Merge pull request pytorch#206 from dominikgrewe/scattergather

CUDA implementations for scatter & gather.

Author: soumith

CMakeLists.txt

@@ -91,6 +91,7 @@ SET(src-cuda
  THCTensorIndex.cu
  THCTensorConv.cu
  THCTensorRandom.cu
+ THCTensorScatterGather.cu
  THCApply.cu
  THCTensorSort.cu
  )

THCTensorMath.h

@@ -130,6 +130,10 @@ THC_API void THCudaTensor_maskedFillByte(THCState* state, THCudaTensor *tensor,
 THC_API void THCudaTensor_maskedCopyByte(THCState* state, THCudaTensor *tensor, THByteTensor *mask, THCudaTensor *src);
 THC_API void THCudaTensor_maskedSelectByte(THCState* state, THCudaTensor *tensor, THCudaTensor *src, THByteTensor *mask);
 
+THC_API void THCudaTensor_gather(THCState* state, THCudaTensor *tensor, THCudaTensor *src, int dim, THCudaTensor *index);
+THC_API void THCudaTensor_scatter(THCState* state, THCudaTensor *tensor, int dim, THCudaTensor *index, THCudaTensor *src);
+THC_API void THCudaTensor_scatterFill(THCState* state, THCudaTensor *tensor, int dim, THCudaTensor *index, float value);
+
 THC_API int THCudaTensor_logicalall(THCState *state, THCudaTensor *self);
 THC_API int THCudaTensor_logicalany(THCState *state, THCudaTensor *self);