Re: End of optimization...

Robert@Knighten.org (Robert L. Knighten)
10 Aug 2003 11:01:51 -0400

          From comp.compilers

Related articles
[7 earlier articles]
Re: End of optimization... nej22@cl.cam.ac.uk (2003-07-25)
Re: End of optimization... fjh@cs.mu.oz.au (Fergus Henderson) (2003-07-25)
Re: End of optimization... Robert@Knighten.org (2003-07-31)
Re: End of optimization... channahr@comcast.net (norm@plumpirate.com) (2003-08-04)
Re: End of optimization... qsmgmt@earthlink.net (Alan Lehotsky) (2003-08-04)
Re: End of optimization... fjh@cs.mu.oz.au (Fergus Henderson) (2003-08-04)
Re: End of optimization... Robert@Knighten.org (2003-08-10)
Re: End of optimization... Robert@Knighten.org (2003-08-10)
| List of all articles for this month |

From: Robert@Knighten.org (Robert L. Knighten)
Newsgroups: comp.compilers
Date: 10 Aug 2003 11:01:51 -0400
Organization: Posted via Supernews, http://www.supernews.com
References: 03-07-026 03-07-068 03-07-151 03-07-179 03-07-200 03-08-019
Keywords: optimize, practice
Posted-Date: 10 Aug 2003 11:01:51 EDT

Fergus Henderson <fjh@cs.mu.oz.au> writes:


> There are also other (probably better) examples out there. For
> instance, see www.retarget.com, which advertises a tool that will take
> a machine description and produce not just a compiler, but also an
> assembler, disassembler, linker, simulator, debugger, etc. They claim
> that their compiler does a variety of optimizations (see the long list
> at www.retarget.com/doc/factsheet.pdf).
>
> The choice this approach offers is either writing your whole
> application in assembler by hand, or writing a fairly small machine
> description and then writing most of your application in C, with just
> the innermost loops hand-coded. For all but very small applications,
> if you hand-code it all in assembler then your total source code size,
> development effort, and time-to-market will all be larger, even though
> you don't have to write a machine description, because assembler is a
> considerably lower-level language than C. Furthermore, it will be
> harder to reuse the results of this effort for your next project.


This looks very much like what was being claimed for retargetable
compiler technology when it was last in vogue fifteen to twenty years
ago. The killer back then was that machine descriptions were quite
time consuming to write and required even more effort to verify,
hardware specific optimizations were hard to fit in to the scheme, and
register management even harder. Retarget.com certainly claims to
have solved all this with their Chess retargetable compiler and nML
machine modeling language. Indeed they claim to have done a very
great deal more. Does anyone here have any experience using any of
their tools?


I'm curious about your remark "a fairly small machine description" as
the only small machine descriptions I've ever seen have been for
extremely simple processors. The description of the instruction set
for the Intel Pentium processor runs over 300 pages, and I know the
actual RTL is considerably longer. Typically the machine description
needed for a high quality optimizing code generator is actually longer
than the ISA descriptions as it provides no information about
performance trade-offs.


-- Bob


--
Robert L. Knighten
Robert@Knighten.org


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.