fix: atomicAdd() is defined twice

daniilidis-group / neural_renderer

A PyTorch port of the Neural 3D Mesh Renderer

Other

1.12k stars 248 forks source link

fix: atomicAdd() is defined twice #60

Open erodola opened 5 years ago

erodola commented 5 years ago

Fixed compilation-breaking issue that arises with CUDA 10.0 (possibly also older versions). The fix is taken verbatim from the file THCAtomics.cuh (line 119) of PyTorch 1.1.0 (full path \Lib\site-packages\torch\include\torch\csrc\api\include\torch), which avoids duplicate definition of atomicAdd(double*, double).