Tarantool development patches archive
 help / color / mirror / Atom feed
From: Sergey Kaplun via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: Maxim Kokryashkin <m.kokryashkin@tarantool.org>,
	Sergey Bronnikov <sergeyb@tarantool.org>
Cc: tarantool-patches@dev.tarantool.org
Subject: [Tarantool-patches] [PATCH luajit] Fix FOLD rule for x-0.
Date: Tue, 31 Oct 2023 15:44:46 +0300	[thread overview]
Message-ID: <20231031124446.17478-1-skaplun@tarantool.org> (raw)

From: Mike Pall <mike>

Reported by XmiliaH.

(cherry-picked from commit 7b994e0ee0399caf6319865bbac88ddf62129a36)

Fold optimization x - (-0) ==> x is INVALID for x = -0 in FP arithmetic.
Its result is -0 instead of +0. This patch allows only x - (+0) ==> x
optimization.

Sergey Kaplun:
* added the description and the test for the problem

Part of tarantool/tarantool#9145
---

Branch: https://github.com/tarantool/luajit/tree/skaplun/lj-783-fix-fold-x-0
Tarantool PR: https://github.com/tarantool/tarantool/pull/9320
Related Issues:
* https://github.com/tarantool/tarantool/issues/9145
* https://github.com/LuaJIT/LuaJIT/pull/783

 src/lj_opt_fold.c                            |  3 +--
 test/tarantool-tests/lj-783-fold--0.test.lua | 28 ++++++++++++++++++++
 2 files changed, 29 insertions(+), 2 deletions(-)
 create mode 100644 test/tarantool-tests/lj-783-fold--0.test.lua

diff --git a/src/lj_opt_fold.c b/src/lj_opt_fold.c
index 09e6c87b..944a9ecc 100644
--- a/src/lj_opt_fold.c
+++ b/src/lj_opt_fold.c
@@ -1005,8 +1005,7 @@ LJFOLDF(simplify_numadd_xneg)
 LJFOLD(SUB any KNUM)
 LJFOLDF(simplify_numsub_k)
 {
-  lua_Number n = knumright;
-  if (n == 0.0)  /* x - (+-0) ==> x */
+  if (ir_knum(fright)->u64 == 0)  /* x - (+0) ==> x */
     return LEFTFOLD;
   return NEXTFOLD;
 }
diff --git a/test/tarantool-tests/lj-783-fold--0.test.lua b/test/tarantool-tests/lj-783-fold--0.test.lua
new file mode 100644
index 00000000..d6b4b493
--- /dev/null
+++ b/test/tarantool-tests/lj-783-fold--0.test.lua
@@ -0,0 +1,28 @@
+local tap = require('tap')
+
+-- Test file to demonstrate LuaJIT's incorrect fold optimization
+-- x - (-0) ==> x.
+-- See also https://github.com/LuaJIT/LuaJIT/issues/783.
+local test = tap.test('lj-783-fold--0'):skipcond({
+  ['Test requires JIT enabled'] = not jit.status(),
+})
+
+test:plan(2)
+
+-- XXX: Use the variable to avoid folding during parsing.
+local minus_zero = -0
+local results = {}
+
+jit.opt.start('hotloop=1')
+
+for i = 1, 4 do
+  results[i] = tostring(minus_zero - (-0))
+end
+
+-- Fold optimization x - (-0) ==> x is INVALID for x = -0 in FP
+-- arithmetic. Its result is -0 instead of +0.
+
+test:is(results[1], '0', 'correct VM value for -0 - (-0)')
+test:samevalues(results, '-0 folding in simplify_numsub_k')
+
+test:done(true)
-- 
2.42.0


             reply	other threads:[~2023-10-31 12:49 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-31 12:44 Sergey Kaplun via Tarantool-patches [this message]
2023-10-31 15:17 ` Maxim Kokryashkin via Tarantool-patches
2023-11-08 19:36 ` Sergey Bronnikov via Tarantool-patches
2023-11-23  6:29 ` Igor Munkin via Tarantool-patches

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=20231031124446.17478-1-skaplun@tarantool.org \
    --to=tarantool-patches@dev.tarantool.org \
    --cc=m.kokryashkin@tarantool.org \
    --cc=sergeyb@tarantool.org \
    --cc=skaplun@tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH luajit] Fix FOLD rule for x-0.' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox