New version of transitive reduction.

Author: gsvgit

cfpq_add_context/add_contexts.py

@@ -27,23 +27,24 @@ def indexed_to_boolean_decomposition(graph, block_count):
 
 def transitive_reduction(assigns, mask):
     result = Matrix(BOOL, assigns.ncols, assigns.ncols, name = "reduced_assigns")
-    result << Matrix.mxm(mask, assigns, "land_lor")
-    total = Matrix(BOOL, assigns.ncols, assigns.ncols, name = "closure")
+    result << Matrix.mxm(mask, assigns, "any_pair")
+    #total = Matrix(BOOL, assigns.ncols, assigns.ncols, name = "closure")
     unused_assigns = Matrix(BOOL, assigns.ncols, assigns.ncols, name = "unused_assigns")
     unused_assigns(~result.S) << assigns
     closure = unused_assigns
-    total("lor") << closure
-    #while closure.nvals > 0:
-    while True:
+    #total("any") << closure
+    while closure.nvals > 0:
+    #while True:
         print ("Closure nvals = ", closure.nvals)
-        closure << Matrix.mxm(closure, unused_assigns, "land_lor")
-        result("lor") << Matrix.mxm(closure, mask, "land_lor")
+        closure << Matrix.mxm(closure, unused_assigns, "any_pair")
+        result("any") << Matrix.mxm(closure, mask, "any_pair")
         new_closure = Matrix(BOOL, assigns.ncols, assigns.ncols, name = "closure")
         new_closure(~result.S) << closure
-        nnz = total.nvals
-        total("lor") << closure
-        if total.nvals == nnz:
-            break
+        #nnz = total.nvals
+     #   total("any") << closure
+        closure = new_closure
+      #  if total.nvals == nnz:
+       #     break
     return result
 
 def to_label_decomposed_graph(graph, automata_size, initial_graph_size):
@@ -56,18 +57,18 @@ def to_label_decomposed_graph(graph, automata_size, initial_graph_size):
     alloc_r << alloc.T
     print("Boolean matrix for alloc_r nvals: ", alloc_r.nvals)
 
-    #print("mask start")
-    #mask_v = Vector(BOOL, graph.ncols, name = "mask_vector")
-    #mask_v(op.lor) << alloc.reduce_columnwise("lor")  
-    #mask_v(op.lor) << alloc.reduce_rowwise("lor")
+    print("mask start")
+    mask_v = Vector(BOOL, graph.ncols, name = "mask_vector")
+    mask_v(op.lor) << alloc.reduce_columnwise("lor")  
+    mask_v(op.lor) << alloc.reduce_rowwise("lor")
 
     #print("entrypoints start")
 
     #entrypoints = Vector(bool,graph.nrows, name="entrypoints")
     #entrypoints << graph.reduce_columnwise(op.lor)
     #mask_v(op.lor) << Vector.from_coo(list(set(range(0,graph.nrows)).difference(entrypoints.to_coo(values=False)[0])), values=True, dtype = BOOL)
 
-    #mask_v(op.lor) << Vector.from_coo([i * automata_size for i in range(0, initial_graph_size)], values=True, dtype = BOOL, size= graph.ncols)
+    mask_v(op.lor) << Vector.from_coo([i * automata_size for i in range(0, initial_graph_size)], values=True, dtype = BOOL, size= graph.ncols)
 
     load_i = Matrix(UINT64, graph.ncols, graph.nrows, name = "load_i_after_intersection")
     load_i << graph.select(graphblas.select.select_load).apply(graphblas.unary.decode_load)
@@ -77,11 +78,11 @@ def to_label_decomposed_graph(graph, automata_size, initial_graph_size):
     store_i << graph.select(graphblas.select.select_store).apply(graphblas.unary.decode_store)
     print("Matrix for store_i nvals: ", store_i.nvals)
 
-    #mask_v(op.lor) << load_i.reduce_columnwise("lor")
-    #mask_v(op.lor) << load_i.reduce_rowwise("lor")
+    mask_v(op.lor) << load_i.reduce_columnwise("lor")
+    mask_v(op.lor) << load_i.reduce_rowwise("lor")
 
-    #mask_v(op.lor) << store_i.reduce_columnwise("lor")
-    #mask_v(op.lor) << store_i.reduce_rowwise("lor")
+    mask_v(op.lor) << store_i.reduce_columnwise("lor")
+    mask_v(op.lor) << store_i.reduce_rowwise("lor")
 
     store_block_count = store_i.reduce_scalar("max").get(0) + 1
     load_block_count = load_i.reduce_scalar("max").get(0) + 1
@@ -109,14 +110,14 @@ def to_label_decomposed_graph(graph, automata_size, initial_graph_size):
 
 
 
-    #assign_mask = mask_v.diag(name = "assign_mask")
+    assign_mask = mask_v.diag(name = "assign_mask")
 
     assign = Matrix(BOOL, graph.ncols, graph.nrows, name = "assign_after_intersection")
     assign << graph.select(graphblas.select.select_assign)
     print("Boolean matrix for assign nvals: ", assign.nvals)
 
 
-    #assign << transitive_reduction(assign, assign_mask)
+    assign << transitive_reduction(assign, assign_mask)
 
     #assign_res = Matrix(BOOL, graph.ncols, graph.nrows, name = "assign_after_transitive_reduction")
     #assign_1 = Matrix.mxm(assign_mask, assign, "land_lor")