[Cs-dev] some random observations on multicore csound OSX

[Cs-dev] some random observations on multicore csound OSX

Date	2011-10-13 00:22
From	Victor Lazzarini
Subject	[Cs-dev] some random observations on multicore csound OSX
	Just testing multicore on OSX 10.5, Intel dual core. 1. xanadu.csd and hommage.csd : there is no major change in timings for single-thread and -j 2, -j 4 , -j 8 (original ksmps and ksmps=441) 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = 100, it's pretty much the same So, in general no gains, but no major losses too even at the original ksmps, which is an improvement. Interestingly even if I shoot up the number of threads to 128, I do not get a loss in performance (possibly lots of threads not doing anything...) I noticed also that with -j 4 I get a better distribution of work between the two processors (at least it's my impression from looking at the graphic CPU monitor) than -j 2. At the end of the score I am also getting this: Csound tidy up: Floating point exception which is a bit worrying. ============== Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:26
From	Steven Yi
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	Just tried with my local build for OSX 10.7, I get segfaults if I use anything besides -j1 with --new-parser. I get the floating point exception if I use -j2 but if I don't turn on the new parser: Dag2 Cache Size: 0 Csound tidy up: Floating point exception I assume then that there may be multicore clean stuff that happens if -j is used that shouldn't be happening if the old parser is on. Victor, did you use --new-parser? On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini wrote: > Just testing multicore on OSX 10.5, Intel dual core. > > 1. xanadu.csd and hommage.csd : there is no major change in timings > for single-thread and -j 2, -j 4 , -j 8 (original ksmps and ksmps=441) > 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = > 100, it's pretty much the same > > So, in general no gains, but no major losses too even at the original > ksmps, which is an improvement. Interestingly even if I shoot up > the number of threads to 128, I do not get a loss in performance > (possibly lots of threads not doing anything...) > > I noticed also that with -j 4 I get a better distribution of work > between the two processors (at least it's my impression from looking at > the graphic CPU monitor) than -j 2. > > At the end of the score I am also getting this: > > Csound tidy up: Floating point exception > > which is a bit worrying. > > ============== > Dr Victor Lazzarini > Senior Lecturer > Dept. of Music > NUI Maynooth Ireland > tel.: +353 1 708 3545 > Victor dot Lazzarini AT nuim dot ie > > > > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel > ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:33
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	yes, I used --new-parser On 13 Oct 2011, at 00:26, Steven Yi wrote: > Just tried with my local build for OSX 10.7, I get segfaults if I use > anything besides -j1 with --new-parser. I get the floating point > exception if I use -j2 but if I don't turn on the new parser: > > Dag2 Cache Size: 0 > Csound tidy up: Floating point exception > > > I assume then that there may be multicore clean stuff that happens if > -j is used that shouldn't be happening if the old parser is on. > > Victor, did you use --new-parser? > > On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini > wrote: >> Just testing multicore on OSX 10.5, Intel dual core. >> >> 1. xanadu.csd and hommage.csd : there is no major change in timings >> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >> ksmps=441) >> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = >> 100, it's pretty much the same >> >> So, in general no gains, but no major losses too even at the original >> ksmps, which is an improvement. Interestingly even if I shoot up >> the number of threads to 128, I do not get a loss in performance >> (possibly lots of threads not doing anything...) >> >> I noticed also that with -j 4 I get a better distribution of work >> between the two processors (at least it's my impression from >> looking at >> the graphic CPU monitor) than -j 2. >> >> At the end of the score I am also getting this: >> >> Csound tidy up: Floating point exception >> >> which is a bit worrying. >> >> ============== >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel >> > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:38
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	actually something must be going wrong. The activity monitor says csound is always using 1 thread only. So I am not using the multicore code. But I am using the new parser alright. Victor On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: > yes, I used --new-parser > > On 13 Oct 2011, at 00:26, Steven Yi wrote: > >> Just tried with my local build for OSX 10.7, I get segfaults if I use >> anything besides -j1 with --new-parser. I get the floating point >> exception if I use -j2 but if I don't turn on the new parser: >> >> Dag2 Cache Size: 0 >> Csound tidy up: Floating point exception >> >> >> I assume then that there may be multicore clean stuff that happens if >> -j is used that shouldn't be happening if the old parser is