Some measurings with profiling enabled, with the command "darcs diff
--last=2000 +RTS -prof". This accounts only for the change of
breakBeforeNthNewline.
Before the patch:
1.
total time = 13.74 secs (13744 ticks @ 1000 us, 1 processor)
total alloc = 15,524,518,472 bytes (excludes profiling overheads)
2.
total time = 9.73 secs (9726 ticks @ 1000 us, 1 processor)
total alloc = 15,523,474,664 bytes (excludes profiling overheads)
3.
total time = 9.63 secs (9631 ticks @ 1000 us, 1 processor)
total alloc = 15,523,645,368 bytes (excludes profiling overheads)
After:
1.
total time = 7.11 secs (7106 ticks @ 1000 us, 1 processor)
total alloc = 5,452,901,552 bytes (excludes profiling overheads)
2.
total time = 7.12 secs (7115 ticks @ 1000 us, 1 processor)
total alloc = 5,447,844,184 bytes (excludes profiling overheads)
3.
total time = 7.18 secs (7184 ticks @ 1000 us, 1 processor)
total alloc = 5,447,981,704 bytes (excludes profiling overheads)
Profiling graphs ( with "darcs diff --last=2000 +RTS -h -i0.1" ) show
the same difference (from ~ 350 MBytes used to ~225 MBytes used).
Now for breakAfterNthNewline I don't have figures, it would be nice to
have a read-only command that uses this function a lot (unlike amend for
instance).
|