How do people map high-level info down to lower-levels?

Joe Hummel <jhummel@cy4.ICS.UCI.EDU>
Wed, 12 Oct 1994 23:20:57 GMT

          From comp.compilers

Related articles
How do people map high-level info down to lower-levels? jhummel@cy4.ICS.UCI.EDU (Joe Hummel) (1994-10-12)
| List of all articles for this month |

Newsgroups: comp.compilers
From: Joe Hummel <jhummel@cy4.ICS.UCI.EDU>
Keywords: optimize, question
Organization: UC Irvine, Department of ICS
Date: Wed, 12 Oct 1994 23:20:57 GMT

Hi all. I'm curious about something, and I'm hoping people can share
what approaches they take; I'll post a summary if people are interested.


When compiler X does analysis at the source-code level, is the info it
collects passed down to the instruction level? If so, how are people
going about this? An obvious goal is to use the info to do better
instruction scheduling. Consider the following, with source code on the
left and the corresponding instructions on the right:


                S1: *a = ... ; ===> I1: store ...,(r1)
                ... ...
                S2: *b = ... ; I2: store ...,(r2)


Suppose the compiler is able to determine at the source-level that
a and b refer to different, non-overlapping memory locations. Is this
info also available at the instruction level? Is the solution simply
to map the instruction back to its original source statement? What
if the instruction is moved?


Thanks in advance for whatever info you can share,


    - joe
--
Joe Hummel
ICS Graduate Student, UC Irvine
Internet: jhummel@ics.uci.edu
--


Post a followup to this message

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