Mplab X | Compiler

void delay_ms(int ms) { for(int i=0; i<ms*1000; i++); } At -O0 , it works. At -O3 , the compiler notices the loop has no side effects. It doesn't just optimize the loop—it deletes the entire function . Your LED now toggles at 100 MHz. Poof.

You write a delay function:

Pro tip: Enable the . It outputs a .map file that reads like a confessional—showing exactly where every variable sinned (i.e., consumed a cycle). 2. Optimization Levels: The "Surgeon" vs. The "Wrecking Ball" The default optimization ( -O1 ) is safe. But -O3 (for XC32) or -Os (for XC8) is where things get interesting. mplab x compiler

Never assume the compiler is stupid. Use volatile strategically, not habitually. The XC32 compiler’s -fno-delete-null-pointer-checks is a lifesaver, but its -faggressive-loop-optimizations is a trap for the unwary. 3. The Mystical __attribute__ Directives This is where the compiler stops being a tool and starts being a wizard. XC compilers support GCC-style attributes plus Microchip-specific ones. __attribute__((persistent)) Place a variable in .persistent memory. It survives a device reset without re-initialization. Perfect for "why did I reboot?" state machines. __attribute__((interrupt(automatic_priority))) XC32 automatically handles the shadow register set and prologue/epilogue. Did you know that writing void __ISR(_TIMER_1_VECTOR, ipl2) my_handler(void) tells the compiler exactly which priority level to use, saving 7 cycles of software context saving? __attribute__((space(prog))) On XC16 (dsPIC), this forces a constant into program memory (flash) instead of RAM. Your 4KB lookup table now costs zero RAM. The compiler generates PSV windows for you automatically. 4. The "Free" Static Analysis You Are Ignoring Open your project properties. Go to MPLAB XCxx Compiler > Diagnostics . Turn on -Wconversion and -Wshadow . void delay_ms(int ms) { for(int i=0; i&lt;ms*1000; i++);

Most developers ignore warnings. They shouldn't. Consider this: Your LED now toggles at 100 MHz

Instead of:

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.