This is the mail archive of the
mailing list for the GCC project.
Re: [RFC] speed up compute_store_table
- From: Roger Sayle <roger at eyesopen dot com>
- To: Richard dot Earnshaw at arm dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sun, 30 Nov 2003 08:22:45 -0700 (MST)
- Subject: Re: [RFC] speed up compute_store_table
On Thu, 27 Nov 2003, Richard Earnshaw wrote:
> <date> Richard Earnshaw <firstname.lastname@example.org>
> * gcse.c (reg_clear_last_set): New function.
> (reg_set_info): If data is non-null, treat it as an sbitmap of
> registers, set the bit for the register being set.
> (compute_store_table): Allocate last_set_in with xcalloc. Do not
> memset this array on each iteration. Pass reg_set_in_block[bb->index]
> to note_stores while computing last_set_in instead of scanning
> last_set_in after the first pass through the insns.
> Clear last_set_in using reg_clear_last_set instead of explicitly
> rescanning after each insn. If checking is enabled, assert that
> last_set_in is completely zeroed after each bb has been processed.
This looks OK for mainline, given that it addresses "3.4 compile-time"
which may be considered a regression from previous releases. If any
other reviewer objects, we can always move the patch to tree-ssa and
revert it from mainline.