This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [new-regalloc-branch] Emit remat information
- From: Michael Matz <matz at suse dot de>
- To: "Mukta Punjani, Noida" <muktap at noida dot hcltech dot com>
- Cc: gcc at gcc dot gnu dot org, Daniel Berlin <dberlin at dberlin dot org>
- Date: Sun, 24 Aug 2003 16:59:43 +0200 (CEST)
- Subject: Re: [new-regalloc-branch] Emit remat information
Hi,
On Sat, 23 Aug 2003, Mukta Punjani, Noida wrote:
> The variable "emitted_remat" (to count the number of remat performed)
> doesn't seem to be updated in the new-regalloc-branch. Hence it is printing
> cost as 0 in the debugging dumps.
It's a fallout when Denis moved the code for rematerialization to another
point. While doing spills itself nothing is really rematerialized.
Instead simple remat webs are later optimized again, to effectively form a
rematerializing load instead of a load/store pair. The counting wasn't
adjusted, though.
Ciao,
Michael.