ODE's multithreading worse than one thread, am I missing something?

[Vaillancourt]

Hello!

I'm currently working on a project that implies (among other things) multiple "complex" vehicles (vehicles with multiple joints, think forklift). 

I'm trying to figure of a way to speed things up a bit in the "dWorldQuickStep" function; I've seen in the demos that there are ways to multithread it (via dWorldSetStepThreadingImplementation), however, if I just try to set up the multithreading like it's done in the demos, using calls to

    dThreadingImplementationID threading = dThreadingAllocateMultiThreadedImplementation();
    dThreadingThreadPoolID pool = dThreadingAllocateThreadPool(4, 0, dAllocateFlagBasicData, NULL);
    dThreadingThreadPoolServeMultiThreadedImplementation(pool, threading);
    dWorldSetStepThreadingImplementation(world, dThreadingImplementationGetFunctions(threading), threading);

the simulation speed is way worse.

Am I missing something with what is implemented in ODE? I'm using ODE 0.16.2, on Windows.

Thanks!

[Oleh Derevenko]

Hi,

I could suggest just fairly trivial checks:

1) Make sure you have enough CPU cores for the number you are passing in the first parameter of dThreadingAllocateThreadPool();

2) Make sure you have enough RAM available so that the application does not start swapping out and in;

3) Make sure you enable optimization for speed with "Inline any suitable" when you are building ODE.

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Vaillancourt <alexandre.vai...@gmail.com>
Sent: Wednesday, December 18, 2024 1:13
To: ode-users <ode-...@googlegroups.com>
Subject: [ode-users] ODE's multithreading worse than one thread, am I missing something?

 

--
You received this message because you are subscribed to the Google Groups "ode-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ode-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ode-users/486bb6ff-029d-41b8-be74-497f0bf4d506n%40googlegroups.com.

---

This e-mail may contain privileged and confidential information. If you are not the intended recipient, be aware that any use, disclosure, copying or distribution of this e-mail or any attachments is prohibited. If you have received this e-mail in error, please notify us immediately by returning it to the sender and delete this copy from your system. Thank you.

[Vaillancourt]

Hello, thanks for the pointers Oleh,

1) I have 8 cores with this machine, tested with 1, 4 and 8 threads; the more threads I add, the slower the simulation is.

2) I'm running the app in 32-bit, I have 32gb of RAM and the app doesn't reach the max limit of ~4gb (linking with /LARGEADDRESSAWARE) (stack size: 15 mb)

3) I checked that and unfortunately, the difference between "inline any suitable" and "inline only __inline" is negligible. 

Setting `dThreadingAllocateThreadPool(4, 0, dAllocateFlagBasicData, NULL);` then `dWorldSetStepIslandsProcessingMaxThreadCount(world, 1);` makes it so as if I did not set multi threading at all.

For the sake of testing this in a minimum environment, I modified the "boxstack" demo:

1) instead of the boxes piling all at the same location, now it creates 10 stacks (i.e. in the hope of creating 10 "islands")

2) bodies don't auto-disable/enable, they are always enabled (because in my situations, the vehicles always move, they are never disabled)

3) max number of boxes increased to 5000 (to actually have nice stacks)

4) multi-threading starts disabled, enable it with 'm' and disable it with 'n'

5) a stop watch surrounds dWorldQuickStep, this is accompanied by a counter; every 1 whole second spent in this function, the frame count is printed and the timer/frame counter are reset, so we have the frame rate just for the stepping function call.

On my machine after spawning 410 objects, I get ~111 fps with multi-thread, while I get ~630 fps without multithread. I have attached the modified demo if anyone is interested in looking/testing. I'm not confident that I know enough about multi-threading and/or the internals of ODE to be able to ever bring a fix to this, so I'll think about other ways to improve our use of ODE for this specific scenario.

Thanks!

Alexandre

[Oleh Derevenko]

Alrigt. I'll have this in my backlog after the box-plane collider improvements and inelastic collision addition proposed earlier.

