C++Now 2019

In my day job, I work on a lot of exotic hardware and highly constrained systems, both in resources and latency/timing. In my domain, C is king and the reason given is almost always performance. I have found, however, that at least for my domain and my needs I cannot get the performance I need from C, I need C++.
Some of the custom assembler generating metaprograms I have written beat industry standard libraries in performance by 10x and more in some benchmarks and are far safer and easier to use than the forest of macros against which they compete.

In this talk, we will discuss the trade-offs of optimization within the compiler vs. optimization implemented in the form of metaprograms in library space for niche use cases on custom hardware. As an example of the success of optimization in library space I will present a DSL I wrote for ARM Cortex SoCs which provide reasonable semantics and incredible performance when talking to the more exotic functionality offered by the hardware and not supported by the optimizer (or the language for that matter).