thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

Added functional optensorsyncDevice and optensorsynclocal

Browse source
This commit is contained in:
Alejandro Saucedo 2020-09-06 15:51:31 +01:00
parent e6f4097acb
commit e68d09dbdc
15 changed files with 258 additions and 50 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/OpTensorSyncLocal.cpp
View file

@ -74,10 +74,17 @@ OpTensorSyncLocal::record()
}
void
OpTensorSyncLocal::postSubmit()
OpTensorSyncLocal::preEval()
{
SPDLOG_DEBUG("Kompute OpTensorSyncLocal postSubmit called");
SPDLOG_DEBUG("Kompute OpTensorSyncLocal preEval called");
}
void
OpTensorSyncLocal::postEval()
{
SPDLOG_DEBUG("Kompute OpTensorSyncLocal postEval called");
SPDLOG_DEBUG("Kompute OpTensorSyncLocal mapping data into tensor local");
for (size_t i = 0; i < this->mTensors.size(); i++) {
if (this->mTensors[i]->tensorType() == Tensor::TensorTypes::eDevice) {
this->mStagingTensors[i]->mapDataFromHostMemory();
@ -86,10 +93,6 @@ OpTensorSyncLocal::postSubmit()
this->mTensors[i]->mapDataFromHostMemory();
}
}
// Remove all staging tensors as they are not required after operation
SPDLOG_DEBUG("Kompute OpTensorSyncLocal destroying staging tensors");
this->mStagingTensors.clear();
}
}