-
Notifications
You must be signed in to change notification settings - Fork 18
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Code that fixes CSTP conductance decay bug #19
Conversation
progress toward bug fix with CSTP copyampasyni function is successfully compiling in snn_cpu_module.cpp copyampasyni function is successfully also compiling in snn_gpu_module.cpp memcpy may be working runtimeDataGPU.AMPA_syn_i is now reading and writing values runtimeDataGPU.AMPA_syn_i is now persistantly storing values throughout the sim ampa current appear to successfully compute with new approach using GPU parallelization progress toward all currents working with new method all currents successfully working with new method progress with getting all synapse processed with new nethod auto sizing of syn_i arrays seems to work better auto sizing of syn_i arrays might be fully working now added ca3cstp_taud unit test ca3cstp_taud unit test corrected possibly synapse id issue is fixed with synId now being used cleaned up extra code code further tidied up fixed synId = 0 positions added synId counter reset function added synId counter reset function updated STP calc fixed kernel_STPInit() bug attempted fix of synId for STP calcs cleaned up new code additions work in progress of cstp bug fix work in progress of cstp bug fix work in progress with cstp bug fix work in progress with cstp bug fix candidate CSTP bug fix that works successfully candidate CSTP bug fix that works successfully candidate CSTP bug fix that works successfully potentially fixed tau_d indexing updated synaptic spike adding current renamed variables syn_i to syn_g because that fits their basis equations better work-in-progress on higher performance and capacity version of the CSTP patch work-in-progress on higher performance and capacity version of the CSTP patch potential higher capacity version working now progress toward high-capacity cstp fix working higher capacity cstp fix version now potentially fully working possibly fully working high-capacity cstp patch version. however more testing including watching for race conditions should be done. changed print flags fixed nmda decay removed some atomicadd operations because they appear unneeded and take longer to compute removed some atomicadd operations because they appear unneeded and take longer to compute removed some no longer used functions and variables removed some no longer used functions and variables condensed some operations code revision that appear to cause a significant speed increase code revision that appear to cause a significant speed increase code revision that appear to cause a significant speed increase reduced the number of computations in the synapse decay code added some recoding to make synapse calc less operations and more clear fixed atomicadd operations fixed atomicadd operations reduced nmda calcs in conductanceUpdate() in synapse conductance pitched memory, post is now row index and pre is column index in 2D array, instead of the other way around. This makes more intuitive sense. fixed cudaMallocPitch memory allocation Combined commits to CSTP bug fix code into one commit
…onductance decay. The default is to use doubles. Doubles uses some more GPU RAM but is a bit more accurate.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Activating CARLSIM_CSTP_DOUBLES for calculations with higher precision
increases the memory requirement and slows down the performance.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Changes refer to CSTP only.
This code is designed to fix issue 18. This fix is for CSTP code incorrectly processing conductance decays in synaptic signals. More details are described on the issue 18 page.