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

Amended workgroup to std::array

Browse source
This commit is contained in:
Alejandro Saucedo 2021-02-14 07:18:33 +00:00
parent 7a71ff0751
commit c8370e0a3a
5 changed files with 23 additions and 28 deletions
Download patch file Download diff file Expand all files Collapse all files

8
src/OpAlgoLhsRhsOut.cpp
View file

@ -14,7 +14,7 @@ OpAlgoLhsRhsOut::OpAlgoLhsRhsOut(
std::shared_ptr<vk::Device> device,
std::shared_ptr<vk::CommandBuffer> commandBuffer,
std::vector<std::shared_ptr<Tensor>> tensors,
KomputeWorkgroup komputeWorkgroup)
const std::array<uint32_t, 3>& komputeWorkgroup)
// The inheritance is initialised with the copyOutputData to false given that
// this depencendant class handles the transfer of data via staging buffers in
// a granular way.
@ -93,9 +93,9 @@ OpAlgoLhsRhsOut::record()
vk::PipelineStageFlagBits::eHost,
vk::PipelineStageFlagBits::eComputeShader);
this->mAlgorithm->recordDispatch(this->mKomputeWorkgroup.x,
this->mKomputeWorkgroup.y,
this->mKomputeWorkgroup.z);
this->mAlgorithm->recordDispatch(this->mKomputeWorkgroup[0],
this->mKomputeWorkgroup[1],
this->mKomputeWorkgroup[2]);
// Barrier to ensure the shader code is executed before buffer read
this->mTensorOutput->recordBufferMemoryBarrier(