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Vaillancourt <alexandre.vai...@gmail.com>

Sent: Wednesday, December 18, 2024 19:28
To: ode-users <ode-...@googlegroups.com>
Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

 

To view this discussion visit https://groups.google.com/d/msgid/ode-users/7430f990-51ab-487c-8794-eff47685b1a3n%40googlegroups.com.

[Oleh Derevenko]

It's funny though, that after 12 years of existence somebody has made a performance test for it, finally.  🙂

Oleh Derevenko

-- Skype with underscore

 

---

From: 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>
Sent: Wednesday, December 18, 2024 23:40
To: ode-...@googlegroups.com <ode-...@googlegroups.com>

To view this discussion visit https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899F9622CC08F729775924DDD052%40AS4PR07MB8899.eurprd07.prod.outlook.com.

[Vaillancourt]

Thanks Oleh, no worries.

I'll check the repo and the list here to see if/when there is an update.

Do you want me to add an issue to the bitbucket repository?

Alexandre Vaillancourt

[Oleh Derevenko]

Not necessary. It's already a big help you made a test case with time measurement.

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Vaillancourt <alexandre.vai...@gmail.com>

Sent: Thursday, December 19, 2024 15:28

To: ode-users <ode-...@googlegroups.com>
Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

Thanks Oleh, no worries.

I'll check the repo and the list here to see if/when there is an update.

Do you want me to add an issue to the bitbucket repository?

[Vaillancourt]

All right!

[Oleh Derevenko]

Merry Christmas!

So, I made some improvements to the threading implementation and made some tests with the provided modified demo.

I was dropping 1000 boxes and trying 4 threads.

By default, with threading off, it was showing 90 fps, with threading enabled — 47 fps.

I tried increasing cooperative processing blocks for preparatory and result retrieval stages (so that threads would not be writing to the same cache line and would not kick it out of cache for each other). But this did not have any significant effect — I'll, probably, leave the code in place as maybe other architectures could show some difference depending on these settings.

What had effect was limiting iterative solving stage thread limit to a single thread (in this case original simpler routine is selected) — that gave ~150 fps.

And what had even more effect, was limiting stepper thread count per island to 1 thread. In this case, direct call branches are selected as well, instead of posting asynchronous calls to the work queue for the thread pool. So, running islands in parallel with 1 thread per island produced ~260 fps.

It appears that the biggest impact on performance is caused by threads entering the kernel for blocking and signaling other threads. It completely destroys all the benefit from parallel computations. Maybe, it would outperform the single threaded variant on some huge scenes — I don't know.

And the second conclusion is that it looks like the dWorldSetStepIslandsProcessingMaxThreadCount(world, ...) function is implemented wrong. Currently it limits the number of threads that iterate islands (and the same limit is used for threads within each island). That is, the function is limiting the number of threads used per world in general. What would be needed would be an ability to define a separate parameter for number of threads within island so that multiple islands could be stepped in parallel, but a single thread could be assigned per an island. I guess we'll need to add a new function for this.

But generally, all the multi-threaded code, other than running separate islands in parallel, seems to be useless, at least on Windows.

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Vaillancourt <alexandre.vai...@gmail.com>

Sent: Friday, December 20, 2024 17:39

To: ode-users <ode-...@googlegroups.com>
Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

[Oleh Derevenko]

Alright.

I've pushed my changes into master.
So far, I've made it that islands are stepped in multiple threads but, by default,  LCP is solved in one thread. This way the performance is the best.

There are new functions dWorldSetSteppingThreadingParameters and dWorldGetSteppingThreadingParameters for accessing new parameters that have been added. 

Oleh Derevenko

-- Skype with underscore

 

---

From: 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>
Sent: Wednesday, December 25, 2024 21:20
To: ode-...@googlegroups.com <ode-...@googlegroups.com>

--
You received this message because you are subscribed to the Google Groups "ode-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ode-users+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/ode-users/AS4PR07MB889998CD0E66346CD884A8EBDD0C2%40AS4PR07MB8899.eurprd07.prod.outlook.com.

