[antlr-interest] java.g does not compile
Jim Idle
jimi at temporal-wave.com
Thu Jan 21 07:00:30 PST 2010
I wouldn't change the default time out as then your project depends on a custom version of NATLR for no good reason. That was just my 6:40AM typo of course :-)
I have a QX9450 and some i7s. I think that the Xeon server versions of 9450 etc might be slower on a single thread. I think a lot of the i7s are faster than Xeon? However I haven't bothered with Xeon myself. But, it depends what you are measuring. Most of the published benchmark programs are worthless.
Jim
> -----Original Message-----
> From: Andrew Haley [mailto:aph at redhat.com]
> Sent: Thursday, January 21, 2010 6:51 AM
> To: Jim Idle
> Cc: antlr-interest at antlr.org
> Subject: Re: [antlr-interest] java.g does not compile
>
> On 01/21/2010 02:40 PM, Jim Idle wrote:
>
> > You are probably right on the limit of the default 10000, or perhaps
> > you are not compiling the exact original?
>
> I haven't touched it. Honestly!
>
> Besides, the default seems to be 1000, not 10000.
>
> $ java -jar Downloads/antlr-3.2.jar -X
> -Xconversiontimeout t set NFA conversion timeout (ms) for each
> decision [1000]
>
> I changed it to 10000, and all is fine:
>
> --- antlr-3.2/tool/src/main/java/org/antlr/analysis/DFA.java~ 2009-
> 09-23 19:36:06.000000000 +0100
> +++ antlr-3.2/tool/src/main/java/org/antlr/analysis/DFA.java 2010-
> 01-21 13:08:32.625782840 +0000
> @@ -53,7 +53,7 @@
> */
>
> /** Set to 0 to not terminate early (time in ms) */
> - public static int MAX_TIME_PER_DFA_CREATION = 1*1000;
> + public static int MAX_TIME_PER_DFA_CREATION = 10*1000;
>
> /** How many edges can each DFA state have before a "special"
> state
> * is created that uses IF expressions instead of a table?
>
> > Try the on in the examples zip and see if there are any
> > differences. However, Xeon's are not as fast as you think on a
> > single thread which is what the analysis phase runs on by default.
>
> Err, how on Earth do you know how fast I think Xeons are? :-)
> But anyway, most users aren't likely to have anything hugely faster.
>
> Andrew.
More information about the antlr-interest
mailing list