This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: O0 cleanup_cfg vs debuggability


James E Wilson wrote:
> Sorry, I got the two ACT debug problems confused.

 No problem. I'm following the other one too, and we're very happy to get
 such constructive feedback.

> Yours is the one with the missing break, not the missing type/variable debug
> info.

 Right.

> It would be nice if there was a way to distinguish between forwarder 
> blocks emitted by the compiler and ones that represent actual code.  Not 
> clear how to do that though, and I'm not sure if it is even possible. 
> Maybe we can keep track of when a block is created, and one created at 
> RTL gen time represents real code, and one created later represents a 
> compiler generated construct.  That doesn't seem very elegant though.

 Isn't that a bit of overkill just to indirectly influence some optimization
 pass in a non optimizing compilation anyway ? 

> A new CLEANUP_* enum would be a reasonable solution once you figure out 
> how to fix this.

 I can look into this, although that might take some time. Especially if
 you're not sure it's even possible :)

 I'm not clear why a mere inhibition of the edge forwarding optimization in
 non-optimizing compilations doesn't seem to be an option.

 Could you please enlighten ?

 Thanks again for your feedback,

 Olivier

 
 
 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]