[sage-devel] Re: does zn_poly normally take a long time to build?
On Jul 16, 2008, at 11:54 PM, tkeller wrote: I may have been imprecise. To clarify, zn_poly built, then displayed this message: Calibrating cycle counter... ok (3.84e+18) Okay, this means that zn_poly thinks your clock speed is around 3.84 billion GHz. (Yes, 3.84 * 10^18 cycles per second.) This means zn_poly thinks it's allowed to spend a bit longer (!) on the tuning process. I assume your machine is not really that fast. It's possible that either my cycle counting code is not working on your machine, or that something like getrusage is not working as I expect on your machine, or maybe some other weird timing fluke. Could you try building again, to see if it's easily reproducible? david --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
Rebuilt 3.0.5 and didn't have any issues this time. Cycle counter gave (1.68e+09) this time around and zn_poly built in about 2 minutes or so. I also rebuilt 3.0.3 and likewise didn't have any problems. Sorry for the phantom bug report. Thomas On Jul 17, 6:36 am, David Harvey [EMAIL PROTECTED] wrote: On Jul 16, 2008, at 11:54 PM, tkeller wrote: I may have been imprecise. To clarify, zn_poly built, then displayed this message: Calibrating cycle counter... ok (3.84e+18) Okay, this means that zn_poly thinks your clock speed is around 3.84 billion GHz. (Yes, 3.84 * 10^18 cycles per second.) This means zn_poly thinks it's allowed to spend a bit longer (!) on the tuning process. I assume your machine is not really that fast. It's possible that either my cycle counting code is not working on your machine, or that something like getrusage is not working as I expect on your machine, or maybe some other weird timing fluke. Could you try building again, to see if it's easily reproducible? david --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
On Jul 17, 2008, at 10:31 AM, tkeller wrote: Rebuilt 3.0.5 and didn't have any issues this time. Cycle counter gave (1.68e+09) this time around and zn_poly built in about 2 minutes or so. I also rebuilt 3.0.3 and likewise didn't have any problems. Sorry for the phantom bug report. I've been looking at the calibration code and I can see one spot where (in extremely rare circumstances) the behaviour you observed could occur. Thanks for the bug report, and let us know if it happens again! david --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
On Jul 16, 8:39 pm, tkeller [EMAIL PROTECTED] wrote: Hi, The last build I built from source (3.0.3) took ~ 3 hours total on my average dell laptop (running kubuntu 8.0.4.1). Building 3.0.5 is ongoing, but has spent the last 5+ hours on zn_poly tuning program. Is this normal? It hasn't stalled, but has effectively tripled the compilation time (at least). This sounds like a build issue to me. zn_poly's build time should be more or less instant. Even with the check target active (which is disabled right now) it should only take 10 to 12 minutes on a decent CPU. Can you save the current tail of install.log and put it up on the web somewhere. Then hit CTRL-C and restart the build process with make. If it hangs again/takes very long we need to poke around. Is this normal for other platforms? No, this is not normal. Thomas Cheers, Michael --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
I may have been imprecise. To clarify, zn_poly built, then displayed this message: Calibrating cycle counter... ok (3.84e+18) KS mul: ... KS sqr: ... Nussbaumer mul: ... Nussbaumer sqr: ... KS/FFT mul: ... KS/FFT sqr: ... midmul FFT: . This was ongoing for the 5ish hours I mentioned, before I accidentally aborted the build process. The actual build process itself for zn_poly went relatively quickly. I'll check the log after this attempt finishes. Regards, Thomas On Jul 16, 10:43 pm, mabshoff [EMAIL PROTECTED] wrote: On Jul 16, 8:39 pm, tkeller [EMAIL PROTECTED] wrote: Hi, The last build I built from source (3.0.3) took ~ 3 hours total on my average dell laptop (running kubuntu 8.0.4.1). Building 3.0.5 is ongoing, but has spent the last 5+ hours on zn_poly tuning program. Is this normal? It hasn't stalled, but has effectively tripled the compilation time (at least). This sounds like a build issue to me. zn_poly's build time should be more or less instant. Even with the check target active (which is disabled right now) it should only take 10 to 12 minutes on a decent CPU. Can you save the current tail of install.log and put it up on the web somewhere. Then hit CTRL-C and restart the build process with make. If it hangs again/takes very long we need to poke around. Is this normal for other platforms? No, this is not normal. Thomas Cheers, Michael --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
On Jul 16, 8:54 pm, tkeller [EMAIL PROTECTED] wrote: Hi Thomas, I may have been imprecise. To clarify, zn_poly built, then displayed this message: Calibrating cycle counter... ok (3.84e+18) KS mul: ... KS sqr: ... Nussbaumer mul: ... Nussbaumer sqr: ... KS/FFT mul: ... KS/FFT sqr: ... midmul FFT: . This was ongoing for the 5ish hours I mentioned, before I accidentally aborted the build process. The actual build process itself for zn_poly went relatively quickly. I'll check the log after this attempt finishes. Yeah, this is pretty much as I expected. The tuning phase should only take a couple minutes, so you either hit a bug in zn_poly or something is wrong with your RAM, compiler, etc. I expect David Harvey (the author of zn_poly) to pop up in the morning (east coast time) and ask you more detailed questions. Regards, Thomas What was the last version of Sage that did build for you? IIRC we did not touch zn_poly since Sage 3.0.2 or so, so unless something changed on your end it seems odd to start failing now. Cheers, Michael --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
[sage-devel] Re: does zn_poly normally take a long time to build?
Sage 3.0.3 definitely built from source for me 3ish weeks ago. I've applied suggested Debian/Ubuntu updates since then (nothing too radical sticks out) , but I'll trying rebuilding 3.0.3 tomorrow morning. Thanks much for your responses. Regards, Thomas On Jul 16, 10:59 pm, mabshoff [EMAIL PROTECTED] wrote: On Jul 16, 8:54 pm, tkeller [EMAIL PROTECTED] wrote: Hi Thomas, I may have been imprecise. To clarify, zn_poly built, then displayed this message: Calibrating cycle counter... ok (3.84e+18) KS mul: ... KS sqr: ... Nussbaumer mul: ... Nussbaumer sqr: ... KS/FFT mul: ... KS/FFT sqr: ... midmul FFT: . This was ongoing for the 5ish hours I mentioned, before I accidentally aborted the build process. The actual build process itself for zn_poly went relatively quickly. I'll check the log after this attempt finishes. Yeah, this is pretty much as I expected. The tuning phase should only take a couple minutes, so you either hit a bug in zn_poly or something is wrong with your RAM, compiler, etc. I expect David Harvey (the author of zn_poly) to pop up in the morning (east coast time) and ask you more detailed questions. Regards, Thomas What was the last version of Sage that did build for you? IIRC we did not touch zn_poly since Sage 3.0.2 or so, so unless something changed on your end it seems odd to start failing now. Cheers, Michael --~--~-~--~~~---~--~~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~--~~~~--~~--~--~---
