🚀 go-pugleaf

I want to do the following many times:

   x = f * y;

Currently x and y are integers and f is floating point number.
In an attempt to speed up a program I tried replacing this with

   x = (a * y)/b;

where a and b are integers.  After compiling a loop that has 100,000
iterations of the code described above I timed the two different
versions.

To my surprise I found that the floating point version returned
quickly - it used about 1.6 seconds of user time - while the integer
version took much longer- about 43 seconds.

After looking at the code with adb it was noticed that the integer
multiply and divide were calls to routines like .mul but were using
the dynamic loading stubs.  We relinked the code using -Bstatic and
running time for the integer version went down to about 10 seconds.
However this still much slower than the floating point version.

What is going on here?  Intuitively, an integer multiply followed by
an integer divide should be faster than a convert int to float,
floating multiply, and convert float to int but this does not seem to
be the case.

Any suggestions on a way to increase performance?  

Is it possible to inline the calls to the library routines?

Do the new sparc chips have integer multiply and divide instructions?

Thanks,


John Shewchuk                                     jps@cs.brown.edu



John Shewchuk                                                jps@cs.brown.edu