|
 |
On 2017-10-14 16:37:26 +0000, pkoning said:
> Yvo Smellenbergh <E-mail: yvo(DOT)s(AT)gmx.net> wrote:
>>
>> ...
>> I see a completely different result, 3.8.0 is 5 to 6 times *slower* than 3.7.0.
>>
>> Benchmark.pov scene with exactly the same settings:
>> 3.8.0 >> 583,35 seconds
>> 3.7.0 >> 114,76 seconds
>
> Yes, I see that too. On my machine, 3.7.0 (the older version, I haven't
> installed the correct one yet) took 200.4 seconds trace time, while 3.8.0 took
> 1118.9 seconds trace time.
>
> The model I used for those earlier results is one with many simple objects
> (about 19k of them), all but about 50 are opaque.
A new version (r2) is now available:
http://megapov.inetart.net/povrayunofficial_mac/alphapov.html
Speed is back to normal...
--
Yvo
--
-------------------------------------------------------------------------------------------
POV-Ray 3.7 unofficial: http://megapov.inetart.net/povrayunofficial_mac/
UberPOV Mac: http://megapov.inetart.net/uberpov_mac/index.html#Mac
MegaPOV: http://megapov.inetart.net
E-mail: yvo(DOT)s(AT)gmx.net
Post a reply to this message
Attachments:
Download 'iso-8859-1' (4 KB)
|
|
|
|
Yvo Smellenbergh <E-mail: yvo(DOT)s(AT)gmx.net> wrote:
> ...
> The slowdown is related to the new compiler version (Xcode 9), somehow
> the compiler isn't optimizing the code for this project.
That would certainly be a problem, but I think there's more going on. Some of
my models are a lot slower with 3.8.0, and I just noticed a big difference. It
also appears in the benchmark model
3.7.0 (the "r9" Mac build) shows this:
Render Statistics
Image Resolution 512 x 512
----------------------------------------------------------------------------
Pixels: 294912 Samples: 776043 Smpls/Pxl: 2.63
Rays: 2454702 Saved: 10620 Max Level: 12/12
3.8.0 has this:
Render Statistics
Image Resolution 512 x 512
----------------------------------------------------------------------------
Pixels: 294912 Samples: 779220 Smpls/Pxl: 2.64
Rays: 8101498 Saved: 66232 Max Level: 12/12
The samples/pixel value is slightly different -- I wonder why. But the Rays
value is VERY different, much larger for 3.8.0. I see that also in my models
where 3.8.0 is slower, though the factor isn't quite as large. The details
below that summary (the intersection numbers) are correspondingly larger for
3.8.0 also.
Tracing 4x the rays would certainly slow things down even without "help" from
the compiler.
Post a reply to this message
|
|
