From: Jakub Jelinek <jakub@redhat.com>
To: Uros Bizjak <ubizjak@gmail.com>
Cc: gcc-patches@gcc.gnu.org
Subject: [PATCH] i386: Fix up ix86_expand_int_sse_cmp [PR107585]
Date: Thu, 10 Nov 2022 10:29:11 +0100 [thread overview]
Message-ID: <Y2zEZ6f0v/74nBbT@tucnak> (raw)
Hi!
The following patch fixes ICE on the testcase. I've used GEN_INT
incorrectly thinking the code punts on the problematic boundaries.
It does, but only for LE and GE, i.e. signed comparisons, for unsigned
the boundaries are 0 and unsigned maximum, so when say unsigned char
adds one to 127 or subtracts one from 128 we need to canonicalize it.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk?
2022-11-10 Jakub Jelinek <jakub@redhat.com>
PR target/107585
PR target/107546
* config/i386/i386-expand.cc (ix86_expand_int_sse_cmp): Use
gen_int_mode rather than GEN_INT.
* gcc.dg/pr107585.c: New test.
--- gcc/config/i386/i386-expand.cc.jj 2022-11-08 12:21:48.704047171 +0100
+++ gcc/config/i386/i386-expand.cc 2022-11-09 14:40:12.157012775 +0100
@@ -4540,7 +4540,8 @@ ix86_expand_int_sse_cmp (rtx dest, enum
rtvec v = rtvec_alloc (n_elts);
for (i = 0; i < n_elts; ++i)
RTVEC_ELT (v, i)
- = GEN_INT (INTVAL (CONST_VECTOR_ELT (cop1, i)) + 1);
+ = gen_int_mode (INTVAL (CONST_VECTOR_ELT (cop1, i)) + 1,
+ eltmode);
cop1 = gen_rtx_CONST_VECTOR (mode, v);
std::swap (cop0, cop1);
code = code == LE ? GT : GTU;
@@ -4584,7 +4585,8 @@ ix86_expand_int_sse_cmp (rtx dest, enum
rtvec v = rtvec_alloc (n_elts);
for (i = 0; i < n_elts; ++i)
RTVEC_ELT (v, i)
- = GEN_INT (INTVAL (CONST_VECTOR_ELT (cop1, i)) - 1);
+ = gen_int_mode (INTVAL (CONST_VECTOR_ELT (cop1, i)) - 1,
+ eltmode);
cop1 = gen_rtx_CONST_VECTOR (mode, v);
code = code == GE ? GT : GTU;
break;
--- gcc/testsuite/gcc.dg/pr107585.c.jj 2022-11-09 14:52:37.554779118 +0100
+++ gcc/testsuite/gcc.dg/pr107585.c 2022-11-09 14:48:24.063258991 +0100
@@ -0,0 +1,13 @@
+/* PR target/107585 */
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+
+typedef unsigned char __attribute__((__vector_size__ (16))) V;
+char c;
+void bar (int);
+
+void
+foo (void)
+{
+ bar (((V) (c <= (V){127}))[2]);
+}
Jakub
next reply other threads:[~2022-11-10 9:29 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-10 9:29 Jakub Jelinek [this message]
2022-11-10 9:41 ` Uros Bizjak
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=Y2zEZ6f0v/74nBbT@tucnak \
--to=jakub@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=ubizjak@gmail.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).