mirror/tracy
1
0
Fork 0
mirror of https://github.com/wolfpld/tracy.git synced 2024-09-20 05:42:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use big allocation mode for Vector's reserve_exact.

Browse Source
This commit is contained in:
Bartosz Taudul 2019-02-15 01:58:23 +01:00
parent 930190f2cb
commit 7b023e533d
2 changed files with 25 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
server/TracyVector.hpp
View File

@ -9,6 +9,7 @@
#include "../common/TracyForceInline.hpp" #include "../common/TracyForceInline.hpp"
#include "TracyMemory.hpp" #include "TracyMemory.hpp"
#include "TracyPopcnt.hpp" #include "TracyPopcnt.hpp"
#include "TracySlab.hpp"
namespace tracy namespace tracy
{ {
@ -46,15 +47,11 @@ public:
~Vector() ~Vector()
{ {
if( m_capacity == std::numeric_limits<uint8_t>::max() ) if( m_capacity != std::numeric_limits<uint8_t>::max() )
{
memUsage.fetch_sub( m_size * sizeof( T ) );
}
else
{ {
memUsage.fetch_sub( Capacity() * sizeof( T ), std::memory_order_relaxed ); memUsage.fetch_sub( Capacity() * sizeof( T ), std::memory_order_relaxed );
delete[] m_ptr;
} }
delete[] m_ptr;
} }
Vector& operator=( const Vector& ) = delete; Vector& operator=( const Vector& ) = delete;
@ -234,13 +231,13 @@ public:
m_size = sz; m_size = sz;
} }
tracy_force_inline void reserve_exact( uint32_t sz ) template<size_t U>
tracy_force_inline void reserve_exact( uint32_t sz, Slab<U>& slab )
{ {
assert( !m_ptr ); assert( !m_ptr );
m_capacity = std::numeric_limits<uint8_t>::max(); m_capacity = std::numeric_limits<uint8_t>::max();
m_size = sz; m_size = sz;
m_ptr = new T[sz]; m_ptr = (T*)slab.AllocBig( sizeof( T ) * sz );
memUsage.fetch_add( sz * sizeof( T ) );
} }
tracy_force_inline void clear() tracy_force_inline void clear()

38
server/TracyWorker.cpp
View File

@ -344,7 +344,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
if( fileVer >= FileVersion( 0, 3, 202 ) ) if( fileVer >= FileVersion( 0, 3, 202 ) )
{ {
f.Read( sz ); f.Read( sz );
m_data.frames.Data().reserve_exact( sz ); m_data.frames.Data().reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
auto ptr = m_slab.AllocInit<FrameData>(); auto ptr = m_slab.AllocInit<FrameData>();
@ -352,7 +352,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
f.Read( &ptr->continuous, sizeof( ptr->continuous ) ); f.Read( &ptr->continuous, sizeof( ptr->continuous ) );
uint64_t fsz; uint64_t fsz;
f.Read( &fsz, sizeof( fsz ) ); f.Read( &fsz, sizeof( fsz ) );
ptr->frames.reserve_exact( fsz ); ptr->frames.reserve_exact( fsz, m_slab );
if( fileVer >= FileVersion( 0, 4, 2 ) ) if( fileVer >= FileVersion( 0, 4, 2 ) )
{ {
int64_t refTime = 0; int64_t refTime = 0;
@ -400,7 +400,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
ptr->name = 0; ptr->name = 0;
ptr->continuous = 1; ptr->continuous = 1;
f.Read( sz ); f.Read( sz );
ptr->frames.reserve_exact( sz ); ptr->frames.reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
f.Read( &ptr->frames[i].start, sizeof( int64_t ) ); f.Read( &ptr->frames[i].start, sizeof( int64_t ) );
@ -413,7 +413,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
flat_hash_map<uint64_t, const char*, nohash<uint64_t>> pointerMap; flat_hash_map<uint64_t, const char*, nohash<uint64_t>> pointerMap;
f.Read( sz ); f.Read( sz );
