TODO.md

To be done for current version

test modes CUDA_Ref and HIP_Ref

remove these test modes (initial and inefficient implementations)

Power usage

Add Watt measures for Baseline test mode

Test mode DPC++

Finalize implementation

Test modes HIP and HIP_Opt

Finalize implementation

testCase_Memory and testCase_Grid.cpp

Add case identical to STREAM triad

Same cases should implemented on both test cases

propagator_Ac2.cpp

Optimal time step is not optimal when spacing grid sampling is not the same in the different axis

Implement 4th order in time

testCase_Propa.cpp

Elapse time includes MPI comm. but it does not allow to know the time spent in MPI communication (propaGpointFD = propaGpointEff, because testCase_time_com = 0)

testCase_Grid.cpp

ApplyBoundaryConditionGB is not exact but a rough estimate

Add an option to select the type of grid points to consider: inner or all points

For accelerators: add function to measure transfer between device to CPU and vice versa

To be done for future versions

FD computations

Computation of spatial derivatives with FFT

Propagator

Propagator with pseudo spectral method (based on FFT)

More complex propagator (such as elastic wave equation)

Test mode OpenAcc

Finalize implementation