compiler: Rework HaloSpot optimization #2477

georgebisbas · 2024-11-04T11:16:47Z

Fix for issue #2448, supplemented with some code edits

devito/passes/iet/mpi.py

tests/test_mpi.py

codecov · 2024-11-04T11:36:02Z

Codecov Report

Attention: Patch coverage is 99.20319% with 2 lines in your changes missing coverage. Please review.

Project coverage is 87.35%. Comparing base (41d04ed) to head (309dd1b).

Files with missing lines	Patch %	Lines
devito/passes/iet/mpi.py	96.77%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2477      +/-   ##
==========================================
+ Coverage   87.29%   87.35%   +0.06%     
==========================================
  Files         238      238              
  Lines       45382    45561     +179     
  Branches     4031     4034       +3     
==========================================
+ Hits        39616    39802     +186     
+ Misses       5083     5078       -5     
+ Partials      683      681       -2

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

review-notebook-app · 2024-11-04T12:58:01Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

devito/ir/iet/nodes.py

EdCaunt · 2024-11-04T13:21:33Z

devito/mpi/halo_scheme.py

+                     frozenset(self.halos),
+                     frozenset(self.dims)))
+
+    def rebuild(self, **kwargs):


Subclass Reconstructable and add rargs?

EdCaunt · 2024-11-04T13:37:49Z

devito/mpi/halo_scheme.py

@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
        self._honored = frozendict(self._honored)

    def __repr__(self):


Could this be attached to a mixin class for both HaloSpot and HaloSchemeEntry?

yes but where would you add it?

I would just drop all these repr it's a lot of code for what at the end of the day?

When you print the IET, you see more useful details, it makes dev/debug life easier!

anyway, Ed has a point, this is redundant code which should somehow be factorized (a private method of HaloScheme, also called by HaloSpot.__repr__? not sure)

I was also thinking of this, but fmapper is the problem, we need to drop it from the entry? and thus be a class itself?
needs some thought, I may restructure more

I don't understand your last comment

What I am trying to say is that in order to factor code out, someone has to factor out the fmapper/functions, but I had failed to do it. Now I think I did it, fix coming

tests/test_mpi.py

devito/ir/iet/nodes.py

FabioLuporini · 2024-11-05T08:04:13Z

devito/mpi/halo_scheme.py

@@ -28,7 +28,39 @@ class HaloLabel(Tag):
 STENCIL = HaloLabel('stencil')


-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:


as Ed wrote below, this should inherit from Reconstructable

FabioLuporini · 2024-11-05T08:04:59Z

devito/mpi/halo_scheme.py

-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:
+
+    def __init__(self, loc_indices, loc_dirs, halos, dims):


can it not inherit, alternatively, from EnrichedTuple?

THere are some 'getters', that look redundant ?
Reconstrcutable seems to work fine

Is that ok now?

that look redundant

with EnrichedTuple, once you properly override __rargs__ and __rkwargs__, you probably don't need to override __init__ and __eq__

I fail to do so, so far,, the already existing code of DimensionTuple does not seem to help me....

so what's the issue?

There's a lot of redundant code here -- we should use EnrichedTuple

Then:

you shouldn't need to override __eq__

you shouldn't need to override __repr__

I have some doubts about overriding __hash__, but maybe also see comment below

I still do not understand this. SHould I still be able to construct HaloSchemeEntry the same way I do it now?
Will I need to specify getters ?
Like e.g. here:

def glb_shape(self): """Shape of the decomposed domain.""" return EnrichedTuple(*self._glb_shape, getters=self.dimensions)

SHould I still be able to construct HaloSchemeEntry the same way I do it now?

what happens if you try (so yes, w/o specifying any getters)

ok, I think I managed to do it, not sure what I was doing wrong earlier

FabioLuporini · 2024-11-05T08:05:57Z

devito/mpi/halo_scheme.py

@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
        self._honored = frozendict(self._honored)

    def __repr__(self):


yes but where would you add it?

I would just drop all these repr it's a lot of code for what at the end of the day?

tests/test_mpi.py

FabioLuporini · 2024-11-05T08:28:24Z

devito/passes/iet/mpi.py

-    # a stopper to halo merging
+    # Loop over the functions in the HaloSpots
+    for f, v in hs1.fmapper.items():
+        # If no time accesses, skip


"E.g., if no time accesses, skip"

you have to say it's an example...

wdym by 'say it's an example' ?

