I can compile successfully withhines wrote:What I don't know is whether you were able to use the
configure statements I indicated above to make a working version. i.e no python and no mpi and no flags
other than the ones I indicated.
Code: Select all
./configure --prefix=`pwd` CC=gcc-4 CXX=g++-4 LDFLAGS='-Wl,--enable-auto-import'
Code: Select all
./configure --prefix=`pwd` --with-nrnpython CC=gcc-4 CXX=g++-4 LDFLAGS='-Wl,--enable-auto-import'
Code: Select all
oc>cvode_cache_efficient(1)
nrniv: cvode_cache_efficient undefined function
near line 14
cvode_cache_efficient(1)
^
cvode_cache_efficient(1 )
Code: Select all
./configure --prefix=`pwd` --with-paranrn --with-nrnpython CC=gcc-4 CXX=g++-4 LDFLAGS='-Wl,--enable-auto-import'
I want to be able to use optimization flags. I do not use paranrn because I need IV. also I do not use nrnpython because the code become ugly with python with long unnecessary commands and also because hoc has what I need.
also I noticed that the "make mswin" copies the gcc-3 to the c:\marshalnrn\bin --> which it would be better to copy gcc-4 instead
also I want to thank you hines for your concern and support and quick reply.
I want you know that I am doing what I can to improve neuron.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
ps: I realized that the your official windows XP (32bit) version of neuron is faster than my self-compiled optimized version of neuron on Linux (64bit)?! I search for it on internet and there is other similar reports with other applications?!!!!!!!
Linux wins the match only the read and write on the hard disk is an issue or when multitasking with several different programs (but windows is more powerful in parallel threads). if you are interested look this link --> Ubuntu 8.04 LTS vs. Windows XP SP3: Application Performance Benchmark
I also noticed that when compiling with paranrn on Linux there is some performance loss.