Re: [Chicken-hackers] encoded-literals

[Peter Bex]

On Wed, Dec 19, 2007 at 10:42:13AM +0100, felix winkelmann wrote:
> Hi!
> 
> I have merged the "encoded-literals" branch into the trunk, in which
> literals in code are decoded at runtime from a machine-independent
> binary format. Also, lambda-info (the stuff you see when you print
> procedures) is statically preallocated. This should give better C
> compile times (the toplevel procedures are smaller), sometimes
> dramatically (hopefully - perhaps someone can perform a timing
> comparison on ppc hardware?). This also makes the compress-literals
> option/declaration obsolete. I tested this on various platforms and found
> no problems yet, but wanted to note this, anyway.

The results of a quick'n dirty benchmark
(ie, 'for i in benchmarks/*.scm; do time csc $i; done') look very good!
I compared Chicken 2.736 to trunk, 2.738.  This is on NetBSD/macppc
4.0RC1, running on an iBook G4 with 512Mb RAM at ~500MHz.  I compiled
chicken with no special optimizations, and DEBUGBUILD=1.

2.736 had the following total time (in seconds) it took to compile the
benchmark tests:

boyer.scm 48.283
browse.scm 7.330
conform.scm 24.612
cpstak.scm 2.083
ctak.scm 2.392
dderiv.scm 4.295
deriv.scm 4.176
destructive.scm 3.090
div-iter.scm 2.220
div-rec.scm 2.168
dynamic.scm 87.35
earley.scm 19.709
fft.scm 6.303
fib.scm 1.627
fibc.scm 2.424
fprint.scm 2.839
fread.scm 2.413
hanoi.scm 2.679
lattice.scm 8.735
maze.scm 51.477
nqueens.scm 2.666
puzzle.scm 8.476
scheme.scm 59.240
tak.scm 1.764
takl.scm 2.508
takr.scm 45.637
traverse.scm 10.555
travinit.scm 10.745
triangl.scm 5.410
=======================
Total: 433.206 seconds

Chicken 2.738 had the following results:

boyer.scm 12.811
browse.scm 4.500
conform.scm 14.803
cpstak.scm 1.530
ctak.scm 1.618
dderiv.scm 2.504
deriv.scm 2.234
destructive.scm 2.239
div-iter.scm 1.558
div-rec.scm 1.508
dynamic.scm 52.529
earley.scm 17.735
fft.scm 8.454
fib.scm 2.562
fibc.scm 3.659
fprint.scm 2.163
fread.scm 1.143
hanoi.scm 1.323
lattice.scm 4.520
maze.scm 28.535
nqueens.scm 2.016
puzzle.scm 5.113
scheme.scm 34.096
tak.scm 1.332
takl.scm 1.753
takr.scm 20.722
traverse.scm 4.966
travinit.scm 4.959
triangl.scm 2.428
=======================
Total: 245.313 seconds

No idea how representative this is, but it looks very good.

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

pgpRjrQECrWZA.pgp
Description: PGP signature