[patch] Don't allocate df_link objects in the df_chain basic block pool

[Steven Bosscher]

Hi,

The df_chain problem currently has no basic block specific data.
Every df problem has an alloc pool for the problem-specific basic
block data, but df_chain allocates df_link objects there.  On the
rtl-fud-branch, the df_chain problem does have problem-specific basic
block data, which holds the PHI nodes for the basic block.  So we have
to allocate df_link objects from another pool, which I've hung from
df_chain->problem_data in the attached patch.

Bootstrapped&tested (trunk and rtl-fud-branch) on ia64-unknown-linux-gnu.
I will commit this on the rtl-fud branch.  Is this OK for the trunk too?

Gr.
Steven
-------------- next part --------------
A non-text attachment was scrubbed...
Name: df_chain_link_pool.diff
Type: text/x-patch
Size: 4612 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20080628/96c46446/attachment.bin>