George Bisbas

compiler: Define Iteration bounds through hierarchy-related rules

2

Definitely needs some improvement in terms of presenting better the theoretic background of it. Main concept : This is a system of rules to handle iteration bounds in a parametric...

compiler: source injection to nested field

3

Opening for discussion. This helps to work around a bug when injecting to more complex fields such as f(t, u(x,y,z)). Helps to implement Listing 3 of https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9460483

TTI compilation error on KNL with gcc

2

TODO: Need to add MFE... (sth with nested-par/gcc) Nested-parallelism OpenMP loops are not digested by GCC on KNLs. () To reproduce with (gcc (GCC) 9.1.0) `DEVITO_LOGGING=DEBUG DEVITO_PLATFORM=knl DEVITO_LANGUAGE=openmp DEVITO_ARCH=gcc python...

linalg.py failing with PGI openacc

3

Failing on the reduction clause to reproduce: pgi+openacc ``` export DEVITO_LANGUAGE=openacc export DEVITO_PLATFORM=nvidiaX export DEVITO_ARCH=pgcc export DEVITO_LOGGING=DEBUG #optional ``` `python3 misc/linalg.py mat-vec` Some discussion: https://devitocodes.slack.com/archives/CQ0AT90R0/p1607003860367600

[JuliaCon/proceedings-review] Performance metrics

4

Hi all, q1) what is the reason behind focusing on T_eff and not on Gpts/s as commonly used in papers reporting stencil performance? q2) Figure 2 shows that using the...

mpi: Add basic2 mode

1

Preallocated buffers using MPIMsg. An MPIMsgEnriched version for send/recv, will follow

misc: Add MPI0 logging level

5

This logging mode helps to reduce the output of MPI runs to only emit performance numbers once per MPI rank. It is mostly aimed to reduce the redundant verbose performance...

mpi: MPI cross-rank assignments fail

1

Cross-rank assignments fail, while expected to work ```python from devito import Grid, TimeFunction, Eq, Operator grid = Grid(shape=(4, 4)) u = TimeFunction(name="u", grid=grid, space_order=2) u.data[0, 1:-1, 1:-1] = 1. u.data[0,...