m_data.stringData.reserve_exact( sz ); m_data.stringData.reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
uint64_t ptr, ssz; uint64_t ptr, ssz;
@ -466,12 +466,12 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
} }
f.Read( sz ); f.Read( sz );
m_data.sourceLocationExpand.reserve_exact( sz ); m_data.sourceLocationExpand.reserve_exact( sz, m_slab );
f.Read( m_data.sourceLocationExpand.data(), sizeof( uint64_t ) * sz ); f.Read( m_data.sourceLocationExpand.data(), sizeof( uint64_t ) * sz );
const auto sle = sz; const auto sle = sz;
f.Read( sz ); f.Read( sz );
m_data.sourceLocationPayload.reserve_exact( sz ); m_data.sourceLocationPayload.reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
auto srcloc = m_slab.Alloc<SourceLocation>(); auto srcloc = m_slab.Alloc<SourceLocation>();
@ -566,7 +566,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
lockmap.threadList.emplace_back( t ); lockmap.threadList.emplace_back( t );
} }
f.Read( tsz ); f.Read( tsz );
lockmap.timeline.reserve_exact( tsz ); lockmap.timeline.reserve_exact( tsz, m_slab );
auto ptr = lockmap.timeline.data(); auto ptr = lockmap.timeline.data();
if( fileVer >= FileVersion( 0, 4, 2 ) ) if( fileVer >= FileVersion( 0, 4, 2 ) )
{ {
@ -677,7 +677,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
f.Read( sz ); f.Read( sz );
if( eventMask & EventType::Messages ) if( eventMask & EventType::Messages )
{ {
m_data.messages.reserve_exact( sz ); m_data.messages.reserve_exact( sz, m_slab );
if( fileVer >= FileVersion( 0, 4, 2 ) ) if( fileVer >= FileVersion( 0, 4, 2 ) )
{ {
int64_t refTime = 0; int64_t refTime = 0;
@ -721,7 +721,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
s_loadProgress.progress.store( LoadProgress::Zones, std::memory_order_relaxed ); s_loadProgress.progress.store( LoadProgress::Zones, std::memory_order_relaxed );
f.Read( sz ); f.Read( sz );
m_data.threads.reserve_exact( sz ); m_data.threads.reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
auto td = m_slab.AllocInit<ThreadData>(); auto td = m_slab.AllocInit<ThreadData>();
@ -749,7 +749,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
f.Read( msz ); f.Read( msz );
if( eventMask & EventType::Messages ) if( eventMask & EventType::Messages )
{ {
td->messages.reserve_exact( msz ); td->messages.reserve_exact( msz, m_slab );
for( uint64_t j=0; j<msz; j++ ) for( uint64_t j=0; j<msz; j++ )
{ {
uint64_t ptr; uint64_t ptr;
@ -768,7 +768,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
s_loadProgress.progress.store( LoadProgress::GpuZones, std::memory_order_relaxed ); s_loadProgress.progress.store( LoadProgress::GpuZones, std::memory_order_relaxed );
f.Read( sz ); f.Read( sz );
m_data.gpuData.reserve_exact( sz ); m_data.gpuData.reserve_exact( sz, m_slab );
for( uint64_t i=0; i<sz; i++ ) for( uint64_t i=0; i<sz; i++ )
{ {
auto ctx = m_slab.AllocInit<GpuCtxData>(); auto ctx = m_slab.AllocInit<GpuCtxData>();
@ -825,7 +825,7 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
f.Read( pd->max ); f.Read( pd->max );
uint64_t psz; uint64_t psz;
f.Read( psz ); f.Read( psz );
pd->data.reserve_exact( psz ); pd->data.reserve_exact( psz, m_slab );
if( fileVer >= FileVersion( 0, 4, 2 ) ) if( fileVer >= FileVersion( 0, 4, 2 ) )
{ {
int64_t refTime = 0; int64_t refTime = 0;
@ -863,13 +863,13 @@ Worker::Worker( FileRead& f, EventType::Type eventMask )