[Oleh Derevenko]

I would be grateful if anyone running UNIX systems could take ODE-Oder branch and test demo_boxstack dropping boxes (e.g. drop 1000 of them) and switching to multi-threaded solver with 'm' key. 

I need to see if it will not crash at least.

Oleh Derevenko

-- Skype with underscore

 

---

From: 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>

Sent: Sunday, January 5, 2025 1:02

To view this discussion visit https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899AE5AFFB63196716D7C84DD162%40AS4PR07MB8899.eurprd07.prod.outlook.com.

[Leopold Palomo-Avellaneda]

Oleh,


El 5/1/25 a les 0:32, 'Oleh Derevenko' via ode-users ha escrit:

> I would be grateful if anyone running UNIX systems could take ODE-Oder
> branch and test demo_boxstack dropping boxes (e.g. drop 1000 of them)
> and switching to multi-threaded solver with 'm' key.
> I need to see if it will not crash at least.

cloned the repo and build in a Debian 12 amd64 arch. Build with cmake,
Release and using libccd from system (double precision).

Tested demo_boxstack, dropping

dWorldQuickStep frames in 1 sec (fps): 28, num of objects: 1265

switch to multithread (pressing m)

"Enabled multi-thread."

No crash by now. The scenario is plenty of objects (boxes and capsules)

Is it ok for you?

Leopold

--
--
Linux User 152692 GPG: 05F4A7A949A2D9AA
Catalonia
-------------------------------------
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[Oleh Derevenko]

Hi Leopold, 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Leopold Palomo-Avellaneda <l...@alaxarxa.net>
Sent: Monday, January 6, 2025 2:25

To: ode-...@googlegroups.com <ode-...@googlegroups.com>

Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

 

[You don't often get email from l...@alaxarxa.net. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Oleh,


El 5/1/25 a les 0:32, 'Oleh Derevenko' via ode-users ha escrit:
> I would be grateful if anyone running UNIX systems could take ODE-Oder
> branch and test demo_boxstack dropping boxes (e.g. drop 1000 of them)
> and switching to multi-threaded solver with 'm' key.
> I need to see if it will not crash at least.

cloned the repo and build in a Debian 12 amd64 arch. Build with cmake,
Release and using libccd from system (double precision).

Tested demo_boxstack, dropping

dWorldQuickStep frames in 1 sec (fps): 28, num of objects: 1265

switch to multithread (pressing m)

"Enabled multi-thread."

No crash by now. The scenario is plenty of objects (boxes and capsules)

Is it ok for you?

Thank very much for checking. I tested it would build on Linux but I do not have graphics set up, so running visual tests is problematic for me.

Oleh Derevenko

-- Skype with underscore

---

[Vaillancourt]

Hi Oleh, 

I have tested this with the box stack demo in your branch and with our software and it seems it did improve the stepping phase time, thanks a lot for having looked into this!

Much appreciated!

Alexandre Vaillancourt

[Oleh Derevenko]

Post us some video link at youtube with your game demo when it's ready.

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Vaillancourt <alexandre.vai...@gmail.com>

Sent: Monday, January 6, 2025 23:30

To view this discussion visit https://groups.google.com/d/msgid/ode-users/4844e2ce-62d3-405a-aa0d-edfa6fdbc444n%40googlegroups.com.

[Leopold Palomo-Avellaneda]

Oleh,

the changes that you made are enough to make a new 0.16.6 version?

Leopold

El 6/1/25 a les 10:51, 'Oleh Derevenko' via ode-users ha escrit:
> Hi Leopold,
>
> ------------------------------------------------------------------------
> *From:* ode-...@googlegroups.com <ode-...@googlegroups.com> on

> behalf of Leopold Palomo-Avellaneda <l...@alaxarxa.net>

> *Sent:* Monday, January 6, 2025 2:25
> *To:* ode-...@googlegroups.com <ode-...@googlegroups.com>
> *Subject:* Re: [ode-users] ODE's multithreading worse than one thread,

> am I missing something?
> [You don't often get email from l...@alaxarxa.net. Learn why this is

