From: waugh@probitas.cs.utas.edu.au (Sam Waugh)
Message-Id: <9410100613.AA68794@probitas.cs.utas.edu.au>
Subject: "Bug" in 0.22 -- order of magnitude slower than hbc
To: glasgow-haskell-bugs@dcs.gla.ac.uk
Date: Mon, 10 Oct 1994 17:13:41 +1000 (EETDT)

Hi.

I've come up with what you might consider to be a bug with ghc-0.22 (unless
I've done something obviously wrong).  I wrote some code to perform a simple
backpropagation neural network simulator (just to see how it would go), and
I have just recompiled it using hbc version 0.999.4.  The ghc executable was
much slower -- even when "optimised".

I've included at the bottom of this message a uuencoded gzipped tar file
(took a while to make) which includes the following documents:

	*.lhs		-- actual code
	makefile	-- makefile for ghc (hbc was compiled by producing
				object files and linking them together.  The
				optimised hbc code used -O.  The optimised
				ghc code was generated with the extra options
				in the makefile).
	temp		-- machine and gcc compiler details.
	compile		-- standard ghc compilation with -v
	out.0.22	-- output from all trials
	err.*		-- the timings of the different trials (ignore the
				.2 on the .999.4.2 files -- it was a second
				try).

Hopefully that is all the files I've included and all the files you need.
Let me know if you have any problems.

On a final note -- when compiling using all the optimisations for ghc there
were an awful lot of warnings.  Is there anyway you can get rid of these
(like by fixing the problems)?  It might be obscuring something important.

Thanks for your time.

Sam.
-- 
Sam Waugh				Phone: +61 02 202962
Department of Computer Science		Fax:   +61 02 202913
University of Tasmania			Email: waugh@cs.utas.edu.au
GPO Box 252C, Hobart Tasmania 7001, Australia

[snip snip]