A workaround for this I found was, rather than appending to an array and then unpacking the array, to create arrays of zeros for each parameter, and then filling the arrays at each iteration. Perhaps that tidbit could clue someone in on what the problem is -- unfortunately, I do not have enough computer science knowledge as of yet.
On Friday, August 8, 2014 1:00:43 PM UTC-4, [email protected] wrote: > > Hi Keno, > > So there was no lag or huge delay for you for the unpack function? > I am using a Mac OSX 10.9.4 system, Julia 0.3.0-RC1. > If you change nMCMC to 500, that should allow the algorithm to reach > unpack function in a reasonable amount of time while still having too long > of a pararray to deal with. At that point, the code hangs on two Macs > that I've tried. > > Thank you, > Wally > > On Thursday, August 7, 2014 10:29:43 PM UTC-4, Keno Fischer wrote: >> >> Hi Wally, I ran your code for a couple of hours today, but it never >> crashed for me. Can you tell me what I should do to reproduce? >> >> On Thu, Aug 7, 2014 at 3:03 PM, <[email protected]> wrote: >> > Edit: I got rid of the collect function in the code, and while very >> slow, >> > the code does what I intended it to do, and no longer returns that >> error. I >> > have not tried larger arrays yet, but tentatively, the issue no longer >> > occurs for smaller pararrays. >> > >> > >> Jeearray,Jiiarray,Jeiarray,Jieiarray,Jiecarray,cs1array,cs2array,cs3array,cs4array,cs0array,chiarray,r2array >> >> >> > = zip(pararray...) >> > >> > >> > >> > On Thursday, August 7, 2014 2:39:50 PM UTC-4, [email protected] >> wrote: >> >> >> >> Hey guys, >> >> >> >> So I've tried to get around the issue by just running the algorithm by >> >> running fewer Markov chain Monte Carlo iterations (labeled in the code >> as >> >> nMCMC), but even in the hundreds of iterations, the issue is cropping >> up, >> >> and when I'm trying to run in the tens of thousands, stopping >> frequently to >> >> update parameters becomes very inefficient. >> >> >> >> Another way around this would maybe be to write a meta script that >> calls >> >> the script for a few tens of iterations, update the parameters, then >> call >> >> the script again, but I do not know how to do that yet, and that also >> sounds >> >> very inefficient. >> >> >> >> Ideally, I could find some way of getting around this issue within one >> >> script. >> >> >> >> Here is the problem line: >> >> >> >> >> >> >> Jeearray,Jiiarray,Jeiarray,Jieiarray,Jiecarray,cs1array,cs2array,cs3array,cs4array,cs0array,chiarray,r2array >> >> >> >> = collect(zip(pararray...)) >> >> >> >> >> >> I can do a similar thing in Python, with >> >> >> >> >> >> >> Jeelist,Jiilist,Jeilist,Jieilist,Jieclist,cs1list,cs2list,cs3list,cs4list,cs0list,chilist,r2list >> >> >> >> = zip(*parlist) >> >> >> >> >> >> even when I have 500,000 entries of tuples stored in the "parlist," so >> I >> >> am not sure where Julia is going wrong. >> >> >> >> On Monday, August 4, 2014 12:13:42 PM UTC-4, [email protected] >> wrote: >> >>> >> >>> No rush. Thanks for taking the time! >> >>> >> >>> On