compiler: support mutli-buffering

Author: mloubout

devito/ir/clusters/algorithms.py

@@ -259,10 +259,17 @@ def guard(clusters):
             # Separate out the indirect ConditionalDimensions, which only serve
             # the purpose of protecting from OOB accesses
             cds = [d for d in cds if not d.indirect]
+            modes = [cd.relation for cd in cds]
+            if len({m == 'strict' for m in modes}) > 1:
+                raise CompilationError("Only one `strict` condition"
+                                       "can be used in an equation")
+            elif 'strict' in modes:
+                mode = 'strict'
+            else:
+                mode = sympy.And if sympy.And in modes else sympy.Or
 
             # Chain together all `cds` conditions from all expressions in `c`
             guards = {}
-            mode = sympy.Or
             for cd in cds:
                 # `BOTTOM` parent implies a guard that lives outside of
                 # any iteration space, which corresponds to the placeholder None
@@ -279,7 +286,6 @@ def guard(clusters):
 
                 # Pull `cd` from any expr
                 condition = guards.setdefault(k, [])
-                mode = mode and cd.relation
                 for e in exprs:
                     try:
                         condition.append(e.conditionals[cd])
@@ -296,7 +302,10 @@ def guard(clusters):
 
             # Combination `mode` is And by default.
             # If all conditions are Or then Or combination `mode` is used.
-            guards = {d: mode(*v, evaluate=False) for d, v in guards.items()}
+            if mode == 'strict':
+                guards = {d: v[0] for d, v in guards.items()}
+            else:
+                guards = {d: mode(*v, evaluate=False) for d, v in guards.items()}
 
             # Construct a guarded Cluster
             processed.append(c.rebuild(exprs=exprs, guards=guards))

devito/ir/equations/equation.py

@@ -237,14 +237,6 @@ def __new__(cls, *args, **kwargs):
                     cond = d.relation(cond, GuardFactor(d))
                 conditionals[d] = cond
 
-        # Replace the ConditionalDimensions in `expr`
-        for d, cond in conditionals.items():
-            # Replace dimension with index
-            index = d.index
-            index = index - relational_min(cond, d.parent)
-            shift = relational_shift(cond, d.parent)
-            expr = uxreplace(expr, {d: IntDiv(index, d.symbolic_factor) + shift})
-
         # Merge conditionals when possible. E.g if we have an implicit_dim
         # and there is a dimension with the same parent, we ca merged
         # its condition
@@ -254,14 +246,23 @@ def __new__(cls, *args, **kwargs):
             for cd in dict(conditionals):
                 if cd.parent == d.parent and cd != d:
                     cond = conditionals.pop(d)
-                    mode = cd.relation and d.relation
-                    if issubclass(mode, sympy.Or):
+                    if d.relation == 'strict':
                         conditionals[d] = cond
                         conditionals.pop(cd)
                     else:
+                        mode = cd.relation and d.relation
                         conditionals[cd] = mode(cond, conditionals[cd])
                     break
 
+        # Replace the ConditionalDimensions in `expr`
+        for d, cond in conditionals.items():
+            # Replace dimension with index
+            index = d.index
+            if d.condition is not None and d in expr.free_symbols:
+                index = index - relational_min(cond, d.parent)
+            shift = relational_shift(cond, d.parent)
+            expr = uxreplace(expr, {d: IntDiv(index, d.symbolic_factor) + shift})
+
         # Lower all Differentiable operations into SymPy operations
         rhs = diff2sympy(expr.rhs)
 

devito/passes/clusters/asynchrony.py

@@ -1,6 +1,6 @@
 from collections import defaultdict
 
-from sympy import true
+from sympy import Mod, true
 
 from devito.ir import (
     Backward, Forward, GuardBoundNext, PrefetchUpdate, Queue, ReleaseLock, SyncArray,
@@ -78,7 +78,8 @@ def callback(self, clusters, prefix):
             d = self.key0(c0)
             if d is not dim:
                 continue
-
+            if d in c0.guards and not c0.guards[d].has(Mod):
+                continue
             protected = self._schedule_waitlocks(c0, d, clusters, locks, syncs)
             self._schedule_withlocks(c0, d, protected, locks, syncs)