This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[patch][Ping #3] PR80929: Realistic PARALLEL cost in seq_cost.
- From: Georg-Johann Lay <avr at gjlay dot de>
- To: gcc-patches <gcc-patches at gcc dot gnu dot org>
- Cc: Jeff Law <law at redhat dot com>, Ian Lance Taylor <ian at airs dot com>
- Date: Wed, 28 Jun 2017 16:04:34 +0200
- Subject: [patch][Ping #3] PR80929: Realistic PARALLEL cost in seq_cost.
- Authentication-results: sourceware.org; auth=none
Ping #3
https://gcc.gnu.org/ml/gcc-patches/2017-06/msg00096.html
On 02.06.2017 09:53, Georg-Johann Lay wrote:
> Hi,
>
> this small addition improves costs of PARALLELs in
> rtlanal.c:seq_cost(). Up to now, these costs are
> assumed to be 1 which gives gross inexact costs for,
> e.g. divmod which is represented as PARALLEL.
>
> The patch just forwards cost computation to insn_rtx_cost
> which uses the cost of the 1st SET (if any) and otherwise
> assign costs of 1 insn.
>
> Bootstrapped & regtested on x86_64.
>
> Moreover, it fixed the division by constant on avr where
> the problem popped up since PR79665.
>
> Ok to install?
>
> Johann
>
> gcc/
> PR middle-end/80929
> * rtlanal.c (seq_cost) [PARALLEL]: Get cost from insn_rtx_cost
> instead of assuming cost of 1.
Index: rtlanal.c
===================================================================
--- rtlanal.c (revision 248745)
+++ rtlanal.c (working copy)
@@ -5300,6 +5300,9 @@ seq_cost (const rtx_insn *seq, bool spee
set = single_set (seq);
if (set)
cost += set_rtx_cost (set, speed);
+ else if (INSN_P (seq)
+ && PARALLEL == GET_CODE (PATTERN (seq)))
+ cost += insn_rtx_cost (PATTERN (seq), speed);
else
cost++;
}