|
From: Nicholas N. <nj...@cs...> - 2005-02-25 22:42:44
|
On Thu, 24 Feb 2005, Jeremy Fitzhardinge wrote: >>> * big-bang shadow allocation causes problems on kernels that don't >>> do overcommitment >>> >>> * a fixed partitioning scheme is less appropriate if we move towards >>> compressed representations of shadow memory, since that compression >>> ratio could be variable >> >> >> Plus all the other problems mentioned in bug #82301, which we've talked >> about at length multiple times in the past. > > The only additional problem in that bug is the difficulty with loading large > symbol files, which won't be helped by the superblock scheme at all; it will > be made worse. But that's OK, because the symtab reader needs to be fixed > anyway. That's not the only additional problem. Sometimes Valgrind tools run out of memory in their space when the client space is not full, and vice versa. N |