f.Read( sz ); f.Read( sz );
if( eventMask & EventType::Memory ) if( eventMask & EventType::Memory )
{ {
m_data.memory.data.reserve_exact( sz ); m_data.memory.data.reserve_exact( sz, m_slab );
if( fileVer >= FileVersion( 0, 3, 201 ) ) if( fileVer >= FileVersion( 0, 3, 201 ) )
{ {
uint64_t activeSz, freesSz; uint64_t activeSz, freesSz;
f.Read2( activeSz, freesSz ); f.Read2( activeSz, freesSz );
m_data.memory.active.reserve( activeSz ); m_data.memory.active.reserve( activeSz );
m_data.memory.frees.reserve_exact( freesSz ); m_data.memory.frees.reserve_exact( freesSz, m_slab );
} }
auto mem = m_data.memory.data.data(); auto mem = m_data.memory.data.data();
s_loadProgress.subTotal.store( sz, std::memory_order_relaxed ); s_loadProgress.subTotal.store( sz, std::memory_order_relaxed );
@ -3124,7 +3124,7 @@ void Worker::ReconstructMemAllocPlot()
plot->name = 0; plot->name = 0;
plot->type = PlotType::Memory; plot->type = PlotType::Memory;
plot->data.reserve_exact( psz ); plot->data.reserve_exact( psz, m_slab );
auto aptr = mem.data.begin(); auto aptr = mem.data.begin();
auto aend = mem.data.end(); auto aend = mem.data.end();
@ -3322,7 +3322,7 @@ void Worker::ReadTimelineUpdateStatistics( ZoneEvent* zone, uint16_t thread )
void Worker::ReadTimeline( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t thread, uint64_t size, int64_t& refTime ) void Worker::ReadTimeline( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t thread, uint64_t size, int64_t& refTime )
{ {
assert( size != 0 ); assert( size != 0 );
vec.reserve_exact( size ); vec.reserve_exact( size, m_slab );
m_data.zonesCnt += size; m_data.zonesCnt += size;
for( uint64_t i=0; i<size; i++ ) for( uint64_t i=0; i<size; i++ )
@ -3341,7 +3341,7 @@ void Worker::ReadTimeline( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t thread
void Worker::ReadTimelinePre042( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t thread, uint64_t size, int fileVer ) void Worker::ReadTimelinePre042( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t thread, uint64_t size, int fileVer )
{ {
assert( size != 0 ); assert( size != 0 );
vec.reserve_exact( size ); vec.reserve_exact( size, m_slab );
m_data.zonesCnt += size; m_data.zonesCnt += size;
for( uint64_t i=0; i<size; i++ ) for( uint64_t i=0; i<size; i++ )
@ -3374,7 +3374,7 @@ void Worker::ReadTimelinePre042( FileRead& f, Vector<ZoneEvent*>& vec, uint16_t
void Worker::ReadTimeline( FileRead& f, Vector<GpuEvent*>& vec, uint64_t size, int64_t& refTime, int64_t& refGpuTime ) void Worker::ReadTimeline( FileRead& f, Vector<GpuEvent*>& vec, uint64_t size, int64_t& refTime, int64_t& refGpuTime )
{ {
assert( size != 0 ); assert( size != 0 );
vec.reserve_exact( size ); vec.reserve_exact( size, m_slab );
for( uint64_t i=0; i<size; i++ ) for( uint64_t i=0; i<size; i++ )
{ {
@ -3406,7 +3406,7 @@ void Worker::ReadTimeline( FileRead& f, Vector<GpuEvent*>& vec, uint64_t size, i
void Worker::ReadTimelinePre042( FileRead& f, Vector<GpuEvent*>& vec, uint64_t size, int fileVer ) void Worker::ReadTimelinePre042( FileRead& f, Vector<GpuEvent*>& vec, uint64_t size, int fileVer )
{ {
assert( size != 0 ); assert( size != 0 );
vec.reserve_exact( size ); vec.reserve_exact( size, m_slab );
for( uint64_t i=0; i<size; i++ ) for( uint64_t i=0; i<size; i++ )
{ {