FabioLuporini · 2024-11-05T08:28:57Z

devito/passes/iet/mpi.py

-        return not any(d in hs.dimensions or dep.distance_mapper[d] is S.Infinity
-                       for d in dep.cause)
+        # If the function is not in both HaloSpots, skip
+        if (*hs0.functions, *hs1.functions).count(f) < 2:


.intersection(...), again as mentioned above?

no needed anymore

devito/passes/iet/mpi.py

FabioLuporini · 2024-11-05T08:36:01Z

devito/passes/iet/mpi.py


-                    candidates = [i.dim._defines for i in iters[n:]]
+            for hs1 in halo_spots[i+1:]:


maybe you could avoid an indentation level but rather considering all pairs

from itertools import combinations ... for hs0, hs1 in combinations(halo_schemes, 2): ...

btw, are we sure that the current code works if there are 2 hoisting opportunities, instead of 1 ?

could we add a test?

I am working on this now

georgebisbas · 2024-11-08T11:15:29Z

devito/mpi/halo_scheme.py

-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:
+
+    def __init__(self, loc_indices, loc_dirs, halos, dims):


Is that ok now?

devito/passes/iet/mpi.py

tests/test_mpi.py

devito/mpi/halo_scheme.py

devito/passes/iet/mpi.py

devito/mpi/halo_scheme.py

FabioLuporini · 2024-11-12T08:20:52Z

devito/mpi/halo_scheme.py

-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:
+
+    def __init__(self, loc_indices, loc_dirs, halos, dims):


that look redundant

with EnrichedTuple, once you properly override __rargs__ and __rkwargs__, you probably don't need to override __init__ and __eq__

FabioLuporini · 2024-11-12T08:22:40Z

devito/mpi/halo_scheme.py

@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
        self._honored = frozendict(self._honored)

    def __repr__(self):


anyway, Ed has a point, this is redundant code which should somehow be factorized (a private method of HaloScheme, also called by HaloSpot.__repr__? not sure)

FabioLuporini · 2024-11-12T08:23:14Z

devito/mpi/halo_scheme.py

-                hse = HaloSchemeEntry(frozendict(loc_indices),
-                                      frozendict(loc_dirs),
-                                      hse0.halos, hse0.dims)
+                hse = hse0._rebuild(loc_indices=frozendict(loc_indices),


maybe the cast to frozendict should be moved to inside HaloSchemeEntry.__init__

should it be for all attributes?

devito/passes/iet/mpi.py

FabioLuporini · 2024-11-12T08:33:29Z

devito/passes/iet/mpi.py

-                   for hs, v in cond_mapper.items()}
+    cond_mapper = _make_cond_mapper(iet)
+
+    mapper = {}

    iter_mapper = MapNodes(Iteration, HaloSpot, 'immediate').visit(iet)


you may introudce a private function to get this iter_mapper too, since you're doing the same exact filtering that appears in hoist_invariant

now that I think about it, you can probably also move the cond_mapper inside such a new utility function, and do the conditional filtering in there directly, rather than repeating it in both passes (it's basically the same stuff repeated at the moment)

I think this can be resolved

devito/passes/iet/mpi.py

tests/test_mpi.py

mloubout · 2024-11-12T13:14:58Z

devito/passes/iet/mpi.py

-        if i is None or len(halo_spots) <= 1:
-            continue
+    # Drop pairs that have keys that are None
+    iter_mapper = {k: v for k, v in iter_mapper.items() if k is not None}


This whole section looks like a plain copy paste of 89-119 especially the loop next, iter_maper filtering and conditions, can't this be. merged or lifted in some common piece?

simplified a bit

georgebisbas · 2024-11-13T19:02:32Z

devito/mpi/halo_scheme.py

@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
        self._honored = frozendict(self._honored)

    def __repr__(self):


I was also thinking of this, but fmapper is the problem, we need to drop it from the entry? and thus be a class itself?
needs some thought, I may restructure more

devito/mpi/halo_scheme.py

georgebisbas · 2024-11-14T10:41:04Z

devito/mpi/halo_scheme.py

-                hse = HaloSchemeEntry(frozendict(loc_indices),
-                                      frozendict(loc_dirs),
-                                      hse0.halos, hse0.dims)
+                hse = hse0._rebuild(loc_indices=frozendict(loc_indices),


georgebisbas · 2024-11-14T10:41:13Z

devito/mpi/halo_scheme.py

-                hse = HaloSchemeEntry(frozendict(loc_indices),
-                                      frozendict(loc_dirs),
-                                      hse0.halos, hse0.dims)
+                hse = hse0._rebuild(loc_indices=frozendict(loc_indices),


should it be for all attributes?

