As part of a university project we optimized a simple string to uppercase function via assembler analysis and asm. Directly optimizing the C code to produce mor

As part of a university project we optimized a simple string to uppercase function via assembler analysis and asm. Directly optimizing the C code to produce more optimized assembly (notably removing branching) was nearly an order of magnitude faster. // uppercase without if *c -= (*c-'a'<26U)<<5; Adding loop unrolling to that C-optimized version via asm provided another 20% performance boost. And I didn't even add vectorization using AVX. While compilers provide a lot of optimization directly, there are huge performance gains in simple functionality by helping the compiler with easier to optimize code or by using asm directly. Also notably the highest GCC optimization level I tested (-O3) reverted the optimization of the optimized C code, while -O2 kept them, resulting in -O2 being much faster. So only using asm guaranteed the performance gain.