CFG

CFG

[Marcio de Oliveira Buss]

	Good Morning (afternoon) to all...
        Could you please help me? I want to know if the CFG constructed
 in the dataflow analysis is updated at each step subsequent steps of
 the compilation, and, in turn, remains until the Final phase of the
 compilation process. I want to transform the CFG into another structure
 but only after all the optimizations have been made over it. I know
 that the almost all optimizations are made over the RTL expressions,
 but I don't know if some optimization touchs the CFG. 
        In other words, I would like to know if the CFG is destructed        
 after the dataflow analysis, i.e., after the informations provided
 by the CFG are recorded for futures uses, or if the CFG remains        
 untouchable.

        Thanks.       

        Best Regards,

        Marcio.

Re: CFG

[Jeffrey A Law]

  In message < Pine.GSO.4.10.10006281746080.16817-100000@xingu.dcc.unicamp.br >yo
u write:
  > 
  > 
  > 	Good Morning (afternoon) to all...
  >         Could you please help me? I want to know if the CFG constructed
  >  in the dataflow analysis is updated at each step subsequent steps of
  >  the compilation, and, in turn, remains until the Final phase of the
  >  compilation process. I want to transform the CFG into another structure
  >  but only after all the optimizations have been made over it. I know
  >  that the almost all optimizations are made over the RTL expressions,
  >  but I don't know if some optimization touchs the CFG. 
  >         In other words, I would like to know if the CFG is destructed      
  >   
  >  after the dataflow analysis, i.e., after the informations provided
  >  by the CFG are recorded for futures uses, or if the CFG remains        
  >  untouchable.
Most of the time we [re]build the cfg when we need it (ie, we don't assume
it's valid, except in a few special cases).  Most of the optimizers are
not using the cfg right now and do not keep it accurate.
jeff

Re: CFG

[Marcio de Oliveira Buss]

	Thus, is there an answer for this question?
	"what is the compilation point where I could take the most
 optimized CFG for a function to build another structure based on that
 CFG?" And which is the file representing this point.

	Thanks.


On Wed, 28 Jun 2000, Jeffrey A Law wrote:

>   >         Could you please help me? I want to know if the CFG
>   >  constructed in the dataflow analysis is updated at each step
>   >  subsequent steps of the compilation, and, in turn, remains until
>   >  the Final phase of the compilation process. I want to transform
>   >  the CFG into another structure but only after all the optimizations
>   >  have been made over it. I know that the almost all optimizations
>   >  are made over the RTL expressions, but I don't know if some
>   >  optimization touchs the CFG. 
>   >         In other words, I would like to know if the CFG is
>   >  destructed after the dataflow analysis, i.e., after the
>   >  informations provided by the CFG are recorded for futures uses, or
>   >  if the CFG remains untouchable.

> Most of the time we [re]build the cfg when we need it (ie, we don't assume
> it's valid, except in a few special cases).  Most of the optimizers are
> not using the cfg right now and do not keep it accurate.
> jeff
 

Re: CFG

[Richard Henderson]

On Wed, Jun 28, 2000 at 06:35:50PM -0300, Marcio de Oliveira Buss wrote:
> 	Thus, is there an answer for this question?
> 	"what is the compilation point where I could take the most
>  optimized CFG for a function to build another structure based on that
>  CFG?" And which is the file representing this point.

You can rebuild the CFG at almost any point before delayed-branch
scheduling.  I.e. before dbr_schedule.  Of course, what you have in
mind fror "build another structure" may also influence where things
may occur.


r~