Re: Compiler support for a faster interrupt response

Ji Hoon Yoon stated:


      Whenever interrupt occurs, it is basically true that
      all global registers should be saved in the interrupt
      prolog and restored in the epilog.


      But in the case of 29K processor, nearly 40 ~ 50 registers
      should be saved and restored at each interrupt, which, of course,
      increses the interrupt response time.


      Thus I want to survey various existing techniques for conquering this
      heavy register saving/restoring problem.
      Would you help me?


I suggest you look at the article:


%A J. Snyder
%A D. B. Whalley
%A T. P. Baker
%T Fast Context Switches: Compiler and Architectural Support for Preemptive Scheduling
%J Microprocessors and Microsystems
%V 19
%N 1
%D February 1995
%P 35-42


We reduced the number of registers that would require saving and restoring
at context switch points. You can access the postscript for this paper
at either ftp.cs.fsu.edu:/pub/whalley/papers/mnm95.ps.Z or in
http://www.cs.fsu.edu/~whalley.


Dave
--

Return to the comp.compilers page.
Search the comp.compilers archives again.