You write a delay function:
__asm__ volatile ("bsf %0, %1" : "=r"(PORT) : "r"(0)); The compiler will allocate the register for you. It won't clobber the WREG. It's civilised.
But what if I told you that the MPLAB X compiler suite (XC8, XC16, XC32) is not just a translator? It is a co-pilot . When wielded correctly, it can predict hardware race conditions, eliminate entire functions at compile time, and even write assembly better than you can.
uint16_t timer = 65000; timer = timer + 1000; // Warning: implicit conversion loses integer precision On an 8-bit PIC, that operation is 6 assembly instructions. On a 32-bit ARM (via XC32), it's one. The warning isn't pedantry—it's telling you that your 16-bit overflow will behave differently on different architectures.
Let’s dive into the dark arts of the XC compiler—the features that separate firmware hackers from embedded artists. Unlike GCC for Linux, Microchip’s XC compilers are deeply married to the silicon. The XC8 compiler, for example, doesn't just see a PIC16F18877 as a generic 8-bit CPU. It knows the exact banking scheme, the access bank, and even the shadow registers.