[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Chicken-hackers] [Chicken-users] Need help to figure out where this

From: Jörg F . Wittenberger
Subject: Re: [Chicken-hackers] [Chicken-users] Need help to figure out where this strange performance impact is coming from
Date: Thu, 14 Jan 2016 13:15:23 +0100
User-agent: Mozilla/5.0 (X11; Linux armv7l; rv:38.0) Gecko/20100101 Icedove/38.4.0


Still somebody having another architecture available is needed to track
this further down.

However I got more infos to share:

Wild guess: this may be some alignment problem.

It does _not_ have to do with anything to do with srfi-1.  It just needs
to have *enough* exports in the program.

The story so far:  Profiling with srfi-1 pulled inline did reveal no
calls to srfi-1.  Turned the inlined srfi-1 into a module with no
exports -> slow down.  Export * -> full speed.  Binary search cutting
the export list -> slow down when less then 17 bindings exported.  Does
not matter which 17 bindings.  Rewrote the test to not import "extras"
-> still fast.  Removed the "(use extras)" -> slow down.  Use random egg
exporting many bindings instead -> fast again.


Am 14.01.2016 um 12:12 schrieb Christian Kellermann:
> * Jörg F. Wittenberger <address@hidden> [160114 11:30]:
>> Tried that too:  on AMD64 (Debian) chicken 4.10.1 from tarball does NOT
>> give any difference.
>> But even if it may be an ARM related problem: how is it even possible??!
> So next let's isolate whether it's the architecture (ARM
> vs. world)type or the word size (32bit vs 64bit). Can anyone with a
> 32bit system try that please?
> Thanks,
> Christian
> --
> May you be peaceful, may you live in safety, may you be free from
> suffering, and may you live with ease.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]