jessea@dynasys.UUCP (Jesse W. Asher) (02/22/90)
I was wondering if programs compiled under unix and then moved to
a unix system would run slower than if they had been compiled
under the unix system. I'm not asking if a program will run
slower or faster on a unix or xenix system. What I'm asking really
is "Does the process of translating xenix programs on a unix system
slow the program down?" I was going to give someone a program I
had compiled on a xenix system, but he refused it saying that he
wanted to compile it on his system (unix) because the program
would run slower. It this true? Thanx in advance.
--
Jesse W. Asher - Dynasys - (901)382-1705
6196-1 Macon Rd., Suite 200, Memphis, TN 38134
UUCP: {uunet,fedeva,rayo}!dynasys!jessearogerk@sco.COM (Roger Knopf 5502) (02/28/90)
In article <57@dynasys.UUCP> jessea@dynasys.UUCP (Jesse W. Asher) writes: >I was wondering if programs compiled under unix and then moved to >a unix system would run slower than if they had been compiled >under the unix system. I'm not asking if a program will run >slower or faster on a unix or xenix system. What I'm asking really >is "Does the process of translating xenix programs on a unix system >slow the program down?" I was going to give someone a program I >had compiled on a xenix system, but he refused it saying that he >wanted to compile it on his system (unix) because the program >would run slower. It this true? Thanx in advance. There is nothing inherent in a Xenix (x.out) 386 binary that would cause it to run slower under SCO Unix. The common assumption stated above, that Xenix programs are "translated" to run under SCO Unix, is absolutely incorrect for 386 programs. SCO went to a lot of trouble to put true native execution of Xenix binaries in SCO Unix. There is no translation or emulation. Full support is there, just like it was in Xenix. 286 binaries need translation for system calls under SCO Unix. Xenix 386 handled system calls from 286 binaries directly in the kernel, but when SCO Unix execs a 286 binary, /bin/i286emul is invoked to run the program. If the program in question does a lot of system calls, it may be slower than under Xenix. If the program is medium or large model 286, system call translation is not nearly as much of an issue as having 286 code - recompiling as 386 would make it faster, but not necessarily because of translation. This is less true for small model programs. Roger Knopf SCO Consulting Services rogerk@sco.com or {uunet,sun}!sco!rogerk
usenet@cps3xx.UUCP (Usenet file owner) (03/02/90)
From article <2750@scorn.sco.COM>, by rogerk@sco.COM (Roger Knopf 5502): $ $ In article <57@dynasys.UUCP> jessea@dynasys.UUCP (Jesse W. Asher) writes: $>I was wondering if programs compiled under unix and then moved to $>a unix system would run slower than if they had been compiled $>under the unix system. I'm not asking if a program will run $>slower or faster on a unix or xenix system. What I'm asking really $>is "Does the process of translating xenix programs on a unix system $>slow the program down?" I was going to give someone a program I $>had compiled on a xenix system, but he refused it saying that he $>wanted to compile it on his system (unix) because the program $>would run slower. It this true? Thanx in advance. From some experiments I did when I first got SCO UNIX the answer to the last question is YES. The reason for this appears to be more compiler specific than OS specific. I found that binaries I generated under XENIX 386/2.3.2 ran slower under UNIX than Xenix. When the same program was compiled under UNIX and created either a Xenix or a UNIX binary it ran faster un UNIX than the one compiled under Xenix. I believe this is more because of improvements in the compiler from MSC v4.0 to MSC v5.1. j |%|John Lawitzke, Dale Computer Corp., R&D |%|UUCP: uunet!frith!dale1!jhl Work |%| uunet!frith!ipecac!jhl Home Inquiring minds just wondering. |%|Internet: jhl@frith.egr.msu.edu