Povilas Kanapickas
Povilas Kanapickas
Extracted from https://github.com/p12tic/libsimdpp/issues/108.
New architectures support not only conversions to/from IEEE754 16-bit float types, but also operations on them. The availability of the API could be indicated by `SIMDPP_HAS_FLOAT16_SIMD`, `SIMDPP_HAS_FLOAT32_TO_FLOAT16_CONVERSION` and `SIMDPP_HAS_FLOAT16_TO_FLOAT32_CONVERSION` macros.
Add support for small vectors, such as `int32`. Currently there's no way to specify that we indeed want to perform operation on small number of vector elements. A possible solution...
This PR improves the positions of symbol bounding boxes in cases when LSTM model is used. Up to 20 times less errors have been observed in sample images. This PR...
The current max line length of 99 was preserved. The command that was used to create this commit is as follows: ``` black -l 99 master worker pkg common smokes...
Currently we disable all logging out of qml by default. This is problematic in case qml loading fails for any reason (e.g. missing dependencies) as the user will be presented...
Eigen is most optimal when using SIMD instructions which require overaligned buffers. In C++14 this is a problem because there's no built-in support for correctly aligned allocations, so this Eigen...
This huge PR is a refactoring that routes I/O accesses through a "virtual" file system. It is currently just a proof of concept to show what changes to the codebase...
This is a simple cleanup PR that removes accidental copies of strings where needed. The performance impact is trivial, but it's still worth to consistently use pass-by-reference lest some function...