1 INFO-VAX	Fri, 22 Jul 2005	Volume 2005 : Issue 406       Contents: accessing TCP/IP port 9880?  Re: accessing TCP/IP port 9880?  Re: accessing TCP/IP port 9880?  Re: accessing TCP/IP port 9880? 6 Re: Anyone Using Multinet NFS to access CMS Libraries?6 Re: Anyone Using Multinet NFS to access CMS Libraries?
 checkpointed? ! RE: Congratulations to Ed Wilts ! ! Re: Congratulations to Ed Wilts ! ! Re: Congratulations to Ed Wilts ! ! Re: Congratulations to Ed Wilts ! ! Re: Congratulations to Ed Wilts ! " DEC Forms Calender and Time Panels Modbus TCP for VMS? / perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx 3 Re: perfomance question about bool variables in cxx ) Re: Performance issue after 7.3-2 upgrade ) Re: Performance issue after 7.3-2 upgrade & Re: printing from MS Excel through VMS& Re: printing from MS Excel through VMS" Re: simple image processing on VMS  F ----------------------------------------------------------------------    Date: 22 Jul 2005 07:56:50 -0700' From: "tadamsmar" <tadamsmar@yahoo.com> $ Subject: accessing TCP/IP port 9880?B Message-ID: <1122044210.886867.63900@g49g2000cwa.googlegroups.com>  9 I am looking at a communications protocol document for an @ instrument that says that "C-link functions can be accessed over! Ethernet using TCP/IP port 9880".   B How does do this on an OpenVMS system?  We have the TCP/IP layered? product, but I have never done any interfacing to get data from % instruments using a TCP/IP interface.   C Otherwise the protocol is a straightforward ASCII protocol that can  also be used over RS232.   ------------------------------  % Date: Fri, 22 Jul 2005 12:04:15 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> ( Subject: Re: accessing TCP/IP port 9880?, Message-ID: <42E118C1.3E61E935@teksavvy.com>   tadamsmar wrote: > ; > I am looking at a communications protocol document for an B > instrument that says that "C-link functions can be accessed over# > Ethernet using TCP/IP port 9880".  > ) > How does do this on an OpenVMS system?    M The TCPIP Servcices Programming Manual has plenty of examples of how to setup K a program that listens on a particular port and then handles incoming calls  and in/out data.  R Yo can use the Unix style socket routines (easier) or the VMS style QIO interface.   ------------------------------    Date: 22 Jul 2005 09:09:23 -0700; From: "johnhreinhardt@yahoo.com" <johnhreinhardt@yahoo.com> ( Subject: Re: accessing TCP/IP port 9880?C Message-ID: <1122048563.187114.171630@g14g2000cwa.googlegroups.com>    JF Mezei wrote: O > The TCPIP Servcices Programming Manual has plenty of examples of how to setup M > a program that listens on a particular port and then handles incoming calls  > and in/out data. >   F If you don't have the hardcopy or CD versions then you can get HTML orE PDF from the HP website at http://h71000.www7.hp.com/doc/tcpip55.html D Just pick your verison of TCP/IP Services (if it's not V5.5 which is the latest with VMS V8.2).   ------------------------------  % Date: Fri, 22 Jul 2005 18:35:37 +0200 , From: Albrecht Schlosser <ajs856@tiscali.de>( Subject: Re: accessing TCP/IP port 9880?, Message-ID: <497rbd.4ei.ln@news.hus-soft.de>   tadamsmar wrote:; > I am looking at a communications protocol document for an B > instrument that says that "C-link functions can be accessed over# > Ethernet using TCP/IP port 9880".  > D > How does do this on an OpenVMS system?  We have the TCP/IP layeredA > product, but I have never done any interfacing to get data from ' > instruments using a TCP/IP interface.  > E > Otherwise the protocol is a straightforward ASCII protocol that can  > also be used over RS232. >    For testing purposes you can do    $ telnet <hostname> 9880  H and see what you get. Your input will be transmitted to the device, the J devices output will be displayed on your terminal. Remember the displayed = "Escape character", you may need it to disconnect the device.    There is also the command   * $ telnet /create_session <hostname> <port>  M It will display the device name of the created virtual device, e.g. TNA2966.  I You may then use TNA2966 in a program like a usual terminal device (with  5 QIO, FORTRAN READ, etc.) to access the TCP/IP device.   J As others pointed out, the better way for productive code would be to use  the socket functions.    Albrecht   ------------------------------  % Date: Fri, 22 Jul 2005 02:43:27 -0700 3 From: "C.W.Holeman II" <cwhii_googlespam@yahoo.com> ? Subject: Re: Anyone Using Multinet NFS to access CMS Libraries? 0 Message-ID: <11e1fu0946r60d0@corp.supernews.com>   Warren Spencer wrote:   7 > The CMS docs say (for the vms command-line client) to E > use DFS (rather than simple remote network access) because the file L > locking that is neccessary CMS library access is not available without it.   There is a DFS on VMS?   --   C.W.Holeman II) cwhii@Julian5Locals5.com remove the fives   http://free.ProHosting.com/cwhii   ------------------------------    Date: 22 Jul 2005 13:39:25 +01006 From: peter@langstoeger.at (Peter 'EPLAN' LANGSTOEGER)? Subject: Re: Anyone Using Multinet NFS to access CMS Libraries? , Message-ID: <42e0f70d$1@news.langstoeger.at>  f In article <11e1fu0946r60d0@corp.supernews.com>, "C.W.Holeman II" <cwhii_googlespam@yahoo.com> writes: >Warren Spencer wrote: > 8 >> The CMS docs say (for the vms command-line client) toF >> use DFS (rather than simple remote network access) because the fileM >> locking that is neccessary CMS library access is not available without it.  >  >There is a DFS on VMS?   & Don't confuse DECdfs with OSF-DCE DFS.  3 DECdfs (based on DECnet) is of course there on VMS.   / 	http://www.compaq.com/info/SP2878/SP2878PF.PDF   M OSF-DCE DFS (based on TCPIP) is by VMS engineering, though per OSF definition J being a part of the product Open Software Foundation Distributed ComputingK Environment (OSF DCE) - which is available for VMS -, being seen as part of J the (unfortunately dead before arrival) product OSF Distributed ManagementJ Environment (OSF DME) - which is not there for VMS (and also for all otherL operating systems - at least it was so last time I looked, some years ago).   H So, DFS is there for other operating systems, but still not for OpenVMS.  / 	http://www.compaq.com/info/SP4305/SP4305PF.PDF   J btw: I always referred to DME as "that thing which makes UNIX an operatingM system". Without it, UNIX is not existent (only Tru64/HP-UX/SOLARIS/AIX/...). I What is left of OSF is an (distributed) RPC API, or in other words a SDK. C But that is a topic for yet another flame war, so please ignore it.    --   Peter "EPLAN" LANGSTOEGER % Network and OpenVMS system specialist  E-mail  peter@langstoeger.atF A-1030 VIENNA  AUSTRIA              I'm not a pessimist, I'm a realist   ------------------------------  + Date: Fri, 22 Jul 2005 07:36:52 +0000 (UTC) P From: helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) Subject: checkpointed?$ Message-ID: <dbq7mj$7v9$1@online.de>  E I have a batch job retained with an (understandable) error.  However, H the queue entry now has "(checkpointed)" after the file name.  What does
 this mean?  4 I also, for the first time, had a job retained with   1    %SYSTEM-F-INSFMEM, insufficient dynamic memory   I Strange, since I haven't changed much in the system.  Is the solution to  I increase CTLPAGES?  It is now at 2000 on all my machines (default is 256  H on ALPHA and 100 on VAX; I haven't changed it, so I assume AUTOGEN did).6 What would be the disadvantage of increasing CTLPAGES?   ------------------------------  % Date: Fri, 22 Jul 2005 19:53:26 +1000 6 From: "O'Brien Paddy" <Paddy.O'Brien@transgrid.com.au>* Subject: RE: Congratulations to Ed Wilts !X Message-ID: <8BAD914A0B8CA84C9E94187103A1AB9E05BE77@EX-TG2-PR.corporate.transgrid.local>  , This is a multi-part message in MIME format.  ' ------_=_NextPart_001_01C58EA3.34CE2DCB . Content-Type: text/plain; charset="iso-8859-1"+ Content-Transfer-Encoding: quoted-printable    -----Original Message-----6 From: healyzh@aracnet.com [mailto:healyzh@aracnet.com] Sent: Wed 7/20/2005 2:39 AM  To: Info-VAX@Mvb.Saic.Com * Subject: Re: Congratulations to Ed Wilts ! =20   Roy Omond <Roy@omond.net> wrote: > Ed Wilts wrote:   ( > > Romeo> say f$getsyi("cluster_ftime") > > 30-MAY-1999 07:28:07.05  > >=20 > > 'nuff said.   0 > C'mon, Ed !  You've just got to participate in > the uptimes project :-)   J Something tells me he just did. :^)  WVNETcluster owned by "mosaic"  3483d 7h 12m  I The one that scares me is the "Windows .NET Server" up for 745 days, that 3 box has to be missing a lot of MS security patches!   # > I'm sure there are others too ...   K Well, if I was, I'd be in 38th place.  