[prev in list] [next in list] [prev in thread] [next in thread] 

List:       grub-devel
Subject:    Re: Creating grub/grub2/grldr.mbr bootrom with makerom
From:       "Yoshinori K. Okuji" <okuji () enbug ! org>
Date:       2007-12-22 9:47:04
Message-ID: 200712221047.05173.okuji () enbug ! org
[Download RAW message or body]

On Saturday 22 December 2007 10:03, Bean wrote:
> On Dec 22, 2007 4:06 PM, Yoshinori K. Okuji <okuji@enbug.org> wrote:
> > On Friday 21 December 2007 20:04, Robert Millan wrote:
> > > How well does compression work for GRUB 2 ?  core.img is already
> > > compressed (with lzo); if LZMA makes better results perhaps it'd be a
> > > good idea to switch.
> >
> > It's not that simple. LZO was chosen instead of gzip, because of the size
> > requirement on PC. To preserve safety, we need to keep the core part less
> > than 31.5KB (63 sectors).
> >
> > The size is the sum of non-compressable bootstrap code, decompression
> > code and compressed code + data. When I made an experiment in PUPA,
> > although gzip had a better compression ratio, due to the decompression
> > code size, LZO won.
> >
> > I don't know precisely, but I suspect that decompression code for LZMA
> > would be slightly larger than gzip's (IIRC, a range coder is likely to
> > require more code and data). So I don't expect that LZMA can replace the
> > current usage of LZO in normal PC so easily.
>
> The decompression code for LZMA is very small, i use -Os option to
> compile LzmaDecode.c, the result is about 2.8K.

Thank you for your information. In my estimate, LZMA will outperform, if the 
ratio, compressed size : original size, is less than 0.52. So LZMA might 
win. :)

Okuji



[prev in list] [next in list] [prev in thread] [next in thread]