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

Updated OpAlgoBase to not copy data as optensorsync operations are introduced

Browse source
This commit is contained in:
Alejandro Saucedo 2020-09-12 09:14:35 +01:00
parent 4171786b6f
commit 9f8508075a
10 changed files with 92 additions and 180 deletions
Download patch file Download diff file Expand all files Collapse all files

1
test/TestOpAlgoLoopsPassingData.cpp
View file

@ -50,7 +50,6 @@ TEST(TestProcessingIterations, IterateThroughMultipleSumAndCopies) {
sq->record<kp::OpAlgoBase<>>(
{ tensorA, tensorB },
true, // Whether to copy output from device
std::vector<char>(shader.begin(), shader.end()));
sq->record<kp::OpTensorCopy>({tensorB, tensorA});