Joseph J. Strout wrote:
Nope. Not a public list, anyway. Out of curiosity, what would you do
with such a list?
A list of optimized functions and standard optimistic compilation
choices should allow developers to implement the most effective program
flow without hours of testing.
If, for example, we know it is faster to convert data to strings before
populating a listbox, it is one less thing to test. If we know how
optimistic the "IF" branching is, we can decide how to implement
conditionals best. Etc, etc...
A question was asked about "OR" and optimistic compilers. In those
cases, I believe the parser inverts the statements in some compilers. It
really depends on how advanced the compiler design was and at how
"abstracted" a level the compiler was developed. If someone is using
lexx/bison or similar tools, I suppose you could get pretty good with
the code. If, however, you are using C++ directly, you are already a bit
removed from some of the decisions and not creating a "low-level" (or
"true") compiler.
I spent part of last summer trying to see how to implement a port of a
compiler to OS X. In the end, I decided it wasn't worth the effort with
all that I have on my plate. Still, experimenting with the examples in
the text I was using (which created a small Pascal compiler) reminded me
of all the issues with optimization decisions.
(I developed my "language" and even my database file format. I wanted
the language to include full SQL support, with something like "BEGIN
SQL.... END;" and most of the functions were matched to SQL functions --
think Oracle or PostgreSQL's external languages. The goal was a
data-centric language with lots of statistical functions.)
If you optimize for one type of IF, you slow the others slightly. Every
choice affects the best way to use your compiler.
Tell us what is fastest and we will use it. That's all...
- Scott
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
|