devito/passes/iet/mpi.py

georgebisbas · 2024-11-14T10:42:47Z

devito/passes/iet/mpi.py

+    iter_mapper = {k: v for k, v in iter_mapper.items() if len(v) > 1}
+
+    for it, halo_spots in iter_mapper.items():
+


georgebisbas · 2024-11-14T10:43:03Z

devito/passes/iet/mpi.py

+            if ensure_control_flow(hs0, hs1, cond_mapper):
+                continue
+
+            # If there are overlapping time accesses, skip


cahnged, is it better?

devito/passes/iet/mpi.py

examples/seismic/viscoacoustic/operators.py

devito/passes/iet/mpi.py

EdCaunt

Few minor comments. Looks very good to me

devito/passes/iet/mpi.py

EdCaunt · 2024-11-18T11:17:41Z

tests/test_mpi.py

+
+        # Test two variants of receiver interpolation
+        nrec = 1
+        rec = SparseTimeFunction(name="rec", grid=grid, npoint=nrec, nt=tn)


Nitpick: You could remove some lines by having npoint=1, nt=30 in here. You can do similar stuff throughout this test (for example Grid(shape=(2,)).

For the rest ok, for shape maybe better to stay as is

tests/test_mpi.py

FabioLuporini

A few more comments in, but it's surely getting closer and closer to a mergeable state

FabioLuporini · 2024-11-26T08:50:03Z

devito/mpi/halo_scheme.py

@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
        self._honored = frozendict(self._honored)

    def __repr__(self):


I don't understand your last comment

FabioLuporini · 2024-11-26T08:50:27Z

devito/mpi/halo_scheme.py

@@ -556,7 +596,7 @@ def process_loc_indices(raw_loc_indices, directions):
    known = set().union(*[i._defines for i in loc_indices])
    loc_dirs = {d: v for d, v in directions.items() if d in known}

-    return frozendict(loc_indices), frozendict(loc_dirs)
+    return loc_indices, loc_dirs


cuz the returned loc_indices and loc_dirs go directly to the constructor where they are frozen there. Inside HaloSchemeEntry init (THought this is enough?)

ok so these two lines above

halos = frozenset(halos) dims = frozenset(dims)

are droppable as well?

I would say yes!

devito/passes/iet/mpi.py

FabioLuporini · 2024-11-26T08:55:28Z

devito/passes/iet/mpi.py

+                    raw_loc_indices = {}
+
+                    for d in hse.loc_indices:
+                        md = hse.loc_indices[d]


you can spare one line

for d, md in hse.loc_indices.items(): ...

Does md stand for ModuloDimension? are we sure (I really don't remember) that at this point it can only be a ModuloDimension ? I don't think so actually, even though in most cases it will be. Anyway, it might be more prudent to call it v

You are right!

FabioLuporini · 2024-11-26T08:55:49Z

devito/passes/iet/mpi.py

+                            md_sub = it.start
+                            raw_loc_indices[d] = md.symbolic_min.subs(it.dim, md_sub)
+                        else:
+                            raw_loc_indices[d] = md


what's this case?

Is was this tutorial only failing..

https://github.com/devitocodes/devito/blob/master/examples/seismic/tutorials/12_time_blocking.ipynb

need to check closer if I can factor out some mfe I guess

tests/test_mpi.py

FabioLuporini · 2024-11-26T09:03:55Z

tests/test_mpi.py

+class TestElastic:
+
+    @pytest.mark.parallel(mode=[(1, 'diag')])
+    def test_elastic_structure(self, mode):


at first glance, it looks like there's a lot of redundant checks that also appear in the huge test above... but I may be wrong

well....I think that there are no similar checks in other tests.
And this structure is something that we really need to ensure that is tested.

We check not only where haloupdates are placed, but also functions in each one of them...something that could possibly be affected by future work in halo-related optimizations.