The downside is in a few days when I L cross 365 days I'm going to have to take the system down for some electricalK work that needs to be done :^(  This is of course assuming the heat doesn't G take it out first (I was wondering if I'd have to shut down yesterday).    			Zane   L I have the same problem with uptime.  Our building usually has a maintenanc=L e outage annually.  Our high-rise was taken down a week ago, so my middle-r=L ise area will go down soon.  In the meantime, it's been over a year, since =L my VAXstation 4000 - 60 has been up for 418 days  (this is now only used fo=L r my desktop transport to other nodes).  Our production ES40 has done 204 d=% ays and my development DS10 127 days.   L Dave Sneddon must have very long UPS if he can weather power outages and bu=; ilding maintenance -- similar to all the other competitors.    Regards, Paddy    G ***********************************************************************   C "This electronic message and any attachments may contain privileged @ and confidential information intended only for the use of the=20D addressees named above.  If you are not the intended recipient of=20C this email, please delete the message and any attachment and advise D the sender.  You are hereby notified that any use, dissemination,=207 distribution, reproduction of this email is prohibited.   C If you have received the email in error, please notify TransGrid=20 C immediately.  Any views expressed in this email are those of the=20 ? individual sender except where the sender expressly and with=20 C authority states them to be the views of TransGrid.  TransGrid uses > virus-scanning software but excludes any liability for viruses contained in any attachment.  < Please note the email address for TransGrid personnel is now$ firstname.lastname@transgrid.com.au"  G ***********************************************************************     ' ------_=_NextPart_001_01C58EA3.34CE2DCB - Content-Type: text/html; charset="iso-8859-1" + Content-Transfer-Encoding: quoted-printable   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">  <HTML> <HEAD>L <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; charset=3Diso-8859-= 1"> K <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version 6.5.7226.0"> 0 <TITLE>RE: Congratulations to Ed Wilts !</TITLE> </HEAD>  <BODY>) <!-- Converted from text/plain format -->   0 <P><FONT SIZE=3D2>-----Original Message-----<BR>L From: healyzh@aracnet.com [<A HREF=3D"mailto:healyzh@aracnet.com">mailto:he= alyzh@aracnet.com</A>]<BR> Sent: Wed 7/20/2005 2:39 AM<BR>  To: Info-VAX@Mvb.Saic.Com<BR> . Subject: Re: Congratulations to Ed Wilts !<BR> <BR>* Roy Omond &lt;Roy@omond.net&gt; wrote:<BR> &gt; Ed Wilts wrote:<BR> <BR>? &gt; &gt; Romeo&gt; say f$getsyi(&quot;cluster_ftime&quot;)<BR> % &gt; &gt; 30-MAY-1999 07:28:07.05<BR> 
 &gt; &gt;<BR>  &gt; &gt; 'nuff said.<BR>  <BR>< &gt; C'mon, Ed !&nbsp; You've just got to participate in<BR>  &gt; the uptimes project :-)<BR> <BR>L Something tells me he just did. :^)&nbsp; WVNETcluster owned by &quot;mosai= c&quot;&nbsp; 3483d<BR> 
 7h 12m<BR> <BR>L The one that scares me is the &quot;Windows .NET Server&quot; up for 745 da= ys, that<BR>7 box has to be missing a lot of MS security patches!<BR>  <BR>* &gt; I'm sure there are others too ...<BR> <BR>L Well, if I was, I'd be in 38th place.&nbsp; The downside is in a few days w=	 hen I<BR> L cross 365 days I'm going to have to take the system down for some electrica= l<BR> L work that needs to be done :^(&nbsp; This is of course assuming the heat do=	 esn't<BR> K take it out first (I was wondering if I'd have to shut down yesterday).<BR>  <BR>L &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&n=> bsp;&nbsp; &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Zane<BR> <BR>L I have the same problem with uptime.&nbsp; Our building usually has a maint=L enance outage annually.&nbsp; Our high-rise was taken down a week ago, so m=L y middle-rise area will go down soon.&nbsp; In the meantime, it's been over=L  a year, since my VAXstation 4000 - 60 has been up for 418 days&nbsp; (this=L  is now only used for my desktop transport to other nodes).&nbsp; Our produ=B ction ES40 has done 204 days and my development DS10 127 days.<BR> <BR>L Dave Sneddon must have very long UPS if he can weather power outages and bu=? ilding maintenance -- similar to all the other competitors.<BR>  <BR> Regards, Paddy<BR> </FONT>  </P>   <FONT SIZE=3D3><BR>  <BR>K ***********************************************************************<BR>  <BR>G "This electronic message and any attachments may contain privileged<BR> B and confidential information intended only for the use of the <BR>F addressees named above.  If you are not the intended recipient of <BR>G this email, please delete the message and any attachment and advise<BR> F the sender.  You are hereby notified that any use, dissemination, <BR>; distribution, reproduction of this email is prohibited.<BR>  <BR>E If you have received the email in error, please notify TransGrid <BR> E immediately.  Any views expressed in this email are those of the <BR> A individual sender except where the sender expressly and with <BR> G authority states them to be the views of TransGrid.  TransGrid uses<BR> B virus-scanning software but excludes any liability for viruses<BR>  contained in any attachment.<BR> <BR>@ Please note the email address for TransGrid personnel is now<BR>( firstname.lastname@transgrid.com.au"<BR> <BR>K ***********************************************************************<BR>  </FONT>  </BODY>  </HTML> ) ------_=_NextPart_001_01C58EA3.34CE2DCB--    ------------------------------    Date: 22 Jul 2005 03:47:08 -0700# From: "Galen" <gltackett@gmail.com> * Subject: Re: Congratulations to Ed Wilts !C Message-ID: <1122029228.099664.286700@o13g2000cwo.googlegroups.com>    Paddy O'Brien wrote:   > J > Dave Sneddon must have very long UPS if he can weather power outages and? > building maintenance -- similar to all the other competitors.  > % Maybe Dave has a nuclear-powered UPS.  :-)    ------------------------------  % Date: Fri, 22 Jul 2005 12:34:34 +0000 - From: David B Sneddon <dbsneddon@bigpond.com> * Subject: Re: Congratulations to Ed Wilts !( Message-ID: <42E0E7DA.90401@bigpond.com>  # O'Brien Paddy was overheard to say:  > T > Dave Sneddon must have very long UPS if he can weather power outages and building   4 maintenance -- similar to all the other competitors. >  > Regards, Paddy >   = We have had two "maintenance" sessions since the start of the ? reported uptime (when we moved buildings) which were handled by C the UPS but we did power down everything we could to make it easier  for the UPS.? We don't seem to have any power problems where we are -- we are @ on the same grid as a major hospital and this part of town seems pretty stable.   Regards, Dave.  --  D David B Sneddon (dbs)  VMS Systems Programmer  dbsneddon@bigpond.comD Sneddo's quick guide ...     http://www.users.bigpond.com/dbsneddon/D DBS freeware     http://www.users.bigpond.com/dbsneddon/software.htm   ------------------------------    Date: 22 Jul 2005 09:28:59 -0700$ From: "Ed Wilts" <ewilts@ewilts.org>* Subject: Re: Congratulations to Ed Wilts !B Message-ID: <1122049739.737201.26980@g43g2000cwa.googlegroups.com>  J > Dave Sneddon must have very long UPS if he can weather power outages and? > building maintenance -- similar to all the other competitors.   F My approach:  the cluster is split between two data centers.  I *have*? lost an entire data center due to a combination of a faulty UPS E combined with a car vs powerpole and again when we needed to do major F power maintenance.  Both times, the remaining half of the cluster keptE us going.  I'm now in the middle of a total migration from CI storage E to EVA and I will *not* take down the cluster to do it.  There's just - too much pride at stake - easy be damned :-).   F My uptime would be longer except that when I split the cluster betweenB the data centers (they were all in the same data center when I gotE here), I screwed it up and it cost me a cluster outage to fix it.  In E any case, Dave still has me beat and would still be on top even I got G the cluster split right.  Congrats to Dave - I know how much work it is F to get to the top of the list!  BTW, Dave, you can remove the links to4 some of the HP research labs - they've just been cutB (http://www.siliconvalley.com/mld/siliconvalley/12187326.htm) :-(.B On a related note, the VMScluster at my old job was up for about 8F years before they needed an outage for power maintenance at the end of 2004.  That must have hurt...    Cheers, 	    .../Ed    ------------------------------  % Date: Fri, 22 Jul 2005 13:06:38 -0400  From: norm.raphael@metso.com* Subject: Re: Congratulations to Ed Wilts !Q Message-ID: <OFA4DD7649.E9F86107-ON85257046.005DE422-85257046.005E3234@metso.com>   ? "Ed Wilts" <ewilts@ewilts.org> wrote on 07/22/2005 12:28:59 PM:   H > > Dave Sneddon must have very long UPS if he can weather power outages and A > > building maintenance -- similar to all the other competitors.  > H > My approach:  the cluster is split between two data centers.  I *have*A > lost an entire data center due to a combination of a faulty UPS G > combined with a car vs powerpole and again when we needed to do major H > power maintenance.  Both times, the remaining half of the cluster keptG > us going.  I'm now in the middle of a total migration from CI storage G > to EVA and I will *not* take down the cluster to do it.  There's just / > too much pride at stake - easy be damned :-).  > H > My uptime would be longer except that when I split the cluster betweenD > the data centers (they were all in the same data center when I gotG > here), I screwed it up and it cost me a cluster outage to fix it.  In G > any case, Dave still has me beat and would still be on top even I got I > the cluster split right.  Congrats to Dave - I know how much work it is H > to get to the top of the list!  BTW, Dave, you can remove the links to6 > some of the HP research labs - they've just been cutD > (http://www.siliconvalley.com/mld/siliconvalley/12187326.htm) :-(.  F Apparently (if he wasn't misquoted, he's not much of a grammarian ;) :  D Hurd told employees HP has become too complex for both customers andH employees. "In the worst case, there were 14 layers between the customer and I."    Aye, aye, aye!  D > On a related note, the VMScluster at my old job was up for about 8H > years before they needed an outage for power maintenance at the end of > 2004.  That must have hurt...  > 	 > Cheers,  >    .../Ed  >    ------------------------------  % Date: Fri, 22 Jul 2005 12:43:39 -0500 ' From: "Earl Lakia" <elakia@hotmail.com> + Subject: DEC Forms Calender and Time Panels 1 Message-ID: <67idnXAy_NfWrXzfRVn-pg@netnitco.net>   ? Anyone developed a Calender Panel or Clock panel for DEC Forms? D Something similar to what is available in the Visual Basic and other	 Microsoft 
 Environments?   D I can't justify writing my own, and it's ok to call me lazy as I am.   -earl    ------------------------------    Date: 22 Jul 2005 07:14:07 -0700' From: "tadamsmar" <tadamsmar@yahoo.com>  Subject: Modbus TCP for VMS?B Message-ID: <1122041647.878734.85250@g49g2000cwa.googlegroups.com>  E I am a developer for a VMS-based control system.  I need to interface E to devices that have a Modbus TCP interface.  We are considering this ( as an alternative to RS-232 interfacing.  A I don't have any experience with control and data acquistion over A TCP/IP, but it seems that just about every new instrument that we ? consider interfacing these days supports TCP/IP communications.   " Looks like I can get this package:  G http://www.ipact.com/Products/Modicon/TCP%20Communication%20Library.htm   5 but it will cost about $15,000 total for our 5 nodes.   D Anyone familiar with a "roll your own" or freeware approach to this?  2 Anyone have any experience with the Ipact product?   ------------------------------  % Date: Fri, 22 Jul 2005 10:36:41 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> 8 Subject: perfomance question about bool variables in cxx6 Message-ID: <42e0a27b$0$4437$9b622d9e@news.freenet.de>   Hello,  E i don't understand why the 64bit bool variable is slower as 8bit bool 	 variable.    Here is my test program code:    - 8bit bool variable       #include <lib$routines.h> 	   bool c;    unsigned __int64 a;   	   main ()    {            lib$init_timer ();  +           for ( a = 1 ; a < 42949672 ; a++)            {                    c = true;            };             lib$show_timer ();   }      - 64bit bool variable      #include <lib$routines.h>      #define false 0    #define true  1      unsigned __int64 c;    unsigned __int64 a;   	   main ()    {            lib$init_timer ();  +           for ( a = 1 ; a < 42949672 ; a++)            {                    c = true;            };             lib$show_timer ();   }   " And here is the perfomance output:   run test_bool_8 J  ELAPSED:    0 00:00:02.45  CPU: 0:00:02.45  BUFIO: 0  DIRIO: 0  FAULTS: 0   run test_bool_64J  ELAPSED:    0 00:00:02.85  CPU: 0:00:02.85  BUFIO: 0  DIRIO: 0  FAULTS: 0  A The focus is only for the bool variable. I'm alone on the system.   1 The compiler option was "cxx /noopt file_name.c".    Have anywhere a hint?   # Thank you very much for your help !            Klaus-D. Bohn    ------------------------------  % Date: Fri, 22 Jul 2005 12:29:02 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> < Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <42e0bcd4$0$12210$9b622d9e@news.freenet.de>   8 "Klaus-D. Bohn" <info@it-bcsb.de> schrieb im Newsbeitrag0 news:42e0a27b$0$4437$9b622d9e@news.freenet.de... > Hello, > G > i don't understand why the 64bit bool variable is slower as 8bit bool  > variable.  >  > Here is my test program code:  >  > - 8bit bool variable >  >     #include <lib$routines.h>  >   bool c;  >   unsigned __int64 a;  >  >   main ()  >   {  >           lib$init_timer (); > - >           for ( a = 1 ; a < 42949672 ; a++) 
 >           {  >                   c = true;  >           }; >  >           lib$show_timer (); >   }  >  >  > - 64bit bool variable  >  >   #include <lib$routines.h>  >  >   #define false 0  >   #define true  1  >  >   unsigned __int64 c;  >   unsigned __int64 a;  >  >   main ()  >   {  >           lib$init_timer (); > - >           for ( a = 1 ; a < 42949672 ; a++) 
 >           {  >                   c = true;  >           }; >  >           lib$show_timer (); >   }  > $ > And here is the perfomance output: >  > run test_bool_8 L >  ELAPSED:    0 00:00:02.45  CPU: 0:00:02.45  BUFIO: 0  DIRIO: 0  FAULTS: 0 >  > run test_bool_64L >  ELAPSED:    0 00:00:02.85  CPU: 0:00:02.85  BUFIO: 0  DIRIO: 0  FAULTS: 0 > C > The focus is only for the bool variable. I'm alone on the system.  > 3 > The compiler option was "cxx /noopt file_name.c".  >  > Have anywhere a hint?  > % > Thank you very much for your help !  >  >         Klaus-D. Bohn  >  >      stop, stop. stop.   H I find myself the solution. I forgett the compiler options "/model=ansi"   Now here are my new output:   > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi test_bool_64.c; > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi test_bool_8.c;  % cxxlink /model = ansi    test_bool_64 $ cxxlink /model = ansi    test_bool_8   run test_bool_8 J  ELAPSED:    0 00:00:00.59  CPU: 0:00:00.60  BUFIO: 0  DIRIO: 0  FAULTS: 0   run test_bool_64J  ELAPSED:    0 00:00:00.55  CPU: 0:00:00.56  BUFIO: 0  DIRIO: 0  FAULTS: 0  4 That's ok. Perfomance difference round about  7,x %.   Thanks       Klaus-D. Bohn    ------------------------------  % Date: Fri, 22 Jul 2005 07:00:19 -0600 4 From: Norman Lastovica <norman.lastovica@oracle.com>< Subject: Re: perfomance question about bool variables in cxx* Message-ID: <42E0EDE3.369D9024@oracle.com>  < /NOOPT?  why are you questioning performance *and* compiling9 your test program telling the compiler to *NOT* optimize? - How about /OPT/ARCH=HOST as a starting point.    "Klaus-D. Bohn" wrote: > : > "Klaus-D. Bohn" <info@it-bcsb.de> schrieb im Newsbeitrag2 > news:42e0a27b$0$4437$9b622d9e@news.freenet.de...
 > > Hello, > > I > > i don't understand why the 64bit bool variable is slower as 8bit bool 
 > > variable.  > > ! > > Here is my test program code:  > >  > > - 8bit bool variable > > ! > >     #include <lib$routines.h> 
 > >   bool c;  > >   unsigned __int64 a;  > > 
 > >   main ()  > >   {   > >           lib$init_timer (); > > / > >           for ( a = 1 ; a < 42949672 ; a++)  > >           {  > >                   c = true;  > >           }; > >   > >           lib$show_timer (); > >   }  > >  > >  > > - 64bit bool variable  > >  > >   #include <lib$routines.h>  > >  > >   #define false 0  > >   #define true  1  > >  > >   unsigned __int64 c;  > >   unsigned __int64 a;  > > 