on. >> >> Victor, did you use --new-parser? >> >> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >> wrote: >>> Just testing multicore on OSX 10.5, Intel dual core. >>> >>> 1. xanadu.csd and hommage.csd : there is no major change in timings >>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>> ksmps=441) >>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = >>> 100, it's pretty much the same >>> >>> So, in general no gains, but no major losses too even at the >>> original >>> ksmps, which is an improvement. Interestingly even if I shoot up >>> the number of threads to 128, I do not get a loss in performance >>> (possibly lots of threads not doing anything...) >>> >>> I noticed also that with -j 4 I get a better distribution of work >>> between the two processors (at least it's my impression from >>> looking at >>> the graphic CPU monitor) than -j 2. >>> >>> At the end of the score I am also getting this: >>> >>> Csound tidy up: Floating point exception >>> >>> which is a bit worrying. >>> >>> ============== >>> Dr Victor Lazzarini >>> Senior Lecturer >>> Dept. of Music >>> NUI Maynooth Ireland >>> tel.: +353 1 708 3545 >>> Victor dot Lazzarini AT nuim dot ie >>> >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel > > Dr Victor Lazzarini > Senior Lecturer > Dept. of Music > NUI Maynooth Ireland > tel.: +353 1 708 3545 > Victor dot Lazzarini AT nuim dot ie > > > > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:39
From	Rory Walsh
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	It's past your bedtime Victor! On 13 October 2011 00:40, Victor Lazzarini wrote: > That is because I had not built with multicore support... idiot. > On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: > >> actually something must be going wrong. The activity monitor says >> csound is always using 1 thread only. >> So I am not using the multicore code. But I am using the new parser >> alright. >> >> Victor >> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >> >>> yes, I used --new-parser >>> >>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>> >>>> Just tried with my local build for OSX 10.7, I get segfaults if I >>>> use >>>> anything besides -j1 with --new-parser. I get the floating point >>>> exception if I use -j2 but if I don't turn on the new parser: >>>> >>>> Dag2 Cache Size: 0 >>>> Csound tidy up: Floating point exception >>>> >>>> >>>> I assume then that there may be multicore clean stuff that happens >>>> if >>>> -j is used that shouldn't be happening if the old parser is on. >>>> >>>> Victor, did you use --new-parser? >>>> >>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>> wrote: >>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>> >>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>> timings >>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>> ksmps=441) >>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = >>>>> 100, it's pretty much the same >>>>> >>>>> So, in general no gains, but no major losses too even at the >>>>> original >>>>> ksmps, which is an improvement. Interestingly even if I shoot up >>>>> the number of threads to 128, I do not get a loss in performance >>>>> (possibly lots of threads not doing anything...) >>>>> >>>>> I noticed also that with -j 4 I get a better distribution of work >>>>> between the two processors (at least it's my impression from >>>>> looking at >>>>> the graphic CPU monitor) than -j 2. >>>>> >>>>> At the end of the score I am also getting this: >>>>> >>>>> Csound tidy up: Floating point exception >>>>> >>>>> which is a bit worrying. >>>>> >>>>> ============== >>>>> Dr Victor Lazzarini >>>>> Senior Lecturer >>>>> Dept. of Music >>>>> NUI Maynooth Ireland >>>>> tel.: +353 1 708 3545 >>>>> Victor dot Lazzarini AT nuim dot ie >>>>> >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >>> Dr Victor Lazzarini >>> Senior Lecturer >>> Dept. of Music >>> NUI Maynooth Ireland >>> tel.: +353 1 708 3545 >>> Victor dot Lazzarini AT nuim dot ie >>> >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >> >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel > > Dr Victor Lazzarini > Senior Lecturer > Dept. of Music > NUI Maynooth Ireland > tel.: +353 1 708 3545 > Victor dot Lazzarini AT nuim dot ie > > > > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel > ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:40
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	That is because I had not built with multicore support... idiot. On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: > actually something must be going wrong. The activity monitor says > csound is always using 1 thread only. > So I am not using the multicore code. But I am using the new parser > alright. > > Victor > On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: > >> yes, I used --new-parser >> >> On 13 Oct 2011, at 00:26, Steven Yi wrote: >> >>> Just tried with my local build for OSX 10.7, I get segfaults if I >>> use >>> anything besides -j1 with --new-parser. I get the floating point >>> exception if I use -j2 but if I don't turn on the new parser: >>> >>> Dag2 Cache Size: 0 >>> Csound tidy up: Floating point exception >>> >>> >>> I assume then that there may be multicore clean stuff that happens >>> if >>> -j is used that shouldn't be happening if the old parser is on. >>> >>> Victor, did you use --new-parser? >>> >>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>> wrote: >>>> Just testing multicore on OSX 10.5, Intel dual core. >>>> >>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>> timings >>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>> ksmps=441) >>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with kr = >>>> 100, it's pretty much the same >>>> >>>> So, in general no gains, but no major losses too even at the >>>> original >>>> ksmps, which is an improvement. Interestingly even if I shoot up >>>> the number of threads to 128, I do not get a loss in performance >>>> (possibly lots of threads not doing anything...) >>>> >>>> I noticed also that with -j 4 I get a better distribution of work >>>> between the two processors (at least it's my impression from >>>> looking at >>>> the graphic CPU monitor) than -j 2. >>>> >>>> At the end of the score I am also getting this: >>>> >>>> Csound tidy up: Floating point exception >>>> >>>> which is a bit worrying. >>>> >>>> ============== >>>> Dr Victor Lazzarini >>>> Senior Lecturer >>>> Dept. of Music >>>> NUI Maynooth Ireland >>>> tel.: +353 1 708 3545 >>>> Victor dot Lazzarini AT nuim dot ie >>>> >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >> >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel > > Dr Victor Lazzarini > Senior Lecturer > Dept. of Music > NUI Maynooth Ireland > tel.: +353 1 708 3545 > Victor dot Lazzarini AT nuim dot ie > > > > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:42
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	true, but it's hard to give up. On 13 Oct 2011, at 00:39, Rory Walsh wrote: > It's past your bedtime Victor! > > On 13 October 2011 00:40, Victor Lazzarini > wrote: >> That is because I had not built with multicore support... idiot. >> On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: >> >>> actually something must be going wrong. The activity monitor says >>> csound is always using 1 thread only. >>> So I am not using the multicore code. But I am using the new parser >>> alright. >>> >>> Victor >>> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >>> >>>> yes, I used --new-parser >>>> >>>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>>> >>>>> Just tried with my local build for OSX 10.7, I get segfaults if I >>>>> use >>>>> anything besides -j1 with --new-parser. I get the floating point >>>>> exception if I use -j2 but if I don't turn on the new parser: >>>>> >>>>> Dag2 Cache Size: 0 >>>>> Csound tidy up: Floating point exception >>>>> >>>>> >>>>> I assume then that there may be multicore clean stuff that happens >>>>> if >>>>> -j is used that shouldn't be happening if the old parser is on. >>>>> >>>>> Victor, did you use --new-parser? >>>>> >>>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>>> wrote: >>>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>>> >>>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>>> timings >>>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>>> ksmps=441) >>>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with >>>>>> kr = >>>>>> 100, it's pretty much the same >>>>>> >>>>>> So, in general no gains, but no major losses too even at the >>>>>> original >>>>>> ksmps, which is an improvement. Interestingly even if I shoot up >>>>>> the number of threads to 128, I do not get a loss in performance >>>>>> (possibly lots of threads not doing anything...) >>>>>> >>>>>> I noticed also that with -j 4 I get a better distribution of work >>>>>> between the two processors (at least it's my impression from >>>>>> looking at >>>>>> the graphic CPU monitor) than -j 2. >>>>>> >>>>>> At the end of the score I am also getting this: >>>>>> >>>>>> Csound tidy up: Floating point exception >>>>>> >>>>>> which is a bit worrying. >>>>>> >>>>>> ============== >>>>>> Dr Victor Lazzarini >>>>>> Senior Lecturer >>>>>> Dept. of Music >>>>>> NUI Maynooth Ireland >>>>>> tel.: +353 1 708 3545 >>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> All the data continuously generated in your IT infrastructure >>>>>> contains a >>>>>> definitive record of customers, application performance, security >>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>> makes >>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>> _______________________________________________ >>>>>> Csound-devel mailing list >>>>>> Csound-devel@lists.sourceforge.net >>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>>> Dr Victor Lazzarini >>>> Senior Lecturer >>>> Dept. of Music >>>> NUI Maynooth Ireland >>>> tel.: +353 1 708 3545 >>>> Victor dot Lazzarini AT nuim dot ie >>>> >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >>> Dr Victor Lazzarini >>> Senior Lecturer >>> Dept. of Music >>> NUI Maynooth Ireland >>> tel.: +353 1 708 3545 >>> Victor dot Lazzarini AT nuim dot ie >>> >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >> >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel >> > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 00:44
From	Rory Walsh