FabioLuporini · 2024-11-26T09:04:49Z

tests/test_mpi.py

+        assert calls[4].arguments[1] is v[1]
+
+
+class TestTTIwMPI:


you can drop "wMPI"

georgebisbas

I realized that I had added comments but were not posted, only drafted...releasing them + some new ones

devito/passes/iet/mpi.py

tests/test_mpi.py

georgebisbas · 2024-11-18T11:47:22Z

tests/test_mpi.py

+
+        # Test two variants of receiver interpolation
+        nrec = 1
+        rec = SparseTimeFunction(name="rec", grid=grid, npoint=nrec, nt=tn)


For the rest ok, for shape maybe better to stay as is

georgebisbas · 2024-11-26T16:05:23Z

devito/passes/iet/mpi.py

+                            md_sub = it.start
+                            raw_loc_indices[d] = md.symbolic_min.subs(it.dim, md_sub)
+                        else:
+                            raw_loc_indices[d] = md


Is was this tutorial only failing..

https://github.com/devitocodes/devito/blob/master/examples/seismic/tutorials/12_time_blocking.ipynb

need to check closer if I can factor out some mfe I guess

devito/passes/iet/mpi.py

examples/seismic/viscoacoustic/operators.py

georgebisbas · 2024-11-26T16:10:09Z

examples/seismic/viscoelastic/operators.py

@@ -64,4 +64,4 @@ def ForwardOperator(model, geometry, space_order=4, save=False, **kwargs):

    # Substitute spacing terms to reduce flops
    return Operator([u_v, u_r, u_t] + src_rec_expr, subs=model.spacing_map,
-                    name='Forward', **kwargs)
+                    name='ViscoElForward', **kwargs)


the new names aren't consistent since here you don't specify the type of media

ViscoIsoElForward

georgebisbas

More comments and fixes to follow tmr

FabioLuporini · 2024-12-05T08:26:28Z

devito/ir/iet/nodes.py

+        return "<%s(%s)>" % (self.__class__.__name__, functions)
+
+
+class HaloSpot(HaloMixin, Node):


after re-thinking about this change, I struggle to understand.

HaloScheme (in mpi.py) has this new __repr__

HaloSpot (here) carries a halo_scheme

Hence, couldn't HaloSpot.__repr__ simply resort to self.halo_scheme.__repr__() (or a wrapper of it if you want to add more information)

I think this thing that the same exact representation is used in two different places doesn't make much sense

Ι did a trick to emit the same repr.
Lemme know what you think

FabioLuporini · 2024-12-05T08:30:08Z

devito/mpi/halo_scheme.py

-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:
+
+    def __init__(self, loc_indices, loc_dirs, halos, dims):


so what's the issue?

There's a lot of redundant code here -- we should use EnrichedTuple

Then:

you shouldn't need to override __eq__

you shouldn't need to override __repr__

I have some doubts about overriding __hash__, but maybe also see comment below

FabioLuporini · 2024-12-05T08:31:07Z

devito/mpi/halo_scheme.py

+                self.halos == other.halos and
+                self.dims == other.dims)
+
+    def __hash__(self):


why do you need to use all these frozentsets?

if halos and dims are mutuable, they should be turned into immutable inside __init__

is this fine now, or I still do not understand?

FabioLuporini · 2024-12-05T08:31:53Z

devito/mpi/halo_scheme.py

@@ -556,7 +596,7 @@ def process_loc_indices(raw_loc_indices, directions):
    known = set().union(*[i._defines for i in loc_indices])
    loc_dirs = {d: v for d, v in directions.items() if d in known}

-    return frozendict(loc_indices), frozendict(loc_dirs)
+    return loc_indices, loc_dirs


devito/passes/iet/mpi.py

tests/test_mpi.py

FabioLuporini · 2024-12-05T08:54:01Z

tests/test_mpi.py

+        # The correct we want
+        assert len(calls) == 5
+
+        assert len(FindNodes(HaloUpdateCall).visit(op.body.body[1].body[1].body[0])) == 1


are you saying these are not covered already in the new test_issue_* that you've added above?

it seems to be the same dependency pattern, hence a redundant test...

