Re: And speaking of fast compilers...

sasdrf@unx.sas.com (Dave Frederick) writes:
>[Had the Brand X Ada compiler passed validation?
> A lot of optimizations are compile-time range and bounds checks.]


I do not know if the Brand X compiler was validated. As the moderator
points out, compiling to C does not imply any loss of type safety (for
that matter, the first validated Ada system was an interpreter...), so it
may well be a validated compiler.


The IBM PL.8 compiler removed a substantial number of the runtime range
and bound checks with only a modest increase in the compile time (the
range-checked code was still 5-10% slower than unchecked code). So I
doubt that removing bound checks explains why the compiler was slow,
especially given that the compilation was slow (40 minutes v. 6) even with
the slower compiler's optimization OFF.


%A Marc Auslander
%A Martin Hopkins
%T An Overview of the PL.8 Compiler
%P 22-31
%J Proceedings of the ACM SIGPLAN '82 Syumposium on Compiler
Construction
%D June 1982
%P 22-31


%A Rajiv Gupta
%T A Fresh Look at Optimizing Array Bound Checking
%J Proceedings of the ACM SIGPLAN '90 conference on Programming
Language Design and Implementation.
%D June 1990
%P 272


And I haven't read it, but it's the common citation:


%A V. Markstein
%A J. Cocke
%A P. Markstein
%T Optimization of Range Checking
%J Proceedings of the ACM SIGPLAN '82 Syumposium on Compiler
Construction


;-D oN ( Bound checked and gaggled ) Pardo
[The PL.8 range checking was quite innovative. They patented it. -John]
--

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