AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-378008: Adding changes to serialize the kernels in rocprofV2

Browse Source 
Change-Id: I82353ba94b3a15fdc5991e6129fe47f6765a9f74


[ROCm/rocprofiler commit: 54f6e2afb7]
This commit is contained in:
Sriraksha Nagaraj
2023-03-27 09:39:18 -05:00
committed by Ammar Elwazir
parent 00962f5862
commit a157bb93b7
8 changed files with 345 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler/src/api/rocprofiler_singleton.cpp
+2 -3
View File
@@ -1,3 +1,4 @@
/* Copyright (c) 2022 Advanced Micro Devices, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy
@@ -135,8 +136,6 @@ rocprofiler_session_id_t ROCProfiler_Singleton::CreateSession(
}
void ROCProfiler_Singleton::DestroySession(rocprofiler_session_id_t session_id) {
while (GetCurrentActiveInterruptSignalsCount() != 0) {
}
{
std::lock_guard<std::mutex> lock(session_map_lock_);
ASSERTM(sessions_.find(session_id.handle) != sessions_.end(),
@@ -145,7 +144,7 @@ void ROCProfiler_Singleton::DestroySession(rocprofiler_session_id_t session_id)
sessions_.erase(session_id.handle);
}
}
profiler_serializer_t& ROCProfiler_Singleton::GetSerializer() { return profiler_serializer; }
bool ROCProfiler_Singleton::FindDeviceProfilingSession(rocprofiler_session_id_t session_id) {
std::lock_guard<std::mutex> lock(device_profiling_session_map_lock_);
return dev_profiling_sessions_.find(session_id.handle) != dev_profiling_sessions_.end();
projects/rocprofiler/src/api/rocprofiler_singleton.h
+23 -3
View File
@@ -42,8 +42,28 @@
#include "src/core/session/session.h"
#include "src/core/session/device_profiling.h"
#include "src/core/hardware/hsa_info.h"
#include "src/core/hsa/queues/queue.h"
namespace rocprofiler {
/*This is a profiler serializer. It should be instantiated
only once for the profiler. The following is the
description of each field.
1. dispatch_queue - The queue to which the currently dispatched kernel
belongs to.
At any given time, in serialization only one kernel
can be executing.
2. dispatch_ready- It is a software data structure which holds
the queues which have a kernel ready to be dispatched.
This stores the queues in FIFO order.
3. serializer_mutex - The mutex is used for thread synchronization
while accessing the singleton instance of this structure.
Currently, in case of profiling kernels are serialized by default.
*/
struct profiler_serializer_t {
queue::Queue* dispatch_queue{nullptr};
std::vector<rocprofiler::queue::Queue*> dispatch_ready;
std::mutex serializer_mutex;
};
class ROCProfiler_Singleton {
public:
@@ -79,7 +99,7 @@ class ROCProfiler_Singleton {
int cpu_agent_index, int gpu_agent_index);
void DestroyDeviceProfilingSession(rocprofiler_session_id_t session_id);
DeviceProfileSession* GetDeviceProfilingSession(rocprofiler_session_id_t session_id);
profiler_serializer_t& GetSerializer();
// Generic
bool CheckFilterData(rocprofiler_filter_kind_t filter_kind,
@@ -99,7 +119,7 @@ class ROCProfiler_Singleton {
std::unordered_map<uint64_t, Agent::DeviceInfo> agent_device_map_;
ROCProfiler_Singleton();
~ROCProfiler_Singleton();
profiler_serializer_t profiler_serializer;
/*
* XXX: Associating PC samples with a running kernel requires an identifier
* that will be unique across all kernel executions. It is not enough to use