[developers] using PET in 64-bit mode

eric-n at is.naist.jp eric-n at is.naist.jp
Mon Aug 18 07:11:50 CEST 2008


Greetings,

With those flag settings, I get the same results for 32bit and 64bit:

eric-n at thyme:~/pet-test [] [12:56:28] [10200]
:) cat tc-12.txt | /usr/bin/cheap $LOGONROOT/dfki/jacy/japanese.grm - 
limit=10000 -packing -results=100 -tsdbdump 64bit
# On 32-bit machine ...
eric at vmware-jeos:~/pet-test [] [13:27:40] [456]
:) cat tc-12.txt | /usr/bin/cheap $LOGONROOT/dfki/jacy/japanese.grm - 
limit=10000 -packing -results=100 -tsdbdump 32bit
# copy files to 64-bit machine ...
eric-n at thyme:~/pet-test [] [14:14:08] [10235]
:) grep 幼なじみ */result
32bit/result:1184 at 0@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (197  
utterance_rule-decl-finite -2.79125 0 5 (194 hf-adj-i-rule -2.77244 0  
5 (184 hf-complement-rule 0.768242 0 3 (183 hf-complement-rule  
0.691078 0 2 (6 kare 0 0 1 ("彼" 0 1)) (25 to-comp1 0.395993 1 2  
("と" 1 2))) (41 wa-comparg 0.266404 2 3 ("は" 2 3))) (192 hf- 
complement-rule -2.88181 3 5 (188 quantify-n-lrule 0.159015 3 4 (45  
osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (191 unary-vstem-vend-rule  
-2.06373 4 5 (190 subj-zpro-ins-lrule -1.33386 4 5 (189 desu-lexeme- 
infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114 4 5 ("です" 4  
5)))))))))@@@@
32bit/result:1184 at 1@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (198  
utterance_rule-decl-finite -3.54891 0 5 (195 hf-adj-i-rule -3.53009 0  
5 (178 hf-complement-rule 0.0105878 0 3 (177 hf-complement-rule  
0.488607 0 2 (6 kare 0 0 1 ("彼" 0 1)) (19 to-advp-exh-1 -0.0586652 1  
2 ("と" 1 2))) (43 wa-parg -0.145294 2 3 ("は" 2 3))) (192 hf- 
complement-rule -2.88181 3 5 (188 quantify-n-lrule 0.159015 3 4 (45  
osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (191 unary-vstem-vend-rule  
-2.06373 4 5 (190 subj-zpro-ins-lrule -1.33386 4 5 (189 desu-lexeme- 
infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114 4 5 ("です" 4  
5)))))))))@@@@
32bit/result:1184 at 2@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (196  
utterance_rule-decl-finite -4.6957 0 5 (193 hf-adj-i-rule -4.67689 0 5  
(187 hf-complement-rule -1.13621 0 3 (186 hf-complement-rule -1.21337  
0 2 (185 frg-np -0.0314627 0 1 (6 kare 0 0 1 ("彼" 0 1))) (22 to-comp- 
quotarg -0.273349 1 2 ("と" 1 2))) (41 wa-comparg 0.266404 2 3 ("は"  
2 3))) (192 hf-complement-rule -2.88181 3 5 (188 quantify-n-lrule  
0.159015 3 4 (45 osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (191 unary- 
vstem-vend-rule -2.06373 4 5 (190 subj-zpro-ins-lrule -1.33386 4 5  
(189 desu-lexeme-infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114  
4 5 ("です" 4 5)))))))))@@@@
64bit/result:1184 at 0@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (245  
utterance_rule-decl-finite -2.79125 0 5 (242 hf-adj-i-rule -2.77244 0  
5 (232 hf-complement-rule 0.768242 0 3 (231 hf-complement-rule  
0.691078 0 2 (6 kare 0 0 1 ("彼" 0 1)) (25 to-comp1 0.395993 1 2  
("と" 1 2))) (41 wa-comparg 0.266404 2 3 ("は" 2 3))) (240 hf- 
complement-rule -2.88181 3 5 (236 quantify-n-lrule 0.159015 3 4 (45  
osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (239 unary-vstem-vend-rule  
-2.06373 4 5 (238 subj-zpro-ins-lrule -1.33386 4 5 (237 desu-lexeme- 
infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114 4 5 ("です" 4  
5)))))))))@@@@
64bit/result:1184 at 1@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (246  
utterance_rule-decl-finite -3.54891 0 5 (243 hf-adj-i-rule -3.53009 0  
5 (228 hf-complement-rule 0.0105878 0 3 (227 hf-complement-rule  
0.488607 0 2 (6 kare 0 0 1 ("彼" 0 1)) (19 to-advp-exh-1 -0.0586652 1  
2 ("と" 1 2))) (43 wa-parg -0.145294 2 3 ("は" 2 3))) (240 hf- 
complement-rule -2.88181 3 5 (236 quantify-n-lrule 0.159015 3 4 (45  
osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (239 unary-vstem-vend-rule  
-2.06373 4 5 (238 subj-zpro-ins-lrule -1.33386 4 5 (237 desu-lexeme- 
infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114 4 5 ("です" 4  
5)))))))))@@@@
64bit/result:1184 at 2@-1 at -1@-1 at -1@-1 at -1@-1 at -1@(utterance-root (244  
utterance_rule-decl-finite -4.6957 0 5 (241 hf-adj-i-rule -4.67689 0 5  
(235 hf-complement-rule -1.13621 0 3 (234 hf-complement-rule -1.21337  
0 2 (233 frg-np -0.0314627 0 1 (6 kare 0 0 1 ("彼" 0 1))) (22 to-comp- 
quotarg -0.273349 1 2 ("と" 1 2))) (41 wa-comparg 0.266404 2 3 ("は"  
2 3))) (240 hf-complement-rule -2.88181 3 5 (236 quantify-n-lrule  
0.159015 3 4 (45 osananajimi_1 0 3 4 ("幼なじみ" 3 4))) (239 unary- 
vstem-vend-rule -2.06373 4 5 (238 subj-zpro-ins-lrule -1.33386 4 5  
(237 desu-lexeme-infl-rule -1.2065 4 5 (46 desu-v-cop-id-stem -1.04114  
4 5 ("です" 4 5)))))))))@@@@

