
device_events,
mem_events,
op_supplement_events);
std::map<uint64_t, std::vector<HostTraceEventNode*>> nodes =
tree.Traverse(true);
EXPECT_EQ(nodes[10].size(), 1u);
Expand Down