-
Notifications
You must be signed in to change notification settings - Fork 46
/
gpa_counter_scheduler_cl.h
29 lines (24 loc) · 1.2 KB
/
gpa_counter_scheduler_cl.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
//==============================================================================
// Copyright (c) 2016-2021 Advanced Micro Devices, Inc. All rights reserved.
/// @author AMD Developer Tools Team
/// @file
/// @brief Class for counter scheduling for CL.
//==============================================================================
#ifndef GPU_PERF_API_COUNTER_GENERATOR_CL_GPA_COUNTER_SCHEDULER_CL_H_
#define GPU_PERF_API_COUNTER_GENERATOR_CL_GPA_COUNTER_SCHEDULER_CL_H_
#include "gpu_perf_api_counter_generator/gpa_counter_scheduler_base.h"
/// @brief Class for counter scheduling for CL.
class GpaCounterSchedulerCl : public GpaCounterSchedulerBase
{
public:
/// @brief Constructor.
GpaCounterSchedulerCl();
protected:
/// @copydoc GpaCounterSchedulerBase::GetPreferredSplittingAlgorithm()
///
/// For CL, the preferred splitting algorithm is the consolidated one.
/// This keeps the overall number of passes down to a reasonable number, but splits the counters up
/// in a manner that allows them to be more consistent.
virtual GpaCounterSplitterAlgorithm GetPreferredSplittingAlgorithm() const override;
};
#endif // GPU_PERF_API_COUNTER_GENERATOR_CL_GPA_COUNTER_SCHEDULER_CL_H_