On Tuesday 01 June 2010 21:25:47 you wrote: > Hi Jason, > > On Wed, Jun 2, 2010 at 4:39 AM, Jason Moxham <[email protected]> wrote: > > can you build make tune with just > > cd tune > > make clean > > make tune > > > > ie not using -j etc , We have had problems before with parallel build for > > make tune , they tend to come and go , > > Retrying, I first checked out test_stuff and made the following > commenting in mpirtest: > > CPUS=1 > COMP="gcc" > #if [ -f /proc/cpuinfo ] ; then > # CPUS=$(cat /proc/cpuinfo | grep -ce ^processor) > #fi > > In all the test I ran on the reported machines, I did this and also > commented out any part of mpirtest that set the number of CPUs to > greater than 1. I used a script to run these in batch mode: check out > trunk, configure, build, check, tune, run test suite. The test suite > would fail if I don't set the number of CPUs to one. Anyway, after > editing mpirtest as above, I set make to not use more than one thread > and then proceeded to build. All this is on rosemary.math. In > particular, after checking out test_stuff and edited mpirtest as > above, I used the following script to check out trunk, configure, > build, check, tune, run test suite: > > #!/usr/bin/env bash > > # no parallel build > export MAKE='make' > CHECKOUT_LOG='checkout.log' > CONFIG_LOG='configure.log' > MPIR='mpir' > MPIR_ROOT= > MPIR_SVN='http://boxen.math.washington.edu/svn/mpir/mpir' > REVISION= > > # Get the current snapshot of MPIR > echo 'Check out MPIR trunk...' > svn checkout $MPIR_SVN/trunk $MPIR 2>&1 | tee -a $CHECKOUT_LOG > #echo 'Check out test scripts...' > #svn checkout $MPIR_SVN/branches/test_stuff test_stuff 2>&1 | tee -a > checkout.log > REVISION=`tail -1 $CHECKOUT_LOG | awk '{split($4, array, "."); print > array[1]}'` echo $REVISION > $CHECKOUT_LOG > mv $MPIR $MPIR"-r"$REVISION > > # Get tuning parameters > BUILD_LOG='build-r'$REVISION'.log' > CHECK_LOG='check-r'$REVISION'.log' > MPIR_ROOT=$MPIR'-r'$REVISION > MPIR_TUNE=$MPIR_ROOT'-tune' > TUNE_LOG='tune-r'$REVISION'.log' > cp -rf $MPIR_ROOT $MPIR_TUNE > cd $MPIR_TUNE > ./configure 2>&1 | tee -a ../$CONFIG_LOG > ./config.guess 2>&1 | tee -a ../$BUILD_LOG > make 2>&1 | tee -a ../$BUILD_LOG > make check 2>&1 | tee -a ../$CHECK_LOG > cd tune > make tune 2>&1 | tee -a ../../$TUNE_LOG > # ./tuneup -f 100000000 2>&1 | tee -a ../$TUNE_LOG > cd ../.. > > # Run the MPIR test suite > MPIR_TEST=$MPIR_ROOT'-test' > MPIR_TEST_ROOT='test_stuff' > TEST_LOG='test-r'$REVISION'.log' > cp -rf $MPIR_ROOT $MPIR_TEST > cp -rf $MPIR_TEST_ROOT $MPIR_TEST_ROOT'-r'$REVISION > cd $MPIR_TEST_ROOT'-r'$REVISION > ./mpirtest ../$MPIR_TEST 2>&1 | tee -a ../$TEST_LOG > cd .. > > > All of the above steps went fine on rosemary.math now that used only > one thread for building, tuning, and running the test suite. The > configure, build, check, tune, and test logs are up at > > http://sage.math.washington.edu/home/mvngu/doc/mpir/r2989/ > > I think I should make a point of building and running the test suite > with only one thread from now on.
I think it should really build properly with a parallel build , the tools all support it and it is the future , plus when I test things I want it to be fast( for the nightly? builds it doesn't matter , noone is waiting for them to finish) . If nfs is not working to well then the time stamps can make the build go bad , eg on t2.math , so build on the local scratch disk. Otherwise it should all work . make tune fails sometimes on a few machines (make speed passes) , it looks like the makefile has some dependency or something mixed up. Jason -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
