Thanks for your suggestions, but the run still stops in the scipy steps. # Sep 15 - no luck, stopped at 10:01 export NINJA_ARGS="-j1" export JOBS=1 make
This time slightly earlier. [scipy-1.12.0] [spkg-install] [1001/1610] Compiling C object scipy/io/matlab/_streams.cpython-312-x86_64-linux-gnu.so.p/meson-generated__streams.c.o [scipy-1.12.0] [spkg-install] [1002/1610] Linking target scipy/io/matlab/_streams.cpython-312-x86_64-linux-gnu.so During the week, I will do the install on a laptop, and see what happens. regs, Kev On Saturday 14 September 2024 at 17:18:37 UTC+10 dim...@gmail.com wrote: > On Sat, Sep 14, 2024 at 7:00 AM Kevin Youren <kyo...@gmail.com> wrote: > > > > Dima, > > > > thanks for the hint, > > > > but > > > > export NINJA_ARGS="-j4" > > make > > > > didn't work. It still 'froze' . > > Could you try -j1 rather than -j4 ? > And also > > export JOBS=1 > > > > > > > Restarting > > > > using > > export NINJA_ARGS="-j4" > > make > > > > is working using the export and the make, but the scipy steps still had > the 12 cpus at 100% together with the fan noise. > > With the excessively many parallel job as the reason for this, > it's sort of expected. > > Machine is running many jobs in parallel; one is crashing; the rest of > the jobs complete. > Restarting, you don't need to redo complete jobs, and so fewer jobs > are started in the 2nd run. > As far as I understand, scipy build gets started from the beginning in > the 2nd run. > Thus it's some other jobs that are overlapping with scipy in the 1st > run, and don't overlap > in the 2nd run. > > > > > > When scipy finished , the machine has gone quiet and is currently in the > documentation steps: > > > > [sagemath_doc_html-none] [spkg-install] sage --docbuild --no-pdf-links > reference/spkg inventory, etc. > > > > > > regs, Kev > > > > > > > > > > > > On Thursday 12 September 2024 at 18:31:55 UTC+10 dim...@gmail.com wrote: > >> > >> scipy itself is not built with configure/make, it's built with meson, > which invokes ninja (a faster replacement for make, in particular > >> it parallelizes the tasks much better - but in your case it goes > overboard with it). > >> You can see it in your log: > >> > >> [spkg-install] Found ninja-1.11.1 at /usr/bin/ninja > >> [spkg-install] + /usr/bin/ninja > >> > >> now, if you invoke "ninja -h" at the shell prompt, it will print, among > other things, > >> > >> -j N run N jobs in parallel (0 means infinity) [default=6 on this > system] > >> > >> (on your system the "default" is likely much bigger) > >> > >> There is no direct way to specify a non-default "-j" value, however it > appears to be possible to do this via meson, > >> which invokes ninja via "meson compile". > https://mesonbuild.com/Commands.html does not make it clear whether > >> > >> "JOBS" or "NINJA_ARGS" are shell environment variables, or just > placeholders for the actual values, > >> > >> but you can try something like > >> > >> export NINJA_ARGS="-j4" > >> > >> make > >> > >> If this does not work (I can see "export NINJA_ARGS .." on the net at > various places, so it seems to work for some people), one can create a > shell script, called ninja, which merely invokes > >> > >> ninja -j 4 > >> > >> and place it first in your PATH, so that it picked up first, and serves > as a replacement for ninja command. > >> > >> HTH > >> > >> Dima > >> > >> > >> > >> On 12 September 2024 00:22:49 BST, Kevin Youren <kyo...@gmail.com> > wrote: > >>> > >>> Thanks for replying, Eric > >>> > >>> I did precisely what you suggested, tried make -j4, but no luck. > >>> > >>> I even opened and read my paper book "GNU Make" by Stallman. > >>> > >>> So, I tried "make" by itself, and it did slow it down a bit. > >>> > >>> However scipy simply took over all 12 cpus, at lightning speed. > >>> > >>> The advantage of the Tower was you don't even have to look at the > >>> System Monitor, the fans make so much noise trying to cool down the > >>> cpus. > >>> > >>> When I restarted the machine, and just typed in "make", they still used > >>> all 12 cpus, but finished OK without over-heating, and the build > >>> finished OK. > >>> > >>> I am thinking about splitting the the makefile into 3 pieces, and see > >>> what that achieves. > >>> > >>> regs, > >>> > >>> Kev > >>> > >>> > >>> On Wednesday 11 September 2024 at 18:58:20 UTC+10 egourg...@gmail.com > wrote: > >>>> > >>>> Hi, > >>>> > >>>> From the log file: > >>>> > >>>> [spkg-install] g++: fatal error: Killed signal terminated program > cc1plus > >>>> > >>>> This points towards a maximum memory reached. You may decrease the > number of threads in the parallel build, e.g. using make -j4 instead of > make -j8. > >>>> > >>>> Best regards, > >>>> > >>>> Eric. > >>>> > > -- > > You received this message because you are subscribed to the Google > Groups "sage-support" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to sage-support...@googlegroups.com. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-support/b1af99d0-cac3-4f40-a80c-ee795275147an%40googlegroups.com > . > -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/afb85a8f-e2cc-4ff7-8441-b3a19036949bn%40googlegroups.com.
