tvm
Public Member Functions | Public Attributes | Static Public Attributes | List of all members
tvm::runtime::profiling::ReportNode Class Reference

Data collected from a profiling run. Includes per-call metrics and per-device metrics. More...

#include <profiling.h>

Inheritance diagram for tvm::runtime::profiling::ReportNode:
Collaboration diagram for tvm::runtime::profiling::ReportNode:

Public Member Functions

String AsCSV () const
 Output calls in CSV format. More...
 
String AsTable (bool sort=true, bool aggregate=true, bool compute_col_sums=true) const
 Create a human readable table of profiling metrics. More...
 
String AsJSON () const
 Convert this report to JSON. More...
 
 TVM_DECLARE_FINAL_OBJECT_INFO (ReportNode, Object)
 
- Public Member Functions inherited from tvm::runtime::Object
uint32_t type_index () const
 
std::string GetTypeKey () const
 
size_t GetTypeKeyHash () const
 
template<typename TargetType >
bool IsInstance () const
 
bool unique () const
 
 Object ()
 
 Object (const Object &other)
 
 Object (Object &&other)
 
Objectoperator= (const Object &other)
 
Objectoperator= (Object &&other)
 

Public Attributes

Array< Map< String, ObjectRef > > calls
 A list of function calls and the metrics recorded for that call. More...
 
Map< String, Map< String, ObjectRef > > device_metrics
 Metrics collected for the entire run of the model on a per-device basis. More...
 
Map< String, ObjectRefconfiguration
 

Static Public Attributes

static constexpr const char * _type_key = "runtime.profiling.Report"
 
- Static Public Attributes inherited from tvm::runtime::Object
static constexpr const char * _type_key = "runtime.Object"
 
static constexpr bool _type_final = false
 
static constexpr uint32_t _type_child_slots = 0
 
static constexpr bool _type_child_slots_can_overflow = true
 
static constexpr bool _type_has_method_visit_attrs = true
 
static constexpr bool _type_has_method_sequal_reduce = false
 
static constexpr bool _type_has_method_shash_reduce = false
 
static constexpr uint32_t _type_index = TypeIndex::kDynamic
 

Additional Inherited Members

- Public Types inherited from tvm::runtime::Object
typedef void(* FDeleter) (Object *self)
 Object deleter. More...
 
using RefCounterType = std::atomic< int32_t >
 
- Static Public Member Functions inherited from tvm::runtime::Object
static std::string TypeIndex2Key (uint32_t tindex)
 Get the type key of the corresponding index from runtime. More...
 
static size_t TypeIndex2KeyHash (uint32_t tindex)
 Get the type key hash of the corresponding index from runtime. More...
 
static uint32_t TypeKey2Index (const std::string &key)
 Get the type index of the corresponding key from runtime. More...
 
static uint32_t _GetOrAllocRuntimeTypeIndex ()
 
static uint32_t RuntimeTypeIndex ()
 
- Protected Member Functions inherited from tvm::runtime::Object
void IncRef ()
 developer function, increases reference counter. More...
 
void DecRef ()
 developer function, decrease reference counter. More...
 
- Static Protected Member Functions inherited from tvm::runtime::Object
static uint32_t GetOrAllocRuntimeTypeIndex (const std::string &key, uint32_t static_tindex, uint32_t parent_tindex, uint32_t type_child_slots, bool type_child_slots_can_overflow)
 Get the type index using type key. More...
 
- Protected Attributes inherited from tvm::runtime::Object
uint32_t type_index_ {0}
 Type index(tag) that indicates the type of the object. More...
 
RefCounterType ref_counter_ {0}
 The internal reference counter. More...
 
FDeleter deleter_ = nullptr
 deleter of this object to enable customized allocation. If the deleter is nullptr, no deletion will be performed. The creator of the object must always set the deleter field properly. More...
 

Detailed Description

Data collected from a profiling run. Includes per-call metrics and per-device metrics.

Member Function Documentation

◆ AsCSV()

String tvm::runtime::profiling::ReportNode::AsCSV ( ) const

Output calls in CSV format.

Note that this does not include device_metrics, it only includes per-call metrics.

◆ AsJSON()

String tvm::runtime::profiling::ReportNode::AsJSON ( ) const

Convert this report to JSON.

Output JSON will be of this format:

{
"calls": [
{
"Duration (us)": {
"microseconds": 12.3
},
"Name": "fused_dense",
"Count": {
"count": 1
},
"Percent": {
"percent": 10.3
}
}
],
"device_metrics": {
"cpu": {
"Duration (us)": {
"microseconds": 334.2
},
"Percent": {
"percent": 100
}
}
}
}

◆ AsTable()

String tvm::runtime::profiling::ReportNode::AsTable ( bool  sort = true,
bool  aggregate = true,
bool  compute_col_sums = true 
) const

Create a human readable table of profiling metrics.

Parameters
aggregateWhether or not to join multiple calls to the same op into a single line.
sortWhether or not to sort call frames by descending duration. If false and if aggregate is false, frames will be sorted by order of appearance in the program. Order is undefined if sort is false and aggregate is true.
compute_col_sumsWhether or not to include sum totals for the Count, Duation, and Percent columns.

◆ TVM_DECLARE_FINAL_OBJECT_INFO()

tvm::runtime::profiling::ReportNode::TVM_DECLARE_FINAL_OBJECT_INFO ( ReportNode  ,
Object   
)

Member Data Documentation

◆ _type_key

constexpr const char* tvm::runtime::profiling::ReportNode::_type_key = "runtime.profiling.Report"
staticconstexpr

◆ calls

Array<Map<String, ObjectRef> > tvm::runtime::profiling::ReportNode::calls

A list of function calls and the metrics recorded for that call.

Each element is a mapping from metric name to value. Some metrics that appear in every call are "Name" (the function name), "Argument Shapes", and "Duration (us)". Values are one of String, PercentNode, DurationNode, or CountNode.

◆ configuration

Map<String, ObjectRef> tvm::runtime::profiling::ReportNode::configuration

Configuration used for this profiling run. Includes number of threads, executor.

Values must be an object type that can be used with device_metrics.

◆ device_metrics

Map<String, Map<String, ObjectRef> > tvm::runtime::profiling::ReportNode::device_metrics

Metrics collected for the entire run of the model on a per-device basis.

device_metrics is indexed by device name then metric.

These metrics may be larger than the sum of the same metric in calls because these metrics include the overhead of the executor.


The documentation for this class was generated from the following file: