THE POST BELOW IS MORE THAN 5 YEARS OLD. RELATED SUPPORT INFORMATION MIGHT BE OUTDATED OR DEPRECATED
On 15/02/2007 at 11:30, xxxxxxxx wrote:
Cinema 4D Version: 9.6
Platform: Mac OSX ;
Language(s) : C++ ;
I have a performance question. My code runs on a Dual core PC with 39fps. Now, when I run the exact same code on a Dual core IntelMAC (really really fast machine), it runs slow as hell! not even 3fps!!!
What is the problem with XCode? Or is it me not being able to set up this demon correctly? Anybody an idea what the problem may be?
One thing is for sure, I officially hate XCode (version is irrelevant). >8(
thanks for any advice
On 15/02/2007 at 13:48, xxxxxxxx wrote:
well I do not like XCode too.
I think you problem is with XCode optimization settings,
it was one time for me that this settings were reseted to minimal with no reason.
But probably you have tested this already...
On 16/02/2007 at 02:28, xxxxxxxx wrote:
Hmm, I have set the optimisation settings to smalles/fastest. Is that the right setting? However, it may be possible that this wasn´t activated for the last version. I will definetly do a recompile with this setting and see what happens.
On 16/02/2007 at 05:46, xxxxxxxx wrote:
smalles sounds strange for me.
For fastest optimization the code can not be small becouse of all inlined functions and other stuff that need more memory.
On 16/02/2007 at 05:50, xxxxxxxx wrote:
But the documentation says, it´s just like "fastest" however trying to create as small as possible output. :-? I don´t know, it all seems strange in this compiler. Anyway, I will also try a compile with "fastest" only and see if this makes any difference.
I don´t think they mean smallest as in very small code, but I think simply smaller than the fastest...hopefully
On 16/02/2007 at 06:19, xxxxxxxx wrote:
Fastest: ...function inlining and register renaming...
Fastest, smallest: ...enables all 'Faster' optimizations that do not typically increase code size.
Well "Fastest, smallest" seems to be wrong, it does not use inlined functions so the code will be slow.
Just use "Fastest", the size of binary is not as important I think.
What is about "Enable SSE3 extensions" switch, can it be used???
On 16/02/2007 at 06:26, xxxxxxxx wrote:
Ah!! Cool, maybe that is the problem. I clearly have misunderstood the explanation (I should read more carefully *cough* ). Great thanks for pointing this out Remo!
I don´t know if it can be used without problems? The processors from the last years should all support SSE so it shouldn´t do anything wrong there but I am not quite in that topic to know for sure.
On 17/02/2007 at 08:31, xxxxxxxx wrote:
Normally you have to turn on SSE in a seperate option. I think only the Intel compiler can automatically vectorize code by using SIMD instructions ( and even this requires some manual tweaking due to alignment and data dependency issues ). So you probably have to use intrinsics, which gcc on the mac should support
On 19/02/2007 at 13:56, xxxxxxxx wrote:
Hmm, I have now set it to Fastest but it´s still as slow as before. It´s unbelievable, does anybody have an idea what the problem could be?
What are the most appropriate settings for a fast compile? I think the default SDK project should at least have this set, so one can see how to set up.
Anybody from the support an idea?
On 20/02/2007 at 07:53, xxxxxxxx wrote:
Where do you have set it to Fastest?
In "Target" -> "ProjectName" ?
This is strange but i I double click on "ProjectName", the first in the list then I will get another settings.
But probably this is not the problem.
What ca be else?
Well about SSE3 support: It is only optional and may not supported by all Intel Mac.
This is also only necessary if one want to use SSE3 intrinsics.
On 20/02/2007 at 08:08, xxxxxxxx wrote:
thanks for answer and yes, I tried setting it in "Target" settings and also in "project name" settings. I also get different settings for both! (the SDK also has different settings).
When I change the settings in the project name dialog, then these settings are transferred to the target settings too! But the other way round, so when I set in the Target settings something, then the project settings don´t get changed.
So which settings are the important ones? And what effect does each have on the built product? Why are there 2 settings at all?
This is absolutely bullsh*t. I hate this compiler. I already had probs with Codewarrior but this fu**** up compiler does never do what I expect. Argh, I am really raging here because it destroys anything like "development dateline" and although MAXON introduced the XCode compiler since 9.6 there is absolutely no information on anything but the switch, which makes working professionally a PITA. Not really what I call supporting a compiler. I wished there was more info on using XCode for C4D development.
And pity with the SSE3 support! Thanks for the info!
On 20/02/2007 at 08:17, xxxxxxxx wrote:
I think "Target" -> "ProjectName" is the important settings, not sure about another one.
The SSE and SSE2 support are enabled per default with XCode
for Intel compile and in most cases SSE and SSE2 intrinsics are enough.
On 20/02/2007 at 08:21, xxxxxxxx wrote:
On 20/02/2007 at 08:37, xxxxxxxx wrote:
Do you have disabled "Fix & Continue" and other debug stuff ?
On 20/02/2007 at 08:53, xxxxxxxx wrote:
Hmm, where do I find the "Fix & Continue" option? I have disabled everything that could possibly add any debug information. What is that Fix & Continue?
On 20/02/2007 at 08:58, xxxxxxxx wrote:
ah I have found it. It´s turned off.
On 20/02/2007 at 10:32, xxxxxxxx wrote:
I now tried every possible option combinations, every optimisation combinations, with same settings as C4D SDK project, custom settings etc..
There is absolutely no speed up noticable.
On 22/02/2007 at 14:31, xxxxxxxx wrote:
Support? I need this resolved. MAXON forces the devs to use this compiler and says it supports only this one, so I should be able to expect some support for it.
Thank you in advance
On 03/03/2007 at 09:53, xxxxxxxx wrote:
Push...and I´ll push until I get a supportive answer.
On 03/03/2007 at 12:19, xxxxxxxx wrote:
I'm with Katachi on this one. I'll be back here soon to have an indepth discussion with dev support and possibly the developers about the continuing random crashes in my plugins - it has been noted on CGTalk that once all plugins are removed (not just mine), stability is good - with plugins bad.
This certainly isn't the fault of plugin developers, is linked pretty heavily to R10 and XCode (but not exclusively XCode), and is not only frustrating but infuriating (how do you debug random crashes that have no direct connection to your plugin - i.e.: the crashes do not even mention your plugin but you know there is a connection?). I haven't even figured out how to do a debug with XCode for C4D plugins - as Katachi will agree.
To sort of steer my post back to the topic, I have noticed the same types of incredibly slow behavioural response on the Intel Mac running R10 UB. AM sliders that show no sign of sluggishness whatsoever in R9.6 UB (!) constantly raise the 'beach ball' in R10 UB. This can't be my code (!!), now can it (hint - it's the same code for both builds).
I apologize for the long rant and diversionary tactics...