Re: [open-cobol-list] displaying max values in comp-5 field

[michael]

Very significant difference! Great info Gary!

$ cobc -x -fnotrunc -o mathtest ./mathtest.cob
$ ./mathtest
USAGE DISPLAY:   1.00 SECONDS
USAGE COMP:      0.06 SECONDS
USAGE COMP-5:    0.06 SECONDS
USAGE BINARY:    0.06 SECONDS
$ cobc -x  -o mathtest ./mathtest.cob
$ ./mathtest
USAGE DISPLAY:   6.61 SECONDS
USAGE COMP:      3.25 SECONDS
USAGE COMP-5:    0.06 SECONDS
USAGE BINARY:    3.23 SECONDS
$ ./mathtest
USAGE DISPLAY:   6.62 SECONDS
USAGE COMP:      3.25 SECONDS
USAGE COMP-5:    0.06 SECONDS
USAGE BINARY:    3.23 SECONDS
$ cobc -x -fnotrunc -o mathtest ./mathtest.cob
$ ./mathtest
USAGE DISPLAY:   1.00 SECONDS
USAGE COMP:      0.06 SECONDS
USAGE COMP-5:    0.06 SECONDS
USAGE BINARY:    0.06 SECONDS


> Have you checked out the compiler's "-fnotrunc" option?  My testing has
> shown that when you compile with this option (which turns off binary field
> truncation), not only does truncation stop but arithmetic operations
> against
> all numeric USAGEs (including DISPLAY) runs MUCH faster.
>
> Try this test program, compiled with and without "-fnotrunc":
>
>        IDENTIFICATION DIVISION.
>        PROGRAM-ID. mathtest.
>       *****************************************************************
>       ** This compares the performance of performing arithmetic op-  **
>       ** Erations against USAGE DISPLAY, COMP and COMP-5 numeric     **
>       ** data.                                                       **
>       **                                                             **
>       ** Each data item will have 7 added to it ten million times.   **
>       ** The time (to one-one-hundtredth of a second will be         **
>       ** retrieved before and after each test and the difference     **
>       ** between the two will be DISPLAYed.                          **
>       **                                                             **
>       ** Compile (and execute) this program twice - once with binary **
>       ** truncation turned off (-fnotrunc) and once with it turned   **
>       ** on (the default); you'll see some AMAZING differences in    **
>       ** execution times!                                            **
>       **                                                             **
>       ** Remember that COBOL is retrieving wall-clock time, not      **
>       ** actual CPU-used time, so other activities taking place on   **
>       ** your PC may influence the timings - run the program multi-  **
>       ** ple times to get your best view of the relative timings.    **
>       **                                                             **
>       ** AUTHOR:       GARY L. CUTLER                                **
>       **               address@hidden                            **
>       **                                                             **
>       ** DATE-WRITTEN: June 10, 2009                                 **
>       **                                                             **
>       *****************************************************************
>       **  DATE  CHANGE DESCRIPTION                                   **
>       ** ====== ==================================================== **
>       ** GC0609 INITIAL CODING.                                      **
>       *****************************************************************
>        DATA DIVISION.
>        WORKING-STORAGE SECTION.
>        01  Binary-Item         COMP    PIC S9(9) VALUE 0.
>
>        01  Comp-Item           COMP    PIC S9(9) VALUE 0.
>
>        01  Comp-5-Item         COMP-5  PIC S9(9) VALUE 0.
>
>        01  Display-Item        DISPLAY PIC S9(9) VALUE 0.
>
>        01  Old-Time.
>            05 OT-Hours                 PIC 9(2).
>            05 OT-Minutes               PIC 9(2).
>            05 OT-Seconds               PIC 9(2).
>            05 OT-Hundredths            PIC 9(2).
>
>        78  Repeat-Count                VALUE 10000000.
>
>        01  The-Time.
>            05 TT-Hours                 PIC 9(2).
>            05 TT-Minutes               PIC 9(2).
>            05 TT-Seconds               PIC 9(2).
>            05 TT-Hundredths            PIC 9(2).
>
>        01  Time-Diff                   PIC ZZ9.99.
>        PROCEDURE DIVISION.
>
>        010-Test-Usage-DISPLAY.
>            ACCEPT Old-Time FROM TIME.
>            PERFORM Repeat-Count TIMES
>                ADD 7 TO Display-Item
>            END-PERFORM.
>       D    DISPLAY 'Display-Item=' Display-Item.
>            PERFORM 100-Determine-Time-Diff.
>            DISPLAY 'USAGE DISPLAY: ' Time-Diff ' SECONDS'.
>
>        020-Test-Usage-COMP.
>            ACCEPT Old-Time FROM TIME.
>            PERFORM Repeat-Count TIMES
>                ADD 7 TO Comp-Item
>            END-PERFORM.
>       D    DISPLAY 'Comp-Item=' Comp-Item.
>            PERFORM 100-Determine-Time-Diff.
>            DISPLAY 'USAGE COMP:    ' Time-Diff ' SECONDS'.
>
>        030-Test-Usage-COMP-5.
>            ACCEPT Old-Time FROM TIME.
>            PERFORM Repeat-Count TIMES
>                ADD 7 TO Comp-5-Item
>            END-PERFORM.
>       D    DISPLAY 'Comp-5-Item=' Comp-5-Item.
>            PERFORM 100-Determine-Time-Diff.
>            DISPLAY 'USAGE COMP-5:  ' Time-Diff ' SECONDS'.
>
>        040-Test-Usage-BINARY.
>            ACCEPT Old-Time FROM TIME.
>            PERFORM Repeat-Count TIMES
