Re: [pcre-dev] JIT fails with NEON instructions

Top Page
Delete this message
Author: Zoltán Herczeg
Date:  
To: Sebastian Pop
CC: pcre-dev@exim.org
Subject: Re: [pcre-dev] JIT fails with NEON instructions
Hi Sebastian,

thank you for the investigation. Could you estimate when you can send the patch?

Regards,
Zoltan
 
-------- Eredeti levél --------
Feladó: Sebastian Pop < sebpop@??? (Link -> mailto:sebpop@gmail.com) >
Dátum: 2019 október 30 23:00:04
Tárgy: Re: [pcre-dev] JIT fails with NEON instructions
Címzett: Zoltán Herczeg < hzmester@??? (Link -> mailto:hzmester@freemail.hu) >
 
The problem occurs on Ubuntu 18.04 and A-72 as well
when configuring with --enable-pcre2-16 and --enable-pcre2-32
each exposes one fail.
I will send a patch to fix these issues.
 
Sebastian
 
On Wed, Oct 30, 2019 at 3:47 PM Sebastian Pop < sebpop@??? (Link -> mailto:sebpop@gmail.com) > wrote:
A previous email to the pcre-dev list shows the log output:
https://lists.exim.org/lurker/message/20191030.143840.c1a0847b.en.html (Link -> https://lists.exim.org/lurker/message/20191030.143840.c1a0847b.en.html)
 
My testing for the patch was on both Ubuntu 18.04 and 19.04 on A1 instance
based on A-72.  I will try to see if I can reproduce the bug on A1 with a Fedora
distribution.
 
Sebastian
 
On Wed, Oct 30, 2019 at 3:37 PM Sebastian Pop < sebpop@??? (Link -> mailto:sebpop@gmail.com) > wrote:
Thanks Zoltán for the heads up.
I can work on fixing the reported issues.
Is there a log of the output and instructions on how to reproduce the problems?
I would need the configure or cmake line and the make command.
 
Thanks,
Sebastian
 
On Wed, Oct 30, 2019 at 1:28 PM Zoltán Herczeg < hzmester@??? (Link -> mailto:hzmester@freemail.hu) > wrote:
Hi Sebastian,
could you check this failure?
Regards,
Zoltan
-------- Eredeti levél --------
Feladó: Petr Pisar via Pcre-dev < pcre-dev@??? (Link -> mailto:pcre-dev@exim.org) ) >
Dátum: 2019 október 30 18:25:42
Tárgy: Re: [pcre-dev] JIT fails with NEON instructions
Címzett: pcre-dev@??? (Link -> mailto:pcre-dev@exim.org) )
On Wed, Oct 30, 2019 at 03:38:45PM +0100, Petr Pisar via Pcre-dev wrote:
> On Thu, Oct 17, 2019 at 05:54:08PM +0100, ph10@??? (Link -> mailto:ph10@hermes.cam.ac.uk) wrote:
> > I have just made available a Release Candidate for PCRE2 10.34 here:
> >
> > https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.gz (Link -> https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.gz)
> > https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.bz2 (Link -> https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.bz2)
> > https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.zip (Link -> https://ftp.pcre.org/pub/pcre/Testing/pcre2-10.34-RC1.tar.zip)
> >
> > NOTE: this is a different FTP site than was used for previous releases.
> >
> I experience two test failures on 64-bit ARM with Fedora 32 distribution:
>

Both failures are triggered by this PCRE2 commit:
commit a6b4b9c7f9e3d20966de682487ee575cba053a20
Author: zherczeg <zherczeg@6239d852-aaf2-0410-a92c-79f79f948069>
Date: Tue Sep 17 06:59:45 2019 +0000
Support NEON based fast forward character search in ARM64. Patch by Sebastian Pop.
git-svn-id: svn:// vcs.exim.org/pcre2/code/trunk@1172 (Link -> http://vcs.exim.org/pcre2/code/trunk@1172) 6239d852-aaf2-0410-a92c-79f79f948
069
I verified it on this machine:
# lscpu
Architecture: aarch64
CPU op-mode(s): 64-bit
Byte Order: Little Endian
CPU(s): 4
On-line CPU(s) list: 0-3
Thread(s) per core: 1
Core(s) per socket: 4
Socket(s): 1
NUMA node(s): 1
Vendor ID: Cavium
Model: 1
Model name: ThunderX2 99xx
Stepping: 0x1
BogoMIPS: 400.00
NUMA node0 CPU(s): 0-3
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Spec store bypass: Vulnerable
Vulnerability Spectre v1: Mitigation; __user pointer sanitization
Vulnerability Spectre v2: Vulnerable
Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics cpuid
asimdrdm
using Fedora 31 (gcc-9.2.1, binutils-2.32).
The compiler defines __ARM_NEON macro on that machine. getauxval(AT_HWCAP) sets
HWCAP_ASIMD bit. So I believe the machine supports NEON.
When I disable the
#if (defined SLJIT_CONFIG_ARM_64 && SLJIT_CONFIG_ARM_64 && (defined __ARM_NEON || defined
__ARM_NEON__))
branch in src/pcre2_jit_compile.c not to emit NEON instructions, all tests
pass on that machine.
This I conclude something is wrong in the NEON support in sljit.
-- Petr