> important at https://aka.ms/LearnAboutSenderIdentification <https://
> aka.ms/LearnAboutSenderIdentification> ]

>
> Oleh,
>
>
> El 5/1/25 a les 0:32, 'Oleh Derevenko' via ode-users ha escrit:
> > I would be grateful if anyone running UNIX systems could take ODE-Oder
> > branch and test demo_boxstack dropping boxes (e.g. drop 1000 of them)
> > and switching to multi-threaded solver with 'm' key.
> > I need to see if it will not crash at least.
>
> cloned the repo and build in a Debian 12 amd64 arch. Build with cmake,
> Release and using libccd from system (double precision).
>
> Tested demo_boxstack, dropping
>
> dWorldQuickStep frames in 1 sec (fps): 28, num of objects: 1265
>
> switch to multithread (pressing m)
>
> "Enabled multi-thread."
>
> No crash by now. The scenario is plenty of objects (boxes and capsules)
>
> Is it ok for you?
>
>
> Thank very much for checking. I tested it would build on Linux but I do
> not have graphics set up, so running visual tests is problematic for me.
>
> Oleh Derevenko
> -- Skype with underscore
>
>

> ------------------------------------------------------------------------

>
> This e-mail may contain privileged and confidential information. If you
> are not the intended recipient, be aware that any use, disclosure,
> copying or distribution of this e-mail or any attachments is prohibited.
> If you have received this e-mail in error, please notify us immediately
> by returning it to the sender and delete this copy from your system.
> Thank you.
>

> --
> You received this message because you are subscribed to the Google
> Groups "ode-users" group.
> To unsubscribe from this group and stop receiving emails from it, send

> an email to ode-users+...@googlegroups.com <mailto:ode-
> users+un...@googlegroups.com>.

> To view this discussion visit https://groups.google.com/d/msgid/ode-
> users/

> AS4PR07MB8899A43DA449833C989986EEDD102%40AS4PR07MB8899.eurprd07.prod.outlook.com <https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899A43DA449833C989986EEDD102%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer>.

[Oleh Derevenko]

Well, yes. We could publish a new version... Other than nobody tested the POSIX implementation changes well yet. 🙂

Oleh Derevenko

-- Skype with underscore

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Leopold Palomo-Avellaneda <l...@alaxarxa.net>
Sent: Monday, January 13, 2025 20:01
To: ode-...@googlegroups.com <ode-...@googlegroups.com>
Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

 

[You don't often get email from l...@alaxarxa.net. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

To unsubscribe from this group and stop receiving emails from it, send an email to ode-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ode-users/7890d5c6-a840-48ce-8d34-0c8343d784c6%40alaxarxa.net.

---

[Oleh Derevenko]

Like, I remember you tried it. But generally, it's somewhat too soon after substantial changes.

Oleh Derevenko

-- Skype with underscore

 

---

From: 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>

Sent: Monday, January 13, 2025 23:20

To view this discussion visit https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899E24956C9D0EC0E14938ADD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com.

[Oleh Derevenko]

Well, I've checked — there are quite a lot of differences between the development branch and 0.16.x already. As far as I remember, there is some major feature unfinished. But since I have a hard time recalling what exactly I wanted to add, I doubt it will ever be. So, maybe let us rather make a completely new release branch 0.17.x from master? Any objections?

Oleh Derevenko

-- Skype with underscore

 

---

From: 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>

Sent: Monday, January 13, 2025 23:22

To view this discussion visit https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899ACCA6BD925919B488688DD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com.

[Vaillancourt]

I may add a tiny bit of a feature with CMake that would make it so that we would not need to modify it on our end when we move to a new version. As usual, I would make it "opt in" and so would not impact folks that don't need it.

[Vaillancourt]

I've created a pull request about it. It is to allow the installation of the pdbs associated with the (static) .lib when using CMake/Visual Studio. 

