quality is not the problem with pcg, but size. he could have used newlib's rand_r(), because that's already provided. or another simple and short prng, like xor

quality is not the problem with pcg, but size. he could have used newlib's rand_r(), because that's already provided. or another simple and short prng, like xoroshiro64s. btw I never needed a malloc for my baremetal firmwares so far. 10KB just for a stack and heap? crazy, that's way too much. recursion is forbidden anyway, and sbrk not provided. some unfortunate libc's need that, but MISRA will not be happy.