Re: [Chicken-hackers] [PATCH] Another exponential case: analyze-expression on LET expressions

[Peter Bex]

On Sat, Jan 28, 2012 at 01:21:12PM +0100, Felix wrote:
> > Meanwhile, "make check" still passes and since there doesn't seem to be
> > an observable difference but it does affect out performance profile in a
> > positive way, I decided to post the patch to the list.  What could
> > possibly go wrong?
> 
> I think this change is not correct - "lambda" and "let" are not
> equivalent at this stage of the compiler: "let" introduces local
> variables and "lambda" creates a procedure. The distinction is
> important when determining whether a variable is "captured", i.e. when
> it is referenced or assigned from a different procedure (lambda)
> body.

Thanks for the feedback.  I think I understand this part of the compiler
a little better now.

> If I understand this patch correctly, every "let" would be
> treated like a nested lambda and references to variables from outer
> "let" expressions would be marked as "captured".

That's exactly what it does.

Do you think you can produce a testcase which triggers a bug caused
by this patch?  It would be a good regression test for further
optimizer/scrutinizer changes.

> The optimizer makes various assumptions about this (and the
> scrutinizer will, in a patch that is currently prepared).

I'll have another look at that patch.  Armed with this info
I might better understand what it does.

Cheers,
Peter
-- 
http://sjamaan.ath.cx
--
"The process of preparing programs for a digital computer
 is especially attractive, not only because it can be economically
 and scientifically rewarding, but also because it can be an aesthetic
 experience much like composing poetry or music."
                                                        -- Donald Knuth