Fabio Luporini

misc: Update docker/README.md

11

- Mainly updates Docker instructions - Split Docker CPU-GPU - Add Devito pip package - Drop duplicated text

cross-loop blocking in staggered tti and *elastic propagators

1

This should improve runtime performance of TTI staggered. The basic infrastructure is already in the codebase (used for example by the CIRE algorithm), but it currently doesn't support the case...

on going work here: https://github.com/devitocodes/devito/tree/timeM-argvalue

MPI halo too large w.r.t. core region should trigger a warning?

1

``` from devito import * grid = Grid(shape=(3, 3)) f = TimeFunction(name='f', grid=grid, save=5) eq = Eq(f.forward, f + 1., subdomain=grid.interior) op = Operator(eq) op(time_M=3) ``` will produce wrong values...

MPI slicing broken

2

Reproducer ``` import numpy as np from devito import Grid, Function, norm nx = 100 ny = 100 shape = (nx,ny) grid = Grid(shape=shape, dtype=np.float32) f1 = Function(name='f1', grid=grid) f1.data[:]...

copy-pasting from slack: We need to document the par-tile opt option. AFAICT, it is only showcased here https://github.com/devitocodes/devito/blob/master/tests/test_gpu_openacc.py#L150 . We need a paragraph here: https://github.com/devitocodes/devito/blob/master/examples/performance/00_overview.ipynb also, par-tile usable on CPUs...

[WIP] tests: Prove that nested time solves is possible

1

fixes #640 time axis still broken but looks fixable with some extra work

Consider: ``` from devito import * from devito.types import _SymbolCache grid = Grid(shape=(4, 4)) f = Function(name='f', grid=grid) f0 = Function(name='f', grid=grid) ``` both `f` and `f0` have the same...

benchmarks should use argparse, not an external dependency like click

6

and this is what you can use for so called "choice" parameters : http://click.pocoo.org/5/options/#basic-value-options I thought about this when I saw the option --noazimuth in the TTI example potentially work...