 > >   main ()  > >   {   > >           lib$init_timer (); > > / > >           for ( a = 1 ; a < 42949672 ; a++)  > >           {  > >                   c = true;  > >           }; > >   > >           lib$show_timer (); > >   }  > > & > > And here is the perfomance output: > >  > > run test_bool_8 N > >  ELAPSED:    0 00:00:02.45  CPU: 0:00:02.45  BUFIO: 0  DIRIO: 0  FAULTS: 0 > >  > > run test_bool_64N > >  ELAPSED:    0 00:00:02.85  CPU: 0:00:02.85  BUFIO: 0  DIRIO: 0  FAULTS: 0 > > E > > The focus is only for the bool variable. I'm alone on the system.  > > 5 > > The compiler option was "cxx /noopt file_name.c".  > >  > > Have anywhere a hint?  > > ' > > Thank you very much for your help !  > >  > >         Klaus-D. Bohn  > >  > >  >  > stop, stop. stop.  > J > I find myself the solution. I forgett the compiler options "/model=ansi" >  > Now here are my new output:  > @ > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi > test_bool_64.c; @ > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi > test_bool_8.c; > ' > cxxlink /model = ansi    test_bool_64 & > cxxlink /model = ansi    test_bool_8 >  > run test_bool_8 L >  ELAPSED:    0 00:00:00.59  CPU: 0:00:00.60  BUFIO: 0  DIRIO: 0  FAULTS: 0 >  > run test_bool_64L >  ELAPSED:    0 00:00:00.55  CPU: 0:00:00.56  BUFIO: 0  DIRIO: 0  FAULTS: 0 > 6 > That's ok. Perfomance difference round about  7,x %. >  > Thanks >  >     Klaus-D. Bohn    --  	 - - - - - 0  opinions expressed here are mine and mine alone.  and certainly are not intended in any way to 0  express or represent any opinions or commitment  of oracle corporation.   *  norman lastovica / oracle rdb engineering   ------------------------------    Date: 22 Jul 2005 08:35:04 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen) < Subject: Re: perfomance question about bool variables in cxx3 Message-ID: <+pQIpHEF+wsD@eisner.encompasserve.org>   ` In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D. Bohn" <info@it-bcsb.de> writes: > Hello, > G > i don't understand why the 64bit bool variable is slower as 8bit bool  > variable.  >  > Here is my test program code:  >  > - 8bit bool variable >  >     #include <lib$routines.h>   G I am not a C programmer, but reasoning from the same question in Ada...   I ...much of what passes for "CPU time" is actually the time spent fetching H from and storing into main memory.  If memory speeds were instantaneous,- "CPU time" would become dramatically shorter.   F So if one is fetching and storing 8 times as much data, it is going toG take 8 times as long.  You may think that fetching 64 bits should be as H fast as fetching 8 bits, since registers are 64 bits wide.t  But in factF there is a lot of caching in modern CPUs (striving toward that goal ofH faster memory access) and especially for sequential access to successiveH memory locations much of the data will come from the faster cache.  WithH 8 times as much space consumed by a boolean, the cache bandwidth is used 8 times as heavily.    ------------------------------  % Date: Fri, 22 Jul 2005 16:52:49 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> < Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <42e0faa8$0$15152$9b622d9e@news.freenet.de>   , This is a multi-part message in MIME format.  + ------=_NextPart_000_004E_01C58EDD.CD47F990  Content-Type: text/plain;  	charset="iso-8859-1" + Content-Transfer-Encoding: quoted-printable     B "Larry Kilgallen" <Kilgallen@SpamCop.net> schrieb im Newsbeitrag =- news:+pQIpHEF+wsD@eisner.encompasserve.org... J > In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D. Bohn" = <info@it-bcsb.de> writes: 
 > > Hello, > >=20F > > i don't understand why the 64bit bool variable is slower as 8bit = bool
 > > variable.  > >=20! > > Here is my test program code:  > >=20 > > - 8bit bool variable > >=20! > >     #include <lib$routines.h>  >=20D > I am not a C programmer, but reasoning from the same question in = Ada... >=20D > ...much of what passes for "CPU time" is actually the time spent = fetching= > from and storing into main memory.  If memory speeds were =  instantaneous,/ > "CPU time" would become dramatically shorter.  >=20H > So if one is fetching and storing 8 times as much data, it is going toH > take 8 times as long.  You may think that fetching 64 bits should be = asG > fast as fetching 8 bits, since registers are 64 bits wide.t  But in =  factH > there is a lot of caching in modern CPUs (striving toward that goal ofA > faster memory access) and especially for sequential access to = 
 successiveG > memory locations much of the data will come from the faster cache.  =  WithG > 8 times as much space consumed by a boolean, the cache bandwidth is =  used > 8 times as heavily.    Yes, it is very interesting.  ? For a new application for OpenVMS im looking for increase the =  perfomance. B So i tested some little program code to look where is the really =# difference between 32bit and 64bit. ! I'm looking for the optimum code.    Here a next example:     - 32 bit   #include <lib$routines.h>    unsigned int a ,b; float ergebnis_1 ,ergebnis_2;    main ()  {          lib$init_timer ();  +         for ( a =3D 1 ; a < 42949672 ; a++) 	         { +                 ergebnis_1 =3D (a * a) / 2; 
         };           ergebnis_2 =3D (a * a);            lib$show_timer (); }      - 64 bit   #include <lib$routines.h>    unsigned __int64 a ,b; double ergebnis_1 ,ergebnis_2;   main ()  {          lib$init_timer ();  +         for ( a =3D 1 ; a < 42949672 ; a++) 	         { +                 ergebnis_1 =3D (a * a) / 2; 
         };           ergebnis_2 =3D (a * a);            lib$show_timer (); }     I cxx /noopt /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D ansi    = 
 test_32.c;I cxx /noopt /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D ansi    = 
 test_64.c;" cxxlink /model =3D ansi    test_32" cxxlink /model =3D ansi    test_64   run test_32 J  ELAPSED:    0 00:00:05.34  CPU: 0:00:05.35  BUFIO: 0  DIRIO: 0  FAULTS: = 0    run test_64 J  ELAPSED:    0 00:00:02.21  CPU: 0:00:02.21  BUFIO: 0  DIRIO: 0  FAULTS: = 1   ' I'm little uncertein if i do all right.               + ------=_NextPart_000_004E_01C58EDD.CD47F990  Content-Type: text/html; 	charset="iso-8859-1" + Content-Transfer-Encoding: quoted-printable   > <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML><HEAD>7 <META http-equiv=3DContent-Type content=3D"text/html; =  charset=3Diso-8859-1">9 <META content=3D"MSHTML 6.00.2800.1106" name=3DGENERATOR>  <STYLE></STYLE>  </HEAD>  <BODY>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>C <DIV><FONT face=3DArial size=3D2>"Larry Kilgallen" &lt;</FONT><A=20 ; href=3D"mailto:Kilgallen@SpamCop.net"><FONT face=3DArial=20 = size=3D2>Kilgallen@SpamCop.net</FONT></A><FONT face=3DArial =  size=3D2>&gt; schrieb im=20  Newsbeitrag </FONT><A = < href=3D"news:+pQIpHEF+wsD@eisner.encompasserve.org"><FONT=20 face=3DArial =F size=3D2>news:+pQIpHEF+wsD@eisner.encompasserve.org</FONT></A><FONT=20& face=3DArial size=3D2>...</FONT></DIV>A <DIV><FONT face=3DArial size=3D2>&gt; In article &lt;</FONT><A=20 @ href=3D"mailto:42e0a27b$0$4437$9b622d9e@news.freenet.de"><FONT = face=3DArial=20 C size=3D2>42e0a27b$0$4437$9b622d9e@news.freenet.de</FONT></A><FONT =  face=3DArial=20 . size=3D2>&gt;, "Klaus-D. Bohn" &lt;</FONT><A =( href=3D"mailto:info@it-bcsb.de"><FONT=20D face=3DArial size=3D2>info@it-bcsb.de</FONT></A><FONT face=3DArial = size=3D2>&gt;=20@ writes:<BR>&gt; &gt; Hello,<BR>&gt; &gt; <BR>&gt; &gt; i don't = understand why=20 = the 64bit bool variable is slower as 8bit bool<BR>&gt; &gt; =  variable.<BR>&gt;=20H &gt; <BR>&gt; &gt; Here is my test program code:<BR>&gt; &gt; <BR>&gt; =	 &gt; -=20 G 8bit bool variable<BR>&gt; &gt; <BR>&gt; &gt;&nbsp;&nbsp;&nbsp;&nbsp; =  #include=20 F &lt;lib$routines.h&gt;<BR>&gt; <BR>&gt; I am not a C programmer, but = reasoning=20J from the same question in Ada...<BR>&gt; <BR>&gt; ...much of what passes = for=20I "CPU time" is actually the time spent fetching<BR>&gt; from and storing =  into=20 F main memory.&nbsp; If memory speeds were instantaneous,<BR>&gt; "CPU = time" would=20I become dramatically shorter.<BR>&gt; <BR>&gt; So if one is fetching and = 
 storing=20J 8 times as much data, it is going to<BR>&gt; take 8 times as long.&nbsp; =
 You may=20E think that fetching 64 bits should be as<BR>&gt; fast as fetching 8 =  bits, since=20J registers are 64 bits wide.t&nbsp; But in fact<BR>&gt; there is a lot of =
 caching=20D in modern CPUs (striving toward that goal of<BR>&gt; faster memory = access) and=20I especially for sequential access to successive<BR>&gt; memory locations = 
 much of=20I the data will come from the faster cache.&nbsp; With<BR>&gt; 8 times as =  much=20 J space consumed by a boolean, the cache bandwidth is used<BR>&gt; 8 times = as=20  heavily.</FONT></DIV> 4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>2 <DIV><FONT face=3DArial size=3D2>Yes, it is very = interesting.</FONT></DIV> 4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>G <DIV><FONT face=3DArial size=3D2>For a new application for OpenVMS im =  looking for=20% increase the perfomance.</FONT></DIV> J <DIV><FONT face=3DArial size=3D2>So i tested some little program code to =
