Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:124569
Feedback-ID: ifab94697:Fastmail
User-Agent: Cyrus-JMAP/3.11.0-alpha0-582-g5a02f8850-fm-20240719.002-g5a02f885
Precedence: bulk
MIME-Version: 1.0
Message-ID: <889a6014-299b-4be5-956e-ae5134bb6dfb@app.fastmail.com>
In-Reply-To: 
 <CAEZPtU4eWkisw2Axax=Q1HAGBe214d1Gcm+kLu6H5GKyqBDBJA@mail.gmail.com>
References: <9828b776-660f-44ad-91df-11bea14230a4@app.fastmail.com>
 <CAEZPtU4eWkisw2Axax=Q1HAGBe214d1Gcm+kLu6H5GKyqBDBJA@mail.gmail.com>
Date: Wed, 24 Jul 2024 14:06:15 +0200
To: "Pierre Joye" <pierre.php@gmail.com>
Cc: internals@lists.php.net
Subject: Re: [PHP-DEV] tsrm question
Content-Type: multipart/alternative;
 boundary=72efbbd699cf4c159b1043599f20cbab
From: rob@bottled.codes ("Rob Landers")

--72efbbd699cf4c159b1043599f20cbab
Content-Type: text/plain;charset=utf-8
Content-Transfer-Encoding: quoted-printable

On Wed, Jul 24, 2024, at 13:54, Pierre Joye wrote:
> Hi Rob,
>=20
>=20
> On Wed, Jul 24, 2024 at 3:17=E2=80=AFPM Rob Landers <rob@bottled.codes=
> wrote:
> >
> > Hello internals,
> >
> > Last night I went down a rabbit hole with Frankenphp: https://github=
.com/dunglas/frankenphp/pull/933
> >
> > It isn't 100% clear to me what values `ts_resource(id)` holds and if=
 it needs to be freed/allocated per request or per thread. The performan=
ce impact is huge to reallocate on every request (mostly due to the glob=
al mutex during allocation). Is anyone familiar with this and could help=
 reviewing the changes there?
>=20
> I suppose you already looked, but if not, maybe it gives you some dire=
ction(s).
>=20
> The key part is kind of documented here:
>=20
> https://github.com/php/php-src/blob/8e93eb2e79cea5fcca6769b46a429de042=
660da9/TSRM/TSRM.c#L417
>=20
> tsrm knows nothing about requests but only threads.Its jobs are about
> threads data.
>=20
> A good start (surely already looked) are apache and litespeed (or
> embed) for php8, however besides the API names change, the behavior or
> goals of each ts_ or tsrm_ functions remain the same as what we had
> with 5.x. One with a very similar modus operandi than Franken was
> ISAPI for IIS. Mind the names changes, dapt to php8's new names (and
> some behaviors) but the basics and core flows have been kept.
>=20
> If you know (~) how many threads franken is going to need, tsrm allows
> you to preallocate the rsc slots using tsrm_startup_ex. It uses
> tsrm_startup, a more powerful version, which lets franken define how
> many rsc per thread will be preallocated. I suppose that could help a
> bit performance wise:
>=20
> https://github.com/php/php-src/blob/8e93eb2e79cea5fcca6769b46a429de042=
660da9/main/main.c#L2716C13-L2716C25
>=20
> As per the last question, about when to free them:
> tsrm_shutdown (franken shutdown) will free all rsc
>=20
> When a thread is detached (franken kills it for example),
> ts_free_thread needs to be called to kill its rsc. They will be freed
> anyway too but good to be clean.
>=20
> The tsrm resource management will also automatically purge orphan rsrc
> (f.e. when a thread dies unexpectedly and ts_free_thread could not be
> called).
>=20
> Long story short, the TSRM API is very flexible, how and when you
> alloc/free rsrc is basically up to you. It is possible to keep some
> around in the root thread (and be used in other threads, given the
> root thread id is known (can be 0 or else depending how franken
> managed them.
>=20
> I hope it helps and did not say too many outdated/wrong explanations :=
).
>=20
> Btw, Welting rewrote that part with Dmitry back then to fix long
> standing issues (and drop TSRM_LS/DC/CC uses), not sure if he is still
> around but he may help.
>=20
> Best,
> --
> Pierre
>=20
> @pierrejoye | http://www.libgd.org
>=20

Hey Pierre,

Thank you! This helps tremendously.

> A good start (surely already looked) are apache and litespeed (or
> embed) for php8, however besides the API names change, the behavior or
> goals of each ts_ or tsrm_ functions remain the same as what we had
> with 5.x. One with a very similar modus operandi than Franken was
> ISAPI for IIS. Mind the names changes, dapt to php8's new names (and
> some behaviors) but the basics and core flows have been kept.

I looked at litespeed but not apache, so I will look to that as well.

> If you know (~) how many threads franken is going to need, tsrm allows
> you to preallocate the rsc slots using tsrm_startup_ex. It uses
> tsrm_startup, a more powerful version, which lets franken define how
> many rsc per thread will be preallocated. I suppose that could help a
> bit performance wise:

The number of threads are static (at least for now), so we should def do=
 this.

> Long story short, the TSRM API is very flexible, how and when you
> alloc/free rsrc is basically up to you. It is possible to keep some
> around in the root thread (and be used in other threads, given the
> root thread id is known (can be 0 or else depending how franken
> managed them.

Ah, so this is where I was getting confused. It's unclear what the "id" =
is and what it is used for. For example, I see the engine has a couple o=
f ids (compiler/execution) but I'm unclear what the sapi should be doing=
 with them. Should these "ids" be different per thread, or is it just an=
 arbitrary key?

