Many operators don't work for integer type

[eric-haibin-lin]

Many operators implicitly cast data to float and return inaccurate results:

>>> mx.nd.array([50000001], dtype='int32')

[50000001]
<NDArray 1 @cpu(0)>

>>> mx.nd.array([50000001], dtype='int32').floor()

[50000000] ------------->  should be [50000001]
<NDArray 1 @cpu(0)> 

>> mx.nd.array([50000001], dtype='int32').round()

[50000000] ------------->  should be [50000001]
<NDArray 1 @cpu(0)>
>>> mx.nd.array([50000001], dtype='int32').ceil()

[50000000] ------------->  should be [50000001]
<NDArray 1 @cpu(0)>
>>> mx.nd.array([50000001], dtype='int32').trunc()

[50000000] ------------->  should be [50000001]
<NDArray 1 @cpu(0)> 

[apeforest]

@mxnet-label-bot add [Good First Issue, Call for Contribution]

[kalyc]

Thanks for reporting the issue @eric-haibin-lin
Could you also add a link to the relevant place in the code where changes need to be made?

[hanzhaogang]

Hard to trace. Can't find the gen_op module in the current package while working on this line:
from .gen_op import * # pylint: disable=unused-wildcard-import.

[eric-haibin-lin]

It's most likely due to the float conversion in https://github.com/apache/incubator-mxnet/blob/f5ba2678f077b58d31d31029f680f93b313e1cea/src/operator/math_functions-inl.h#L43-L61

[Vikas-kum]

@anirudh2290 Please tag PR-work-in-progress