Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 26 Dec 2005 10:26:45 -0800
From:      Jason Evans <jasone@freebsd.org>
To:        Max Khon <fjoe@samodelkin.net>
Cc:        freebsd-current@freebsd.org
Subject:   Re: New malloc ready, take 42
Message-ID:  <B65EC3BB-6A5A-4299-BD86-A7D2AB381941@freebsd.org>
In-Reply-To: <20051226101825.GA92082@samodelkin.net>
References:  <E1E22E2D-A6D5-49CC-9649-C37C50F0443B@freebsd.org> <20051226101825.GA92082@samodelkin.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Dec 26, 2005, at 2:18 AM, Max Khon wrote:
>
> Do you plan to provide public API for malloc arenas
> similar to SGI amalloc(3)?
>
> http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi? 
> coll=0650&db=man&fname=/usr/share/catman/p_man/cat3p/amallinfo.z

I hadn't planned to do so.  It's not too hard to think of scenarios  
in which manual management of arenas might be useful, but adding non- 
standard public APIs to libc is something that I'm reluctant to do  
unless there is a compelling reason to do so.  Arena-specific APIs  
would work well with jemalloc, but they may not make sense for  
whatever comes after jemalloc, yet we'd be stuck with supporting the  
APIs.

In fact, if I were making API changes to non-standard libc  
interfaces, the first thing I'd do would be to remove reallocf(3), or  
at least rename it to reallocf_np().

Are there compelling use cases you can think of that might sway us  
toward providing malloc arena APIs?

Thanks,
Jason



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?B65EC3BB-6A5A-4299-BD86-A7D2AB381941>