In the example below, Solaris5.10 might be SparcSolaris5.10 or
Solaris5.10-sparc to make it a full platform specification. And Linux
might be Linux-x86 or SLES15-x86, etc. I agree with the decision not to
enforce an exact naming system, but our examples could be better.
We might have gone overboard in providing the compiler hook. But it could
indeed be a place to say something about not only the compiler used but
also switches. For example gcc3.2.3-g for the debuggable object and
gcc3.2.3-o2 for the optimized one. It also could give a hint about
required runtime libraries.
Mike
From: Mirmak, Michael <michael.mirmak@xxxxxxxxx>
Sent: Tuesday, July 3, 2018 1:46 PM
To: Mike LaBonte <mlabonte@xxxxxxxxxx>; ibis-editorial@xxxxxxxxxxxxx
Subject: RE: [ibis-editorial] Another editorial issue with 6.1
Mike,
Thanks! Would you say that "platform" conceptually should include the
processor architecture in some way? Or something else?
I am also assuming that "compiler" really can be any arbitrary string and
that no strict format applies to the contents, outside of the
dash-vs-underscore rules. I have been assuming that the entire objective
of supporting multiple compilers for the same architecture and operating
system is to allow different potential optimizations to be made available
to the tool user at simulation time, such as:
Solaris5.10_HyPrecizionSlowBinaryCompiler2.3.3_64
Solaris5.10_SpeeDeeBinaryCompiler7.2_64
Correct?
* MM
From: Mike LaBonte <mlabonte@xxxxxxxxxx <mailto:mlabonte@xxxxxxxxxx> >
Sent: Tuesday, July 03, 2018 10:30 AM
To: Mirmak, Michael <michael.mirmak@xxxxxxxxx
<mailto:michael.mirmak@xxxxxxxxx> >; ibis-editorial@xxxxxxxxxxxxx
<mailto:ibis-editorial@xxxxxxxxxxxxx>
Subject: RE: [ibis-editorial] Another editorial issue with 6.1
Sounds good, and I've added it to ver6_1_known_issues.docx
<http://ibis.org/ver6.1/ver6_1_known_issues.docx> . At first I was even
wondering if we could stop using the term "platform" altogether, but I see
that Platform_Compiler_Bits is used explicitly as part of Executable. And
yet, none of the examples we give spell out a complete platform:
Linux_gcc3.2.3_32
Solaris5.10_gcc4.1.1_64
Solaris_cc5.7_32
Windows_VisualStudio7.1.3088_32
HP-UX_accA.03.52_32
But we don't mandate exact platform or compiler specifications, so at the
end of the day tools probably just look at the files to see which one
looks like it might load and run.
Mike
From: ibis-editorial-bounce@xxxxxxxxxxxxx
<mailto:ibis-editorial-bounce@xxxxxxxxxxxxx>
<ibis-editorial-bounce@xxxxxxxxxxxxx
<mailto:ibis-editorial-bounce@xxxxxxxxxxxxx> > On Behalf Of Mirmak,
Michael
Sent: Tuesday, July 3, 2018 1:15 PM
To: ibis-editorial@xxxxxxxxxxxxx <mailto:ibis-editorial@xxxxxxxxxxxxx>
Subject: [ibis-editorial] Another editorial issue with 6.1
I believe there's a minor editorial issue with the usage of "platform" vs.
"operating system" and "architecture" in IBIS 6.1. While there are a few
fixes that could be made in this area, the most striking one is on page.
172:
The EDA tool will check for the compiler information and verify if the
executable model file is compatible with the operating system and
platform.
Multiple occurrences, without duplication, of Executable are permitted to
allow for providing executable model files for as many combinations of
operating system platforms and compilers for the same algorithmic model.
I believe this should be:
The EDA tool will check for the compiler information and verify if the
executable model file is compatible with the platform (operating system
and architecture).
Multiple occurrences, without duplication, of Executable are permitted, to
support executable model files for many combinations of operating system,
architecture, and compiler for the same algorithmic model.
* MM
