My research in runtime parallelization has touched on both the inspector and the executor.  I have proposed two ways to speed up the inspector.  This work appeared in the Fourth ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming.  The paper is also available as technical report <A HREF="ftp://ftp.cs.washington.edu/tr/1992/10/UW-CSE-92-10-05.PS.Z"> 92-12-05</A>.  I have also studied various forms of the executor to improve its performance and extend its applicability to complex dependence patterns.  (This research is reported in technical report  <A HREF="ftp://ftp.cs.washington.edu/tr/1992/10/UW-CSE-95-01-08.PS.Z"> 95-01-08</A>.)  My experiments on the KSR1, a shared address space multiprocessor, show that false sharing and poor spatial locality could seriously degrade executor performance.  I have proposed and experimentally evaluated two simple techniques to address these problems by restructuring arrays according to the parallel execution schedule.  (This research is reported in technical report  <A HREF="ftp://ftp.cs.washington.edu/tr/1992/10/UW-CSE-94-02-01.PS.Z"> 94-02-01</A>.)  <!-- Standard footer-->  <HR> <ADDRESS>        <A HREF="http://www.cs.washington.edu/homes/shuntak/"> 	Shun-Tak A. Leung</A><BR>
