Re: finally... BITWIG STUDIO! beta 'soon'...
Posted: Mon Jan 28, 2013 5:09 pm
Thanks for posting!
Ordinarily it would be, but if they're using modern code it will be fine, the fact that they're developing for Linux shows that they're bare metal coders.ian_halsall wrote:sandboxing vsts in their own process sounds slooow
On that note, this exchange is interesting:ian_halsall wrote:sandboxing vsts in their own process sounds slooow

Well it's not unix only, I don't think they're taking that approach, they're not going to use OS specific code... I think they're probably just using modern design patterns, they're competing against Ableton for performance so they've got quite a bit of overhead to use. Maybe it won't be as fast as say Cubase or Sonar but it's not hard to be faster than Ableton, Live has an antiquated codebase, it wasn't fast to begin with but now it's very old and you can tell by how hard it was convert to 64 bit... I think Bitwig will perform very well based on what I've read.ian_halsall wrote:yeah - I forgot about that - I suppose if it's unix only then they can use shared memory, pipes, signals, semaphores etc.
but if they do that then it's difficult to stop the processes crashing each other
sorry this doesn't sound right to me, surely these would be loaded into memory? I don't think you'd be piping via a disk.crumhorn wrote:generally speaking it involves piping data via a file on disk. getting data across a process boundary is an overhead no matter what design patterns you apply, but compared to all the other processing a vst has to do maybe it's not such a great CPU overhead. I'd be more concerned about how small the buffer could be before disk latency could become an issue - causing drop outs.
That's what I thought, DFD samplers stream from disk.agent314 wrote:Yeah, I wouldn't expect disk latency to be an issue during realtime VST performance, since most of that is calculation/RAM-based (with the exception of some samplers, as I understand it)
But then again, my knowledge of hardware/software interactions is minimal at best
I used cubase for years - Live is faster.sdfak1234 wrote:Well it's not unix only, I don't think they're taking that approach, they're not going to use OS specific code... I think they're probably just using modern design patterns, they're competing against Ableton for performance so they've got quite a bit of overhead to use. Maybe it won't be as fast as say Cubase or Sonar but it's not hard to be faster than Ableton, Live has an antiquated codebase, it wasn't fast to begin with but now it's very old and you can tell by how hard it was convert to 64 bit... I think Bitwig will perform very well based on what I've read.ian_halsall wrote:yeah - I forgot about that - I suppose if it's unix only then they can use shared memory, pipes, signals, semaphores etc.
but if they do that then it's difficult to stop the processes crashing each other
We are talking about system resources? I'm sorry you are completely wrong, I don't know the last version of Cubase you used but Cubase is A LOT more efficient than Live, it wipes the floor with live.ian_halsall wrote:I used cubase for years - Live is faster.sdfak1234 wrote:Well it's not unix only, I don't think they're taking that approach, they're not going to use OS specific code... I think they're probably just using modern design patterns, they're competing against Ableton for performance so they've got quite a bit of overhead to use. Maybe it won't be as fast as say Cubase or Sonar but it's not hard to be faster than Ableton, Live has an antiquated codebase, it wasn't fast to begin with but now it's very old and you can tell by how hard it was convert to 64 bit... I think Bitwig will perform very well based on what I've read.ian_halsall wrote:yeah - I forgot about that - I suppose if it's unix only then they can use shared memory, pipes, signals, semaphores etc.
but if they do that then it's difficult to stop the processes crashing each other
pipes can be implemented however you like - on Unix they are not necessarily files although the os may choose to put them to disk
you can open a pipe from c or c++ easily on unix - shared memory is preferred although this is more difficult
lowshelf wrote:On that note, this exchange is interesting:ian_halsall wrote:sandboxing vsts in their own process sounds slooow
