[llvm-commits] SSI patch

Andre Tavares andrelct at dcc.ufmg.br
Mon Jun 29 12:39:55 PDT 2009

Eli Friedman wrote:
> On Fri, Jun 26, 2009 at 6:34 AM, Andre Tavares<andrelct at dcc.ufmg.br> wrote:
>> Changed them all from <vector> to SmallVector, and so on. I need
>> SSI::created to have a record of what variables I have transformed to
>> SSI, and if another pass asks to transform it, I will be able to not do
>> anything because I already know it was transformed before. Without this
>> I would have to try to transform it and see that it is not possible.
> Is attempting to transform a variable that was already transformed
> expensive?  If not, I don't see the point of caching it.
It is not that expensive. It would run for every use of that variable 
and return. The problem is that I'm not sure that my insertion of sigma 
and phi is minimal. If it is not minimal it would include lots of sigma 
for the variable it is asked multiple times. I have to verify that my 
creation of sigma and phi is minimal before this.

>> If we decide to make it an utility class, should I put all content in a
>> .h file and forget about the .cpp?
> Any functions with an implementation longer than a couple of lines
> should go into a .cpp file; this reduces compile times and code size,
> and keeps the header cleaner.

> -Eli
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

Andre Tavares
Master Student in Computer Science - UFMG - Brasil

More information about the llvm-commits mailing list