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	tell me about it! On 13 October 2011 00:42, Victor Lazzarini wrote: > true, but it's hard to give up. > On 13 Oct 2011, at 00:39, Rory Walsh wrote: > >> It's past your bedtime Victor! >> >> On 13 October 2011 00:40, Victor Lazzarini >> wrote: >>> That is because I had not built with multicore support... idiot. >>> On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: >>> >>>> actually something must be going wrong. The activity monitor says >>>> csound is always using 1 thread only. >>>> So I am not using the multicore code. But I am using the new parser >>>> alright. >>>> >>>> Victor >>>> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >>>> >>>>> yes, I used --new-parser >>>>> >>>>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>>>> >>>>>> Just tried with my local build for OSX 10.7, I get segfaults if I >>>>>> use >>>>>> anything besides -j1 with --new-parser. I get the floating point >>>>>> exception if I use -j2 but if I don't turn on the new parser: >>>>>> >>>>>> Dag2 Cache Size: 0 >>>>>> Csound tidy up: Floating point exception >>>>>> >>>>>> >>>>>> I assume then that there may be multicore clean stuff that happens >>>>>> if >>>>>> -j is used that shouldn't be happening if the old parser is on. >>>>>> >>>>>> Victor, did you use --new-parser? >>>>>> >>>>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>>>> wrote: >>>>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>>>> >>>>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>>>> timings >>>>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>>>> ksmps=441) >>>>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with >>>>>>> kr = >>>>>>> 100, it's pretty much the same >>>>>>> >>>>>>> So, in general no gains, but no major losses too even at the >>>>>>> original >>>>>>> ksmps, which is an improvement. Interestingly even if I shoot up >>>>>>> the number of threads to 128, I do not get a loss in performance >>>>>>> (possibly lots of threads not doing anything...) >>>>>>> >>>>>>> I noticed also that with -j 4 I get a better distribution of work >>>>>>> between the two processors (at least it's my impression from >>>>>>> looking at >>>>>>> the graphic CPU monitor) than -j 2. >>>>>>> >>>>>>> At the end of the score I am also getting this: >>>>>>> >>>>>>> Csound tidy up: Floating point exception >>>>>>> >>>>>>> which is a bit worrying. >>>>>>> >>>>>>> ============== >>>>>>> Dr Victor Lazzarini >>>>>>> Senior Lecturer >>>>>>> Dept. of Music >>>>>>> NUI Maynooth Ireland >>>>>>> tel.: +353 1 708 3545 >>>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> ------------------------------------------------------------------------------ >>>>>>> All the data continuously generated in your IT infrastructure >>>>>>> contains a >>>>>>> definitive record of customers, application performance, security >>>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>>> makes >>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>> _______________________________________________ >>>>>>> Csound-devel mailing list >>>>>>> Csound-devel@lists.sourceforge.net >>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> All the data continuously generated in your IT infrastructure >>>>>> contains a >>>>>> definitive record of customers, application performance, security >>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>> makes >>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>> _______________________________________________ >>>>>> Csound-devel mailing list >>>>>> Csound-devel@lists.sourceforge.net >>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>>> Dr Victor Lazzarini >>>>> Senior Lecturer >>>>> Dept. of Music >>>>> NUI Maynooth Ireland >>>>> tel.: +353 1 708 3545 >>>>> Victor dot Lazzarini AT nuim dot ie >>>>> >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>>> Dr Victor Lazzarini >>>> Senior Lecturer >>>> Dept. of Music >>>> NUI Maynooth Ireland >>>> tel.: +353 1 708 3545 >>>> Victor dot Lazzarini AT nuim dot ie >>>> >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >>> Dr Victor Lazzarini >>> Senior Lecturer >>> Dept. of Music >>> NUI Maynooth Ireland >>> tel.: +353 1 708 3545 >>> Victor dot Lazzarini AT nuim dot ie >>> >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel > > Dr Victor Lazzarini > Senior Lecturer > Dept. of Music > NUI Maynooth Ireland > tel.: +353 1 708 3545 > Victor dot Lazzarini AT nuim dot ie > > > > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel > ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 01:10
From	Steven Yi
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	Yeah, I should probably go to bed sometime myself! :P BTW: when you have a chance to test with multicore, are you getting errors like this: Parsing successful! Inserting Parallelism Constructs into AST Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] Inserting Parallelism Constructs into AST [End Inserting Parallelism Constructs into AST] [End Inserting Parallelism Constructs into AST] Csound tidy up: Segmentation fault Looks like an issue with: Program received signal EXC_BAD_ACCESS, Could not access memory. Reason: KERN_INVALID_ADDRESS at address: 0x0000000001301040 0x00000001000a1ad4 in csound_orc_expand_expressions () (gdb) bt #0 0x00000001000a1ad4 in csound_orc_expand_expressions () #1 0x00000001000a71fa in new_orc_parser () I'll try to test tomorrow with my Debug build