intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-29 00:58:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

Dont use Packets without profiling data to calculate kernel duration

Browse Source 
Change-Id: I710348835f8884a3b244502f53ff4e4980441654
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
This commit is contained in:
Bartosz Dunajski
2020-09-29 14:35:23 +02:00
committed by sys_ocldev
parent 77f88f9a96
commit 44a6d70ced
6 changed files with 43 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
opencl/source/command_queue/enqueue_common.h
View File

@@ -519,6 +519,7 @@ void CommandQueueHw<GfxFamily>::processDispatchForCacheFlush(Surface **surfaces,
uint64_t postSyncAddress = 0;
if (getGpgpuCommandStreamReceiver().peekTimestampPacketWriteEnabled()) {
auto timestampPacketNodeForPostSync = timestampPacketContainer->peekNodes().at(0);
timestampPacketNodeForPostSync->setProfilingCapable(false);
postSyncAddress = TimestampPacketHelper::getContextEndGpuAddress(*timestampPacketNodeForPostSync);
}

1
opencl/source/command_queue/hardware_interface_base.inl
View File

@@ -143,6 +143,7 @@ void HardwareInterface<GfxFamily>::dispatchWalker(
uint64_t postSyncAddress = 0;
if (commandQueue.getGpgpuCommandStreamReceiver().peekTimestampPacketWriteEnabled()) {
auto timestampPacketNodeForPostSync = currentTimestampPacketNodes->peekNodes().at(currentDispatchIndex);
timestampPacketNodeForPostSync->setProfilingCapable(false);
postSyncAddress = TimestampPacketHelper::getContextEndGpuAddress(*timestampPacketNodeForPostSync);
}
HardwareCommandsHelper<GfxFamily>::programCacheFlushAfterWalkerCommand(commandStream, commandQueue, mainKernel, postSyncAddress);