chux ceiling div

#include <inttypes.h>
#include <stdbool.h>

static inline uint64_t lg(uint64_t x) {
  return 63U - (uint64_t)__builtin_clzl(x);
}

static inline uint32_t lg32(uint32_t x) {
  return 31U - (uint32_t)__builtin_clz(x);
}

static inline uint64_t divide_chux(uint64_t p, uint64_t q) {
  bool round_up = p & (q-1);
  return (p >> lg(q)) + round_up;
}

static inline uint32_t divide_chux_32(uint32_t p, uint32_t q) {
  bool round_up = p & (q-1);
  return (p >> lg32(q)) + round_up;
}