to see if I can get any further information. On Thu, Oct 13, 2011 at 12:44 AM, Rory Walsh wrote: > tell me about it! > > On 13 October 2011 00:42, Victor Lazzarini wrote: >> true, but it's hard to give up. >> On 13 Oct 2011, at 00:39, Rory Walsh wrote: >> >>> It's past your bedtime Victor! >>> >>> On 13 October 2011 00:40, Victor Lazzarini >>> wrote: >>>> That is because I had not built with multicore support... idiot. >>>> On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: >>>> >>>>> actually something must be going wrong. The activity monitor says >>>>> csound is always using 1 thread only. >>>>> So I am not using the multicore code. But I am using the new parser >>>>> alright. >>>>> >>>>> Victor >>>>> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >>>>> >>>>>> yes, I used --new-parser >>>>>> >>>>>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>>>>> >>>>>>> Just tried with my local build for OSX 10.7, I get segfaults if I >>>>>>> use >>>>>>> anything besides -j1 with --new-parser. I get the floating point >>>>>>> exception if I use -j2 but if I don't turn on the new parser: >>>>>>> >>>>>>> Dag2 Cache Size: 0 >>>>>>> Csound tidy up: Floating point exception >>>>>>> >>>>>>> >>>>>>> I assume then that there may be multicore clean stuff that happens >>>>>>> if >>>>>>> -j is used that shouldn't be happening if the old parser is on. >>>>>>> >>>>>>> Victor, did you use --new-parser? >>>>>>> >>>>>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>>>>> wrote: >>>>>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>>>>> >>>>>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>>>>> timings >>>>>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>>>>> ksmps=441) >>>>>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with >>>>>>>> kr = >>>>>>>> 100, it's pretty much the same >>>>>>>> >>>>>>>> So, in general no gains, but no major losses too even at the >>>>>>>> original >>>>>>>> ksmps, which is an improvement. Interestingly even if I shoot up >>>>>>>> the number of threads to 128, I do not get a loss in performance >>>>>>>> (possibly lots of threads not doing anything...) >>>>>>>> >>>>>>>> I noticed also that with -j 4 I get a better distribution of work >>>>>>>> between the two processors (at least it's my impression from >>>>>>>> looking at >>>>>>>> the graphic CPU monitor) than -j 2. >>>>>>>> >>>>>>>> At the end of the score I am also getting this: >>>>>>>> >>>>>>>> Csound tidy up: Floating point exception >>>>>>>> >>>>>>>> which is a bit worrying. >>>>>>>> >>>>>>>> ============== >>>>>>>> Dr Victor Lazzarini >>>>>>>> Senior Lecturer >>>>>>>> Dept. of Music >>>>>>>> NUI Maynooth Ireland >>>>>>>> tel.: +353 1 708 3545 >>>>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> ------------------------------------------------------------------------------ >>>>>>>> All the data continuously generated in your IT infrastructure >>>>>>>> contains a >>>>>>>> definitive record of customers, application performance, security >>>>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>>>> makes >>>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>>> _______________________________________________ >>>>>>>> Csound-devel mailing list >>>>>>>> Csound-devel@lists.sourceforge.net >>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>>>> >>>>>>> >>>>>>> ------------------------------------------------------------------------------ >>>>>>> All the data continuously generated in your IT infrastructure >>>>>>> contains a >>>>>>> definitive record of customers, application performance, security >>>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>>> makes >>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>> _______________________________________________ >>>>>>> Csound-devel mailing list >>>>>>> Csound-devel@lists.sourceforge.net >>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>> >>>>>> Dr Victor Lazzarini >>>>>> Senior Lecturer >>>>>> Dept. of Music >>>>>> NUI Maynooth Ireland >>>>>> tel.: +353 1 708 3545 >>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> All the data continuously generated in your IT infrastructure >>>>>> contains a >>>>>> definitive record of customers, application performance, security >>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>> makes >>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>> _______________________________________________ >>>>>> Csound-devel mailing list >>>>>> Csound-devel@lists.sourceforge.net >>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>>> Dr Victor Lazzarini >>>>> Senior Lecturer >>>>> Dept. of Music >>>>> NUI Maynooth Ireland >>>>> tel.: +353 1 708 3545 >>>>> Victor dot Lazzarini AT nuim dot ie >>>>> >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>>> Dr Victor Lazzarini >>>> Senior Lecturer >>>> Dept. of Music >>>> NUI Maynooth Ireland >>>> tel.: +353 1 708 3545 >>>> Victor dot Lazzarini AT nuim dot ie >>>> >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >> >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel >> > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel > ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 01:15
