tensorflowlite.dll is huge (compared to other platforms)

Opened on behalf of @sztomi, Issue is current as of this writing. Previous Issue: https://github.com/tensorflow/tensorflow/issues/48118

Please make sure that this is a build/installation issue. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:build_template

System information

OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10
TensorFlow installed from (source or binary): Source
TensorFlow version: https://github.com/tensorflow/tensorflow/commit/5d6cc7bf97a226c1e6a73ad4fc391c154dd622ac
Python version: n/a
Installed using virtualenv? pip? conda?: n/a
Bazel version (if compiling from source): 3.7.2
GCC/Compiler version (if compiling from source): MSVC 2019 (cl: 19.28.29334)
CUDA/cuDNN version: n/a
GPU model and memory: n/a

Describe the problem

Provide the exact sequence of commands / steps that you executed before running into the problem

bazel build -c opt //tensorflow/lite/tensorflowlite.dll

This yields a 15MB statically linked binary when compiled for 32-bit windows (i.e. the 64-bit version will be even larger). Compared to other platforms, this is a 5x increase that is difficult to explain, especially in lieu of tools like bloaty on Windows. Unfortunately I don't know where to start poking this at all - but I do believe that even accounting for platform differences, a 5x size difference is unexpected. The same issue was noted here: https://github.com/tensorflow/tensorflow/issues/33634#issuecomment-620645664 (with a suggestion to use the C API instead, but the C API is a binding that cannot be used independent from the C++ binary).

Any other info / logs

Let me know if I can include any more info.

google-ai-edge / LiteRT