[Octave-bug-tracker] [bug #51871] loading '-ascii' format files is slow

[Dan Sebald]

Follow-up Comment #6, bug #51871 (project octave):

OK, nice work.  A bit much for me to follow right now, but I just wanted to
note some coding nuances.  In this hunk:


+      // Error during conversion.
+      if (str_chunk.size () == 2 &&
+          std::tolower (str_chunk[0]) == 'n' &&
+          std::tolower (str_chunk[1]) == 'a')
+        {
+          // It is Octave NA.
+          val = octave::numeric_limits<T>::NA ();
+        }


would it be more efficient (since speed for large data is important) to not
use std::tolower(), which I assume is a function call (is it a library
function, or does the C++ header for the library know to use an inline?).  The
function call will take extra time for the stack/jump/return.  The ::tolower
has to do some tests for range and then subtract a value (unless it is going
to use some internal 2^8-length lookup table for speed).

Plus, I'm wondering if the str_chunk.size() is extraneous because what affects
the size value is str_chunk[0] == '\0', str_chunk[1] == '\0' or str_chunk[2]
== '\0', which is effectively checked by testing str_chunk[0] being 'n' or
'N'.  In other words, the following is equivalent:


+      // Error during conversion.
+      if ((str_chunk[0]) == 'n' || str_chunk[0] == 'N')) &&
+          (str_chunk[1]) == 'a' || str_chunk[1] == 'A')) &&
+          str_chunk[2] == '\0')
+        {
+          // It is Octave NA.
+          val = octave::numeric_limits<T>::NA ();
+        }



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?51871>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/