Support for implicit solvers

[BritishPiper]

[FabioLuporini]

my overarching question ATM can be summarized as follows:

why, instead of class Break, don't we just add a condition kwarg/property to Eq ? wouldn't this just vastly simplify the compilation pipeline changes that this PR is trying to bring in?

[codecov[bot]]

Codecov Report

Merging #1139 (396aa50) into master (bd5a114) will decrease coverage by 27.66%. The diff coverage is 60.67%.

@@             Coverage Diff             @@
##           master    #1139       +/-   ##
===========================================
- Coverage   86.30%   58.63%   -27.67%     
===========================================
  Files         217      200       -17     
  Lines       31332    29745     -1587     
  Branches     4170     3956      -214     
===========================================
- Hits        27040    17442     -9598     
- Misses       3802    11706     +7904     
- Partials      490      597      +107     

Impacted Files	Coverage Δ
devito/core/autotuning.py	59.54% <ø> (-30.46%)	:arrow_down:
devito/core/cpu.py	77.64% <ø> (-22.36%)	:arrow_down:
devito/core/gpu.py	70.94% <ø> (-23.50%)	:arrow_down:
devito/ir/iet/utils.py	67.85% <0.00%> (-7.59%)	:arrow_down:
devito/operator/operator.py	68.54% <0.00%> (-21.90%)	:arrow_down:
devito/passes/iet/languages/openmp.py	95.60% <ø> (-1.79%)	:arrow_down:
devito/types/array.py	79.31% <ø> (-15.57%)	:arrow_down:
devito/types/parallel.py	71.55% <ø> (-19.60%)	:arrow_down:
tests/test_pickle.py	29.38% <ø> (-70.14%)	:arrow_down:
tests/test_dse.py	14.39% <1.92%> (-85.23%)	:arrow_down:
... and 180 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update bd5a114...f25850e. Read the comment docs.

[BritishPiper]

Changed basically everything to share the compilation pipeline of the Break equation with the ConditionalDimension. Now there are two types of conditions (break conditions and normal conditions), which can be changed by the parameter brk in ConditionalDimension. Break conditions will cause the loop to break whenever the conditions stops being met.

[FabioLuporini]

This is stale and now obsolete, unfortunately. Closing. @edsml-zl5621 is currently working on implicit solvers. @BritishPiper : should you ever get back to this, please get in touch. We always welcome collaborations, and your work here was great!