Changeset 3960

Timestamp:

Aug 5, 2014, 12:38:42 PM (5 years ago)

Author:

cameron

Message:

Eliminating some common subexpressions

Location:

proto/charsetcompiler

Files:

• CC_compiler.py (modified) (2 diffs)
• unicode_category_compiler.py (modified) (3 diffs)

proto/charsetcompiler/CC_compiler.py

@@ -20 +20 @@
         predeclared = [self.bit_var(i) for i in range(0, self.mEncoding.bits)]
         for sym in predeclared: self.common_expression_map[sym] = sym
+
+    def add_symbol_to_map(self, sym): 
+        self.common_expression_map[sym] = sym
+
+    def add_common_expressions(self, enclosing_cgo):
+        for sym in enclosing_cgo.common_expression_map.keys():
+            self.common_expression_map[sym] = enclosing_cgo.common_expression_map[sym]
 
     def bit_var(self, n):
@@ -231 +238 @@
             if isinstance(expr, TrueLiteral): return '-1'
             elif isinstance(expr, FalseLiteral): return '0'
-            elif isinstance(expr, Var): return expr.varname
+            elif isinstance(expr, Var): 
+               # This is a hack.
+               self.common_expression_map[expr.varname] = expr.varname
+               return expr.varname
             elif isinstance(expr, Not):
                e = self.expr_string_to_variable(self.expr2py(expr.operand))

proto/charsetcompiler/unicode_category_compiler.py

@@ -196 +196 @@
      utf8_range_compiler(cgo, lo, hi, range_var)
      inner_cgo = CC_compiler(UTF8(), range_var + '_tmp%i', False, '')
+     inner_cgo.add_common_expressions(cgo)
      generateCharClassDefsInIfHierarchy(inner_cgo, (lo, hi), ifRangeList, charClassMap)
      cgo.add_if_stmt(Var(range_var), inner_cgo.generated_code)
@@ -207 +208 @@
    for k in charClassMap.keys():
      subcc1 = rangeIntersect(charClassMap[k], lo, hi)
-     CC_var = "CC_%s_%x_%x" % (k, lo, hi)
      # Divide by UTF-8 length
      for byte_range in [(0, 0x7F), (0x80, 0x7FF), (0x800, 0xFFFF), (0x10000, 0x10FFFF)]:
-        subcc2 = rangeIntersect(charClassMap[k], byte_range[0], byte_range[1])
-        ulen = utf8_length(byte_range[0])
+        (lo1, hi1) = byte_range
+        subcc2 = rangeIntersect(subcc1, lo1, hi1)
+        ulen = utf8_length(lo1)
         for subrange in subcc2:
-           matched_sequence_compiler(cgo, subrange[0], subrange[1], 1, ulen, CC_var)
-     cgo.add_assignment("struct_%s.cc" % k, cgo.expr2py(make_or(Var("struct_%s.cc" % k), Var(CC_var))))
+           (lo2, hi2) = subrange
+           CC_var = "CC_%s_%x_%x" % (k, lo2, hi2)
+           matched_sequence_compiler(cgo, lo2, hi2, 1, ulen, CC_var)
+           cgo.add_assignment("struct_%s.cc" % k, cgo.expr2py(make_or(Var("struct_%s.cc" % k), Var(CC_var))))
 
 def rangeIntersect(ccList, lo, hi):
@@ -227 +230 @@
 
 defaultIfRangeList = [(0,0x7FF), (0, 0x7F), (0x80, 0x3FF), (0x400,0x7FF), (0x800, 0xFFFF), (0x10000, 0x10FFFF)]
+
+#defaultIfRangeList = [(0x80,0x10FFFF), (0x80,0x7FF), (0x800,0xFFFF), (0x10000, 0x10FFFF)]