[Simh] The new Makefile
Larry Baker
baker at usgs.gov
Fri Sep 19 15:24:30 EDT 2008
I have not compared the SIMH V3.8 Makefile with the replacement
proposed by Philipp. I have converted the SIMH V3.4 descrip.mms file
(the OpenVMS equivalent of a Makefile) to a DCL command procedure (the
OpenVMS equivalent of a shell script or batch file), make.com, for
compiling with DEC C on OpenVMS/Alpha and OpenVMS/VAX. I compiled
everything using our 333 MHz AlphaServer 1000A 5/333 (OpenVMS V7.2-1)
and MicroVAX 4000-100 (OpenVMS V6.2) cluster. As you know, these are
VERY slow computers compared to any modern PC or Mac. I did not find
the time it took to compile objectionable. I liked that I could see
everything that went into each simulator in one command line.
My experience is that I recompile the simulator I use (MicroVAX) maybe
once or twice when a new release looks interesting. I don't usually
have to do that more than once a year or two. I am not concerned with
the time that takes. I am concerned about the performance of the
simulator. My preference is to provide the compiler with every
opportunity at compile time to optimize the generated code,
specifically, by aggressively inlining procedures and, when that is
not possible, by making full use of inter-procedural optimizations
(IPOs). My impression is that the practice of separately compiling
every module makes this more difficult, if not impossible, which makes
me favor the current Makefile method. Of course this assumes SIMH
benefits from inlining and IPOs. My bias, when choices have to be
made, is to prefer alternatives that make the SIMH executable run
faster, even if that increases the compilation time.
Larry Baker
US Geological Survey
650-329-5608
baker at usgs.gov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20080919/7a60c48f/attachment-0003.html>
More information about the Simh
mailing list