public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Jakub Jelinek <jakub@redhat.com>
To: Richard Biener <rguenther@suse.de>,
	"Joseph S. Myers" <josmyers@redhat.com>
Cc: gcc-patches@gcc.gnu.org
Subject: [PATCH] libgcc: Fix quotient and/or remainder negation in __divmodbitint4 [PR114327]
Date: Fri, 15 Mar 2024 09:35:26 +0100	[thread overview]
Message-ID: <ZfQIToN7FocwQcRj@tucnak> (raw)

Hi!

While for __mulbitint3 we actually don't negate anything and perform the
multiplication in unsigned style always, for __divmodbitint4 if the operands
aren't unsigned and are negative, we negate them first and then try to
negate them as needed at the end.
quotient is negated if just one of the operands was negated and the other
wasn't or vice versa, and remainder is negated if the first operand was
negated.
The case which doesn't work correctly is if due to limited range of the
operands we perform the division/modulo in some smaller number of limbs
and then extend it to the desired precision of the quotient and/or
remainder results.  If they aren't negated, the extension is done with
memset to 0, if they are negated, the extension was done with memset
to -1.  The problem is that if the quotient or remainder is zero,
then bitint_negate negates it again to zero (that is ok), but we should
then extend with memset to 0, not memset to -1.

The following patch achieves that by letting bitint_negate also check if
the negated operand is zero and changes the memset argument based on that.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2024-03-15  Jakub Jelinek  <jakub@redhat.com>

	PR libgcc/114327
	* libgcc2.c (bitint_negate): Return UWtype bitwise or of all the limbs
	before negation rather than void.
	(__divmodbitint4): Determine whether to fill in the upper limbs after
	negation based on whether bitint_negate returned 0 or non-zero, rather
	then always filling with -1.

	* gcc.dg/torture/bitint-63.c: New test.

--- libgcc/libgcc2.c.jj	2024-02-02 22:14:14.946684874 +0100
+++ libgcc/libgcc2.c	2024-03-14 13:37:38.227332706 +0100
@@ -1642,19 +1642,22 @@ __mulbitint3 (UBILtype *ret, SItype retp
 #ifdef L_divmodbitint4
 /* D = -S.  */
 
-static void
+static UWtype
 bitint_negate (UBILtype *d, const UBILtype *s, SItype n)
 {
   UWtype c = 1;
+  UWtype r = 0;
   do
     {
       UWtype sv = *s, lo;
+      r |= sv;
       s += BITINT_INC;
       c = __builtin_add_overflow (~sv, c, &lo);
       *d = lo;
       d += BITINT_INC;
     }
   while (--n);
+  return r;
 }
 
 /* D -= S * L.  */
@@ -1977,10 +1980,10 @@ __divmodbitint4 (UBILtype *q, SItype qpr
 	    n = qn;
 	  else
 	    n = un - vn + 1;
-	  bitint_negate (q + BITINT_END (qn - 1, 0),
-			 q2 + BITINT_END (un - vn, 0), n);
+	  SItype c = bitint_negate (q + BITINT_END (qn - 1, 0),
+				    q2 + BITINT_END (un - vn, 0), n) ? -1 : 0;
 	  if (qn > n)
-	    __builtin_memset (q + BITINT_END (0, n), -1,
+	    __builtin_memset (q + BITINT_END (0, n), c,
 			      (qn - n) * sizeof (UWtype));
 	}
       else
@@ -1999,11 +2002,11 @@ __divmodbitint4 (UBILtype *q, SItype qpr
       if (uprec < 0)
 	{
 	  /* Negative remainder.  */
-	  bitint_negate (r + BITINT_END (rn - 1, 0),
-			 r + BITINT_END (rn - 1, 0),
-			 rn > vn ? vn : rn);
+	  SItype c = bitint_negate (r + BITINT_END (rn - 1, 0),
+				    r + BITINT_END (rn - 1, 0),
+				    rn > vn ? vn : rn) ? -1 : 0;
 	  if (rn > vn)
-	    __builtin_memset (r + BITINT_END (0, vn), -1,
+	    __builtin_memset (r + BITINT_END (0, vn), c,
 			      (rn - vn) * sizeof (UWtype));
 	}
       else
--- gcc/testsuite/gcc.dg/torture/bitint-63.c.jj	2024-03-14 13:46:31.591938158 +0100
+++ gcc/testsuite/gcc.dg/torture/bitint-63.c	2024-03-14 13:45:58.306399642 +0100
@@ -0,0 +1,30 @@
+/* PR libgcc/114327 */
+/* { dg-do run { target bitint } } */
+/* { dg-options "-std=c23" } */
+/* { dg-skip-if "" { ! run_expensive_tests }  { "*" } { "-O0" "-O2" } } */
+/* { dg-skip-if "" { ! run_expensive_tests } { "-flto" } { "" } } */
+
+#if __BITINT_MAXWIDTH__ >= 256
+_BitInt(256)
+foo (_BitInt(256) b, _BitInt(256) c)
+{
+  return b % c;
+}
+
+_BitInt(256)
+bar (_BitInt(256) b, _BitInt(256) c)
+{
+  return b / c;
+}
+#endif
+
+int
+main ()
+{
+#if __BITINT_MAXWIDTH__ >= 256
+  if (foo (-0x9e9b9fe60wb, 1wb))
+    __builtin_abort ();
+  if (bar (1wb, -0x9e9b9fe60wb))
+    __builtin_abort ();
+#endif
+}

	Jakub


             reply	other threads:[~2024-03-15  8:35 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-15  8:35 Jakub Jelinek [this message]
2024-03-15 17:05 ` Joseph Myers

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=ZfQIToN7FocwQcRj@tucnak \
    --to=jakub@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=josmyers@redhat.com \
    --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).