public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
To: GCC Patches <gcc-patches@gcc.gnu.org>
Cc: nd <nd@arm.com>
Subject: [PATCH] Fix PR46932: Block auto increment on frame pointer
Date: Thu, 20 Jul 2017 14:49:00 -0000	[thread overview]
Message-ID: <AM5PR0802MB2610C93E52AAFEFD3A1DA5B283A70@AM5PR0802MB2610.eurprd08.prod.outlook.com> (raw)

Block auto increment on frame pointer references.  This is never
beneficial since the SFP expands into SP+C or FP+C during register
allocation.  The generated code for the testcase is now as expected:

	str	x30, [sp, -32]!
	strb	w0, [sp, 31]
	add	x0, sp, 31
	bl	foo3
	ldr	x30, [sp], 32
	ret

On SPEC2017 codesize improves uniformly across the board.

ChangeLog:
2017-07-20  Wilco Dijkstra  <wdijkstr@arm.com>

	PR middle-end/46932
	* auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.

gcc/testsuite/

	* gcc.dg/pr46932.c: New testcase.

--

diff --git a/gcc/auto-inc-dec.c b/gcc/auto-inc-dec.c
index 91fa5cf0bbe04b8a2c2c2b9209d245a327de0ffd..db1bd5bba2cee9fbf24d6d522505ac292688aab9 100644
--- a/gcc/auto-inc-dec.c
+++ b/gcc/auto-inc-dec.c
@@ -769,6 +769,12 @@ parse_add_or_inc (rtx_insn *insn, bool before_mem)
   inc_insn.pat = pat;
   inc_insn.reg_res = SET_DEST (pat);
   inc_insn.reg0 = XEXP (SET_SRC (pat), 0);
+
+  /* Block any auto increment of the frame pointer since it expands into
+     an addition and cannot be removed by copy propagation.  */
+  if (inc_insn.reg0 == frame_pointer_rtx)
+    return false;
+
   if (rtx_equal_p (inc_insn.reg_res, inc_insn.reg0))
     inc_insn.form = before_mem ? FORM_PRE_INC : FORM_POST_INC;
   else
diff --git a/gcc/testsuite/gcc.dg/pr46932.c b/gcc/testsuite/gcc.dg/pr46932.c
new file mode 100644
index 0000000000000000000000000000000000000000..b96febcc095fbec2cc4e02d1155871220d17cc99
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr46932.c
@@ -0,0 +1,13 @@
+/* { dg-options "-O2 -fdump-rtl-auto_inc_dec" } */
+
+/* Check that accesses based on the frame pointer do not
+   use auto increment.  */
+
+extern void foo (char*);
+void t01 (char t)
+{
+  char c = t;
+  foo (&c);
+}
+
+/* { dg-final { scan-rtl-dump-not "success" "auto_inc_dec" } } */

             reply	other threads:[~2017-07-20 14:49 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-07-20 14:49 Wilco Dijkstra [this message]
2017-07-25 20:42 ` Jeff Law
2017-07-25 21:25   ` Wilco Dijkstra
2017-07-25 22:29     ` Jeff Law

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=AM5PR0802MB2610C93E52AAFEFD3A1DA5B283A70@AM5PR0802MB2610.eurprd08.prod.outlook.com \
    --to=wilco.dijkstra@arm.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=nd@arm.com \
    /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).