Landon Thomas Jr., NYT, A New Breed of Trader on Wall Street: Coders With a Ph.D., here. Goldman has Erlang HFT so in some sense that’s even better.
Jane Street was founded at the beginning of the previous decade, when a couple of option traders and a computer expert left Susquehanna to start their own business.
Harnessing Ph.D.-toting mathematicians to the most powerful computers money can buy has become the accepted way for hedge funds and banks to get a trading edge these days, but Jane Street takes this marriage of high tech and high intellect to a new level.
Writing computer code, or at the least being conversant in the firm’s program of choice, OCaml, is a requisite for all traders. Indeed, new traders must complete a monthlong OCaml boot camp before they start trading.
Joel Hruska, Extreme Tech, How L1 and L2 CPU caches work, and why they’re an essential part of modern chips, here.
In the real world, an L1 cache typically has a hit rate between 95% and 97%, but the performance impact of those two values in our simple example isn’t 2% — it’s 14%. Keep in mind, we’re assuming the missed data is always sitting in the L2 cache. If the data has been evicted from the cache and is sitting in main memory, with an access latency of 80-120ns, the performance difference between a 95% and 97% hit rate could nearly double the total time needed to execute the code.
Sort a Zoolander “the files are in the computer” moment, Hruska says the L1 hit rate is, say, 97% but does not say anything about what is in the LI that is getting the high hit rate. Could be Ocaml recursive function call stack parameters or the bid price of the ETF from 300 ns ago in Mahwah. Maybe your hot ETF bid price does not make it into a bunch of high hit rate L1 cache lines that are needed to finish the recursive stack calls. Let us pause to reflect on how Rick Perry wisely summarized a similar situation long ago “Oops.” Then again it is not Hruska’s job to help keep your L1 clean.
Pink Iguana 