[Shootout-list] errors in clean benchmarks

[Brent Fulgham]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On Mar 29, 2005, at 2:24 PM, Diederik van Arkel wrote:

> fasta, mandelbrot, nsieve_bits, and revcomp are all missing
> the -nr runtime option.
>

Thanks -- I should have seen those myself, but didn't look closely.

> nsieve_bits currently has a discrepancy between the file name
> and the module name as specified in that file, in clead this
> leads to a compiler error

Fixed.

> for revcomp I've submitted a fixed implementation which doesn't
> use the substring feature as that is currently not enabled on
> linux.

This has a problem (see output on website).

> For nsieve and sum-file I'm unable to determine what's going
> wrong, they work for me here and the build log doesn't give any
> useful information. Maybe Brent could have a look and give me a
> hint what needs to be changed.

sum-file was a problem with an errant DOS line-ending added to the 
template used to generate the input file.  Nsieve was running out of 
heap space.  I bumped the runtime argument for this and it now works.

> As a general remark: the clean language installation on the test
> machine still appears to be off. Most builds are showing inlining
> failures which shouldn't occur in a properly installed clean system.
> Generally this indicates that the system modules from StdEnv need
> to be compiled another two times (twice to get the cross-module
> inlining between system files right), the other possibility might
> be that the existing abc files for the StdEnv don't have unix eoln
> format. Existing clean compilers might then have a problem with
> inlining (fixed in HEAD).
>

I'll e-mail you tomorrow off-list to go over this stuff.  My Linux 
package is probably not the greatest (I've not played with it much 
since I set aside the GTK binding a while back) and I should update it 
to current.

Now that I'm a mostly-Mac OS guy I want to get back into Clean a bit 
more... :-)

> Brent, I see that pidigits still hasn't appeared. I presume you
> haven't gotten around to looking at the BugNum library yet but
> could you at least show the program on the site so that others
> don't waste their time implementing it?
>

It should be on the site shortly.  We'll get the Debian package squared 
away so all that good stuff is included.

> Finally, I should be submitting fannkuch and k-nucleotide tonight
> or tomorrow. These have been ready some time but I've been staring
> at the generated code for some time in disgust. Fresh motivation to
> add some floating point optimisation to the code generator :-)

I'm looking forward to it.  I love it when the shootout prompts people 
to make their languages even cooler!  :-)

- -Brent
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iD8DBQFCSldEzGDdrzfvUpURAuuiAKCFFolc+oEhEGVlLcEu9mlEEvMo/wCfTI/G
iFMFdzyNwy8SajbGcmE4pkk=
=5/SB
-----END PGP SIGNATURE-----