about speed of latest revisions

[marcomaggi]

Is the latest mosh significantly slower than some months ago at
expanding programs for known reasons? In the last months I changed my
code base to use macros extensively; when running my test suite I am
seeing worst cases like:

===> test file ../tests/test-xlists.sps with Mosh

*** testing xlists

; *** checks *** : 79 correct, 0 failed.
real 53.10
user 52.84
sys 0.13

which was not happening before; this is after having precompiled all
the libraries. Notice that:

===> test file ../tests/test-xlists.sps with Petite Chez

*** testing xlists

; *** checks *** : 79 correct, 0 failed.
real 7.00
user 6.89
sys 0.07

===> test file ../tests/test-xlists.sps with Vicare

; *** checks *** : 79 correct, 0 failed.
real 26.00
user 25.60
sys 0.16

Petite and Vicare are slow, too, but not so much.
--
Marco Maggi

[Yuki Okumura]

Could you try your script with nmosh ?
(nmosh is another version of mosh interpreter and may be installed
with mosh)
It's slightly faster at your (nausicaa) libraries.

nmosh still not have the SRFI-37 and SRFI-38. So, if your scripts
won't run,
please send its diagnostic output by
$ nmosh --verbose --guru-mode YOUR-SCRIPT.sps

# never mind if your script won't run upon nmosh..
# it's under heavy development :)

--

[higepon]

Hi.

Run the tests with profiler option -p will give you some clue.

Cheers.

> --
> You received this message because you are subscribed to the Google Groups "Mosh Developer Disscus" group.
> To post to this group, send email to mosh-develo...@googlegroups.com.
> To unsubscribe from this group, send email to mosh-developer-di...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/mosh-developer-discuss?hl=en.
>
>

[marcomaggi]

On Oct 21, 7:10 am, Yuki Okumura <oku...@gmail.com> wrote:
> Could you try your script with nmosh ?

Faster!

*** testing xlists

; *** checks *** : 79 correct, 0 failed.

real 3.24
user 3.14
sys 0.08

... at running, while precompiling libraries is much slower.
This timing is after letting it compile the needed

libraries.

> nmosh still not have the SRFI-37 and SRFI-38.

With the exception of parameters, I do not use the SRFIs
from the underlying Scheme implementation.

I had to redefine PARAMETERIZE from the (system) library,
enclosing the body in a LET; it looks like the PARAMETERIZE
from mosh allows definitions+expressions in its body, while
the one from nmosh allows only expressions: is this for a
reason?

While running my compile-all script with nmosh I had to
shut down the process because of excess memory usage: after
half an hour of compiling libraries, it allocates all my 1GB
RAM then my OS starts swapping and I get tired of waiting.
This is on an i686-pc-linux-gnu.
--
Marco Maggi

[marcomaggi]

If you want to try it, I have uploaded a development tarball:

<http://github.com/downloads/marcomaggi/nausicaa/nausicaa-0.2d2-
src.tar.bz2>
--
Marco Maggi

[marcomaggi]

On Oct 21, 10:06 am, marcomaggi <mrc....@gmail.com> wrote:
> If you want to try it, I have uploaded a development tarball:
>
> <http://github.com/downloads/marcomaggi/nausicaa/nausicaa-0.2d2-src.tar.bz2>

Sorry I forgot, to use it: unpack the package then:

cd nausicaa-0.2d2
mkdir "=build"
cd "=build"
../configure --enable-mosh
make mfasl MOSH='nmosh --verbose --guru-mode'
--
Marco Maggi

[Yuki Okumura]

> On Oct 21, 10:06 am, marcomaggi <mrc....@gmail.com> wrote:
>
> > If you want to try it, I have uploaded a development tarball:

Thanks for the code and I will check it later.
(my damn computer(win/mac) cannot configure
properly some of your library and I will take some Linux host.)

> it looks like the PARAMETERIZE
> from mosh allows definitions+expressions in its body, while
> the one from nmosh allows only expressions: is this for a
> reason?

It's nothing more than my simple mistake.. I've forgot to implement
it.

nmosh cannot utilize underlaying mosh macro which defined in
boot/baselib.scm.
For example, time macro is implemented in
boot/runtimes/srfi-mosh/lib.boot/mosh.nmosh.ss
I must do same for parameterize within
boot/runtimes/srfi-mosh/lib.boot/system.nmosh.ss
I will fix this soon.

> While running my compile-all script with nmosh I had to
> shut down the process because of excess memory usage:

nmosh generates much more debugging information when compared
with psyntax-mosh. And it will consume some of memory area..
I should add switches for this. (--disable-debug?)

nmosh had also reported some mysterious memory consumption. See Issue
164
http://code.google.com/p/mosh-scheme/issues/detail?id=164
I couldn't manage to reproduce that problem though.


Currently, I'm rewriting whole nmosh expander in R6RS and
current nmosh expander will be dropped soon. So, I recommend
not to do significant work for current nmosh.