 look where=20 > is the really difference between 32bit and 64bit.</FONT></DIV>> <DIV><FONT face=3DArial size=3D2>I'm looking for the optimum = code.</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>. <DIV><FONT face=3DArial size=3D2>Here&nbsp;a = next&nbsp;example:</FONT></DIV> 4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>6 <DIV><FONT face=3DArial size=3D2>- 32 bit</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>+ <DIV><FONT face=3DArial size=3D2>#include = # &lt;lib$routines.h&gt;</FONT></DIV>  <DIV>&nbsp;</DIV> J <DIV><FONT face=3DArial size=3D2>unsigned int a ,b;<BR>float ergebnis_1=20 ,ergebnis_2;</FONT></DIV>  <DIV>&nbsp;</DIV> ( <DIV><FONT face=3DArial size=3D2>main=20G ()<BR>{<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; lib$init_timer=20  ();</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> <DIV><FONT face=3DArial = > size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; for ( a=20 =3D 1 ; a &lt; 42949672 ; = 5 a++)<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20 J {<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&n= bsp;&nbsp;&nbsp;&nbsp;=20  ergebnis_1 =3D (a * a) / =3 2;<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20  };</FONT></DIV>  <DIV>&nbsp;</DIV>  <DIV><FONT face=3DArial = 6 size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20$ ergebnis_2 =3D (a * a);</FONT></DIV> <DIV>&nbsp;</DIV>  <DIV><FONT face=3DArial = 6 size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20$ lib$show_timer ();<BR>}</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>6 <DIV><FONT face=3DArial size=3D2>- 64 bit</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>+ <DIV><FONT face=3DArial size=3D2>#include = # &lt;lib$routines.h&gt;</FONT></DIV>  <DIV>&nbsp;</DIV> C <DIV><FONT face=3DArial size=3D2>unsigned __int64 a ,b;<BR>double = 
 ergebnis_1=20  ,ergebnis_2;</FONT></DIV>  <DIV>&nbsp;</DIV> ( <DIV><FONT face=3DArial size=3D2>main=20G ()<BR>{<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; lib$init_timer=20  ();</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> <DIV><FONT face=3DArial = > size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; for ( a=20 =3D 1 ; a &lt; 42949672 ; = 5 a++)<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20 J {<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&n= bsp;&nbsp;&nbsp;&nbsp;=20  ergebnis_1 =3D (a * a) / =3 2;<BR>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20  };</FONT></DIV> 4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> <DIV><FONT face=3DArial = 6 size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20$ ergebnis_2 =3D (a * a);</FONT></DIV> <DIV>&nbsp;</DIV>  <DIV><FONT face=3DArial = 6 size=3D2>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20$ lib$show_timer ();<BR>}</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>C <DIV><FONT face=3DArial size=3D2>cxx /noopt /member /define =3D ( =  __NEW_STARLET=3D1 )=209 /model =3D ansi&nbsp;&nbsp;&nbsp; test_32.c;</FONT></DIV> C <DIV><FONT face=3DArial size=3D2>cxx /noopt /member /define =3D ( =  __NEW_STARLET=3D1 )=209 /model =3D ansi&nbsp;&nbsp;&nbsp; test_64.c;</FONT></DIV> : <DIV><FONT face=3DArial size=3D2>cxxlink&nbsp;/model =3D = ansi&nbsp;&nbsp;&nbsp;=20  test_32</FONT></DIV>: <DIV><FONT face=3DArial size=3D2>cxxlink&nbsp;/model =3D = ansi&nbsp;&nbsp;&nbsp;=20  test_64</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>& <DIV><FONT face=3DArial size=3D2>run =0 test_32<BR>&nbsp;ELAPSED:&nbsp;&nbsp;&nbsp; 0=20G 00:00:05.34&nbsp; CPU: 0:00:05.35&nbsp; BUFIO: 0&nbsp; DIRIO: 0&nbsp; = 
 FAULTS:=20 0</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>& <DIV><FONT face=3DArial size=3D2>run =0 test_64<BR>&nbsp;ELAPSED:&nbsp;&nbsp;&nbsp; 0=20G 00:00:02.21&nbsp; CPU: 0:00:02.21&nbsp; BUFIO: 0&nbsp; DIRIO: 0&nbsp; = 
 FAULTS:=20 1</FONT></DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>D <DIV><FONT face=3DArial size=3D2>I'm little uncertein if i do all=20 right.</FONT></DIV> 4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>4 <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>B <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV></BODY></HTML>  - ------=_NextPart_000_004E_01C58EDD.CD47F990--    ------------------------------  % Date: Fri, 22 Jul 2005 07:15:30 -0700 # From: "Tom Linden" <tom@kednos.com> < Subject: Re: perfomance question about bool variables in cxx( Message-ID: <opsubi74apzgicya@hyrrokkin>  J On Fri, 22 Jul 2005 16:52:49 +0200, Klaus-D. Bohn <info@it-bcsb.de> wrote:   > D > "Larry Kilgallen" <Kilgallen@SpamCop.net> schrieb im Newsbeitrag  / > news:+pQIpHEF+wsD@eisner.encompasserve.org... K >> In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D. Bohn"    >> <info@it-bcsb.de> writes: >> > Hello,  >> >J >> > i don't understand why the 64bit bool variable is slower as 8bit bool >> > variable. >> >" >> > Here is my test program code: >> > >> > - 8bit bool variable  >> >" >> >     #include <lib$routines.h> >>J >> I am not a C programmer, but reasoning from the same question in Ada... >>E >> ...much of what passes for "CPU time" is actually the time spent    >> fetching K >> from and storing into main memory.  If memory speeds were instantaneous, 0 >> "CPU time" would become dramatically shorter. >>I >> So if one is fetching and storing 8 times as much data, it is going to J >> take 8 times as long.  You may think that fetching 64 bits should be asK >> fast as fetching 8 bits, since registers are 64 bits wide.t  But in fact I >> there is a lot of caching in modern CPUs (striving toward that goal of K >> faster memory access) and especially for sequential access to successive K >> memory locations much of the data will come from the faster cache.  With K >> 8 times as much space consumed by a boolean, the cache bandwidth is used  >> 8 times as heavily. >  > Yes, it is very interesting. > K > For a new application for OpenVMS im looking for increase the perfomance. D > So i tested some little program code to look where is the really  % > difference between 32bit and 64bit. # > I'm looking for the optimum code.  >  > Here a next example: >  > 
 > - 32 bit >  > #include <lib$routines.h>  >  > unsigned int a ,b; > float ergebnis_1 ,ergebnis_2;  > 	 > main ()  > {  >         lib$init_timer (); > + >         for ( a = 1 ; a < 42949672 ; a++)  >         { + >                 ergebnis_1 = (a * a) / 2;  >         }; >  >         ergebnis_2 = (a * a);  >  >         lib$show_timer (); > }  >  > 
 > - 64 bit >  > #include <lib$routines.h>  >  > unsigned __int64 a ,b;  > double ergebnis_1 ,ergebnis_2; > 	 > main ()  > {  >         lib$init_timer (); > + >         for ( a = 1 ; a < 42949672 ; a++)  >         { + >                 ergebnis_1 = (a * a) / 2;  >         }; >  >         ergebnis_2 = (a * a);  >  >         lib$show_timer (); > }  >  > E > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi       > test_32.c;E > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi       > test_64.c;" > cxxlink /model = ansi    test_32" > cxxlink /model = ansi    test_64 > 
 > run test_32 L >  ELAPSED:    0 00:00:05.34  CPU: 0:00:05.35  BUFIO: 0  DIRIO: 0  FAULTS:   > 0  > 
 > run test_64 L >  ELAPSED:    0 00:00:02.21  CPU: 0:00:02.21  BUFIO: 0  DIRIO: 0  FAULTS:   > 1  > ) > I'm little uncertein if i do all right.  >  > J It could also be the conversion routines.  Have you looked at the assembly	 listings?    ------------------------------   Date: 22 Jul 2005 14:34:08 GMT$ From: "Doc." <doc@openvms-rocks.com>< Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <Xns969BA8A931D17dcovmsrox@212.100.160.126>   ) %NEWS-I-NEWMSG, Norman Lastovica wrote in " news:42E0EDE3.369D9024@oracle.com   > > /NOOPT?  why are you questioning performance *and* compiling; > your test program telling the compiler to *NOT* optimize? / > How about /OPT/ARCH=HOST as a starting point.   H Slight problem there - the compiler is so smart it optimises the 64 bit 3 program in Klaus' second test example to nothing...   ? $ cxx /noopt/member/define=( __NEW_STARLET=1)/model=ansi test.c  $ cxxlink/model=ansi test  $ r testI ELAPSED:    0 00:00:03.14  CPU: 0:00:03.15  BUFIO: 0  DIRIO: 0  FAULTS: 1   E $ cxx/opt/arch=host/member/define=(__NEW_STARLET=1)/model=ansi test.c  $ cxxlink/model=ansi test  $ r testI ELAPSED:    0 00:00:00.00  CPU: 0:00:00.00  BUFIO: 0  DIRIO: 0  FAULTS: 0    :-)      Doc. --  G OpenVMS:     Eight out of ten hackers prefer *other* operating systems. G http://www.openvms-rocks.com    Deathrow Public-Access OpenVMS Cluster.    ------------------------------  % Date: Fri, 22 Jul 2005 08:46:44 -0700 # From: "Tom Linden" <tom@kednos.com> < Subject: Re: perfomance question about bool variables in cxx( Message-ID: <opsubnf6nqzgicya@hyrrokkin>  J On Fri, 22 Jul 2005 18:47:46 +0200, Klaus-D. Bohn <info@it-bcsb.de> wrote:  8 > "Tom Linden" <tom@kednos.com> schrieb im Newsbeitrag  $ > news:opsubi74apzgicya@hyrrokkin...H >> On Fri, 22 Jul 2005 16:52:49 +0200, Klaus-D. Bohn <info@it-bcsb.de>  	 >> wrote:  >> >> >E >> > "Larry Kilgallen" <Kilgallen@SpamCop.net> schrieb im Newsbeitrag 2 >> > news:+pQIpHEF+wsD@eisner.encompasserve.org...H >> >> In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D.   >> Bohn" >> >> <info@it-bcsb.de> writes:  >> >> > Hello, >> >> > J >> >> > i don't understand why the 64bit bool variable is slower as 8bit   >> bool  >> >> > variable.  >> >> > % >> >> > Here is my test program code:  >> >> >  >> >> > - 8bit bool variable >> >> > % >> >> >     #include <lib$routines.h>  >> >> H >> >> I am not a C programmer, but reasoning from the same question in  	 >> Ada...  >> >> F >> >> ...much of what passes for "CPU time" is actually the time spent >> >> fetchingA >> >> from and storing into main memory.  If memory speeds were    >> instantaneous, 3 >> >> "CPU time" would become dramatically shorter.  >> >> K >> >> So if one is fetching and storing 8 times as much data, it is going    >> to L >> >> take 8 times as long.  You may think that fetching 64 bits should be   >> as K >> >> fast as fetching 8 bits, since registers are 64 bits wide.t  But in    >> fact K >> >> there is a lot of caching in modern CPUs (striving toward that goal    >> of E >> >> faster memory access) and especially for sequential access to   
