From: Maksim Kokryashkin via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: tarantool-patches@dev.tarantool.org, sergos@tarantool.org,
skaplun@tarantool.org, m.kokryashkin@tarantool.org
Subject: [Tarantool-patches] [PATCH luajit v3 4/4] Fix IR_RENAME snapshot number. Follow-up fix for a32aeadc.
Date: Tue, 28 Mar 2023 03:03:17 +0300 [thread overview]
Message-ID: <20230328000317.33238-5-max.kokryashkin@gmail.com> (raw)
In-Reply-To: <20230328000317.33238-1-max.kokryashkin@gmail.com>
From: Mike Pall <mike>
Reported by Victor Bombi, analyzed by XmiliaH. Thanks!
(cherry-picked from commit bf51d3535109c4745bfbbe19a5587a9eac00259a)
If the `snapalloc` flag is set, then the allocation hasn't
occurred yet, meaning that rename is applied to the next
snapshot. Otherwise, refs are already allocated and rename
is applied to the current snapshot.
Maxim Kokryashkin:
* added the description and the test for the problem
Part of tarantool/tarantool#7745
Part of tarantool/tarantool#8069
---
src/lj_asm.c | 9 ++++++++-
.../gh-7745-ir-rename.test.lua | 20 +++++++++++++++++++
2 files changed, 28 insertions(+), 1 deletion(-)
create mode 100644 test/tarantool-tests/gh-7745-ir-rename.test.lua
diff --git a/src/lj_asm.c b/src/lj_asm.c
index adfaf286..929a6da6 100644
--- a/src/lj_asm.c
+++ b/src/lj_asm.c
@@ -682,7 +682,14 @@ static void ra_rename(ASMState *as, Reg down, Reg up)
RA_DBGX((as, "rename $f $r $r", regcost_ref(as->cost[up]), down, up));
emit_movrr(as, ir, down, up); /* Backwards codegen needs inverse move. */
if (!ra_hasspill(IR(ref)->s)) { /* Add the rename to the IR. */
- ra_addrename(as, down, ref, as->snapno);
+ /*
+ ** The rename is effective at the subsequent (already emitted) exit
+ ** branch. This is for the current snapshot (as->snapno). Except if we
+ ** haven't yet allocated any refs for the snapshot (as->snapalloc == 1),
+ ** then it belongs to the next snapshot.
+ ** See also the discussion at asm_snap_checkrename().
+ */
+ ra_addrename(as, down, ref, as->snapno + as->snapalloc);
}
}
diff --git a/test/tarantool-tests/gh-7745-ir-rename.test.lua b/test/tarantool-tests/gh-7745-ir-rename.test.lua
new file mode 100644
index 00000000..d03f6da5
--- /dev/null
+++ b/test/tarantool-tests/gh-7745-ir-rename.test.lua
@@ -0,0 +1,20 @@
+local tap = require('tap')
+local test = tap.test('IR_RENAME on snapshot allocation'):skipcond({
+ ['Test requires JIT enabled'] = not jit.status(),
+})
+test:plan(1)
+
+jit.opt.start('hotloop=1')
+
+local vals = {-0.1, 0.1, -0.1, 0.1}
+local i = 1
+local _
+while i < 5 do
+ assert(i ~= 1.1)
+ local l1 = vals[i]
+ _ = l1 > 0
+ i = i + 1
+end
+
+test:ok(true, 'IR_RENAME is fine')
+os.exit(test:check())
--
2.37.1 (Apple Git-137.1)
next prev parent reply other threads:[~2023-03-28 0:05 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-28 0:03 [Tarantool-patches] [PATCH luajit v3 0/4] jit: add exception unwinding Maksim Kokryashkin via Tarantool-patches
2023-03-28 0:03 ` [Tarantool-patches] [PATCH luajit v3 1/4] Handle on-trace OOM errors from helper functions Maksim Kokryashkin via Tarantool-patches
2023-04-05 6:32 ` Sergey Kaplun via Tarantool-patches
2023-03-28 0:03 ` [Tarantool-patches] [PATCH luajit v3 2/4] Disable unreliable assertion for external frame unwinding Maksim Kokryashkin via Tarantool-patches
2023-05-03 9:31 ` sergos via Tarantool-patches
2023-03-28 0:03 ` [Tarantool-patches] [PATCH luajit v3 3/4] OSX: " Maksim Kokryashkin via Tarantool-patches
2023-03-28 0:03 ` Maksim Kokryashkin via Tarantool-patches [this message]
2023-04-05 5:33 ` [Tarantool-patches] [PATCH luajit v3 4/4] Fix IR_RENAME snapshot number. Follow-up fix for a32aeadc Sergey Kaplun 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=20230328000317.33238-5-max.kokryashkin@gmail.com \
--to=tarantool-patches@dev.tarantool.org \
--cc=m.kokryashkin@tarantool.org \
--cc=max.kokryashkin@gmail.com \
--cc=sergos@tarantool.org \
--cc=skaplun@tarantool.org \
--subject='Re: [Tarantool-patches] [PATCH luajit v3 4/4] Fix IR_RENAME snapshot number. Follow-up fix for a32aeadc.' \
/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