Article View: bit.listserv.asm370
Article #203Re: Linkage convention
From: Craddock, Chris
Date: Mon, 17 Oct 2005 16:40
Date: Mon, 17 Oct 2005 16:40
38 lines
3533 bytes
3533 bytes
> The GETMAIN code has been rock solid for decades, and has also been > highly optimized in its execution paths. Well yes, that's basically true. However, it is also true that GETMAIN and its cousins are among the most expensive system functions in terms of raw instruction path length. They are heavy on serialization and on searching the various structures used for accounting and managing virtual storage. > If you need to do tens of thousands of GETMAINs per second in disabled > code, then maybe you might want to think about eliminating them. Disabled code can only obtain or release page-fixed, or disabled reference common storage. Doing any more than the barest, of bare minimum storage management, while running disabled, is an utterly appalling idea. > But if you have well-debugged infrastructure code to use software stacks, > by all means use it. But don't bother trying to build your own such > code. I don't think it's worth the development/debug/support effort. The funny thing is, everyone who finds themselves in the business of writing software that runs in these sorts of environments DOES end up writing some form of stack management infrastructure, or they wear the cost of less-than-stellar performance. > And you can reduce the overhead of GETMAINs by using the PC entry form > rather than the SVC form. Paradoxically, even though a PC is a lot faster than an SVC, the STORAGE macro supports a lot more flexible entry conditions and does a lot more state checking than GETMAIN and as a result it is slightly (but measurably) slower than GETMAIN. CC
Technical Details
Message-ID:
<E9DE7963E5EA6546B42A979EC28B4D011565DD18@hou-ex-02.adprod.bmc.com>Path:
arc1-ca.pugleaf.net!g2news1.google.com!news3.google.com!border1.nntp.dca.giganews.com!nntp.giganews.com!wns14feed!worldnet.att.net!4.24.21.153!newsfeed3.dallas1.level3.net!news.level3.com!newsfeed1.easynews.com!easynews.com!easynews!sn-xit-02!sn-xit-12!sn-xit-09!sn-xit-06!sn-xit-14!supernews.com!149.173.6.116.MISMATCH!foggy!vm.sas.com!bmc.com!Chris_CraddockShow Full Headers
Complete Article Headers
X-Google-Thread: fcdcb,91cbe6b24531a1d8
X-Google-Attributes: gidfcdcb,public
X-Google-Language: ENGLISH,ASCII-7-bit
Path: arc1-ca.pugleaf.net!g2news1.google.com!news3.google.com!border1.nntp.dca.giganews.com!nntp.giganews.com!wns14feed!worldnet.att.net!4.24.21.153!newsfeed3.dallas1.level3.net!news.level3.com!newsfeed1.easynews.com!easynews.com!easynews!sn-xit-02!sn-xit-12!sn-xit-09!sn-xit-06!sn-xit-14!supernews.com!149.173.6.116.MISMATCH!foggy!vm.sas.com!bmc.com!Chris_Craddock
From: "Craddock, Chris" <Chris_Craddock@bmc.com>
Newsgroups: bit.listserv.asm370
Subject: Re: Linkage convention
Date: Mon, 17 Oct 2005 11:40:23 -0500
Organization: SAS Inc.
Lines: 38
Sender: IBM Mainframe Assembler List <ASSEMBLER-LIST@LISTSERV.UGA.EDU>
Message-ID: <E9DE7963E5EA6546B42A979EC28B4D011565DD18@hou-ex-02.adprod.bmc.com>
NNTP-Posting-Host: mozart.unx.sas.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Trace: foggy.unx.sas.com 1129567579 30316 10.19.150.41 (17 Oct 2005 16:46:19 GMT)
X-Complaints-To: usenet@unx.sas.com
NNTP-Posting-Date: Mon, 17 Oct 2005 16:46:19 +0000 (UTC)
Comments: Gated by NETNEWS@VM.SAS.COM
Return-Path: <owner-assembler-list@LISTSERV.UGA.EDU>
X-MimeOLE: Produced By Microsoft Exchange V6.0.6603.0
Content-class: urn:content-classes:message
Thread-Topic: Re: Linkage convention
Thread-Index: AcXTNdks+mQRcI/MTzGOwH8ib01lrgAAkPuw
X-OriginalArrivalTime: 17 Oct 2005 16:40:24.0666 (UTC)
FILETIME=[79016BA0:01C5D339]
X-Spamicity: Yes, tests=bogofilter, spamicity=1.000000, version=0.94.14
X-MIME-Autoconverted: from quoted-printable to 8bit by listserv.cc.uga.edu id
j9HGeVld029831
Approved-By: "Craddock, Chris" <Chris_Craddock@BMC.COM>
Return-Path: owner-assembler-list@LISTSERV.UGA.EDU
X-Mime-Autoconverted: from 8bit to 7bit by mime827
Originator: daemon@mozart.unx.sas.com