 >> successive K >> >> memory locations much of the data will come from the faster cache.     >> With K >> >> 8 times as much space consumed by a boolean, the cache bandwidth is    >> used  >> >> 8 times as heavily.  >> >! >> > Yes, it is very interesting.  >> >D >> > For a new application for OpenVMS im looking for increase the   >> perfomance.E >> > So i tested some little program code to look where is the really ( >> > difference between 32bit and 64bit.& >> > I'm looking for the optimum code. >> > >> > Here a next example:  >> > >> >
 >> > - 32 bit  >> > >> > #include <lib$routines.h> >> > >> > unsigned int a ,b; " >> > float ergebnis_1 ,ergebnis_2; >> > >> > main () >> > { >> >         lib$init_timer ();  >> >. >> >         for ( a = 1 ; a < 42949672 ; a++) >> >         {. >> >                 ergebnis_1 = (a * a) / 2; >> >         };  >> >" >> >         ergebnis_2 = (a * a); >> > >> >         lib$show_timer ();  >> > } >> > >> >
 >> > - 64 bit  >> > >> > #include <lib$routines.h> >> > >> > unsigned __int64 a ,b; # >> > double ergebnis_1 ,ergebnis_2;  >> > >> > main () >> > { >> >         lib$init_timer ();  >> >. >> >         for ( a = 1 ; a < 42949672 ; a++) >> >         {. >> >                 ergebnis_1 = (a * a) / 2; >> >         };  >> >" >> >         ergebnis_2 = (a * a); >> > >> >         lib$show_timer ();  >> > } >> > >> >C >> > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi  >> > test_32.c; C >> > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi  >> > test_64.c; % >> > cxxlink /model = ansi    test_32 % >> > cxxlink /model = ansi    test_64  >> > >> > run test_32G >> >  ELAPSED:    0 00:00:05.34  CPU: 0:00:05.35  BUFIO: 0  DIRIO: 0    
 >> FAULTS: >> > 0 >> > >> > run test_64G >> >  ELAPSED:    0 00:00:02.21  CPU: 0:00:02.21  BUFIO: 0  DIRIO: 0    
 >> FAULTS: >> > 1 >> >, >> > I'm little uncertein if i do all right. >> > >> >F >> It could also be the conversion routines.  Have you looked at the   >> assembly  >> listings? >>+ > No, i haven't look for the assembly code. H > 64bit variables are faster than 32bit variables. Please see my short  B > test programs. The only difference is the variables definitions. > I hope my test is ok.   E Not quite, in the first you are converting 32 bit int to float and in F the second 64 bit int to double.  Now this may not be significant, but it could be.   ------------------------------  % Date: Fri, 22 Jul 2005 18:47:46 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> < Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <42e11595$0$15047$9b622d9e@news.freenet.de>   , This is a multi-part message in MIME format.  + ------=_NextPart_000_0095_01C58EED.DA3029B0  Content-Type: text/plain;  	charset="iso-8859-15"+ Content-Transfer-Encoding: quoted-printable   6 "Tom Linden" <tom@kednos.com> schrieb im Newsbeitrag =" news:opsubi74apzgicya@hyrrokkin...G > On Fri, 22 Jul 2005 16:52:49 +0200, Klaus-D. Bohn <info@it-bcsb.de> =  wrote: >=20 > > H > > "Larry Kilgallen" <Kilgallen@SpamCop.net> schrieb im Newsbeitrag =201 > > news:+pQIpHEF+wsD@eisner.encompasserve.org... G > >> In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D. = 	 Bohn" =20  > >> <info@it-bcsb.de> writes:
 > >> > Hello,  > >> >I > >> > i don't understand why the 64bit bool variable is slower as 8bit =  bool > >> > variable. > >> >$ > >> > Here is my test program code: > >> > > >> > - 8bit bool variable  > >> >$ > >> >     #include <lib$routines.h> > >>G > >> I am not a C programmer, but reasoning from the same question in =  Ada... > >>I > >> ...much of what passes for "CPU time" is actually the time spent =20 
 > >> fetching @ > >> from and storing into main memory.  If memory speeds were = instantaneous,2 > >> "CPU time" would become dramatically shorter. > >>J > >> So if one is fetching and storing 8 times as much data, it is going = toH > >> take 8 times as long.  You may think that fetching 64 bits should = be as J > >> fast as fetching 8 bits, since registers are 64 bits wide.t  But in = factJ > >> there is a lot of caching in modern CPUs (striving toward that goal = ofD > >> faster memory access) and especially for sequential access to =
 successiveJ > >> memory locations much of the data will come from the faster cache.  = WithJ > >> 8 times as much space consumed by a boolean, the cache bandwidth is = used > >> 8 times as heavily. > >   > > Yes, it is very interesting. > > C > > For a new application for OpenVMS im looking for increase the =  perfomance. H > > So i tested some little program code to look where is the really =20' > > difference between 32bit and 64bit. % > > I'm looking for the optimum code.  > >  > > Here a next example: > >  > >  > > - 32 bit > >  > > #include <lib$routines.h>  > >  > > unsigned int a ,b;! > > float ergebnis_1 ,ergebnis_2;  > >  > > main ()  > > {  > >         lib$init_timer (); > > / > >         for ( a =3D 1 ; a < 42949672 ; a++) 
 > >         { / > >                 ergebnis_1 =3D (a * a) / 2;  > >         }; > > # > >         ergebnis_2 =3D (a * a);  > >  > >         lib$show_timer (); > > }  > >  > >  > > - 64 bit > >  > > #include <lib$routines.h>  > >  > > unsigned __int64 a ,b;" > > double ergebnis_1 ,ergebnis_2; > >  > > main ()  > > {  > >         lib$init_timer (); > > / > >         for ( a =3D 1 ; a < 42949672 ; a++) 
 > >         { / > >                 ergebnis_1 =3D (a * a) / 2;  > >         }; > > # > >         ergebnis_2 =3D (a * a);  > >  > >         lib$show_timer (); > > }  > >  > > J > > cxx /noopt /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D ansi =    =20 > > test_32.c;J > > cxx /noopt /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D ansi =    =20 > > test_64.c;& > > cxxlink /model =3D ansi    test_32& > > cxxlink /model =3D ansi    test_64 > >  > > run test_32 F > >  ELAPSED:    0 00:00:05.34  CPU: 0:00:05.35  BUFIO: 0  DIRIO: 0  = FAULTS: =20  > > 0  > >  > > run test_64 F > >  ELAPSED:    0 00:00:02.21  CPU: 0:00:02.21  BUFIO: 0  DIRIO: 0  = FAULTS: =20  > > 1  > > + > > I'm little uncertein if i do all right.  > >  > > E > It could also be the conversion routines.  Have you looked at the =  assembly > listings?  >=20) No, i haven't look for the assembly code. F 64bit variables are faster than 32bit variables. Please see my short =@ test programs. The only difference is the variables definitions. I hope my test is ok.=20  + ------=_NextPart_000_0095_01C58EED.DA3029B0  Content-Type: text/html; 	charset="iso-8859-15"+ Content-Transfer-Encoding: quoted-printable   > <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML><HEAD>7 <META http-equiv=3DContent-Type content=3D"text/html; =  charset=3Diso-8859-15"> 9 <META content=3D"MSHTML 6.00.2800.1106" name=3DGENERATOR>  <STYLE></STYLE>  </HEAD>  <BODY>> <DIV><FONT face=3DArial size=3D2>"Tom Linden" &lt;</FONT><A=204 href=3D"mailto:tom@kednos.com"><FONT face=3DArial=20D size=3D2>tom@kednos.com</FONT></A><FONT face=3DArial size=3D2>&gt; =
 schrieb im=20 F Newsbeitrag </FONT><A href=3D"news:opsubi74apzgicya@hyrrokkin"><FONT = face=3DArial=20 H size=3D2>news:opsubi74apzgicya@hyrrokkin</FONT></A><FONT face=3DArial=20 size=3D2>...</FONT></DIV> D <DIV><FONT face=3DArial size=3D2>&gt; On Fri, 22 Jul 2005 16:52:49 = +0200, Klaus-D.=20H Bohn &lt;</FONT><A href=3D"mailto:info@it-bcsb.de"><FONT face=3DArial=20E size=3D2>info@it-bcsb.de</FONT></A><FONT face=3DArial size=3D2>&gt; =  wrote:<BR>&gt;=20 = <BR>&gt; &gt;<BR>&gt; &gt; "Larry Kilgallen" &lt;</FONT><A=20 ; href=3D"mailto:Kilgallen@SpamCop.net"><FONT face=3DArial=20 = size=3D2>Kilgallen@SpamCop.net</FONT></A><FONT face=3DArial =  size=3D2>&gt; schrieb im=20 , Newsbeitrag&nbsp; <BR>&gt; &gt; </FONT><A=20I href=3D"news:+pQIpHEF+wsD@eisner.encompasserve.org"><FONT face=3DArial=20 E size=3D2>news:+pQIpHEF+wsD@eisner.encompasserve.org</FONT></A><FONT =  face=3DArial=20 9 size=3D2>...<BR>&gt; &gt;&gt; In article &lt;</FONT><A=20 @ href=3D"mailto:42e0a27b$0$4437$9b622d9e@news.freenet.de"><FONT = face=3DArial=20 C size=3D2>42e0a27b$0$4437$9b622d9e@news.freenet.de</FONT></A><FONT =  face=3DArial=20 G size=3D2>&gt;, "Klaus-D. Bohn"&nbsp; <BR>&gt; &gt;&gt; &lt;</FONT><A=20 5 href=3D"mailto:info@it-bcsb.de"><FONT face=3DArial=20 E size=3D2>info@it-bcsb.de</FONT></A><FONT face=3DArial size=3D2>&gt; =  writes:<BR>&gt;=20J &gt;&gt; &gt; Hello,<BR>&gt; &gt;&gt; &gt;<BR>&gt; &gt;&gt; &gt; i don't =  G understand why the 64bit bool variable is slower as 8bit bool<BR>&gt; =  &gt;&gt;=20 G &gt; variable.<BR>&gt; &gt;&gt; &gt;<BR>&gt; &gt;&gt; &gt; Here is my =  test=20 H program code:<BR>&gt; &gt;&gt; &gt;<BR>&gt; &gt;&gt; &gt; - 8bit bool=201 variable<BR>&gt; &gt;&gt; &gt;<BR>&gt; &gt;&gt; =  &gt;&nbsp;&nbsp;&nbsp;&nbsp;=20 H #include &lt;lib$routines.h&gt;<BR>&gt; &gt;&gt;<BR>&gt; &gt;&gt; I am =
 not a C=20E programmer, but reasoning from the same question in Ada...<BR>&gt;=20 D &gt;&gt;<BR>&gt; &gt;&gt; ...much of what passes for "CPU time" is = actually the=20 G time spent&nbsp; <BR>&gt; &gt;&gt; fetching<BR>&gt; &gt;&gt; from and = 
 storing=20F into main memory.&nbsp; If memory speeds were instantaneous,<BR>&gt; = &gt;&gt;=20 H "CPU time" would become dramatically shorter.<BR>&gt; &gt;&gt;<BR>&gt; = &gt;&gt;=20 E So if one is fetching and storing 8 times as much data, it is going = 
 to<BR>&gt;=20 J &gt;&gt; take 8 times as long.&nbsp; You may think that fetching 64 bits =	 should=20 H be as<BR>&gt; &gt;&gt; fast as fetching 8 bits, since registers are 64 = bits=20 H wide.t&nbsp; But in fact<BR>&gt; &gt;&gt; there is a lot of caching in =	 modern=20 C CPUs (striving toward that goal of<BR>&gt; &gt;&gt; faster memory =  access) and=20H especially for sequential access to successive<BR>&gt; &gt;&gt; memory = locations=20F much of the data will come from the faster cache.&nbsp; With<BR>&gt; =
 &gt;&gt; 8=20 C times as much space consumed by a boolean, the cache bandwidth is =  used<BR>&gt;=20 I &gt;&gt; 8 times as heavily.<BR>&gt; &gt;<BR>&gt; &gt; Yes, it is very=20 J interesting.<BR>&gt; &gt;<BR>&gt; &gt; For a new application for OpenVMS = im=20 D looking for increase the perfomance.<BR>&gt; &gt; So i tested some =	 little=20 I program code to look where is the really&nbsp; <BR>&gt; &gt; difference = 
 between=20J 32bit and 64bit.<BR>&gt; &gt; I'm looking for the optimum code.<BR>&gt;=20J &gt;<BR>&gt; &gt; Here a next example:<BR>&gt; &gt;<BR>&gt; &gt;<BR>&gt; =	 &gt; -=20 J 32 bit<BR>&gt; &gt;<BR>&gt; &gt; #include &lt;lib$routines.h&gt;<BR>&gt; =  E &gt;<BR>&gt; &gt; unsigned int a ,b;<BR>&gt; &gt; float ergebnis_1=20 H ,ergebnis_2;<BR>&gt; &gt;<BR>&gt; &gt; main ()<BR>&gt; &gt; {<BR>&gt;=20E &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; lib$init_timer =  ();<BR>&gt;=20I &gt;<BR>&gt; &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; for ( =  a =3D 1 ;=20! a &lt; 42949672 ; a++)<BR>&gt;=20 A &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; {<BR>&gt;=20 J &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nb= sp;&nbsp;&nbsp;&nbsp;&nbsp;=20& ergebnis_1 =3D (a * a) / 2;<BR>&gt;=20A &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; };<BR>&gt; =  &gt;<BR>&gt;=20 J &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; ergebnis_2 =3D (a * =   a);<BR>&gt; &gt;<BR>&gt; =7 &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20 J lib$show_timer ();<BR>&gt; &gt; }<BR>&gt; &gt;<BR>&gt; &gt;<BR>&gt; &gt; = - 64=20 H bit<BR>&gt; &gt;<BR>&gt; &gt; #include &lt;lib$routines.h&gt;<BR>&gt;=20J &gt;<BR>&gt; &gt; unsigned __int64 a ,b;<BR>&gt; &gt; double ergebnis_1=20H ,ergebnis_2;<BR>&gt; &gt;<BR>&gt; &gt; main ()<BR>&gt; &gt; {<BR>&gt;=20E &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; lib$init_timer =  ();<BR>&gt;=20I &gt;<BR>&gt; &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; for ( =  a =3D 1 ;=20! a &lt; 42949672 ; a++)<BR>&gt;=20 A &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; {<BR>&gt;=20 J &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nb= sp;&nbsp;&nbsp;&nbsp;&nbsp;=20& ergebnis_1 =3D (a * a) / 2;<BR>&gt;=20A &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; };<BR>&gt; =  &gt;<BR>&gt;=20 J &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; ergebnis_2 =3D (a * =   a);<BR>&gt; &gt;<BR>&gt; =7 &gt;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20 J lib$show_timer ();<BR>&gt; &gt; }<BR>&gt; &gt;<BR>&gt; &gt;<BR>&gt; &gt; = cxx=20> /noopt /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D=20H ansi&nbsp;&nbsp;&nbsp;&nbsp; <BR>&gt; &gt; test_32.c;<BR>&gt; &gt; cxx =	 /noopt=20 6 /member /define =3D ( __NEW_STARLET=3D1 ) /model =3D = ansi&nbsp;&nbsp;&nbsp;&nbsp;=20 : <BR>&gt; &gt; test_64.c;<BR>&gt; &gt; cxxlink /model =3D = ansi&nbsp;&nbsp;&nbsp;=20 @ test_32<BR>&gt; &gt; cxxlink /model =3D ansi&nbsp;&nbsp;&nbsp; = test_64<BR>&gt;=202 &gt;<BR>&gt; &gt; run test_32<BR>&gt; &gt;&nbsp; = ELAPSED:&nbsp;&nbsp;&nbsp; 0=20 H 00:00:05.34&nbsp; CPU: 0:00:05.35&nbsp; BUFIO: 0&nbsp; DIRIO: 0&nbsp;=20= FAULTS:&nbsp; <BR>&gt; &gt; 0<BR>&gt; &gt;<BR>&gt; &gt; run =  test_64<BR>&gt;=20@ &gt;&nbsp; ELAPSED:&nbsp;&nbsp;&nbsp; 0 00:00:02.21&nbsp; CPU: = 0:00:02.21&nbsp;=20 E BUFIO: 0&nbsp; DIRIO: 0&nbsp; FAULTS:&nbsp; <BR>&gt; &gt; 1<BR>&gt; =  &gt;<BR>&gt;=20 C &gt; I'm little uncertein if i do all right.<BR>&gt; &gt;<BR>&gt; =  &gt;<BR>&gt;=20 I It could also be the conversion routines.&nbsp; Have you looked at the=20 0 assembly<BR>&gt; listings?<BR>&gt; </FONT></DIV>G <DIV><FONT face=3DArial size=3D2>No, i haven't look for the assembly=20  code.</FONT></DIV>H <DIV><FONT face=3DArial size=3D2>64bit variables are faster than 32bit =
 variables.=20 J Please see my short test programs. The only difference is the variables=20 definitions.</FONT></DIV> D <DIV><FONT face=3DArial size=3D2>I hope my test is ok. </FONT></DIV>B <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV></BODY></HTML>  - ------=_NextPart_000_0095_01C58EED.DA3029B0--    ------------------------------  % Date: Fri, 22 Jul 2005 19:01:33 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> < Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <42e118d0$0$15037$9b622d9e@news.freenet.de>   5 "Doc." <doc@openvms-rocks.com> schrieb im Newsbeitrag 1 news:Xns969BA8A931D17dcovmsrox@212.100.160.126... + > %NEWS-I-NEWMSG, Norman Lastovica wrote in # > news:42E0EDE3.369D9024@oracle.com  > @ > > /NOOPT?  why are you questioning performance *and* compiling= > > your test program telling the compiler to *NOT* optimize? 1 > > How about /OPT/ARCH=HOST as a starting point.  > I > Slight problem there - the compiler is so smart it optimises the 64 bit 5 > program in Klaus' second test example to nothing...  > A > $ cxx /noopt/member/define=( __NEW_STARLET=1)/model=ansi test.c  > $ cxxlink/model=ansi test 
 > $ r testK > ELAPSED:    0 00:00:03.14  CPU: 0:00:03.15  BUFIO: 0  DIRIO: 0  FAULTS: 1  > G > $ cxx/opt/arch=host/member/define=(__NEW_STARLET=1)/model=ansi test.c  > $ cxxlink/model=ansi test 
 > $ r testK > ELAPSED:    0 00:00:00.00  CPU: 0:00:00.00  BUFIO: 0  DIRIO: 0  FAULTS: 0  >  > :-)  >  >  > Doc. > --I > OpenVMS:     Eight out of ten hackers prefer *other* operating systems. I > http://www.openvms-rocks.com    Deathrow Public-Access OpenVMS Cluster.   
 Hello Doc,  K Yes, that's ok. The compiler is very intelligent. So, you must compile with 9 the option "/nooptimize" to force the short program code.   J cxx /opt  /member /define = ( __NEW_STARLET=1 ) /model = ansi    test_64.cJ cxx /opt  /member /define = ( __NEW_STARLET=1 ) /model = ansi    test_32.c! cxxlink  /model = ansi    test_64 ! cxxlink  /model = ansi    test_32  run test_32 J  ELAPSED:    0 00:00:00.00  CPU: 0:00:00.00  BUFIO: 0  DIRIO: 0  FAULTS: 0 run test_64 J  ELAPSED:    0 00:00:00.00  CPU: 0:00:00.00  BUFIO: 0  DIRIO: 0  FAULTS: 0   Klaus    ------------------------------  % Date: Fri, 22 Jul 2005 19:07:12 +0200 ' From: "Klaus-D. Bohn" <info@it-bcsb.de> < Subject: Re: perfomance question about bool variables in cxx7 Message-ID: <42e11a24$0$15073$9b622d9e@news.freenet.de>   4 "Tom Linden" <tom@kednos.com> schrieb im Newsbeitrag" news:opsubnf6nqzgicya@hyrrokkin...L > On Fri, 22 Jul 2005 18:47:46 +0200, Klaus-D. Bohn <info@it-bcsb.de> wrote: > 8 > > "Tom Linden" <tom@kednos.com> schrieb im Newsbeitrag& > > news:opsubi74apzgicya@hyrrokkin...H > >> On Fri, 22 Jul 2005 16:52:49 +0200, Klaus-D. Bohn <info@it-bcsb.de> > >> wrote:  > >> > >> >G > >> > "Larry Kilgallen" <Kilgallen@SpamCop.net> schrieb im Newsbeitrag 4 > >> > news:+pQIpHEF+wsD@eisner.encompasserve.org...H > >> >> In article <42e0a27b$0$4437$9b622d9e@news.freenet.de>, "Klaus-D.
 > >> Bohn"! > >> >> <info@it-bcsb.de> writes:  > >> >> > Hello,	 > >> >> > J > >> >> > i don't understand why the 64bit bool variable is slower as 8bit	 > >> bool  > >> >> > variable. 	 > >> >> > ' > >> >> > Here is my test program code: 	 > >> >> >  > >> >> > - 8bit bool variable	 > >> >> > ' > >> >> >     #include <lib$routines.h>  > >> >> H > >> >> I am not a C programmer, but reasoning from the same question in > >> Ada...  > >> >> H > >> >> ...much of what passes for "CPU time" is actually the time spent > >> >> fetchingA > >> >> from and storing into main memory.  If memory speeds were  > >> instantaneous, 5 > >> >> "CPU time" would become dramatically shorter.  > >> >> K > >> >> So if one is fetching and storing 8 times as much data, it is going  > >> to L > >> >> take 8 times as long.  You may think that fetching 64 bits should be > >> as K > >> >> fast as fetching 8 bits, since registers are 64 bits wide.t  But in 	 > >> fact K > >> >> there is a lot of caching in modern CPUs (striving toward that goal  > >> of E > >> >> faster memory access) and especially for sequential access to  > >> successive J > >> >> memory locations much of the data will come from the faster cache.	 > >> With K > >> >> 8 times as much space consumed by a boolean, the cache bandwidth is 	 > >> used  > >> >> 8 times as heavily.  > >> ># > >> > Yes, it is very interesting.  > >> >D > >> > For a new application for OpenVMS im looking for increase the > >> perfomance.G > >> > So i tested some little program code to look where is the really * > >> > difference between 32bit and 64bit.( > >> > I'm looking for the optimum code. > >> > > >> > Here a next example:  > >> > > >> > > >> > - 32 bit  > >> >  > >> > #include <lib$routines.h> > >> > > >> > unsigned int a ,b; $ > >> > float ergebnis_1 ,ergebnis_2; > >> > > >> > main () > >> > {! > >> >         lib$init_timer ();  > >> >0 > >> >         for ( a = 1 ; a < 42949672 ; a++) > >> >         {0 > >> >                 ergebnis_1 = (a * a) / 2; > >> >         };  > >> >$ > >> >         ergebnis_2 = (a * a); > >> >! > >> >         lib$show_timer ();  > >> > } > >> > > >> > > >> > - 64 bit  > >> >  > >> > #include <lib$routines.h> > >> > > >> > unsigned __int64 a ,b; % > >> > double ergebnis_1 ,ergebnis_2;  > >> > > >> > main () > >> > {! > >> >         lib$init_timer ();  > >> >0 > >> >         for ( a = 1 ; a < 42949672 ; a++) > >> >         {0 > >> >                 ergebnis_1 = (a * a) / 2; > >> >         };  > >> >$ > >> >         ergebnis_2 = (a * a); > >> >! > >> >         lib$show_timer ();  > >> > } > >> > > >> >E > >> > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi  > >> > test_32.c; E > >> > cxx /noopt /member /define = ( __NEW_STARLET=1 ) /model = ansi  > >> > test_64.c; ' > >> > cxxlink /model = ansi    test_32 ' > >> > cxxlink /model = ansi    test_64  > >> > > >> > run test_32F > >> >  ELAPSED:    0 00:00:05.34  CPU: 0:00:05.35  BUFIO: 0  DIRIO: 0 > >> FAULTS: > >> > 0 > >> > > >> > run test_64F > >> >  ELAPSED:    0 00:00:02.21  CPU: 0:00:02.21  BUFIO: 0  DIRIO: 0 > >> FAULTS: > >> > 1 > >> >. > >> > I'm little uncertein if i do all right. > >> > > >> >F > >> It could also be the conversion routines.  Have you looked at the
 > >> assembly  > >> listings? > >>- > > No, i haven't look for the assembly code. H > > 64bit variables are faster than 32bit variables. Please see my shortD > > test programs. The only difference is the variables definitions. > > I hope my test is ok.  > G > Not quite, in the first you are converting 32 bit int to float and in H > the second 64 bit int to double.  Now this may not be significant, but > it could be. >  That's correctly.   J in the first program I'm only using 32bit variables. The second program is1 using 64bit variables.That's only the difference. F That's was my test to see the difference between clean 32bit and 64bit
 program code.    ------------------------------    Date: 22 Jul 2005 02:18:03 -0700% From: "Rob" <ratkinson@tbs-ltd.co.uk> 2 Subject: Re: Performance issue after 7.3-2 upgradeC Message-ID: <1122023883.173377.141860@f14g2000cwb.googlegroups.com>   ? Jeff - so you know you're not alone, we're experiencing similar C problems, except our HSG's completely crash as well as other MNTVFY  type problems.   Rob.   ------------------------------    Date: 22 Jul 2005 02:17:56 -0700% From: "Rob" <ratkinson@tbs-ltd.co.uk> 2 Subject: Re: Performance issue after 7.3-2 upgradeC Message-ID: <1122023876.199306.308800@f14g2000cwb.googlegroups.com>   ? Jeff - so you know you're not alone, we're experiencing similar C problems, except our HSG's completely crash as well as other MNTVFY  type problems.   Rob.   ------------------------------    Date: 22 Jul 2005 02:30:46 -0700 From: dooleys@snowy.net.au/ Subject: Re: printing from MS Excel through VMS C Message-ID: <1122024646.106983.284650@z14g2000cwz.googlegroups.com>   ? I think I must have "mis-remembered" what I had done in similar 1 circumstances (PCL file output by an application) 6 and you are correct that the system logical is needed.2 The thing to do is move the "odd" queue to the end5 of tcpip_telnetsym$startup.com and define the logical  just before starting the queue.  Phil   ------------------------------    Date: 22 Jul 2005 08:15:36 -0700# From: "WhoDat?" <whohe@whoever.com> / Subject: Re: printing from MS Excel through VMS C Message-ID: <1122045336.286980.219330@z14g2000cwz.googlegroups.com>    denny wrote:0 > We are using TCPIP services for VMS (from HP). > A >   Compaq TCP/IP Services for OpenVMS Alpha Version V5.3 - ECO 4 0 >   on a AlphaServer ES40 running OpenVMS V7.3-1 > " > The symbiont is TCPIP$TELNETSYM. > # > We do not have DCPS at this site.  > + > Postscript is not an option at this site.  > ? > In the TCPIP manual, Chapter 25, it says (in part) "...unless I > TCPIP$TELNETSYM_RAW_TCP is defined for the queue."  How do I define one 3 > of these control logicals for a particular queue?     ? Define the logicals you need for each queue before starting it.    ------------------------------  # Date: Fri, 22 Jul 2005 17:06:08 GMT " From:   VAXman-  @SendSpamHere.ORG+ Subject: Re: simple image processing on VMS 0 Message-ID: <00A47246.96F523B1@SendSpamHere.ORG>  Z In article <dbp5io$opi$1@news01.intel.com>, Ken Fairfield <my.full.name@intel.com> writes: >Alphaman wrote:2 >> Phillip Helbig---remove CLOTHES to reply wrote: >>  J >>> I'm looking for an easy-to-use program for simple image processing on J >>> VMS.  In other words, something I can load a .JPG into and produce as J >>> output another .JPG, but with a different resolution and/or different I >>> size, gamma correction etc.  I also need the possibility to crop the  D >>> image.  That's all I really need, nothing fancy.  Best would be J >>> something which is already installed on VMS by default as part of the F >>> DECwindows apps or something.  Freeware and/or something from the B >>> hobbyist CD would also be an option.  What do folks recommend? >>> E >>> I really need just the bare minimum of features.  What I want is  5 >>> something which is easy to use and easy to learn.  >>>  >>   >>   >>  K >> Although a lot of other folks have recommended ImageMagick, I've got to  D >> put my opinion in, especially since you asked for "easy-to-use". F >> ImageMagick has a lot of potential, but after having spent 2 weeks E >> trying to get it built on OpenVMS, I have to say it's a nightmare.  >  ><Rest of complaint elided...> > C >Didn't VAXMAN post earlier in this thread that he uses ImageMagick ; >to process (semi-automatically) his digital camera images?  > B >Brian, what did you need to do to get ImageMagick working on VMS? >Can you share?   E Oh boy... tough question.  I'd an old release and I don't recall what F I needed to get it built.  It's done all that I want or need so I have" not looked at any recent releases.   $ run identify7 Version: @(#)ImageMagick 3.4 94/11/01 cristy@dupont.com    --  K VAXman- A Bored Certified VMS Kernel Mode Hacker   VAXman(at)TMESIS(dot)COM              5   "Well my son, life is like a beanstalk, isn't it?"     ------------------------------   End of INFO-VAX 2005.406 ************************