 |
|
|
|
|
|
| |
| |
|
|
From: yesbird
Subject: Photon mapping calculation limited to 8 threads
Date: 6 May 2024 07:54:42
Message: <6638c502@news.povray.org>
|
|
|
| |
| |
|
|
Hi, all !
While rendering large number of images on server with photon mapping
techniques I discovered, that only 8 kernels are used in time of
photons calculation step, while on tracing step all 40 kernels are
loaded.
Is there a way to use more threads for photon calculation to improve
performance ?
Thanks in advance.
--
YB.
Post a reply to this message
|
|
| |
| |
|
|
From: Alain Martel
Subject: Re: Photon mapping calculation limited to 8 threads
Date: 6 May 2024 11:05:42
Message: <6638f1c6$1@news.povray.org>
|
|
|
| |
| |
|
|
Le 2024-05-06 à 07:54, yesbird a écrit :
> Hi, all !
>
> While rendering large number of images on server with photon mapping
> techniques I discovered, that only 8 kernels are used in time of
> photons calculation step, while on tracing step all 40 kernels are
> loaded.
>
> Is there a way to use more threads for photon calculation to improve
> performance ?
>
> Thanks in advance.
> --
> YB.
>
>
>
>
>
So, you have one light and 8 target objects, or 2 lights and 4 target
objects, or 4 lights and 2 target objects or 8 lights and 1 target object.
The maximum number of threads used while photon mapping is the product
of the number of light_source by the number of objects with a target block.
To the best of my knowledge : Currently, it is not possible to divide
that process between several threads. The algorithm used is strictly
linear and can't get split between multiple threads.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 06/05/2024 18:05, Alain Martel wrote:
> To the best of my knowledge : Currently, it is not possible to divide
> that process between several threads. The algorithm used is strictly
> linear and can't get split between multiple threads.
Thanks, Alan, I will take it into account.
--
YB
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Alain Martel <kua### [at] videotron ca> wrote:
> So, you have one light and 8 target objects, or 2 lights and 4 target
> objects, or 4 lights and 2 target objects or 8 lights and 1 target object.
>
> The maximum number of threads used while photon mapping is the product
> of the number of light_source by the number of objects with a target block.
> To the best of my knowledge : Currently, it is not possible to divide
> that process between several threads. The algorithm used is strictly
> linear and can't get split between multiple threads.
Makes sense, thanks for clarifying.
However, POV-Ray 3.7 seems to log this incorrectly. With 16 CPU cores, 1 light
source and 3 target objects, I can clearly see 300% CPU usage during photon
mapping (so 3 cores are used), however, 19 threads instead of 3 threads are
logged.
Render Time:
Photon Time: 0 hours 2 minutes 53 seconds (173.948 seconds)
using 19 thread(s) with 456.834 CPU-seconds total
Radiosity Time: No radiosity
Trace Time: 0 hours 11 minutes 12 seconds (672.855 seconds)
using 16 thread(s) with 10100.757 CPU-seconds total
Regards,
Thomas.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
