RE: Supercompiled Perl: HP's Dynamo Project

HP was able to run a native HP-RISC program in an HP-RISC instruction set
interpreter (virtual machine) running on an HP-RISC processor and achieve a
20% increase in performance. It may be hard to imagine running a program
running in an interpreter for an instruction set on the same processor and
yet running FASTER, but HP did it. This research was the basis for the Aries
IA-32 interpreter designed to run on the Itanium IA-64 which provides IA-32
binary compatibility. (See IEEE Computer cover story, March 2000). Not only
does this technique work, it's in service.

I have some more references, papers, theses, etc. if you're interested.



-----Original Message-----
From: Paris Sinclair [mailto:[email protected]]
Sent: Friday, August 10, 2001 8:36 PM
To: Rick Rankin
Cc: [email protected]
Subject: Re: Supercompiled Perl: HP's Dynamo Project


On Fri, 10 Aug 2001, Rick Rankin wrote:

> Dynamo does not depend on any special programming language, compiler,
> operating system or hardware support. The program binary is not
> instrumented and is left untouched during Dynamo's operation. Dynamo
> observes the program's behavior through interpretation to dynamically
> select hot instruction traces from the running program. The hot traces
> are optimized using low-overhead optimization techniques and emitted
> into a software code cache. Subsequent instances of these traces cause
> the cached version to be executed, resulting in a performance boost.
>
> Contrary to intuition, we demonstrate that it is possible to use a
> piece of software to improve the performance of a native, statically
> optimized program binary, while it is executing. Dynamo not only
> speeds up real application programs, its performance improvement is
> often quite significant. For example, the performance of many +O2
> optimized SPECint95 binaries running under Dynamo is comparable to the
> performance of their +O4 optimized version running without Dynamo."
>
> Transparent Dynamic Optimization;Vasanth Bala, Evelyn Duesterwald, Sanjeev
> Banerjia;HP Laboratories Cambridge;June, 1999

So then, I think it is worth asking, can AI in Perl be used to speed up
this sort of optimization? I would be surprised if the gains are larger
than the losses. If that is the case, then it is clearly a dead end from
the perspective of Perl AI, excepting that anything that speeds up perl
speeds up Perl.

--
Paris