Why should we? ---------------- - SPI improvements? - Allowing to `make` for just one architecture? (stm32/f1) - Anything else that could be interesting? Things to consider ------------------- - Our current `stm32f103x8.ld` file [may be different in the latest version of libopencm3](https://github.com/libopencm3/libopencm3/blob/master/lib/stm32/f1/stm32f103x8.ld)?