>                ADD 7 TO Binary-Item
>            END-PERFORM.
>       D    DISPLAY 'Binary-Item=' Comp-5-Item.
>            PERFORM 100-Determine-Time-Diff.
>            DISPLAY 'USAGE BINARY:  ' Time-Diff ' SECONDS'.
>
>        099-Done.
>            STOP RUN.
>
>        100-Determine-Time-Diff.
>            ACCEPT The-Time FROM TIME.
>            COMPUTE Time-Diff =
>               ((  TT-Hours * 360000
>                 + TT-Minutes * 6000
>                 + TT-Seconds * 100
>                 + TT-Hundredths)
>               -
>                (  OT-Hours * 360000
>                 + OT-Minutes * 6000
>                 + OT-Seconds * 100
>                 + OT-Hundredths)) / 100.
>
> In my experience, by the way, it appears that the truncation happens when
> you refer to the field, not when you save a value into it - it appears
> that
> full binary precision is maintained.
>
>
>
>
>
>
> -----Original Message-----
> From: Bill Klein [mailto:address@hidden
> Sent: Monday, July 27, 2009 23:00
> To: 'vince coen'; address@hidden
> Subject: Re: [open-cobol-list] displaying max values in comp-5 field
>
> Vince,
>   Do you know of any compilers that do truncation with COMP-5?  I know
> LOTS
> that do with COMP or BINARY.  But almost all compilers that I know of that
> have COMP-5 as an extension, have it explicitly so it will NOT truncate
> based on picture clause (i.e. base 10).  Most, but not all, also have
> COMP-5
> in "operations system - best - binary" format, i.e. little-endian /
> big-endian automatically done based not he O/S on which the compiler is
> run.
>
> Again,
>   If the compiler allows for
>
> 01  aGroup.
>  05  Num1  Pic S9(4) Com-5.
>           ...
> Move High-Values to aGroup
> Display Num1
>
> then it SHOULD (IMHO) display the full value of the data item.  If the
> display doesn't work that way, then the compiler should reject it.
>
>> -----Original Message-----
>> From: vince coen [mailto:address@hidden
>> Sent: Monday, July 27, 2009 6:24 AM
>> To: address@hidden
>> Subject: Re: [open-cobol-list] displaying max values in comp-5 field
>>
>> Hi
>> On Monday 27 July 2009, Sergey Kashyrin wrote:
>> > Hi,
>> >
>> > I think it's not about that
>> > It's about DISPLAY which actually not suppose to display
>> ANY binary or
>> > comp-3/4/5 fields. I don't see any contradictions with formal Cobol.
>>
>> Oops, I 'assumed' that you had moved it to a numeric display
>> field prior to
>> display, but my argument is still valid against various
>> compilers in that
>> auto truncation can occur and manipulating a s9(4) field
>> after over flow any
>> thing can and no doubt will, happen so it is still verifyable
>> to fix by
>> change to s9(5) as needed.
>>
>> There again I have programmed in the basis of a specific
>> compilers behaviour
>> as against what the language specs say. It is only over the
>> last few years
>> that I have changed my programming style to match the
>> language specs knowing
>> that the compiler will no doubt change during the life of the
>> app which will
>> be longer than I have anticipated as proven with code written
>> in the 60's and
>> 70's now being converted from MF workbench to run under Open Cobol.
>>
>> Vince
>>
>> >
>> > Not going to do any fixes on that, correct me if I wrong
>> >
>> > Regards
>> > SKA
>> >
>> > ----- Original Message -----
>> > From: "vince coen" <address@hidden>
>> > To: <address@hidden>
>> > Sent: Sunday, July 26, 2009 7:05 AM
>> > Subject: [open-cobol-list] displaying max values in comp-5 field
>> >
>> > > Hi;
>> > >
>> > > On Saturday 25 July 2009, Robert Keane wrote:
>> > >> Hello All,
>> > >>
>> > >>
>> > >> I've found a problem when displaying the maximum value
>> in a comp-5
>> > >> field. If an s9(4) comp-5 can hold +32767 or -32768.
>> After initializing
>> > >> or moving +32767 or -32768 to the s9(4) comp-5 field the
>> display gives
>> > >> +2767 or -2768, the high order digit is dropped. However
>> if the comp-5
>> > >> field is moved to a field defined with enough z's the
>> display is ok.
>> > >
>> > > A dropped leading digit does not supprise me.  Moving
>> above 9,999 to a
>> > > pic s9 (4) will result in truncation. Solution is to
>> increase picture
>> > > size to reflect the maximum value your field will get to
>> in this case pic
>> > > s9(5) comp-5.
>> > >
>> > > Or better still use binary-short, its more effective for
>> arithmetic etc
>> > > in OC.
>> > >
>> > > Vince
>> > >
>> >
>>
>> --------------------------------------------------------------
>> ----------------
>> _______________________________________________
>> open-cobol-list mailing list
>> address@hidden
>> https://lists.sourceforge.net/lists/listinfo/open-cobol-list
>>
>
>
> ----------------------------------------------------------------------------
> --
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008
> 30-Day
> trial. Simplify your report design, integration and deployment - and focus
> on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> open-cobol-list mailing list
> address@hidden
> https://lists.sourceforge.net/lists/listinfo/open-cobol-list
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008
> 30-Day
> trial. Simplify your report design, integration and deployment - and focus
> on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> open-cobol-list mailing list
> address@hidden
> https://lists.sourceforge.net/lists/listinfo/open-cobol-list
>