On Wednesday, 25 December 2013 at 08:27:53 UTC, Philippe Sigaud
wrote:
Out of curiosity, do you know which one of his 3 suggestions
brought you
the highest speed boost? What happens if you do not disable the
GC, for
example?
Good question, I did test each one incrementally:
1. Switching from "split+to!size_t" to "parse" (and commenting
out the "union") reduced running time from ~26s to ~20s (on
average).
2. Switching from "byLine" to "byLineFast" (and commenting out
the "union") reduced running time from ~20s to ~14s (on average).
3. With "parse" and "byLineFast", and with GC still on, and
populating the "union" the program took about 2m45s .
4. Adding "GC.disable" brought it down to 25s.
HTH,
-gordon
