memory.md

How much memory?

We can estimate the amount of memory required for the simulations considering the biggest struct Run, i.e. the states Started and Ended.

Considering Started

The main thing is EcDNADistributionNPlus which is a vec of DNACopy (u16). Given a number of iterations n, we have: 16 bit * cells * 0.125 byte/bit, where cells is the number of nplus cells. The number of nplus cells depends on the proliferation rate of the cells w/ ecDNA $\rho_1$ strenght.

If $\rho_1$ is low (around 1), then cells will be approximately 0.1 * n, so the memory required for one run will be approximately: 16 * 0.1 * n * 0.125 = 0.2n.

If $\rho_1$ is higher than 1.4, then cells will be approximately 0.1 * n, so the memory required in the worst-case scenario for one run will be approximately: 16 * n * 0.125 = 2n.

Considering Ended

Once the run has been simulated, its state switches to Ended and the data in Started is summarized by Ended. The biggest struct here is EcDNADistribution, which I think should be negigeable compared to EcDNADistributionNPlus since it's a sparse histogram with the number of entries being the number of k copies present in the data.

Dynamics settings

In the dynamics settings, we should add also the dynamical measurements, that keep track of the state of the system e.g. the mean, nplus cells count per iteration... This depends on the dynamics that we are simulating, e.g. mean is a vec of f32 wherease nplus is a vec of u64. I think we can say that we can add to the previous estimations n bytes.

Example: nplus/nminus. u64 * n * timepoints * 0.125 bytes/bit = 64 bit * 0.125 bytes/bit * n * timepoints = 8 * bytes * n * timepoints.