I am using a slightly older SVN pull of Jacy:

eric-n at thyme:~/logon/dfki/jacy [] [14:24:51] [10245]
:) svn info
Path: .
URL: https://jacy.opendfki.de/repos/trunk
Repository Root: https://jacy.opendfki.de/repos
Repository UUID: 9f3be1c0-4e39-0410-8dd3-8e784bbfab8c
Revision: 376
Node Kind: directory
Schedule: normal
Last Changed Author: francis_bond
Last Changed Rev: 376
Last Changed Date: 2008-06-26 18:30:27 +0900 (Thu, 26 Jun 2008)

Also, I am using the new cheap packages compiled without the LKB MRS  
lisp code:

eric-n at thyme:~/logon/dfki/jacy [] [14:24:54] [10246]
:) dpkg -l | grep pet
ii  pet-cheap                                   
0.99.14-1~20080808svn-1nlp1~0hardy1      The Cheap unification parser
ii  pet-doc                                     
0.99.14-1~20080808svn-1nlp1              Documentation for PET
ii  pet-flop                                    
0.99.14-1~20080808svn-1nlp1~0hardy1      The Flop unification grammar  
preprocessor

I will try switching back to the old cheap and see if I get different  
results.

Eric Nichols

On Aug 18, 2008, at 11:51 AM, Francis Bond wrote:

> G'day,
>
> 2008/8/15  <eric-n at is.naist.jp>:
>> Just to clarify: are you using the exact same settings on 32-bit  
>> and 64-bit?
>
> Yes.
>
>> Could you post your cheap command with flags?
>
> cat tc-12.txt | /usr/bin/cheap /home/bond/svn/fresh/japanese.grm
> -limit=10000 -packing -results=100 -tsdbdump 64bit
>
> and dumping to 32bit on the 32bit machine.
>
>
> -- 
> Francis Bond <http://www2.nict.go.jp/x/x161/en/member/bond/>
> NICT Language Infrastructure Group





More information about the developers mailing list