[Leopold Palomo-Avellaneda]

El 13/1/25 a les 23:53, 'Oleh Derevenko' via ode-users ha escrit:

> Well, I've checked — there are quite a lot of differences between the
> development branch and 0.16.x already. As far as I remember, there is some major
> feature unfinished. But since I have a hard time recalling what exactly I wanted
> to add, I doubt it will ever be. So, maybe let us rather make a completely new
> release branch 0.17.x from master? Any objections?

Well, it is binary compatible? If it is not, in Debian we need to make a
transition process. We can do it because we are not in freeze period.

Leopold

> Oleh Derevenko
>
> -- Skype with underscore
>
>

> --------------------------------------------------------------------------------
> *From:* 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>
> *Sent:* Monday, January 13, 2025 23:22

> *To:* ode-...@googlegroups.com <ode-...@googlegroups.com>
> *Subject:* Re: [ode-users] ODE's multithreading worse than one thread, am I
> missing something?

> Like, I remember you tried it. But generally, it's somewhat too soon after
> substantial changes.
>
>
> Oleh Derevenko
>
> -- Skype with underscore
>
>

> --------------------------------------------------------------------------------
> *From:* 'Oleh Derevenko' via ode-users <ode-...@googlegroups.com>
> *Sent:* Monday, January 13, 2025 23:20

> *To:* ode-...@googlegroups.com <ode-...@googlegroups.com>
> *Subject:* Re: [ode-users] ODE's multithreading worse than one thread, am I
> missing something?

> Well, yes. We could publish a new version... Other than nobody tested the POSIX
> implementation changes well yet. 🙂
>
> Oleh Derevenko
>
> -- Skype with underscore
>
>

> --------------------------------------------------------------------------------

> *From:* ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of
> Leopold Palomo-Avellaneda <l...@alaxarxa.net>

> *Sent:* Monday, January 13, 2025 20:01

> *To:* ode-...@googlegroups.com <ode-...@googlegroups.com>
> *Subject:* Re: [ode-users] ODE's multithreading worse than one thread, am I
> missing something?
> [You don't often get email from l...@alaxarxa.net. Learn why this is important at
> https://aka.ms/LearnAboutSenderIdentification <https://aka.ms/
> LearnAboutSenderIdentification> ]
>
> Oleh,
>
> the changes that you made are enough to make a new 0.16.6 version?
>
> Leopold
>
> El 6/1/25 a les 10:51, 'Oleh Derevenko' via ode-users ha escrit:
>> Hi Leopold,
>>
>> ------------------------------------------------------------------------
>> *From:* ode-...@googlegroups.com <ode-...@googlegroups.com> on
>> behalf of Leopold Palomo-Avellaneda <l...@alaxarxa.net>
>> *Sent:* Monday, January 6, 2025 2:25
>> *To:* ode-...@googlegroups.com <ode-...@googlegroups.com>
>> *Subject:* Re: [ode-users] ODE's multithreading worse than one thread,
>> am I missing something?
>> [You don't often get email from l...@alaxarxa.net. Learn why this is
>> important at https://aka.ms/LearnAboutSenderIdentification <https://aka.ms/

> LearnAboutSenderIdentification> <https://

>> To view this discussion visit https://groups.google.com/d/msgid/ode- <https://groups.google.com/d/msgid/ode->
>> users/
>> AS4PR07MB8899A43DA449833C989986EEDD102%40AS4PR07MB8899.eurprd07.prod.outlook.com <https://groups.google.com/d/msgid/ode-users/
> AS4PR07MB8899A43DA449833C989986EEDD102%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer <https://groups.google.com/d/msgid/ode-users/AS4PR07MB8899A43DA449833C989986EEDD102%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer>>.

