Date: Mon, 12 Dec 2005 10:17:09 +0000 From: Doug Rabson <dfr@nlsystems.com> To: freebsd-current@freebsd.org Cc: Johan Bucht <bucht@acc.umu.se>, Jason Evans <jasone@canonware.com>, current@freebsd.org Subject: Re: New libc malloc patch Message-ID: <200512121017.10220.dfr@nlsystems.com> In-Reply-To: <439CEB74.9080505@acc.umu.se> References: <20051212014852.GA8775@shaka.acc.umu.se> <9FAD2B4B-C167-42D7-A8E7-BE03F4C07543@canonware.com> <439CEB74.9080505@acc.umu.se>
next in thread | previous in thread | raw e-mail | index | archive | help
On Monday 12 December 2005 03:16, Johan Bucht wrote: > Isn't 8 byte alignment expected by some applications? > How do you know if a allocation is huge if you don't have a tag? > Something more to read up on i guess. =) Actually, I'm glad you pointed that out. My own applications use SSE2 primitives a lot and those guys need to be allocated on 16-byte boundaries. I currently use phkmalloc which has the nice property that small allocations are aligned to the next greater power-of-2 boundary which is (for me) always at least 16. Since you are targetting modern architectures with this allocator, could you please set the minimum alignment on all i386 and amd64 processors to 16. For what its worth, when gcc is using SSE instructions, it assumes this for all memory, stack or malloc.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200512121017.10220.dfr>