Re: Re: [RFC] lang/pypy

2013-03-10 Thread David Naylor
On Sunday, 10 March 2013 00:37:15 poyop...@puripuri.plala.or.jp wrote:
 Hi, good work, David.

Hi Kuro

 It compiles, packages and works flawlessly here for replacement of 1.9
 so far for a week.

Great, thanks.  Good to know :-)
 
 On my compile box, amd64/Athlon64 5050e 2.6GHz 2 core/8GB,
 memory requirement for translation processes is far less than
 warning shown. It prevents build with
 
 | warn: this system has insufficient memory, expected at least 9227MiB RAM
 
 however my translation processes under -DPYPY_IGNORE_MEMORY take 2GB
 for normal binary and 2.5GB for sandboxed one so they aggregate 4.5GB
 to run parallel. 

This is good news.  Could you please detail how you measured peak memory?  I 
might need to retest the port.  

 This is far less than expected, no page thrashing,
 no hang, no stuttering. It does not matter being with pypy1.9 or pypy2.0
 (yes, I built twice for 2.0 self hosting. Not tried with cPython.)
 
 So I think this warning is a bit excessive that makes everyone just put
 PYPY_IGNORE_MEMORY=1 in their make.conf. Just in my case.

I'll disable the test for now and revise my estimation.  Thanks for reporting 
back.  

Regards

signature.asc
Description: This is a digitally signed message part.


Re: [RFC] lang/pypy

2013-03-10 Thread poyopoyo
Hi David,

At Sun, 10 Mar 2013 21:18:12 +0300,
David Naylor wrote:
  however my translation processes under -DPYPY_IGNORE_MEMORY take 2GB
  for normal binary and 2.5GB for sandboxed one so they aggregate 4.5GB
  to run parallel. 
 
 This is good news.  Could you please detail how you measured peak memory?  I 
 might need to retest the port.  

I heuristically measured them with top(1) RESources, amount of ZFS
ARC and swap increased. Oh and free memory after one of translation
process finised. I'm afraid I couldn't provide reproducible method for
environments for others.

 I'll disable the test for now and revise my estimation.  Thanks for reporting 
 back.  

I wonder where so much difference of memory usage between yours
and mine comes from. 5.5GB vs 2.5GB on the same platform (64bit/pypy)
is not so negligible.

-- 
Kuro poyop...@puripuri.plala.or.jp

PS. like this, please. thanks for the offer.
___
freebsd-ports@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ports
To unsubscribe, send any mail to freebsd-ports-unsubscr...@freebsd.org


Re: [RFC] lang/pypy

2013-03-09 Thread poyopoyo
Hi, good work, David.

It compiles, packages and works flawlessly here for replacement of 1.9
so far for a week.

On my compile box, amd64/Athlon64 5050e 2.6GHz 2 core/8GB,
memory requirement for translation processes is far less than
warning shown. It prevents build with
| warn: this system has insufficient memory, expected at least 9227MiB RAM
however my translation processes under -DPYPY_IGNORE_MEMORY take 2GB
for normal binary and 2.5GB for sandboxed one so they aggregate 4.5GB
to run parallel. This is far less than expected, no page thrashing,
no hang, no stuttering. It does not matter being with pypy1.9 or pypy2.0
(yes, I built twice for 2.0 self hosting. Not tried with cPython.)

So I think this warning is a bit excessive that makes everyone just put
PYPY_IGNORE_MEMORY=1 in their make.conf. Just in my case.

-- 
kuro
___
freebsd-ports@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ports
To unsubscribe, send any mail to freebsd-ports-unsubscr...@freebsd.org