aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/convolution
AgeCommit message (Expand)Author
2019-04-16COMPMID-1995: Improves DepthwiseConvolution for qasymm8.Georgios Pinitas
2019-04-16COMPMID-2063: New Winograd implementationPablo Tello
2019-04-09COMPMID-1995: Fix 32-bit NEDepthwiseConvolution errors.Georgios Pinitas
2019-03-29COMPMID-1975: Update depthwise convolution.Georgios Pinitas
2018-11-02COMPMID-1623: NEWinograd reduce the number of output tiles.Pablo Tello
2018-11-02COMPMID-1600: Reduce number of tile specialisations.Pablo Tello
2018-11-02COMPMID-1564: Add NEDepthwiseConvolution3x3 for QASYMM8Georgios Pinitas
2018-11-02COMPMID-1532: Add DepthwiseConvolution3x3 FP16 on NEONGeorgios Pinitas
2018-11-02COMPMID-1552: support kernels sizes 1x7, 7x1, 1x5, 5x1 in NEWinogradPablo Tello
2018-11-02COMPMID-1550: Winograd integrate RSH changes.Pablo Tello
2018-11-02COMPMID-1247:Integrate kernel size 1x3 & 3x1 support in NEWinogradLayer.Pablo Tello
2018-11-02COMPMID-1112: Enabled multithreading transforms in Winograd.Pablo Tello
2018-11-02COMPMID-1010: Remove RSH profiler headerGeorgios Pinitas
2018-11-02COMPMID-1040: Added support for nullptr bias tensor in NEWinogradLayerAndrew Mundy
2018-11-02COMPMID-1005: Update Depthwise Convolution form RSHGeorgios Pinitas
2018-11-02COMPMID-976: Fix warnings for ndk-r16bAnthony Barbier
2018-11-02COMPMID-784: Fixed SAME paddding in WinogradLayerPablo Tello
2018-11-02COMPMID-873: Integrate RSH NEON Depthwise Convolution routineGeorgios Pinitas
generated by cgit v1.2.1 (git 2.23.0) at 2024-05-15 00:15:43 +0000