Merge branch 'armhf' into 'master'

fix: change the armlinux abi from armeabi-v7a to armhf

See merge request deep-computing/mace!1260

Merge branch 'lp_normalization' into 'master'

add lpnorm、mvnorm op for caffe, enhance biasadd、reshape op

See merge request !1224

1. Remove unimplemented gpu matmul.

2. Fix the length of abbreviated commit id in MACE version.
3. Fix some bugs.

Big change for tools and performance.

Handle runtime failure

Add gpu float support