Monday, August 4, 2014 12:01:22 PM UTC-4, Keno Fischer wrote: >> >>>> >> >>>> Not yet, sorry. Will get to it today. >> >>>> >> >>>> On Mon, Aug 4, 2014 at 11:27 AM, <[email protected]> wrote: >> >>>> > Hi Keno, >> >>>> > >> >>>> > Was just wondering if you were able to take a look? >> >>>> > >> >>>> > Thanks, >> >>>> > Wally >> >>>> > >> >>>> > On Friday, August 1, 2014 5:37:44 PM UTC-4, Keno Fischer wrote: >> >>>> >> >> >>>> >> The code looks perfectly fine to me, so it certainly shouldn't >> crash. >> >>>> >> I'll take a look. >> >>>> >> >> >>>> >> On Fri, Aug 1, 2014 at 4:35 PM, <[email protected]> wrote: >> >>>> >> > Hi guys, >> >>>> >> > >> >>>> >> > So in short, I coded a model-fitting Markov chain to fit some >> >>>> >> > parameters >> >>>> >> > of >> >>>> >> > a neural network to existing data. >> >>>> >> > With few amounts of Markov chain iterations, the code works >> fine, >> >>>> >> > but >> >>>> >> > when I >> >>>> >> > run for longer, the code dies before it completes my function >> of >> >>>> >> > finding >> >>>> >> > the >> >>>> >> > parameters matching the lowest chi value and yields the error >> >>>> >> > message: >> >>>> >> > >> >>>> >> > LLVM ERROR: Cannot select: 0x7fb67eaf3510: f64,ch = load >> >>>> >> > 0x7fb679d2c438, >> >>>> >> > 0x7fb67cc5b610, 0x7fb67cc57710<LD8[%13]> [ORD=49678] >> [ID=620013] >> >>>> >> > 0x7fb67cc5b610: i64 = add 0x7fb67c8cb010, 0x7fb67cc57e10 >> >>>> >> > [ORD=49676] >> >>>> >> > [ID=500012] >> >>>> >> > 0x7fb67c8cb010: i64,ch = load 0x7fb679d2c438, >> 0x7fb67eb07a10, >> >>>> >> > 0x7fb67cc57710<LD8[%9]> [ORD=49674] [ID=380011] >> >>>> >> > 0x7fb67eb07a10: i64 = add 0x7fb67eb0af10, 0x7fb67cc58210 >> >>>> >> > [ORD=49673] >> >>>> >> > [ID=260011] >> >>>> >> > 0x7fb67eb0af10: i64,ch = load 0x7fb679d2c438, >> >>>> >> > 0x7fb67eb0a510, >> >>>> >> > 0x7fb67cc57710<LD8[%6]> [ORD=49671] [ID=140011] >> >>>> >> > 0x7fb67eb0a510: i64 = add 0x7fb67c90fc10, >> 0x7fb67cc58210 >> >>>> >> > [ORD=49670] [ID=130011] >> >>>> >> > 0x7fb67c90fc10: i64,ch = load 0x7fb679d2c438, >> >>>> >> > 0x7fb67cde5710, >> >>>> >> > 0x7fb67cc57710<LD8[%3]> [ORD=49668] [ID=120011] >> >>>> >> > 0x7fb67cde5710: i64,ch = CopyFromReg >> 0x7fb679d2c438, >> >>>> >> > 0x7fb67cc5b810 [ORD=49667] [ID=120009] >> >>>> >> > 0x7fb67cc5b810: i64 = Register %vreg1 >> [ORD=49667] >> >>>> >> > [ID=1] >> >>>> >> > 0x7fb67cc57710: i64 = undef [ORD=49668] [ID=2] >> >>>> >> > 0x7fb67cc58210: i64 = Constant<16> [ORD=49670] >> [ID=3] >> >>>> >> > 0x7fb67cc57710: i64 = undef [ORD=49668] [ID=2] >> >>>> >> > 0x7fb67cc58210: i64 = Constant<16> [ORD=49670] [ID=3] >> >>>> >> > 0x7fb67cc57710: i64 = undef [ORD=49668] [ID=2] >> >>>> >> > 0x7fb67cc57e10: i64 = Constant<8> [ORD=49676] [ID=4] >> >>>> >> > 0x7fb67cc57710: i64 = undef [ORD=49668] [ID=2] >> >>>> >> > In function: jlcall__mkZip;20079 >> >>>> >> > >> >>>> >> > Is this a memory issue? >> >>>> >> > >> >>>> >> > Anyhow, I have attached my code for your perusal. Thank you for >> >>>> >> > your >> >>>> >> > time >> >>>> >> > and help. >> >>>> >> > >> >>>> >> > Regards, >> >>>> >> > Wally >> >
