new times

Suresh Jagannathan suresh@research.nj.nec.com
Fri, 26 Feb 1999 18:13:01 -0500


These are new times of the recursive inliner running with polyvariance turned
on.  Unfortunately, above threshold 0 (i.e., 5 rounds of 0), the c compiler
seems to blow up on mlyacc.  The code space increase over the current inliner
appears to be modest (< 30%) on most of the benchmarks.  Running times are
comparable to the current inliner except for lexgen (31.3 vs. 47) and nucleic
(13.7 vs 21.5),


Times.
-------
                             0

            fft           36.3

     barnes-hut           12.7

matrix-multiply           13.9

   knuth-bendix           34.8

         lexgen           31.3

   count-graphs           18.0

           life           56.5

          logic           52.1

     mandelbrot           16.3

         mlyacc           23.1

        nucleic           13.7

         simple           13.0

	    tsp           29.0

Code space
------------   

-rwxr-xr-x   1 suresh   CSI         38753 Feb 26 17:43 fft
-rwxr-xr-x   1 suresh   CSI         58986 Feb 26 17:44 barnes-hut
-rwxr-xr-x   1 suresh   CSI         20601 Feb 26 17:44 matrix-multiply
-rwxr-xr-x   1 suresh   CSI         80692 Feb 26 17:45 knuth-bendix
-rwxr-xr-x   1 suresh   CSI        243279 Feb 26 17:46 lexgen
-rwxr-xr-x   1 suresh   CSI         66221 Feb 26 17:47 count-graphs
-rwxr-xr-x   1 suresh   CSI         49573 Feb 26 17:47 life
-rwxr-xr-x   1 suresh   CSI        167743 Feb 26 17:49 logic
-rwxr-xr-x   1 suresh   CSI         20039 Feb 26 17:50 mandelbrot
-rwxr-xr-x   1 suresh   CSI        830501 Feb 26 17:59 mlyacc
-rwxr-xr-x   1 suresh   CSI         93280 Feb 26 18:00 nucleic
-rwxr-xr-x   1 suresh   CSI        411564 Feb 26 18:06 simple
-rwxr-xr-x   1 suresh   CSI         42835 Feb 26 18:07 tsp