I did a simple test with the compiled bock, I'm copying 5 packed primitives to 100,000 points, performance monitor gives me 3.1s on an i7, 4 core @ 3.6 with HT, I noticed that if you visualize any node inside the compiled block, the cooking time is longer. Also measuring the with compiled blocks on the performance editor should only show the compiled block nodes, it doesn't any node inside of them.
On Fri, Sep 29, 2017 at 11:34 AM, Olivier Jeannel <facialdel...@gmail.com> wrote: > We're not talking about "Packed Disk Primitive" right ? just plain packed > prim. > I opened his scene, upped the number of points to 350 000 , opened a > performance Monitor, hit record and turned the blue (display/render) on. > I red 32.4s on the Total Statistic. (it varies a bit, between 34s and 30s) > And this doesn't include the scatter computation. > My machine is a dual xeon of 2013 or 2014 at 2*6 core at 2.3Ghz 64GB ram. > Not the most recent maybe ? > Am I totally off ? > > 2017-09-29 11:25 GMT+02:00 Dan Yargici <danyarg...@gmail.com>: > >> I haven't, but I've done other very similar setups with compiled foreach >> block and found the performance to be fantastic. >> >> >> >> On Fri, Sep 29, 2017 at 10:19 AM, Olivier Jeannel <facialdel...@gmail.com >> > wrote: >> >>> Have you tried his setup ? >>> I'm talking about that one : >>> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.toadstorm.com_blog_wp-2Dcontent_uploads_2017_09_par&d=DwIFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=AXeT7qlVXjyaCTUsZjNxb5JeWbUrW4w9IJmPDRj2A3w&s=XxkiJ1uozp9qDGqDps3_767qR9iaUAcdaNGPLwnRUD8&e= >>> >>> t8_compiled.png >>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.toadstorm.com_blog_wp-2Dcontent_uploads_2017_09_part8-5Fcompiled.png&d=DwMFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=X3jUNTpd_yay0EWyazkRT8iIUdx9ddwwjFHRFeLYZ40&s=n_3UBEFLkytkZJbe3GWp_KDLcZF62K46L26sRotCYY0&e=> >>> >>> 2017-09-29 11:12 GMT+02:00 Dan Yargici <danyarg...@gmail.com>: >>> >>>> I'm very surprised by this result then! >>>> >>>> Unless you're running Houdini on a laptop from 2005 and not telling >>>> us... :) >>>> >>>> >>>> >>>> On Fri, Sep 29, 2017 at 9:46 AM, Olivier Jeannel < >>>> facialdel...@gmail.com> wrote: >>>> >>>>> Yes I did remember :D >>>>> >>>>> 2017-09-29 10:25 GMT+02:00 Dan Yargici <danyarg...@gmail.com>: >>>>> >>>>>> Did you remember to tick "Multihread when compiled" on your >>>>>> foreach_end? >>>>>> >>>>>> >>>>>> On Fri, Sep 29, 2017 at 8:38 AM, Olivier Jeannel < >>>>>> facialdel...@gmail.com> wrote: >>>>>> >>>>>>> Mmm, unless I made a mistake replicating his copy + switch + >>>>>>> compile, I'm getting a rough 33s to copy 10 object other 350000 points. >>>>>>> >>>>>>> 2017-09-29 8:34 GMT+02:00 Olivier Jeannel <facialdel...@gmail.com>: >>>>>>> >>>>>>>> Nice read. I'll try to replicate the last setup to compare specs. >>>>>>>> I really wonder how much a compile block speeds up the computation. >>>>>>>> >>>>>>>> >>>>>>>> On Thursday, September 28, 2017, Jonathan Moore < >>>>>>>> jonathan.moo...@gmail.com> wrote: >>>>>>>> >>>>>>>>> Just came across this great blog post from Toadstorm entitled "A >>>>>>>>> Long-Winded Guide to Houdini Instancing" I'd label it "everything you >>>>>>>>> wanted to know about Houdini instancing but were afraid to ask!" >>>>>>>>> myself. >>>>>>>>> It's a damn fine read, and great for those just starting out in >>>>>>>>> Houdini too. >>>>>>>>> >>>>>>>>> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.toadstorm.com_blog_-3Fp-3D493&d=DwIFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=AXeT7qlVXjyaCTUsZjNxb5JeWbUrW4w9IJmPDRj2A3w&s=jGFN8cWg9wLvCcbTCny2kt2aqAPT4fYet89f17EJfjA&e= >>>>>>>>> >>>>>>>>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.toadstorm.com_blog_-3Fp-3D493&d=DwMFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=NIPJF89o1z2Sf-yJSnls9v63QFl5b0y4nhmUBf8UWkg&s=h4hreQntQUMdMiO3LeMrJyHXHEyKLnKAKoxAMLdA0yo&e=> >>>>>>>>> >>>>>>>>> Enjoy. >>>>>>>>> >>>>>>>>> >>>>>>>>> On 25 September 2017 at 16:22, Olivier Jeannel < >>>>>>>>> facialdel...