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	yes, I know I am tired etc but there is something not quite right here. I have made sure csound was built with multicore support, used -j 2 etc and still only 1 thread is reported by the activity monitor. Strange thing is that I consistenly get 4.125 secs for -j 2 and below 4 secs not using -j. I wonder if the difference is to do with the init work of the parser/ multicore code, that ends up not spawning any threads. Victor On 13 Oct 2011, at 00:44, Rory Walsh wrote: > tell me about it! > > On 13 October 2011 00:42, Victor Lazzarini > wrote: >> true, but it's hard to give up. >> On 13 Oct 2011, at 00:39, Rory Walsh wrote: >> >>> It's past your bedtime Victor! >>> >>> On 13 October 2011 00:40, Victor Lazzarini >>> wrote: >>>> That is because I had not built with multicore support... idiot. >>>> On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: >>>> >>>>> actually something must be going wrong. The activity monitor says >>>>> csound is always using 1 thread only. >>>>> So I am not using the multicore code. But I am using the new >>>>> parser >>>>> alright. >>>>> >>>>> Victor >>>>> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >>>>> >>>>>> yes, I used --new-parser >>>>>> >>>>>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>>>>> >>>>>>> Just tried with my local build for OSX 10.7, I get segfaults >>>>>>> if I >>>>>>> use >>>>>>> anything besides -j1 with --new-parser. I get the floating >>>>>>> point >>>>>>> exception if I use -j2 but if I don't turn on the new parser: >>>>>>> >>>>>>> Dag2 Cache Size: 0 >>>>>>> Csound tidy up: Floating point exception >>>>>>> >>>>>>> >>>>>>> I assume then that there may be multicore clean stuff that >>>>>>> happens >>>>>>> if >>>>>>> -j is used that shouldn't be happening if the old parser is on. >>>>>>> >>>>>>> Victor, did you use --new-parser? >>>>>>> >>>>>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>>>>> wrote: >>>>>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>>>>> >>>>>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>>>>> timings >>>>>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>>>>> ksmps=441) >>>>>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with >>>>>>>> kr = >>>>>>>> 100, it's pretty much the same >>>>>>>> >>>>>>>> So, in general no gains, but no major losses too even at the >>>>>>>> original >>>>>>>> ksmps, which is an improvement. Interestingly even if I shoot >>>>>>>> up >>>>>>>> the number of threads to 128, I do not get a loss in >>>>>>>> performance >>>>>>>> (possibly lots of threads not doing anything...) >>>>>>>> >>>>>>>> I noticed also that with -j 4 I get a better distribution of >>>>>>>> work >>>>>>>> between the two processors (at least it's my impression from >>>>>>>> looking at >>>>>>>> the graphic CPU monitor) than -j 2. >>>>>>>> >>>>>>>> At the end of the score I am also getting this: >>>>>>>> >>>>>>>> Csound tidy up: Floating point exception >>>>>>>> >>>>>>>> which is a bit worrying. >>>>>>>> >>>>>>>> ============== >>>>>>>> Dr Victor Lazzarini >>>>>>>> Senior Lecturer >>>>>>>> Dept. of Music >>>>>>>> NUI Maynooth Ireland >>>>>>>> tel.: +353 1 708 3545 >>>>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> ------------------------------------------------------------------------------ >>>>>>>> All the data continuously generated in your IT infrastructure >>>>>>>> contains a >>>>>>>> definitive record of customers, application performance, >>>>>>>> security >>>>>>>> threats, fraudulent activity and more. Splunk takes this data >>>>>>>> and >>>>>>>> makes >>>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>>> _______________________________________________ >>>>>>>> Csound-devel mailing list >>>>>>>> Csound-devel@lists.sourceforge.net >>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>>>> >>>>>>> >>>>>>> ------------------------------------------------------------------------------ >>>>>>> All the data continuously generated in your IT infrastructure >>>>>>> contains a >>>>>>> definitive record of customers, application performance, >>>>>>> security >>>>>>> threats, fraudulent activity and more. Splunk takes this data >>>>>>> and >>>>>>> makes >>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>> _______________________________________________ >>>>>>> Csound-devel mailing list >>>>>>> Csound-devel@lists.sourceforge.net >>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>> >>>>>> Dr Victor Lazzarini >>>>>> Senior Lecturer >>>>>> Dept. of Music >>>>>> NUI Maynooth Ireland >>>>>> tel.: +353 1 708 3545 >>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> All the data continuously generated in your IT infrastructure >>>>>> contains a >>>>>> definitive record of customers, application performance, security >>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>> makes >>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>> _______________________________________________ >>>>>> Csound-devel mailing list >>>>>> Csound-devel@lists.sourceforge.net >>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>>> Dr Victor Lazzarini >>>>> Senior Lecturer >>>>> Dept. of Music >>>>> NUI Maynooth Ireland >>>>> tel.: +353 1 708 3545 >>>>> Victor dot Lazzarini AT nuim dot ie >>>>> >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>>> Dr Victor Lazzarini >>>> Senior Lecturer >>>> Dept. of Music >>>> NUI Maynooth Ireland >>>> tel.: +353 1 708 3545 >>>> Victor dot Lazzarini AT nuim dot ie >>>> >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >> >> Dr Victor Lazzarini >> Senior Lecturer >> Dept. of Music >> NUI Maynooth Ireland >> tel.: +353 1 708 3545 >> Victor dot Lazzarini AT nuim dot ie >> >> >> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel >> > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net

Date	2011-10-13 01:16
From	Victor Lazzarini
Subject	Re: [Cs-dev] some random observations on multicore csound OSX
	I get those messages here but there is no crash. It runs to the end of performance. Victor On 13 Oct 2011, at 01:10, Steven Yi wrote: > Yeah, I should probably go to bed sometime myself! :P BTW: when you > have a chance to test with multicore, are you getting errors like > this: > > Parsing successful! > Inserting Parallelism Constructs into AST > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > Inserting Parallelism Constructs into AST > [End Inserting Parallelism Constructs into AST] > [End Inserting Parallelism Constructs into AST] > Csound tidy up: Segmentation fault > > Looks like an issue with: > > Program received signal EXC_BAD_ACCESS, Could not access memory. > Reason: KERN_INVALID_ADDRESS at address: 0x0000000001301040 > 0x00000001000a1ad4 in csound_orc_expand_expressions () > (gdb) bt > #0 0x00000001000a1ad4 in csound_orc_expand_expressions () > #1 0x00000001000a71fa in new_orc_parser () > > I'll try to test tomorrow with my Debug build to see if I can get any > further information. > > On Thu, Oct 13, 2011 at 12:44 AM, Rory Walsh wrote: >> tell me about it! >> >> On 13 October 2011 00:42, Victor Lazzarini >> wrote: >>> true, but it's hard to give up. >>> On 13 Oct 2011, at 00:39, Rory Walsh wrote: >>> >>>> It's past your bedtime Victor! >>>> >>>> On 13 October 2011 00:40, Victor Lazzarini >>>> wrote: >>>>> That is because I had not built with multicore support... idiot. >>>>> On 13 Oct 2011, at 00:38, Victor Lazzarini wrote: >>>>> >>>>>> actually something must be going wrong. The activity monitor >>>>>> says >>>>>> csound is always using 1 thread only. >>>>>> So I am not using the multicore code. But I am using the new >>>>>> parser >>>>>> alright. >>>>>> >>>>>> Victor >>>>>> On 13 Oct 2011, at 00:33, Victor Lazzarini wrote: >>>>>> >>>>>>> yes, I used --new-parser >>>>>>> >>>>>>> On 13 Oct 2011, at 00:26, Steven Yi wrote: >>>>>>> >>>>>>>> Just tried with my local build for OSX 10.7, I get segfaults >>>>>>>> if I >>>>>>>> use >>>>>>>> anything besides -j1 with --new-parser. I get the floating >>>>>>>> point >>>>>>>> exception if I use -j2 but if I don't turn on the new parser: >>>>>>>> >>>>>>>> Dag2 Cache Size: 0 >>>>>>>> Csound tidy up: Floating point exception >>>>>>>> >>>>>>>> >>>>>>>> I assume then that there may be multicore clean stuff that >>>>>>>> happens >>>>>>>> if >>>>>>>> -j is used that shouldn't be happening if the old parser is on. >>>>>>>> >>>>>>>> Victor, did you use --new-parser? >>>>>>>> >>>>>>>> On Thu, Oct 13, 2011 at 12:22 AM, Victor Lazzarini >>>>>>>> wrote: >>>>>>>>> Just testing multicore on OSX 10.5, Intel dual core. >>>>>>>>> >>>>>>>>> 1. xanadu.csd and hommage.csd : there is no major change in >>>>>>>>> timings >>>>>>>>> for single-thread and -j 2, -j 4 , -j 8 (original ksmps and >>>>>>>>> ksmps=441) >>>>>>>>> 2. trapped.csd : slightly slower with -j 2, -j 4, -j 8 ; with >>>>>>>>> kr = >>>>>>>>> 100, it's pretty much the same >>>>>>>>> >>>>>>>>> So, in general no gains, but no major losses too even at the >>>>>>>>> original >>>>>>>>> ksmps, which is an improvement. Interestingly even if I >>>>>>>>> shoot up >>>>>>>>> the number of threads to 128, I do not get a loss in >>>>>>>>> performance >>>>>>>>> (possibly lots of threads not doing anything...) >>>>>>>>> >>>>>>>>> I noticed also that with -j 4 I get a better distribution of >>>>>>>>> work >>>>>>>>> between the two processors (at least it's my impression from >>>>>>>>> looking at >>>>>>>>> the graphic CPU monitor) than -j 2. >>>>>>>>> >>>>>>>>> At the end of the score I am also getting this: >>>>>>>>> >>>>>>>>> Csound tidy up: Floating point exception >>>>>>>>> >>>>>>>>> which is a bit worrying. >>>>>>>>> >>>>>>>>> ============== >>>>>>>>> Dr Victor Lazzarini >>>>>>>>> Senior Lecturer >>>>>>>>> Dept. of Music >>>>>>>>> NUI Maynooth Ireland >>>>>>>>> tel.: +353 1 708 3545 >>>>>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> ------------------------------------------------------------------------------ >>>>>>>>> All the data continuously generated in your IT infrastructure >>>>>>>>> contains a >>>>>>>>> definitive record of customers, application performance, >>>>>>>>> security >>>>>>>>> threats, fraudulent activity and more. Splunk takes this >>>>>>>>> data and >>>>>>>>> makes >>>>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>>>> _______________________________________________ >>>>>>>>> Csound-devel mailing list >>>>>>>>> Csound-devel@lists.sourceforge.net >>>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>>>>> >>>>>>>> >>>>>>>> ------------------------------------------------------------------------------ >>>>>>>> All the data continuously generated in your IT infrastructure >>>>>>>> contains a >>>>>>>> definitive record of customers, application