Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:109847
To: "internals@lists.php.net" <internals@lists.php.net>
Thread-Topic: Optimizing assembly for spl_object_id(), fdiv(), etc in the JIT?
Thread-Index: AQHWGxz5+Txr028sTUe2R2cg2slM+A==
Date: Sat, 25 Apr 2020 16:43:46 +0000
Message-ID:
 <DM6PR07MB661819D3D9D654287317A56DF9D10@DM6PR07MB6618.namprd07.prod.outlook.com>
Accept-Language: en-CA, en-US
Content-Language: en-CA
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-MS-Exchange-CrossTenant-RMS-PersistedConsumerOrg: 00000000-0000-0000-0000-000000000000
X-MS-Exchange-CrossTenant-Network-Message-Id: c5f373c9-d987-4dbe-df83-08d7e937d2a7
X-MS-Exchange-CrossTenant-rms-persistedconsumerorg: 00000000-0000-0000-0000-000000000000
X-MS-Exchange-CrossTenant-originalarrivaltime: 25 Apr 2020 16:43:46.2474
 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Internet
X-MS-Exchange-CrossTenant-id: 84df9e7f-e9f6-40af-b435-aaaaaaaaaaaa
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN8NAM12HT032
Subject: Optimizing assembly for spl_object_id(), fdiv(), etc in the JIT?
From: tysonandre775@hotmail.com (tyson andre)

Hi internals,=0A=
=0A=
I was wondering if the maintainers of the Opcache Jit would be open to work=
 on optimizing functions such as intdiv(), fdiv(), and spl_object_id() in t=
he JIT to assembly code when it was safe to do so.=0A=
These functions have simple C implementations, so the performance overhead =
of calling PHP functions in the generated assembly=0A=
would be noticeable compared to emitting optimized assembly.=0A=
=0A=
I'd expect that it'd be safe to do so under the following circumstances:=0A=
=0A=
- All arguments are Compiled Variables(CV) in the opcodes (i.e. $var)=0A=
  (http://nikic.github.io/2017/04/14/PHP-7-Virtual-machine.html#variable-ty=
pes)=0A=
- The types and count of the function arguments are known to be strictly co=
rrect, e.g.=0A=
  - Opcache infers they will not throw a TypeError or emit undefined variab=
le warnings=0A=
  - maybe strictly accept floats but not integers for fdiv=0A=
- The return value of the expression won't throw=0A=
  (e.g. don't perform this optimization for `$y =3D spl_object_id($x)` when=
 a CV is the return value)=0A=
- The function in question exists and is not disabled.=0A=
=0A=
For example, these operands would be seen by the JIT for this sequence of o=
pcodes=0A=
=0A=
```php=0A=
/*=0A=
0000 CV0($x) =3D RECV 1=0A=
0001 INIT_FCALL 1 96 string("spl_object_id")=0A=
0002 SEND_VAR CV0($x) 1=0A=
0003 V2 =3D DO_ICALL=0A=
0004 T1 =3D INIT_ARRAY 1 (packed) CV0($x) V2=0A=
0005 RETURN T1=0A=
*/=0A=
function create_set(stdClass $x) : array {=0A=
    return [spl_object_id($x) =3D> $x];=0A=
}=0A=
```=0A=
=0A=
I expect it to be technically feasible to check for the sequence of opcodes=
 INIT_FCALL, SEND_VAR (repeated), and DO_ICALL, in ext/opcache/jit/zend_jit=
.c. The resulting assembly would be smaller and much faster.=0A=
(this would be done before emitting any assembly - skip over the opcodes fo=
r INIT_ARRAY and SEND_VAR and DO_ICALL)=0A=
=0A=
- What do the maintainers of the JIT module think of this idea?=0A=
  (e.g. are there concerns about this making it harder to understand the JI=
T codebase, adding potential bugs, or slowing down the generation of assemb=
ly)=0A=
- Are there any general guidelines (or talks/articles) you'd recommend for =
new contributors to the Opcache JIT? I couldn't find a README and https://w=
iki.php.net/rfc/jit doesn't seem to make recommendations that apply to my q=
uestion.=0A=
=0A=
```=0A=
// core fdiv C implementation (does not throw for a divisor of 0)=0A=
	RETURN_DOUBLE(dividend / divisor);=0A=
// core of spl_object_id C implementation=0A=
	RETURN_LONG((zend_long)Z_OBJ_HANDLE_P(obj));=0A=
```=0A=
=0A=
I started looking into this because I work on an application that heavily u=
ses spl_object_id()=0A=
and would see a small performance benefit from this,=0A=
and was reminded of this when looking at the get_resource_id() proposal.=0A=
(https://github.com/phan/phan/search?q=3Dspl_object_id&unscoped_q=3Dspl_obj=
ect_id)=0A=
=0A=
Thanks,=0A=
- Tyson=