Calculate per-thread source location counts.

This commit is contained in:
Bartosz Taudul 2023-12-20 16:42:37 +01:00
parent e9911ab127
commit d01113150d
No known key found for this signature in database
GPG Key ID: B7FE2008B7575DF3
2 changed files with 30 additions and 1 deletions

View File

@ -4898,9 +4898,11 @@ void Worker::ProcessZoneEnd( const QueueZoneEnd& ev )
const auto timeSpan = timeEnd - zone->Start(); const auto timeSpan = timeEnd - zone->Start();
if( timeSpan > 0 ) if( timeSpan > 0 )
{ {
const auto ctid = CompressThread( td->id );
ZoneThreadData ztd; ZoneThreadData ztd;
ztd.SetZone( zone ); ztd.SetZone( zone );
ztd.SetThread( CompressThread( td->id ) ); ztd.SetThread( ctid );
auto slz = GetSourceLocationZones( zone->SrcLoc() ); auto slz = GetSourceLocationZones( zone->SrcLoc() );
slz->zones.push_back( ztd ); slz->zones.push_back( ztd );
if( slz->min > timeSpan ) slz->min = timeSpan; if( slz->min > timeSpan ) slz->min = timeSpan;
@ -4911,6 +4913,7 @@ void Worker::ProcessZoneEnd( const QueueZoneEnd& ev )
if( slz->selfMin > selfSpan ) slz->selfMin = selfSpan; if( slz->selfMin > selfSpan ) slz->selfMin = selfSpan;
if( slz->selfMax < selfSpan ) slz->selfMax = selfSpan; if( slz->selfMax < selfSpan ) slz->selfMax = selfSpan;
slz->selfTotal += selfSpan; slz->selfTotal += selfSpan;
if( !isReentry ) if( !isReentry )
{ {
slz->nonReentrantCount++; slz->nonReentrantCount++;
@ -4922,6 +4925,16 @@ void Worker::ProcessZoneEnd( const QueueZoneEnd& ev )
{ {
td->childTimeStack.back() += timeSpan; td->childTimeStack.back() += timeSpan;
} }
auto it = slz->threadCnt.find( ctid );
if( it == slz->threadCnt.end() )
{
slz->threadCnt.emplace( ctid, 1 );
}
else
{
it->second++;
}
} }
else else
{ {
@ -7465,15 +7478,18 @@ void Worker::ReconstructZoneStatistics( uint8_t* countMap, ZoneEvent& zone, uint
{ {
auto it = m_data.sourceLocationZones.find( zone.SrcLoc() ); auto it = m_data.sourceLocationZones.find( zone.SrcLoc() );
assert( it != m_data.sourceLocationZones.end() ); assert( it != m_data.sourceLocationZones.end() );
ZoneThreadData ztd; ZoneThreadData ztd;
ztd.SetZone( &zone ); ztd.SetZone( &zone );
ztd.SetThread( thread ); ztd.SetThread( thread );
auto& slz = it->second; auto& slz = it->second;
slz.zones.push_back( ztd ); slz.zones.push_back( ztd );
if( slz.min > timeSpan ) slz.min = timeSpan; if( slz.min > timeSpan ) slz.min = timeSpan;
if( slz.max < timeSpan ) slz.max = timeSpan; if( slz.max < timeSpan ) slz.max = timeSpan;
slz.total += timeSpan; slz.total += timeSpan;
slz.sumSq += double( timeSpan ) * timeSpan; slz.sumSq += double( timeSpan ) * timeSpan;
if( countMap[uint16_t(zone.SrcLoc())] == 0 ) if( countMap[uint16_t(zone.SrcLoc())] == 0 )
{ {
slz.nonReentrantCount++; slz.nonReentrantCount++;
@ -7481,6 +7497,7 @@ void Worker::ReconstructZoneStatistics( uint8_t* countMap, ZoneEvent& zone, uint
if( slz.nonReentrantMax < timeSpan ) slz.nonReentrantMax = timeSpan; if( slz.nonReentrantMax < timeSpan ) slz.nonReentrantMax = timeSpan;
slz.nonReentrantTotal += timeSpan; slz.nonReentrantTotal += timeSpan;
} }
if( zone.HasChildren() ) if( zone.HasChildren() )
{ {
auto& children = GetZoneChildren( zone.Child() ); auto& children = GetZoneChildren( zone.Child() );
@ -7492,9 +7509,20 @@ void Worker::ReconstructZoneStatistics( uint8_t* countMap, ZoneEvent& zone, uint
timeSpan -= childSpan; timeSpan -= childSpan;
} }
} }
if( slz.selfMin > timeSpan ) slz.selfMin = timeSpan; if( slz.selfMin > timeSpan ) slz.selfMin = timeSpan;
if( slz.selfMax < timeSpan ) slz.selfMax = timeSpan; if( slz.selfMax < timeSpan ) slz.selfMax = timeSpan;
slz.selfTotal += timeSpan; slz.selfTotal += timeSpan;
auto tit = slz.threadCnt.find( thread );
if( tit == slz.threadCnt.end() )
{
slz.threadCnt.emplace( thread, 1 );
}
else
{
tit->second++;
}
} }
} }

View File

@ -205,6 +205,7 @@ private:
int64_t nonReentrantMin = std::numeric_limits<int64_t>::max(); int64_t nonReentrantMin = std::numeric_limits<int64_t>::max();
int64_t nonReentrantMax = std::numeric_limits<int64_t>::min(); int64_t nonReentrantMax = std::numeric_limits<int64_t>::min();
int64_t nonReentrantTotal = 0; int64_t nonReentrantTotal = 0;
unordered_flat_map<uint16_t, uint64_t> threadCnt;
}; };
struct GpuSourceLocationZones struct GpuSourceLocationZones