Merge pull request #7142 from drbaggy/master

Tweaked for C speed

2 files changed, 34 insertions, 3 deletions

diff --git a/challenge-192/james-smith/README.md b/challenge-192/james-smith/README.md
index 7c47f95f78..76dac905f8 100644
--- a/challenge-192/james-smith/README.md
+++ b/challenge-192/james-smith/README.md
@@ -47,11 +47,14 @@ This can also be done by converting to a string and then coverting back again.
 
 ```
 sub string_flip {
-  oct '0b'.sprintf('%b',$_[0])=~tr/01/10/r;
+  $_[0] ? oct '0b'.sprintf('%b',$_[0])=~tr/01/10/r : 0;
+}
 ```
 
 We use `tr` with the `r` option to return the result of the translation...
 
+Note we have to check whether the input is `0` as in this case the output is also `0` as there is no leading 1.
+
 ### Performance...
 
 Well this is where Perl is uber fast when it comes to strings - the string solution is faster than the bit manipulation. This is probably due to the overhead of each separate operation in the numeric version. For a test example of "12345678" (`1011 1100 0110 0001 0100 1110`) the string method is 8x faster than the binary method.
@@ -74,6 +77,22 @@ int c_flip(int n) {
 
 Now - when comparing this to the other two: The C version is 4.5 times faster than the string version OR 35x faster than the equivalent Perl version.
 
+A further re-write of the C gives:
+
+```C
+int c2(int n) {
+  int o = n;
+  int m = 0;
+  while(o) {
+    m<<=1;
+    m++;
+    o>>=1
+  }
+  return n^m;
+}
+```
+
+Here we compute a mask of `1`s as long as the binary representation of the number so for `25` = `11001` we have a mask of `11111` and so doing a bitwize XOR operation gives us `00110` or `6`. Which is even faster as it only does the XOR (`^`) once. {approx 5 times faster than the regex version}
 
 # Task 2 - Equal Distribution
 
diff --git a/challenge-192/james-smith/perl/ch-1.pl b/challenge-192/james-smith/perl/ch-1.pl
index b2e18ab4be..89e9c957ac 100644
--- a/challenge-192/james-smith/perl/ch-1.pl
+++ b/challenge-192/james-smith/perl/ch-1.pl
@@ -13,16 +13,17 @@ my @TESTS = (  [5,2],[4,3],[6,1] );
 is( binary_flip( $_->[0] ), $_->[1] ) for @TESTS;
 is( string_flip( $_->[0] ), $_->[1] ) for @TESTS;
 is( c_flip(      $_->[0] ), $_->[1] ) for @TESTS;
+is( c2_flip(     $_->[0] ), $_->[1] ) for @TESTS;
 done_testing();
 
 sub string_flip {
-  oct '0b'.sprintf('%b',$_[0])=~tr/01/10/r;
+  $_[0] ? oct '0b'.sprintf('%b',$_[0])=~tr/01/10/r : 0
 }
 
 sub binary_flip {
   my($r,$k,$n) = (0,1,shift);
   $r|=(~$n&1)<<$k++, $n>>=1 while $n;
-  $r;
+  $r
 }
 
 __END__
@@ -36,3 +37,14 @@ int c_flip(int n) {
   }
   return r;
 }
+
+int c2_flip(int n) {
+  int o = n;
+  int m = 0;
+  while(o) {
+    m<<=1;
+    m++;
+    o>>=1;
+  }
+  return n^m;
+}