warning: implicit conversion from 'const size_t' (aka 'const unsigned int') to 'float' changes value from 2147483647 to 2147483648

[yurivict]

clang-12 complains:

/wrkdirs/usr/ports/misc/mxnet/work/incubator-mxnet-1.9.0/src/operator/quantization/quantized_elemwise_mul.cc:141:31: warning: implicit conversion from 'const size_t' (aka 'const unsigned int') to 'float' changes value from 2147483647 to 2147483648 [-Wimplicit-const-int-float-conversion]
    float output_data_range = kInt32Range;
          ~~~~~~~~~~~~~~~~~   ^~~~~~~~~~~

This looks like a real bug - the value get accidentally changed.

Version: 1.9.0
OS: FreeBSD 13

[anko-intel]

You are right.
But I am not sure if it is worth to fix it. Error in final results coming from it is minimal (or even it will be invisible), but for fixing it I have to change the type of output_data_range to double - so a couple of calculation below will be made on double - it takes more time.

[anko-intel]

I have decided to remove compilation warning in #20854. The rounding error is to small to have infuence on final results.
@yurivict could you close the issue?