Update Build Status Location

Fix precision for using std.math

Fix illegal instruction errors on Windows

Ensure deterministic iteration over edges during serde

Fix crosscompile to raspberry pi3 (arm ca7)

size_t has only 32 bit on the ARM platform. So we need some explicit
casts at the right place. Changed _currInd to be size_t, which matches
best here as it needs the least number of conversions later.

Fixup ddoc comment, so that ddox build runs without error resultcode

vectorflow.math: Enable flushing denormals to zero

align(...) is not respected for stack variables.

Using a regular function to print a compile-time message still causes
the compiler to emit an empty function in the object file; a
compile-time construct is better suited for compile-time side effects.

Use a mixin template to allow usage both inside and outside function
scopes.

Removing the unnecessary static constructor in vectorflow.neuralnet
fixes the cyclic dependency issue with vectorflow.math in non-release
builds.

simplified template

conv.to instead of cast

32bits support

Update README for ubuntu

Typo Fixes

Just quick fixes, awesome readme!