Because C allows for some straightforward mechanical sympathy. As Stutter said, C makes you care about memory layout and cache hierarchy a little, so you win by using all the hardware niceties (prefetchers). I'd say C has a good pareto index. You can go faster but it will require far? more work (to either write or learn a different platform/paradigm).
ps: for instance in some cases, jitted code will go faster than C, some Jitted kernels (never tested personally) allowed for a good 30% performance increase.
ps: for instance in some cases, jitted code will go faster than C, some Jitted kernels (never tested personally) allowed for a good 30% performance increase.