lmo@lsr-vax.UUCP ("Lance M. Optican - LMO") (01/31/91)
Following TFM, I diligently incluided the "-align16" flag in my compilations. Now I see info-iris mail saying you don't need this flag. What is the truth?! ----------------------------------------------+-------------------------- Lance M. Optican | uunet.uu.net!lsr-vax!lmo Laboratory of Sensorimotor Research | lsr-vax!lmo@uunet.uu.net National Eye Institute, NIH, Bethesda, MD | (301) 496-3549 ----------------------------------------------+--------------------------
Dan Karron@UCBVAX.BERKELEY.EDU (01/31/91)
I happly used the -align16 flag and did not notice anything. Then I started getting very weird crashes with innocent operations with chars' and shorts. DBX would indicate a buss error and not show anything wrong at the source line. Serious hacking showed that the errors were alignment errors that were caused by attempting to assign unaligned shorts to something. My queries to sgi indicated that the -align16, -align32 options were buggy. Then the question arose why use the -alignNN options, and I pointed to the man page. After more up and back with sgi, it was determined that you DON"T need the -alignNN option. You do need 16 BYTE alignment for somethings, but I don't know what for. This was a weird one, and many people apparently can live very happly without ever having problems with the -alignNN options. I appreciate sgi being honest and admiting the problem. >Date: Thu, 31 Jan 91 09:10:57 EST >From: "Lance M. Optican - LMO" <uunet.uu.net!lsr-vax!lmo> >Message-Id: <9101311410.AA02066@> >To: uunet.uu.net!uunet!brl.mil!info-iris >Subject: Parallel Programming > > > Following TFM, I diligently incluided the "-align16" flag >in my compilations. Now I see info-iris mail saying you don't need >this flag. > > What is the truth?! > >----------------------------------------------+-------------------------- > Lance M. Optican | uunet.uu.net!lsr-vax!lmo > Laboratory of Sensorimotor Research | lsr-vax!lmo@uunet.uu.net > National Eye Institute, NIH, Bethesda, MD | (301) 496-3549 >----------------------------------------------+-------------------------- > +-----------------------------------------------------------------------------+ | karron@nyu.edu (E-mail alias that will always find me) | | Fax: 212 263 7190 * Dan Karron, Research Associate | | . . . . . . . . . . . . . . * New York University Medical Center | | 560 First Avenue \*\ Pager <1> (212) 397 9330 | | New York, New York 10016 \**\ <2> 10896 <3> <your-number-here> | | (212) 263 5210 \***\_________________________________________ | | Main machine: karron.med.nyu.edu (128.122.135.3) IRIS 85GT | +-----------------------------------------------------------------------------+ NOTE PHONE NUMBER CHANGE: The Med Ctr has changed from 340 to 263 exchange.
davea@quasar.wpd.sgi.com (David B.Anderson) (02/01/91)
In article <9101311410.AA02066@> lmo@lsr-vax.UUCP ("Lance M. Optican - LMO") writes: > > Following TFM, I diligently incluided the "-align16" flag >in my compilations. Now I see info-iris mail saying you don't need >this flag. > > What is the truth?! The truth is that there was a miscommunication between the compiler group and the people preparing the manual: Parallel Programming on Silicon Graphics Computer Systems 007-0770-010 at: Understanding the Basics of Parallel Programming page 2-9, 2.2.2 Compiler Flags for Multi-Thread Code The section is wrong. There is no applicable compiler flag. Please do not use -align16! Dan Karron (karron@nyu.edu) noticed that there was something wrong and brought it to our attention December 26, 1990. Thanks Dan! We are sorry about this mistake. Regards, [ David B. Anderson Silicon Graphics (415)335-1548 davea@sgi.com ] [``What can go wrong?'' --Calvin to Hobbes]