What is better here, is that it has more dimensions, closer to the >1d example taher than the 1d mfe of test_issue_*.
This leads to also check merge_halospots doing the right thing as in haloupdate(v_x, v_y....) and not only v_x....

tests/test_mpi.py

FabioLuporini · 2024-12-10T11:05:01Z

devito/ir/iet/nodes.py

@@ -1492,6 +1495,9 @@ def body(self):
    def functions(self):
        return tuple(self.fmapper)

+    def __repr__(self):


show stay below __init__, just like all other special __X methods

FabioLuporini · 2024-12-10T11:06:12Z

devito/mpi/halo_scheme.py


 Halo = namedtuple('Halo', 'dim side')

 OMapper = namedtuple('OMapper', 'core owned')


-class HaloScheme:
+class HaloScheme():


FabioLuporini · 2024-12-10T11:08:44Z

devito/ir/iet/nodes.py

@@ -1492,6 +1495,9 @@ def body(self):
    def functions(self):
        return tuple(self.fmapper)

+    def __repr__(self):
+        funcs = self.halo_scheme.__reprfuncs__()


again, it's a useless complication. Just return something along the lines of HaloSpot(f,g) and if the user really cares about the underlying halo_scheme, they can always do hs.halo_scheme and this way get the textual information they are after

so, IOW, drop __reprfuncs__

ok, then I can just return to the old one I guess

FabioLuporini · 2024-12-10T11:09:52Z

devito/mpi/halo_scheme.py

-HaloSchemeEntry = namedtuple('HaloSchemeEntry', 'loc_indices loc_dirs halos dims')
+class HaloSchemeEntry:
+
+    def __init__(self, loc_indices, loc_dirs, halos, dims):


SHould I still be able to construct HaloSchemeEntry the same way I do it now?

what happens if you try (so yes, w/o specifying any getters)

FabioLuporini · 2024-12-10T11:10:07Z

devito/mpi/halo_scheme.py

@@ -93,9 +120,17 @@ def __init__(self, exprs, ispace):
            self._honored[i.root] = frozenset([(ltk, rtk)])
        self._honored = frozendict(self._honored)

+    def __reprfuncs__(self):


as explained above, avoid; put it back inside __repr__

FabioLuporini · 2024-12-10T11:16:48Z

devito/passes/iet/mpi.py

+                        else:
+                            raw_loc_indices[d] = v
+
+                    hse = hse._rebuild(loc_indices=raw_loc_indices)


again you might not need this extra variable if u just use loc_indices

I thnk it is good here to differentiate since we have both raw and processed

FabioLuporini · 2024-12-10T11:17:14Z

devito/passes/iet/mpi.py

+
+
+def _make_cond_mapper(iet):
+


no blank line; add simple docstring

FabioLuporini · 2024-12-10T11:19:25Z

devito/passes/iet/mpi.py

+    for hs, v in MapHaloSpots().visit(iet).items():
+        conditionals = set()
+        for i in v:
+            if i.is_Conditional and not isinstance(i.condition, GuardFactorEq):


this smells like a hack. What are you trying to accomplish here?

Is this a recent addition or is it something I've missed all along until now?

also, it should have been a set comprehension --

conditionals = {i for i in v if i.is_Conditional ...}

But again, this line seems a bit dodgy to me

It is only factored-out code; I did not write this, just to avoid duplication.

Did comprehension

FabioLuporini · 2024-12-10T11:20:46Z

tests/test_mpi.py

@@ -1220,7 +1223,7 @@ def test_avoid_fullmode_if_crossloop_dep(self, mode):
        assert np.all(f.data[:] == 2.)

    @pytest.mark.parallel(mode=2)
-    def test_avoid_haloudate_if_flowdep_along_other_dim(self, mode):
+    def test_avoid_haloupdate_if_flowdep_along_other_dim(self, mode):


why is this being tagged as a diff?

typo in haloupdate, letter p mising

FabioLuporini · 2024-12-10T11:21:25Z

tests/test_mpi.py

@@ -1333,8 +1336,10 @@ def test_merge_haloupdate_if_diff_locindices_v1(self, mode):

        * the second and third Eqs cannot be fused in the same loop

-        In the IET we end up with *one* HaloSpots, placed right before the
-        second Eq. The third Eq will seamlessy find its halo up-to-date.
+        In the IET we end up with *two* HaloSpots, one placed before the


does the name of the test still make sense or should it change as well?

