squased commits from internal repository
authorMarkus Wittmann <markus.wittmann@fau.de>
Tue, 22 May 2018 08:24:08 +0000 (10:24 +0200)
committerMarkus Wittmann <markus.wittmann@fau.de>
Tue, 22 May 2018 08:24:08 +0000 (10:24 +0200)
commit9e0051cb083e4d8575cbd9f4a41d11552358e151
tree5917b368910fbb85f42efd79dbe8f86f6a8675eb
parent8cafd9ea08a6b1103eab29811227a7ae536dffa6
squased commits from internal repository

- doc: skylakesp2 results now with AVX512 intrinsics

- add AVX512 support for single precision intrinsics without
  gather/scatter

  The AVX512 intrinsics are divided into "pure" AVX512
  (load/store/compute) and AVX512-GATHER which include gather/scatter.

  This enables us to support at least AVX512 single precision intrinsics
  for all kernels which do not require gather/scatter support.

- fix test.sh: reenabled dp tests

- fix pull-split-nt: adjusted tmp array size to different vector
  lenghts
doc/images/benchmark-skylakesp2-dp.png
doc/images/benchmark-skylakesp2-sp.png
doc/main.html
doc/main.rst
src/BenchKernelD3Q19ListPullSplitNtCommon.c
src/KernelFunctions.h
src/Main.c
src/Makefile
src/Vector.h
src/test.sh
This page took 0.260441 seconds and 5 git commands to generate.