[Csnd] Question on table sizes

Date	2013-07-12 16:03
From	lppier
Subject	[Csnd] Question on table sizes
	Hi, I am trying the 2 sine waves below that are used by my project's oscili's and tablei opcode's, I think I hear a difference. The one with the larger table size sounds less harsh to me at higher frequencies. f1 0 16384 10 1 ;sine f1 0 65536 10 1 ;sine I've also read somewhere that with larger table sizes, one can use oscil instead of oscili now, without the interpolation to save on cpu. Is this true? Thanks. -- View this message in context: http://csound.1045644.n5.nabble.com/Question-on-table-sizes-tp5725465.html Sent from the Csound - General mailing list archive at Nabble.com.

Date	2013-07-12 16:21
From	Michael Gogins
Subject	Re: [Csnd] Question on table sizes
	Yes, larger function tables are inherently less noisy. There are mathematical reasons for this. I am not surprised you can hear the difference, so can I. I always use 65536. Maybe I should use even bigger ones. If you have anything like a recent computer, you don't need to worry about the amount of memory. The second point also is true, more or less, but I would just go ahead and use the interpolation anyway. I haven't done any speed tests but I'd be surprised if this made much of a difference. If you do some speed tests, let us know what you find. Hope this helps, Mike =========================== Michael Gogins Irreducible Productions http://michaelgogins.tumblr.com Michael dot Gogins at gmail dot com On Fri, Jul 12, 2013 at 11:03 AM, lppier <madstrum@gmail.com> wrote: Hi, I am trying the 2 sine waves below that are used by my project's oscili's and tablei opcode's, I think I hear a difference. The one with the larger table size sounds less harsh to me at higher frequencies. f1 0 16384 10 1 ;sine f1 0 65536 10 1 ;sine I've also read somewhere that with larger table sizes, one can use oscil instead of oscili now, without the interpolation to save on cpu. Is this true? Thanks. -- View this message in context: http://csound.1045644.n5.nabble.com/Question-on-table-sizes-tp5725465.html Sent from the Csound - General mailing list archive at Nabble.com. Send bugs reports to the Sourceforge bug tracker https://sourceforge.net/tracker/?group_id=81968&atid=564599 Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

Date	2013-07-12 16:35
From	Justin Smith
Subject	Re: [Csnd] Question on table sizes
	An important tradeoff, on the architecture level, is the usage of cache vs. CPU cycles. Moving data from RAM into cache can actually be more expensive than interpolating between samples (and it gets more complicated when you consider that the level of influence this has varies based on how much your other algorithms also want to populate the cache). The only way to be sure is to benchmark on your specific architecture with different tables sizes with and without interpolation, and for best results you would want some representative usage of other opcodes. On Fri, Jul 12, 2013 at 8:21 AM, Michael Gogins <michael.gogins@gmail.com> wrote: Yes, larger function tables are inherently less noisy. There are mathematical reasons for this. I am not surprised you can hear the difference, so can I. I always use 65536. Maybe I should use even bigger ones. If you have anything like a recent computer, you don't need to worry about the amount of memory. The second point also is true, more or less, but I would just go ahead and use the interpolation anyway. I haven't done any speed tests but I'd be surprised if this made much of a difference. If you do some speed tests, let us know what you find. Hope this helps, Mike =========================== Michael Gogins Irreducible Productions http://michaelgogins.tumblr.com Michael dot Gogins at gmail dot com On Fri, Jul 12, 2013 at 11:03 AM, lppier <madstrum@gmail.com> wrote: Hi, I am trying the 2 sine waves below that are used by my project's oscili's and tablei opcode's, I think I hear a difference. The one with the larger table size sounds less harsh to me at higher frequencies. f1 0 16384 10 1 ;sine f1 0 65536 10 1 ;sine I've also read somewhere that with larger table sizes, one can use oscil instead of oscili now, without the interpolation to save on cpu. Is this true? Thanks. -- View this message in context: http://csound.1045644.n5.nabble.com/Question-on-table-sizes-tp5725465.html Sent from the Csound - General mailing list archive at Nabble.com. Send bugs reports to the Sourceforge bug tracker https://sourceforge.net/tracker/?group_id=81968&atid=564599 Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"

Date	2013-07-12 16:37
From	Michael Gogins
Subject	Re: [Csnd] Question on table sizes
	True. On Jul 12, 2013 11:36 AM, "Justin Smith" <noisesmith@gmail.com> wrote: An important tradeoff, on the architecture level, is the usage of cache vs. CPU cycles. Moving data from RAM into cache can actually be more expensive than interpolating between samples (and it gets more complicated when you consider that the level of influence this has varies based on how much your other algorithms also want to populate the cache). The only way to be sure is to benchmark on your specific architecture with different tables sizes with and without interpolation, and for best results you would want some representative usage of other opcodes. On Fri, Jul 12, 2013 at 8:21 AM, Michael Gogins <michael.gogins@gmail.com> wrote: Yes, larger function tables are inherently less noisy. There are mathematical reasons for this. I am not surprised you can hear the difference, so can I. I always use 65536. Maybe I should use even bigger ones. If you have anything like a recent computer, you don't need to worry about the amount of memory. The second point also is true, more or less, but I would just go ahead and use the interpolation anyway. I haven't done any speed tests but I'd be surprised if this made much of a difference. If you do some speed tests, let us know what you find. Hope this helps, Mike =========================== Michael Gogins Irreducible Productions http://michaelgogins.tumblr.com Michael dot Gogins at gmail dot com On Fri, Jul 12, 2013 at 11:03 AM, lppier <madstrum@gmail.com> wrote: Hi, I am trying the 2 sine waves below that are used by my project's oscili's and tablei opcode's, I think I hear a difference. The one with the larger table size sounds less harsh to me at higher frequencies. f1 0 16384 10 1 ;sine f1 0 65536 10 1 ;sine I've also read somewhere that with larger table sizes, one can use oscil instead of oscili now, without the interpolation to save on cpu. Is this true? Thanks. -- View this message in context: http://csound.1045644.n5.nabble.com/Question-on-table-sizes-tp5725465.html Sent from the Csound - General mailing list archive at Nabble.com. Send bugs reports to the Sourceforge bug tracker https://sourceforge.net/tracker/?group_id=81968&atid=564599 Discussions of bugs and features can be posted here To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"