@gmail.com> wrote: >>>>>>>>> >>>>>>>>>> Kind of rooky question that I already asked but... >>>>>>>>>> >>>>>>>>>> So I have that "problem" with the copy sop (copy to point, >>>>>>>>>> whatever) when I have to do a copy with multiple input objects. >>>>>>>>>> It was something really easy within ice with high performances. >>>>>>>>>> >>>>>>>>>> I watched the sidefx tutorial, and so far I undertstood they >>>>>>>>>> deprecated the stamp method for a copy within a loop.It's rather >>>>>>>>>> slow. >>>>>>>>>> >>>>>>>>>> I also saw a method where it copies all the multiple objects on >>>>>>>>>> each point and then delete according to the ptnum the undesired >>>>>>>>>> copies. it >>>>>>>>>> gets very slow when you have many different objects to copy and it >>>>>>>>>> eats >>>>>>>>>> memory. >>>>>>>>>> >>>>>>>>>> So I made my "own" method >>>>>>>>>> The scenario was >>>>>>>>>> working with packed >>>>>>>>>> 10 different objects(platonic, torus, sphere, etc) >>>>>>>>>> I have some scattered point >>>>>>>>>> I have a noise (float) on the scattered points >>>>>>>>>> I want to distribute my 10 objects according to the values of >>>>>>>>>> that noise >>>>>>>>>> >>>>>>>>>> so I divided my noise in 10 zones (based on grey float values) >>>>>>>>>> I counted howmany of each object there was per zone(@numbperzone) >>>>>>>>>> I copied each object the good amount >>>>>>>>>> I placed each copy at its corresponding scattered point >>>>>>>>>> >>>>>>>>>> et voilà >>>>>>>>>> >>>>>>>>>> now for 10 objects distributed on 350000 points it takes under >>>>>>>>>> 10s. >>>>>>>>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__drive.google.com_open-3Fid-3D0B1QqhXD7Y15qVnBqMVhKeWl0dnc&d=DwIFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=AXeT7qlVXjyaCTUsZjNxb5JeWbUrW4w9IJmPDRj2A3w&s=JRSQRYo10A4UJankxBxS-OdQ_qIyhVeYzLbrvC98EOU&e= >>>>>>>>>> >>>>>>>>>> <https://urldefense.proofpoint.com/v2/url?u=https-3A__drive.google.com_open-3Fid-3D0B1QqhXD7Y15qVnBqMVhKeWl0dnc&d=DwMFaQ&c=76Q6Tcqc-t2x0ciWn7KFdCiqt6IQ7a_IF9uzNzd_2pA&r=GmX_32eCLYPFLJ529RohsPjjNVwo9P0jVMsrMw7PFsA&m=NIPJF89o1z2Sf-yJSnls9v63QFl5b0y4nhmUBf8UWkg&s=ObZ8gSpC-gUJkqCl3yDXGzsY0smzMBKUfJXxB92qo80&e=> >>>>>>>>>> >>>>>>>>>> That's my best method for now, but I have now idea if it's worth >>>>>>>>>> something. >>>>>>>>>> >>>>>>>>>> So how do you guys usually do ? >>>>>>>>>> >>>>>>>>>> ------ >>>>>>>>>> Softimage Mailing List. >>>>>>>>>> To unsubscribe, send a mail to softimage-requ...@listproc.aut >>>>>>>>>> odesk.com with "unsubscribe" in the subject, and reply to >>>>>>>>>> confirm. >>>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>> >>>>>>> ------ >>>>>>> Softimage Mailing List. >>>>>>> To unsubscribe, send a mail to softimage-requ...@listproc.aut >>>>>>> odesk.com with "unsubscribe" in the subject, and reply to confirm. >>>>>>> >>>>>> >>>>>> >>>>>> ------ >>>>>> Softimage Mailing List. >>>>>> To unsubscribe, send a mail to softimage-requ...@listproc.aut >>>>>> odesk.com with "unsubscribe" in the subject, and reply to confirm. >>>>>> >>>>> >>>>> >>>>> ------ >>>>> Softimage Mailing List. >>>>> To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com >>>>> with "unsubscribe" in the subject, and reply to confirm. >>>>> >>>> >>>> >>>> ------ >>>> Softimage Mailing List. >>>> To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com >>>> with "unsubscribe" in the subject, and reply to confirm. >>>> >>> >>> >>> ------ >>> Softimage Mailing List. >>> To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com >>> with "unsubscribe" in the subject, and reply to confirm. >>> >> >> >> ------ >> Softimage Mailing List. >> To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com >> with "unsubscribe" in the subject, and reply to confirm. >> > > > ------ > Softimage Mailing List. > To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com > with "unsubscribe" in the subject, and reply to confirm. >
------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.
