[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Time to create :ACCESSORs and:READERs
- To: CommonLoops.pa@Xerox.COM
- Subject: Time to create :ACCESSORs and:READERs
- From: kanderso@WILMA.BBN.COM
- Date: Tue, 09 Feb 88 13:21:47 -0500
- Cc: kanderson@WILMA.BBN.COM, rshapiro@WILMA.BBN.COM, mthome@WILMA.BBN.COM, wsalter@WILMA.BBN.COM
- Redistributed: CommonLoops.pa
Using the 8/27/87 version of PCL on a Symbolics G7.1 machine we have the
following timings for compiling a DEFCLASS with 219 instance variables.
In the first case, :ACCESSOR is specified for each slot:
(time
(DEFCLASS SCHEDULE-D-2
(TAXFORM)
((D-BBN-ID-NUM :INITFORM NIL :ACCESSOR BBN-ID-NUM)
(D-NAME :INITFORM NIL :ALLOCATION :INSTANCE :ACCESSOR NAME)
(D-SSN :INITFORM NIL :ALLOCATION :INSTANCE :ACCESSOR SSN :TYPE T)
...
(D-48 :INITFORM NIL :ALLOCATION :INSTANCE :ACCESSOR TOTAL-BART-INCOME))
(:METACLASS AIIS-CLASS)))
Evaluation of (DEFCLASS SCHEDULE-D-2 (TAXFORM) (# # # # # ...)
(:METACLASS AIIS-CLASS)) took 156.522006 seconds of elapsed time
including 7.070 seconds waiting for the disk for 7 faults.
The garbage collector has flipped, so consing was not measured.
#<Aiis-Class SCHEDULE-D-2 320061670>
In the second case, :READER was specified for each slot:
(time
(DEFCLASS SCHEDULE-D-3
(TAXFORM)
((D-BBN-ID-NUM :INITFORM NIL :READER BBN-ID-NUM)
(D-NAME :INITFORM NIL :ALLOCATION :INSTANCE :READER NAME)
(D-SSN :INITFORM NIL :ALLOCATION :INSTANCE :READER SSN :TYPE T)
...
(D-48 :INITFORM NIL :ALLOCATION :INSTANCE :READER TOTAL-BART-INCOME))
(:METACLASS AIIS-CLASS)))
Evaluation of (DEFCLASS SCHEDULE-D-3 (TAXFORM) (# # # # # ...)
(:METACLASS AIIS-CLASS)) took 14.114444 seconds of elapsed time
including 1.361 seconds waiting for the disk for 33 faults.
16,824 list, 9,082 structure words consed in WORKING-STORAGE-AREA.
9 list words consed in PROPERTY-LIST-AREA.
7 structure words consed in PNAME-AREA.
5 structure words consed in SYMBOL-AREA.
#<Aiis-Class SCHEDULE-D-3 25732766>
This might not be the best way to perform such a test, but we're suprised that
it takes an order of magnitude longer to create the SETF functions.
Sorry, i haven't been able to try this in the current version yet.
k