devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Improved the performance of full reductions.

Browse Source 
AFTER:
BM_fullReduction/10        4541       4543     154017  21.0M items/s
BM_fullReduction/64        5191       5193     100000  752.5M items/s
BM_fullReduction/512       9588       9588      71361  25.5G items/s
BM_fullReduction/4k      244314     244281       2863  64.0G items/s
BM_fullReduction/5k      359382     359363       1946  64.8G items/s

BEFORE:
BM_fullReduction/10        9085       9087      74395  10.5M items/s
BM_fullReduction/64        9478       9478      72014  412.1M items/s
BM_fullReduction/512      14643      14646      46902  16.7G items/s
BM_fullReduction/4k      260338     260384       2678  60.0G items/s
BM_fullReduction/5k      385076     385178       1818  60.5G items/s
This commit is contained in:
Benoit Steiner
2016-06-03 17:27:08 -07:00
parent 8d97ba6b22
commit c2a102345f
3 changed files with 76 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h
View File

@@ -316,7 +316,7 @@ struct OuterReducer {
#if defined(EIGEN_USE_GPU) && defined(__CUDACC__)
template <int B, int N, typename S, typename R, typename I>
__global__ void FullReductionKernel(R, const S, I, typename S::CoeffReturnType*);
__global__ void FullReductionKernel(R, const S, I, typename S::CoeffReturnType*, unsigned int*);
#ifdef EIGEN_HAS_CUDA_FP16
@@ -616,7 +616,7 @@ struct TensorEvaluator<const TensorReductionOp<Op, Dims, ArgType>, Device>
template <typename S, typename O, bool V> friend struct internal::FullReducerShard;
#endif
#if defined(EIGEN_USE_GPU) && defined(__CUDACC__)
template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernel(R, const S, I, typename S::CoeffReturnType*);
template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernel(R, const S, I, typename S::CoeffReturnType*, unsigned int*);
#ifdef EIGEN_HAS_CUDA_FP16
template <typename S, typename R, typename I> friend void internal::ReductionInitFullReduxKernelHalfFloat(R, const S, I, half2*);
template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernelHalfFloat(R, const S, I, half*, half2*);