talkanbaev-artur / com392-final-back-team Goto Github PK

Document the algo structure and the project workflow

Add documentation on human and virus stats, their roles and behaviour. All first-pass (basic) stats are in first section, all second-pass (future) stats are in the second section. All first-pass stats have to be documented well and be clear. Second-pass stats are optional.

Our code is a mess, and we need to fix some bug-prone points.

Please go to this tab


And fix as much as possible. Thanks

Each day is a single call to a kernel from the cpu/gpu code. The layout for it should be ready, so we can add/remove functions and efficiently iterate over the simulation development.

Clean the main function. Modify the crack to use better cli args (not like -fSN.bmp, but something better) and remove dead code. Add docks to it. It would be great to remove all the dead code from other parts of interface.h and interface.cpp.

Please move all the code into the src directory and rewrite the makefile to compile the project correctly. Additionally set the utilities for the make, and create two versions: one for last version of cuda, another for G30 machines.

The bottom of the screen (blocks at the end) has a very weird behaviour: they get their status too quickly in every run.

This is unintended behaviour, and should be fixed.

It would be nice to have a yml conf instead of crack.h. Please find some easy to use libs and implement an extendable yaml conf reader.

We need some statistics going on. I would suggest using SIR as the main struct for stats. Google the SIR for epidemiology and classify all our statuses to this model.

After creating the struct and api for handling these stats (printing to a file, storing somewhere or whatever), implement a way to collect these stats. You can make a device function, which would be run at the end of the day and calculate these stats. It is important to understand that due to the population being large, you would need to write a sum reduction kernel.

Kernel would be launched on each person and you would need to use some kind of memory unifier to output the final SIR struct. Think about the shared memory.