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

Updated access masks for the pipeline barriers

Browse source
This commit is contained in:
Alejandro Saucedo 2021-03-09 08:52:43 +00:00
parent 263f392cbb
commit 0ea3a79123
2 changed files with 8 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

12
src/OpTensorSyncLocal.cpp
View file

@ -31,13 +31,13 @@ OpTensorSyncLocal::record(const vk::CommandBuffer& commandBuffer)
for (size_t i = 0; i < this->mTensors.size(); i++) {
if (this->mTensors[i]->tensorType() == Tensor::TensorTypes::eDevice) {
this->mTensors[i]->recordCopyFromDeviceToStaging(commandBuffer);
this->mTensors[i]->recordPrimaryBufferMemoryBarrier(commandBuffer,
vk::AccessFlagBits::eShaderWrite,
vk::AccessFlagBits::eTransferRead,
vk::PipelineStageFlagBits::eComputeShader,
vk::PipelineStageFlagBits::eTransfer);
this->mTensors[i]->recordStagingBufferMemoryBarrier(commandBuffer,
vk::AccessFlagBits::eTransferWrite,
vk::AccessFlagBits::eHostRead,
vk::PipelineStageFlagBits::eTransfer,
vk::PipelineStageFlagBits::eHost);
this->mTensors[i]->recordCopyFromDeviceToStaging(commandBuffer);
}
}
}