+/* The below is perhaps overboard, but this allows us to save a test at the
+ * expense of a mask. This is because on both EBCDIC and ASCII machines, 'A'
+ * and 'a' differ by a single bit; the same with the upper and lower case of
+ * all other ASCII-range alphabetics. On ASCII platforms, they are 32 apart;
+ * on EBCDIC, they are 64. This uses an exclusive 'or' to find that bit and
+ * then inverts it to form a mask, with just a single 0, in the bit position
+ * where the upper- and lowercase differ. XXX There are about 40 other
+ * instances in the Perl core where this micro-optimization could be used.
+ * Should decide if maintenance cost is worse, before changing those
+ *
+ * Returns a boolean as to whether or not 'v' is either a lowercase or
+ * uppercase instance of 'c', where 'c' is in [A-Za-z]. If 'c' is a
+ * compile-time constant, the generated code is better than some optimizing
+ * compilers figure out, amounting to a mask and test. The results are
+ * meaningless if 'c' is not one of [A-Za-z] */
+#define isARG2_lower_or_UPPER_ARG1(c, v) \
+ (((v) & ~('A' ^ 'a')) == ((c) & ~('A' ^ 'a')))
+