right, renamed

georgebisbas added MPI mpi-related compiler labels Nov 4, 2024

georgebisbas requested review from FabioLuporini, mloubout and EdCaunt November 4, 2024 11:16

georgebisbas self-assigned this Nov 4, 2024

georgebisbas linked an issue Nov 4, 2024 that may be closed by this pull request

compiler: Redundant haloupdate #2448

Open

georgebisbas mentioned this pull request Nov 4, 2024

[RFC] compiler: Add new HaloSpot optimization #2475

Closed

georgebisbas commented Nov 4, 2024

View reviewed changes

devito/passes/iet/mpi.py Show resolved Hide resolved

tests/test_mpi.py Outdated Show resolved Hide resolved

EdCaunt reviewed Nov 4, 2024

View reviewed changes

FabioLuporini reviewed Nov 5, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch 4 times, most recently from ca5251a to 5951eaf Compare November 8, 2024 11:04

georgebisbas commented Nov 8, 2024

View reviewed changes

georgebisbas mentioned this pull request Nov 8, 2024

tests: Add a test-only init for issue 2448 #2465

Closed

EdCaunt reviewed Nov 11, 2024

View reviewed changes

devito/mpi/halo_scheme.py Outdated Show resolved Hide resolved

devito/passes/iet/mpi.py Outdated Show resolved Hide resolved

FabioLuporini reviewed Nov 12, 2024

View reviewed changes

mloubout reviewed Nov 12, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch from 9f23d51 to 57b0464 Compare November 14, 2024 11:04

georgebisbas commented Nov 14, 2024

View reviewed changes

mloubout reviewed Nov 15, 2024

View reviewed changes

devito/passes/iet/mpi.py Outdated Show resolved Hide resolved

EdCaunt reviewed Nov 18, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch from d6ed1a7 to eb1435a Compare November 18, 2024 12:53

FabioLuporini reviewed Nov 26, 2024

View reviewed changes

georgebisbas commented Nov 26, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch from eb1435a to 819ba0f Compare November 26, 2024 16:25

georgebisbas commented Nov 26, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch from 6ac6eb7 to 5d94c64 Compare November 27, 2024 12:18

georgebisbas requested review from EdCaunt, FabioLuporini and mloubout November 28, 2024 09:30

FabioLuporini reviewed Dec 5, 2024

View reviewed changes

georgebisbas force-pushed the halo_opt_revamp_II branch 3 times, most recently from 3c19ade to 2fbed4d Compare December 9, 2024 19:29

FabioLuporini reviewed Dec 10, 2024

View reviewed changes

georgebisbas added 16 commits December 11, 2024 16:24

compiler: Initialize additional halo opt pass

2e77a8f

compiler: Improve candidate selection and mapper application

3858519

compiler: Drop duplicated code

666b558

compiler: Concretize halo_opt_revamp, tests, drop redundant code

5b66aa0

examples: Update MPI notebook output

f13875a

examples: Update with edge case of 03_PML notebook

bec7357

compiler: Inline halo_to_halo

7c0eb9c

compiler: class HaloSchemeEntry(Reconstructable)

ba18de0

compiler: Rework terminology, minor cleanup, comments

350dd59

compiler: add frozendict to __init__

4294d7b

tests: Add backward test and drop redundancies

b0ceb0e

compiler: Improve HaloSpot with reviews

3858ea9

compiler: Further split tests

f6aaf39

compiler: Fix misc review comments

d0fc07a

compiler: Revert __repr__, add more minor fixes

cdf29a4

compiler: class HaloSchemeEntry(EnrichedTuple)

309dd1b

georgebisbas force-pushed the halo_opt_revamp_II branch from d069233 to 309dd1b Compare December 11, 2024 16:24

		@@ -95,8 +127,16 @@ def __init__(self, exprs, ispace):
		self._honored = frozendict(self._honored)

		def __repr__(self):


		candidates = [i.dim._defines for i in iters[n:]]
		for hs1 in halo_spots[i+1:]:

		iter_mapper = {k: v for k, v in iter_mapper.items() if len(v) > 1}

		for it, halo_spots in iter_mapper.items():

		return "<%s(%s)>" % (self.__class__.__name__, functions)


		class HaloSpot(HaloMixin, Node):

compiler: Rework HaloSpot optimization #2477

Are you sure you want to change the base?

compiler: Rework HaloSpot optimization #2477

Conversation

georgebisbas commented Nov 4, 2024

codecov bot commented Nov 4, 2024 • edited Loading

Codecov Report

review-notebook-app bot commented Nov 4, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

EdCaunt left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

FabioLuporini left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

georgebisbas left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov bot commented Nov 4, 2024 •

edited

Loading