Fix CUDA int8x4 vectorize (#3928)
authornoituIover <qazazsw@gmail.com>
Fri, 13 Sep 2019 01:04:52 +0000 (09:04 +0800)
committerWuwei Lin <wuwei@apache.org>
Fri, 13 Sep 2019 01:04:52 +0000 (21:04 -0400)
commit195973c0a136c739c15f238fb4db6b433bf505c6
treefba04abb23e59ee2a0842206375b090cf3851c09
parent880c26039431ee9069e6fa79a323ea8ba2ab6c17
Fix CUDA int8x4 vectorize (#3928)

* Fix int8x4 vectorize

* Fix gpu shared/local memory accumulate

* Add test_shared_memory for int8x4

* Adjust test format

* Fix cpplint
src/codegen/codegen_cuda.cc
src/pass/verify_gpu_code.cc
tests/python/unittest/test_codegen_cuda.py
tests/python/unittest/test_pass_verify_gpu_code.py