[aerosayan]

My home cooked solvers are explicit FVM compressible Euler solvers, since I like shockwaves, and explicit solvers are easy to optimize.


But naive implementation of explicit solvers turn out to be slow as heck. Using local timestepping and geometric multigrid acceleration seemed to help improve the performance very much.


However, implicit solvers seem to be very fast even without multigrid acceleration, and we have the added advantage of CFL acceleration.


Should I focus on writing explicit or implicit solvers since my areas of interest all have transient physics?




My areas of interest are : rocket engine combustion, supersonic and hypersonic aerodynamics, supercavitating ammunition design

[arjun]

I am actually thinking that multi-time stepping explict solvers may be better specially they are easy to make parallel efficient.