All of lore.kernel.org
 help / color / mirror / Atom feed
From: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
To: linux-sparse@vger.kernel.org
Cc: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Subject: [PATCH 08/10] cmps: canonicalize signed compares with constant
Date: Tue, 26 Jan 2021 23:04:30 +0100	[thread overview]
Message-ID: <20210126220432.58265-9-luc.vanoostenryck@gmail.com> (raw)
In-Reply-To: <20210126220432.58265-1-luc.vanoostenryck@gmail.com>

Modify the constants to canonicalize (x < C) to (x <= (C-1)) and
(x <= C) to (x > (C-1)). This choice is partially arbitrary but
1) it's the one with the smallest positive constants,
2) it eliminates all OP_SET_LT & OP_SET_GE with a constant.
A disadvantage of this choice is that we lost some compares with 0:
(x < 0) is now canonicalized into (x <= -1).

Note: Another good choice would be to canonicalize using the
      smallest absolute constants. This would keep compares with 0
      but would also keep the 4 kinds of comparison.

Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
---
 simplify.c                        | 2 ++
 validation/optim/canonical-cmps.c | 1 -
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/simplify.c b/simplify.c
index ad3adb11595e..b29f5d5b2444 100644
--- a/simplify.c
+++ b/simplify.c
@@ -1177,6 +1177,7 @@ static int simplify_compare_constant(struct instruction *insn, long long value)
 			return replace_opcode(insn, OP_SET_NE);
 		if (value == sign_bit(size) + 1)// (x < SMIN + 1) --> (x == SMIN)
 			return replace_binop_value(insn, OP_SET_EQ, sign_bit(size));
+		changed |= replace_binop_value(insn, OP_SET_LE, (value - 1) & bits);
 		break;
 	case OP_SET_LE:
 		if (value == sign_mask(size))	// (x <= SMAX) --> 1
@@ -1193,6 +1194,7 @@ static int simplify_compare_constant(struct instruction *insn, long long value)
 			return replace_opcode(insn, OP_SET_EQ);
 		if (value == sign_bit(size) + 1)// (x >= SMIN + 1) --> (x != SMIN)
 			return replace_binop_value(insn, OP_SET_NE, sign_bit(size));
+		changed |= replace_binop_value(insn, OP_SET_GT, (value - 1) & bits);
 		break;
 	case OP_SET_GT:
 		if (value == sign_mask(size))	// (x >  SMAX) --> 0
diff --git a/validation/optim/canonical-cmps.c b/validation/optim/canonical-cmps.c
index f42664b21e04..42801cdce520 100644
--- a/validation/optim/canonical-cmps.c
+++ b/validation/optim/canonical-cmps.c
@@ -10,7 +10,6 @@ _Bool ge_x(int a) { return (a >= 1234) == (a > 1233); }
 /*
  * check-name: canonical-cmps
  * check-command: test-linearize -Wno-decl $file
- * check-known-to-fail
  *
  * check-output-ignore
  * check-output-returns: 1
-- 
2.30.0


  parent reply	other threads:[~2021-01-27  3:46 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-26 22:04 [PATCH 00/10] simplify and canonicalize signed compares Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 01/10] cmps: make clearer we're using the operands' size Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 02/10] cmps: fix simplification of sext(x) + signed compare of {SMAX,SMIN} Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 03/10] cmpu: fix canonicalization of unsigned (x {<,>=} C) --> (x {<=,>} C-1) Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 04/10] cmps: add testcases for simplification of signed compares Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 05/10] cmps: simplify signed compares with SMIN or SMAX Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 06/10] cmps: canonicalize signed compares with SMIN/SMAX Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 07/10] cmps: canonicalize SMIN/SMAX +- 1 --> EQ/NE Luc Van Oostenryck
2021-01-26 22:04 ` Luc Van Oostenryck [this message]
2021-01-26 22:04 ` [PATCH 09/10] cmps: canonicalize SEL(x {<,<=} y, a, b) --> SEL(x {>=,>} y, b, a) Luc Van Oostenryck
2021-01-26 22:04 ` [PATCH 10/10] cmps: canonicalize SEL(x > 0, a, -a) --> SEL(x >= 0, a, -a) Luc Van Oostenryck
2021-04-17 17:16 ` [PATCH 00/10] simplify and canonicalize signed compares Linus Torvalds
2021-04-17 18:20   ` Luc Van Oostenryck

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=20210126220432.58265-9-luc.vanoostenryck@gmail.com \
    --to=luc.vanoostenryck@gmail.com \
    --cc=linux-sparse@vger.kernel.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.