=E2=80=94 Rob
--72efbbd699cf4c159b1043599f20cbab
Content-Type: text/html;charset=utf-8
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE html><html><head><title></title><style type=3D"text/css">p.Mso=
Normal,p.MsoNoSpacing{margin:0}</style></head><body><div>On Wed, Jul 24,=
 2024, at 13:54, Pierre Joye wrote:<br></div><blockquote type=3D"cite" i=
d=3D"qt" style=3D""><div>Hi Rob,<br></div><div><br></div><div><br></div>=
<div>On Wed, Jul 24, 2024 at 3:17=E2=80=AFPM Rob Landers &lt;<a href=3D"=
mailto:rob@bottled.codes">rob@bottled.codes</a>&gt; wrote:<br></div><div=
>&gt;<br></div><div>&gt; Hello internals,<br></div><div>&gt;<br></div><d=
iv>&gt; Last night I went down a rabbit hole with Frankenphp:&nbsp;<a hr=
ef=3D"https://github.com/dunglas/frankenphp/pull/933">https://github.com=
/dunglas/frankenphp/pull/933</a><br></div><div>&gt;<br></div><div>&gt; I=
t isn't 100% clear to me what values `ts_resource(id)` holds and if it n=
eeds to be freed/allocated per request or per thread. The performance im=
pact is huge to reallocate on every request (mostly due to the global mu=
tex during allocation). Is anyone familiar with this and could help revi=
ewing the changes there?<br></div><div><br></div><div>I suppose you alre=
ady looked, but if not, maybe it gives you some direction(s).<br></div><=
div><br></div><div>The key part is kind of documented here:<br></div><di=
v><br></div><div><a href=3D"https://github.com/php/php-src/blob/8e93eb2e=
79cea5fcca6769b46a429de042660da9/TSRM/TSRM.c#L417">https://github.com/ph=
p/php-src/blob/8e93eb2e79cea5fcca6769b46a429de042660da9/TSRM/TSRM.c#L417=
</a><br></div><div><br></div><div>tsrm knows nothing about requests but =
only threads.Its jobs are about<br></div><div>threads data.<br></div><di=
v><br></div><div>A good start (surely already looked) are apache and lit=
espeed (or<br></div><div>embed) for php8, however besides the API names =
change, the behavior or<br></div><div>goals of each ts_ or tsrm_ functio=
ns remain the same as what we had<br></div><div>with 5.x. One with a ver=
y similar modus operandi than Franken was<br></div><div>ISAPI for IIS. M=
ind the names changes, dapt to php8's new names (and<br></div><div>some =
behaviors) but the basics and core flows have been kept.<br></div><div><=
br></div><div>If you know (~) how many threads franken is going to need,=
 tsrm allows<br></div><div>you to preallocate the rsc slots using tsrm_s=
tartup_ex. It uses<br></div><div>tsrm_startup, a more powerful version, =
which lets franken define how<br></div><div>many rsc per thread will be =
preallocated. I suppose that could help a<br></div><div>bit performance =
wise:<br></div><div><br></div><div><a href=3D"https://github.com/php/php=
-src/blob/8e93eb2e79cea5fcca6769b46a429de042660da9/main/main.c#L2716C13-=
L2716C25">https://github.com/php/php-src/blob/8e93eb2e79cea5fcca6769b46a=
429de042660da9/main/main.c#L2716C13-L2716C25</a><br></div><div><br></div=
><div>As per the last question, about when to free them:<br></div><div>t=
srm_shutdown (franken shutdown) will free all rsc<br></div><div><br></di=
v><div>When a thread is detached (franken kills it for example),<br></di=
v><div>ts_free_thread needs to be called to kill its rsc. They will be f=
reed<br></div><div>anyway too but good to be clean.<br></div><div><br></=
div><div>The tsrm resource management will also automatically purge orph=
an rsrc<br></div><div>(f.e. when a thread dies unexpectedly and ts_free_=
thread could not be<br></div><div>called).<br></div><div><br></div><div>=
Long story short, the TSRM API is very flexible, how and when you<br></d=
iv><div>alloc/free rsrc is basically up to you. It is possible to keep s=
ome<br></div><div>around in the root thread (and be used in other thread=
s, given the<br></div><div>root thread id is known (can be 0 or else dep=
ending how franken<br></div><div>managed them.<br></div><div><br></div><=
div>I hope it helps and did not say too many outdated/wrong explanations=
 :).<br></div><div><br></div><div>Btw, Welting rewrote that part with Dm=
itry back then to fix long<br></div><div>standing issues (and drop TSRM_=
LS/DC/CC uses), not sure if he is still<br></div><div>around but he may =
help.<br></div><div><br></div><div>Best,<br></div><div>--<br></div><div>=
Pierre<br></div><div><br></div><div>@pierrejoye |&nbsp;<a href=3D"http:/=
/www.libgd.org">http://www.libgd.org</a><br></div><div><br></div></block=
quote><div><br></div><div>Hey Pierre,<br></div><div><br></div><div>Thank=
 you! This helps tremendously.<br></div><div><br></div><blockquote type=3D=
"cite" id=3D"qt" style=3D""><div>A good start (surely already looked) ar=
e apache and litespeed (or<br></div><div>embed) for php8, however beside=
s the API names change, the behavior or<br></div><div>goals of each ts_ =
or tsrm_ functions remain the same as what we had<br></div><div>with 5.x=
. One with a very similar modus operandi than Franken was<br></div><div>=
ISAPI for IIS. Mind the names changes, dapt to php8's new names (and<br>=
</div><div>some behaviors) but the basics and core flows have been kept.=
<br></div></blockquote><div><br></div><div>I looked at litespeed but not=
 apache, so I will look to that as well.<br></div><div><br></div><blockq=
uote type=3D"cite" id=3D"qt" style=3D""><div>If you know (~) how many th=
reads franken is going to need, tsrm allows<br></div><div>you to preallo=
cate the rsc slots using tsrm_startup_ex. It uses<br></div><div>tsrm_sta=
rtup, a more powerful version, which lets franken define how<br></div><d=
iv>many rsc per thread will be preallocated. I suppose that could help a=
<br></div><div>bit performance wise:<br></div></blockquote><div><br></di=
v><div>The number of threads are static (at least for now), so we should=
 def do this.<br></div><div><br></div><blockquote type=3D"cite" id=3D"qt=
" style=3D""><div>Long story short, the TSRM API is very flexible, how a=
nd when you<br></div><div>alloc/free rsrc is basically up to you. It is =
possible to keep some<br></div><div>around in the root thread (and be us=
ed in other threads, given the<br></div><div>root thread id is known (ca=
n be 0 or else depending how franken<br></div><div>managed them.<br></di=
v></blockquote><div><br></div><div>Ah, so this is where I was getting co=
nfused. It's unclear what the "id" is and what it is used for. For examp=
le, I see the engine has a couple of ids (compiler/execution) but I'm un=
clear what the sapi should be doing with them. Should these "ids" be dif=
ferent per thread, or is it just an arbitrary key?</div><div><br></div><=
div id=3D"sig121229152">=E2=80=94 Rob<br></div></body></html>
--72efbbd699cf4c159b1043599f20cbab--