[MLton-devel] streamIO

[Matthew Fluet]

Here are the benchmark results:

MLton0 -- mlton-stable
MLton1 -- mlton.cvs.HEAD
compile time
benchmark     MLton0 MLton1
wc-input1       4.09   4.78
wc-scanStream   4.18   5.09
run time
benchmark     MLton0 MLton1
wc-input1      70.14  69.66
wc-scanStream  75.53  89.38
run time ratio
benchmark     MLton1
wc-input1       0.99
wc-scanStream   1.18
size
benchmark     MLton0 MLton1
wc-input1     87,749 91,389
wc-scanStream 88,485 94,069


So, much better than previously.

I'm still curious about one anomaly.  With the old IO, I get:

[fluet@localhost stable.time]$ mlprof -thresh 5 -raw true wc-scanStream
mlmon.out
78.02 seconds of CPU time (2.12 seconds GC)
     function       cur    raw    stack   raw     GC    raw
------------------ ----- -------- ----- -------- ---- -------
<Posix_IO_read>    18.0% (14.46s) 18.0% (14.46s) 0.0%  (0.0s)


While with the new IO, I have:

[fluet@localhost cvs.HEAD.time]$ mlprof -thresh 5 -raw true wc-scanStream
mlmon.out
97.79 seconds of CPU time (4.85 seconds GC)
     function       cur    raw    stack   raw     GC    raw
------------------ ----- -------- ----- -------- ---- -------
<Posix_IO_read>    25.5% (26.13s) 25.5% (26.13s) 0.0%  (0.0s)

Now, the number of calls to Posix_IO_read and the arguments to it are the
same in both programs.  The only thing I can think of is that if the new
IO is just a little slower, then the difference in time between calls is
enough to cause a difference in the system call response (caching, paging,
etc.).




-------------------------------------------------------
This SF.NET email is sponsored by: Thawte.com - A 128-bit supercerts will
allow you to extend the highest allowed 128 bit encryption to all your 
clients even if they use browsers that are limited to 40 bit encryption. 
Get a guide here:http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0030en
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel