Thread View: bit.listserv.asm370
6 messages
6 total messages
Started by Miklos Szigetvari
Thu, 28 Sep 2006 11:27
Baseless code
Author: Miklos Szigetvari
Date: Thu, 28 Sep 2006 11:27
Date: Thu, 28 Sep 2006 11:27
28 lines
2719 bytes
2719 bytes
Hi I have just converted a large assembler application to baseless code: - brach relative instructions - data completly separated from instruction part of the code - SYSSTATE ARCHLVL=2 compiled For my big surprise in our z/800 machine the firts test shows two times higher CPU consumtion for the baseless code as in the original code -- Mit freundlichen Grüßen / Best regards Miklos Szigetvari ISIS Information Systems GmbH Alter Wienerweg 12 2344 Maria Enzersdorf Austria --------------------------------------------------------------- This e-mail is only intended for the recipient and not legally binding. Unauthorised use, publication, reproduction or disclosure of the content of this e-mail is not permitted. This email has been checked for known viruses, but ISIS accepts no responsibility for malicious or inappropriate content. ---------------------------------------------------------------
Re: Baseless code
Author: Martin Trübner
Date: Thu, 28 Sep 2006 11:58
Date: Thu, 28 Sep 2006 11:58
24 lines
2355 bytes
2355 bytes
Miklos, >> ...conversion ...two times higher CPU ...<< I can not believe this. Would you elaborate? relative instructions should have about the same timings as their bddd brothers (and sisters) - what else did you do? seaparate data and instructions.... that should not have any consequences (I do understand the result of a pagefault and the effect of an instruction next to the data avoiding the pagefault vs. data far far away (and causing pagefault on first ref). ARCHLEVEL=2---sorry I am more a person from the other opsys- does this have major impact on the pathlength? -- Martin -- XML2PDF - the way to get all features of PDF into your documents on mainframe or PC systems; more at http://www.pi-sysprog.de
Re: Baseless code
Author: Miklos Szigetvari
Date: Thu, 28 Sep 2006 12:16
Date: Thu, 28 Sep 2006 12:16
58 lines
3868 bytes
3868 bytes
Martin Trübner wrote: > Miklos, > > >> ...conversion ...two times higher CPU ...<< > > I can not believe this. > I thought maybe the z/800 machine is not so "smart". I will check again code, I think I made somewhere a mistake. It is a long running STC server, and I have compared the "startup" CPU times. As far as I remember , I don't made any significant change. The conversion was quite "mechanic" . With ARCHLEVEL=2 most of the system macros compiled well with baseless code. > Would you elaborate? > > relative instructions should have about the same timings as their bddd > brothers (and sisters) - what else did you do? > > seaparate data and instructions.... that should not have any > consequences (I do understand the result of a pagefault and the effect > of an instruction next to the data avoiding the pagefault vs. data far > far away (and causing pagefault on first ref). > > ARCHLEVEL=2---sorry I am more a person from the other opsys- does this > have major impact on the pathlength? > > -- > Martin > -- > XML2PDF - the way to get all features of PDF into your documents > on mainframe or PC systems; more at http://www.pi-sysprog.de > -- Mit freundlichen Grüßen / Best regards Miklos Szigetvari ISIS Information Systems GmbH Alter Wienerweg 12 2344 Maria Enzersdorf Austria --------------------------------------------------------------- This e-mail is only intended for the recipient and not legally binding. Unauthorised use, publication, reproduction or disclosure of the content of this e-mail is not permitted. This email has been checked for known viruses, but ISIS accepts no responsibility for malicious or inappropriate content. ---------------------------------------------------------------
Re: Baseless code
Author: Edward Jaffe
Date: Thu, 28 Sep 2006 13:27
Date: Thu, 28 Sep 2006 13:27
23 lines
2309 bytes
2309 bytes
Miklos Szigetvari wrote: > I have just converted a large assembler application to baseless code: > - brach relative instructions > - data completly separated from instruction part of the code > - SYSSTATE ARCHLVL=2 compiled > > For my big surprise in our z/800 machine the firts test shows two times > higher CPU consumtion for the baseless code as in the original code Hard to believe. I've done many similar conversions and have never experienced anything like this. Any chance you used instructions with long displacements? Though supported, they run extremely slowly on a z800 machine. -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 edjaffe@phoenixsoftware.com http://www.phoenixsoftware.com/
Re: Baseless code
Author: Miklos Szigetvari
Date: Thu, 28 Sep 2006 13:30
Date: Thu, 28 Sep 2006 13:30
6 lines
2091 bytes
2091 bytes
Hi Sorry, sorry everything o.k, In the baseless version there was a DEBUG flag switched on. Now the baseless code needs 0.4 second CPU and the old code 0.49 CPU.
Re: Baseless code
Author: Miklos Szigetvari
Date: Thu, 28 Sep 2006 14:21
Date: Thu, 28 Sep 2006 14:21
54 lines
3784 bytes
3784 bytes
Edward Jaffe wrote: > Miklos Szigetvari wrote: > >> I have just converted a large assembler application to baseless code: >> - brach relative instructions >> - data completly separated from instruction part of the code >> - SYSSTATE ARCHLVL=2 compiled >> >> For my big surprise in our z/800 machine the firts test shows two times >> higher CPU consumtion for the baseless code as in the original code > > > Hard to believe. I've done many similar conversions and have never > experienced anything like this. > > Any chance you used instructions with long displacements? Though > supported, they run extremely slowly on a z800 machine. > I'm just thinking about this. ( I run out of some work area BASE , and the option was to use long displacements) But at least Ed pointed out a hardware difference. Sorry again, in my case it was a wrong DEBUG flag in the baseless code > -- > Edward E Jaffe > Phoenix Software International, Inc > 5200 W Century Blvd, Suite 800 > Los Angeles, CA 90045 > 310-338-0400 x318 > edjaffe@phoenixsoftware.com > http://www.phoenixsoftware.com/ > -- Mit freundlichen Grüßen / Best regards Miklos Szigetvari ISIS Information Systems GmbH Alter Wienerweg 12 2344 Maria Enzersdorf Austria --------------------------------------------------------------- This e-mail is only intended for the recipient and not legally binding. Unauthorised use, publication, reproduction or disclosure of the content of this e-mail is not permitted. This email has been checked for known viruses, but ISIS accepts no responsibility for malicious or inappropriate content. ---------------------------------------------------------------
Thread Navigation
This is a paginated view of messages in the thread with full content displayed inline.
Messages are displayed in chronological order, with the original post highlighted in green.
Use pagination controls to navigate through all messages in large threads.
Back to All Threads