4 ms·
I$ is short for instruction cache ($ -> "cash" -> cache). Since kdb has so little code, more of it fits into the instruction cache at once.
by durumu 3y ago
I$ is short for instruction cache ($ -> "cash" -> cache). Since kdb has so little code, more of it fits into the instruction cache at once.
- omnicognate 3y agoI've encountered this idea that k's terseness somehow improves instruction cache use before. Can you explain further? It seems nonsensical, since instruction caching is about machine code, not source code. Why should it use the instruction cache better than any other JIT? Or is it interpreted, in which case "the terseness of the language improves cache use" might seem more of an admission than a boast... :-)
- mlochbaum 3y agoI say it's nonsensical (and yes, Ks are bytecode interpreted). https://mlochbaum.github.io/BQN/implementation/kclaims.html#instruction-cache https://mlochbaum.github.io/BQN/implementation/kclaims.html#...