From: Xionghu Luo <luoxhu@linux.ibm.com>
To: Richard Biener <rguenther@suse.de>
Cc: gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] tree-optimization/102155 - fix LIM fill_always_executed_in CFG walk
Date: Fri, 10 Sep 2021 21:54:55 +0800 [thread overview]
Message-ID: <44039d9b-d481-2a11-fcdc-493915833ae9@linux.ibm.com> (raw)
In-Reply-To: <8on1r65r-n4qr-9555-s3q6-75783p6622@fhfr.qr>
On 2021/9/9 18:55, Richard Biener wrote:
> diff --git a/gcc/tree-ssa-loop-im.c b/gcc/tree-ssa-loop-im.c
> index 5d6845478e7..4b187c2cdaf 100644
> --- a/gcc/tree-ssa-loop-im.c
> +++ b/gcc/tree-ssa-loop-im.c
> @@ -3074,15 +3074,13 @@ fill_always_executed_in_1 (class loop *loop, sbitmap contains_call)
> break;
>
> if (bb->loop_father->header == bb)
> - {
> - if (!dominated_by_p (CDI_DOMINATORS, loop->latch, bb))
> - break;
> -
> - /* In a loop that is always entered we may proceed anyway.
> - But record that we entered it and stop once we leave it
> - since it might not be finite. */
> - inn_loop = bb->loop_father;
> - }
> + /* Record that we enter into a subloop since it might not
> + be finite. */
> + /* ??? Entering into a not always executed subloop makes
> + fill_always_executed_in quadratic in loop depth since
> + we walk those loops N times. This is not a problem
> + in practice though, see PR102253 for a worst-case testcase. */
> + inn_loop = bb->loop_father;
Yes your two patches extracted the get_loop_body_in_dom_order out and removed
the inn_loop break logic when it doesn't dominate outer loop. Confirmed the replacement
could improve for saving ~10% build time due to not full DOM walker and marked the previously
ignored ALWAYS_EXECUTED bbs.
But if we don't break for inner loop again, why still keep the *inn_loop* variable?
It seems unnecessary and confusing, could we just remove it and restore the original
infinte loop check in bb->succs for better understanding?
diff --git a/gcc/tree-ssa-loop-im.c b/gcc/tree-ssa-loop-im.c
index d1e2104233b..82a0509e0c4 100644
--- a/gcc/tree-ssa-loop-im.c
+++ b/gcc/tree-ssa-loop-im.c
@@ -3200,7 +3200,6 @@ fill_always_executed_in_1 (class loop *loop, sbitmap contains_call)
{
basic_block bb = NULL, last = NULL;
edge e;
- class loop *inn_loop = loop;
if (ALWAYS_EXECUTED_IN (loop->header) == NULL)
{
@@ -3213,17 +3212,6 @@ fill_always_executed_in_1 (class loop *loop, sbitmap contains_call)
edge_iterator ei;
bb = worklist.pop ();
- if (!flow_bb_inside_loop_p (inn_loop, bb))
- {
- /* When we are leaving a possibly infinite inner loop
- we have to stop processing. */
- if (!finite_loop_p (inn_loop))
- break;
- /* If the loop was finite we can continue with processing
- the loop we exited to. */
- inn_loop = bb->loop_father;
- }
-
if (dominated_by_p (CDI_DOMINATORS, loop->latch, bb))
last = bb;
@@ -3232,8 +3220,15 @@ fill_always_executed_in_1 (class loop *loop, sbitmap contains_call)
/* If LOOP exits from this BB stop processing. */
FOR_EACH_EDGE (e, ei, bb->succs)
+ {
if (!flow_bb_inside_loop_p (loop, e->dest))
break;
+ /* Or we enter a possibly non-finite loop. */
+ if (flow_loop_nested_p (bb->loop_father,
+ e->dest->loop_father)
+ && ! finite_loop_p (e->dest->loop_father))
+ break;
+ }
if (e)
break;
@@ -3242,15 +3237,6 @@ fill_always_executed_in_1 (class loop *loop, sbitmap contains_call)
if (bb->flags & BB_IRREDUCIBLE_LOOP)
break;
- if (bb->loop_father->header == bb)
- /* Record that we enter into a subloop since it might not
- be finite. */
- /* ??? Entering into a not always executed subloop makes
- fill_always_executed_in quadratic in loop depth since
- we walk those loops N times. This is not a problem
- in practice though, see PR102253 for a worst-case testcase. */
- inn_loop = bb->loop_father;
-
/* Walk the body of LOOP sorted by dominance relation. Additionally,
if a basic block S dominates the latch, then only blocks dominated
by S are after it.
>
> /* Walk the body of LOOP sorted by dominance relation. Additionally,
> if a basic block S dominates the latch, then only blocks dominated
--
Thanks,
Xionghu
next prev parent reply other threads:[~2021-09-10 13:55 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-01 9:58 Richard Biener
2021-09-02 5:09 ` Xionghu Luo
2021-09-02 8:00 ` Richard Biener
2021-09-02 8:50 ` Richard Biener
2021-09-02 10:00 ` Xionghu Luo
2021-09-02 10:37 ` Richard Biener
2021-09-09 6:40 ` Xionghu Luo
2021-09-09 9:16 ` Richard Biener
2021-09-09 10:55 ` Richard Biener
2021-09-10 13:54 ` Xionghu Luo [this message]
2021-09-13 2:28 ` Xionghu Luo
2021-09-13 8:17 ` Richard Biener
2021-09-14 1:15 ` Xionghu Luo
2021-09-14 7:30 ` Richard Biener
2021-09-13 6:55 ` Richard Biener
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=44039d9b-d481-2a11-fcdc-493915833ae9@linux.ibm.com \
--to=luoxhu@linux.ibm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=rguenther@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).