performance, >>>>>>>> security >>>>>>>> threats, fraudulent activity and more. Splunk takes this data >>>>>>>> and >>>>>>>> makes >>>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>>> _______________________________________________ >>>>>>>> Csound-devel mailing list >>>>>>>> Csound-devel@lists.sourceforge.net >>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>>> >>>>>>> Dr Victor Lazzarini >>>>>>> Senior Lecturer >>>>>>> Dept. of Music >>>>>>> NUI Maynooth Ireland >>>>>>> tel.: +353 1 708 3545 >>>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> ------------------------------------------------------------------------------ >>>>>>> All the data continuously generated in your IT infrastructure >>>>>>> contains a >>>>>>> definitive record of customers, application performance, >>>>>>> security >>>>>>> threats, fraudulent activity and more. Splunk takes this data >>>>>>> and >>>>>>> makes >>>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>>> _______________________________________________ >>>>>>> Csound-devel mailing list >>>>>>> Csound-devel@lists.sourceforge.net >>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>>> >>>>>> Dr Victor Lazzarini >>>>>> Senior Lecturer >>>>>> Dept. of Music >>>>>> NUI Maynooth Ireland >>>>>> tel.: +353 1 708 3545 >>>>>> Victor dot Lazzarini AT nuim dot ie >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> All the data continuously generated in your IT infrastructure >>>>>> contains a >>>>>> definitive record of customers, application performance, security >>>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>>> makes >>>>>> sense of it. Business sense. IT sense. Common sense. >>>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>>> _______________________________________________ >>>>>> Csound-devel mailing list >>>>>> Csound-devel@lists.sourceforge.net >>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>>> Dr Victor Lazzarini >>>>> Senior Lecturer >>>>> Dept. of Music >>>>> NUI Maynooth Ireland >>>>> tel.: +353 1 708 3545 >>>>> Victor dot Lazzarini AT nuim dot ie >>>>> >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------------ >>>>> All the data continuously generated in your IT infrastructure >>>>> contains a >>>>> definitive record of customers, application performance, security >>>>> threats, fraudulent activity and more. Splunk takes this data and >>>>> makes >>>>> sense of it. Business sense. IT sense. Common sense. >>>>> http://p.sf.net/sfu/splunk-d2d-oct >>>>> _______________________________________________ >>>>> Csound-devel mailing list >>>>> Csound-devel@lists.sourceforge.net >>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> All the data continuously generated in your IT infrastructure >>>> contains a >>>> definitive record of customers, application performance, security >>>> threats, fraudulent activity and more. Splunk takes this data and >>>> makes >>>> sense of it. Business sense. IT sense. Common sense. >>>> http://p.sf.net/sfu/splunk-d2d-oct >>>> _______________________________________________ >>>> Csound-devel mailing list >>>> Csound-devel@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >>> Dr Victor Lazzarini >>> Senior Lecturer >>> Dept. of Music >>> NUI Maynooth Ireland >>> tel.: +353 1 708 3545 >>> Victor dot Lazzarini AT nuim dot ie >>> >>> >>> >>> >>> ------------------------------------------------------------------------------ >>> All the data continuously generated in your IT infrastructure >>> contains a >>> definitive record of customers, application performance, security >>> threats, fraudulent activity and more. Splunk takes this data and >>> makes >>> sense of it. Business sense. IT sense. Common sense. >>> http://p.sf.net/sfu/splunk-d2d-oct >>> _______________________________________________ >>> Csound-devel mailing list >>> Csound-devel@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/csound-devel >>> >> >> ------------------------------------------------------------------------------ >> All the data continuously generated in your IT infrastructure >> contains a >> definitive record of customers, application performance, security >> threats, fraudulent activity and more. Splunk takes this data and >> makes >> sense of it. Business sense. IT sense. Common sense. >> http://p.sf.net/sfu/splunk-d2d-oct >> _______________________________________________ >> Csound-devel mailing list >> Csound-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/csound-devel >> > > ------------------------------------------------------------------------------ > All the data continuously generated in your IT infrastructure > contains a > definitive record of customers, application performance, security > threats, fraudulent activity and more. Splunk takes this data and > makes > sense of it. Business sense. IT sense. Common sense. > http://p.sf.net/sfu/splunk-d2d-oct > _______________________________________________ > Csound-devel mailing list > Csound-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/csound-devel Dr Victor Lazzarini Senior Lecturer Dept. of Music NUI Maynooth Ireland tel.: +353 1 708 3545 Victor dot Lazzarini AT nuim dot ie ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Csound-devel mailing list Csound-devel@lists.sourceforge.net