This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [trans-mem] merge from trunk @ 178608
- From: Richard Henderson <rth at redhat dot com>
- To: Aldy Hernandez <aldyh at redhat dot com>
- Cc: Torvald Riegel <triegel at redhat dot com>, gcc at gcc dot gnu dot org, MARLIER Patrick <patrick dot marlier at gmail dot com>
- Date: Wed, 14 Sep 2011 13:10:42 -0700
- Subject: Re: [trans-mem] merge from trunk @ 178608
- References: <4E7107F5.1030502@redhat.com>
On 09/14/2011 01:00 PM, Aldy Hernandez wrote:
> This was very painful, and it's still not over. We hadn't merged in
> almost 1.5 years, and we're paying for it now...
To be fair, the pause in merging was to have a stable base in which
to finish the Velox project. Of course, that ended 9 months ago...
> transaction_invariant_address_p (const_tree mem, basic_block region_entry_block)
> {
> - if ((TREE_CODE (mem) == INDIRECT_REF
> - || TREE_CODE (mem) == MISALIGNED_INDIRECT_REF)
> + if (TREE_CODE (mem) == INDIRECT_REF
> && TREE_CODE (TREE_OPERAND (mem, 0)) == SSA_NAME)
...
> @@ -1432,7 +1433,7 @@ requires_barrier (basic_block entry_bloc
> switch (TREE_CODE (x))
> {
> case INDIRECT_REF:
> - case MISALIGNED_INDIRECT_REF:
> + /* case MISALIGNED_INDIRECT_REF: */
We're missing support for MEM_REF in these places.
I don't see anything else obviously wrong.
r~