search

openppl-public / ppl.cv

ppl.cv is a high-performance image processing library of openPPL supporting various platforms.
Apache License 2.0
484 stars 108 forks source link

trivial case in CUDA resize not synchronized with stream argument #70

Closed lzhangzz closed 2 years ago

lzhangzz commented 2 years ago

The stream argument is not passed to the following cudaMemcpyAsync when handling trivial cases. Causing the copy to be performed on the default stream instead of the passed one. https://github.com/openppl-public/ppl.cv/blob/eebdbd8ef1e831fa09079cf061752a8695b5dd02/src/ppl/cv/cuda/resize.cu#L1034-L1035 https://github.com/openppl-public/ppl.cv/blob/eebdbd8ef1e831fa09079cf061752a8695b5dd02/src/ppl/cv/cuda/resize.cu#L1228-L1229

jimurk commented 2 years ago

thanks, we will fix this in the next update.