On Thu, Jul 26, 2001 at 10:07:20PM +0100, David Starner wrote: > From: Ben Collins <[EMAIL PROTECTED]> > > On Thu, Jul 26, 2001 at 08:27:14PM +0200, Petr Vandrovec wrote: > > > Package: gcc-3.0 > > > Version: 3.0.1-0pre010723 > > > > > > Hi, > > > gcc-3.0.1 contains some really unreasonable default > for -flimit-inline - > > > probably something around 100, but I did not look at sources. > > > > Why not have the kernel Makefile pass "-flimit-inline=10000" explicitly? > > We should look upstream before adding any ill advised patches, as the > problem is known (limit-inline = 100 because before, gcc was trashing > machines with 4 GB of memory. A search for a better default is being made.) > There should be a fix out with 3.0.1-final.
It seems that Linus agrees with this limit, and is converting some extern inlines to static, and so is reducing the problem. I agree. Let them sort it out, and trickle down to us. The original problems seems to be that with the old limit, compiling c++ templated code caused a great deal of overhead with this original 10000, and so it was reduced to something sane. Ben -- .----------=======-=-======-=========-----------=====------------=-=-----. / Ben Collins -- ...on that fantastic voyage... -- Debian GNU/Linux \ ` [EMAIL PROTECTED] -- [EMAIL PROTECTED] -- [EMAIL PROTECTED] ' `---=========------=======-------------=-=-----=-===-======-------=--=---'