>
>
> --
> --
> Linux User 152692     GPG: 05F4A7A949A2D9AA
> Catalonia
> -------------------------------------
> A: Because it messes up the order in which people normally read text.
> Q: Why is top-posting such a bad thing?
> A: Top-posting.
> Q: What is the most annoying thing in e-mail?
>
> --
> You received this message because you are subscribed to the Google Groups "ode-
> users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to ode-users+...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/ode-

> users/7890d5c6-a840-48ce-8d34-0c8343d784c6%40alaxarxa.net <https://
> groups.google.com/d/msgid/ode-users/7890d5c6-
> a840-48ce-8d34-0c8343d784c6%40alaxarxa.net>.
>
> --------------------------------------------------------------------------------

>
> This e-mail may contain privileged and confidential information. If you are not
> the intended recipient, be aware that any use, disclosure, copying or
> distribution of this e-mail or any attachments is prohibited. If you have
> received this e-mail in error, please notify us immediately by returning it to
> the sender and delete this copy from your system. Thank you.
>
> --
> You received this message because you are subscribed to the Google Groups "ode-
> users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to ode-users+...@googlegroups.com <mailto:ode-
> users+un...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/ode-users/

> AS4PR07MB8899E24956C9D0EC0E14938ADD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com
> <https://groups.google.com/d/msgid/ode-users/
> AS4PR07MB8899E24956C9D0EC0E14938ADD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer>.
>
> --------------------------------------------------------------------------------

>
> This e-mail may contain privileged and confidential information. If you are not
> the intended recipient, be aware that any use, disclosure, copying or
> distribution of this e-mail or any attachments is prohibited. If you have
> received this e-mail in error, please notify us immediately by returning it to
> the sender and delete this copy from your system. Thank you.
>
> --
> You received this message because you are subscribed to the Google Groups "ode-
> users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to ode-users+...@googlegroups.com <mailto:ode-
> users+un...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/ode-users/

> AS4PR07MB8899ACCA6BD925919B488688DD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com
> <https://groups.google.com/d/msgid/ode-users/
> AS4PR07MB8899ACCA6BD925919B488688DD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer>.
>
> --------------------------------------------------------------------------------

>
> This e-mail may contain privileged and confidential information. If you are not
> the intended recipient, be aware that any use, disclosure, copying or
> distribution of this e-mail or any attachments is prohibited. If you have
> received this e-mail in error, please notify us immediately by returning it to
> the sender and delete this copy from your system. Thank you.
>
> --
> You received this message because you are subscribed to the Google Groups "ode-
> users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to ode-users+...@googlegroups.com <mailto:ode-
> users+un...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/ode-users/

> AS4PR07MB8899F4C6AA4197593A6B204EDD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com
> <https://groups.google.com/d/msgid/ode-users/
> AS4PR07MB8899F4C6AA4197593A6B204EDD1F2%40AS4PR07MB8899.eurprd07.prod.outlook.com?utm_medium=email&utm_source=footer>.

[Oleh Derevenko]

 

---

From: ode-...@googlegroups.com <ode-...@googlegroups.com> on behalf of Leopold Palomo-Avellaneda <l...@alaxarxa.net>
Sent: Thursday, January 16, 2025 9:48
To: ode-...@googlegroups.com <ode-...@googlegroups.com>
Subject: Re: [ode-users] ODE's multithreading worse than one thread, am I missing something?

 

[You don't often get email from l...@alaxarxa.net. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

El 13/1/25 a les 23:53, 'Oleh Derevenko' via ode-users ha escrit:
> Well, I've checked — there are quite a lot of differences between the
> development branch and 0.16.x already. As far as I remember, there is some major
> feature unfinished. But since I have a hard time recalling what exactly I wanted
> to add, I doubt it will ever be. So, maybe let us rather make a completely new
> release branch 0.17.x from master? Any objections?

Well, it is binary compatible? If it is not, in Debian we need to make a
transition process. We can do it because we are not in freeze period.

Yes, there were method additions only (if not to consider renaming of some internal static inline helpers which were exposed in public headers). So, nominally, it's a minor version increase. But if to start 0.17.x branch I would increment the major version speculatively to keep minor version increase possibility if anyone needed another release from 0.16.x.
But I'm hesitating yet. There are some obvious small bugs in demos that would be good to clean up before a major release. And it would be good to incorporate this multi-threaded boxstack demo. 

Oleh Derevenko
-- Skype with underscore