3 ms·
> Does anybody know some specific case where this direction is advantageous? Very slow memory. Decades ago, a lot of interpreters did the same trick internally
by baybal2 8y ago
> Does anybody know some specific case where this direction is advantageous?
Very slow memory. Decades ago, a lot of interpreters did the same trick internally:
1. tokenise code, 2. convert to some compact binary serialisation format used internally by the rest of interpreter.
The trick was to make interpreter code to remain in CPU cache as long as possible while the code being interpreted did not require fetching from RAM