[trunk] patch to clean up simulation of insns in df.

Kenneth Zadeck zadeck@naturalbridge.com
Fri May 16 15:27:00 GMT 2008


Paolo Bonzini wrote:
>> I renamed df_simulate_artificial_refs_at_end to 
>> df_simulate_artificial_refs_at_end_backwards and renamed 
>> df_simulate_artificial_refs_at_top to 
>> df_simulate_artificial_refs_at_top_backwards.
>
> The idea is obviously okay.  However I'd rather avoid the _backwards 
> suffix since right now we only support backwards scanning.  I mean, if 
> the suffix had been there before, I'd have removed it together with 
> df_simulate_one_insn_forwards.    The comment you added tells that 
> df_simulate_artificial_refs_at_* is okay only for backwards scanning.
>
> In practice, this would reduce the patch to these hunks only:
>
>    (df_simulate_artificial_refs_at_top):
>    Reversed scanning of defs and uses.
>    (df_simulate_one_insn_forwards): Removed.
>
> which would have been a tad simpler to review too.
>
> Paolo
Actually your suggestion does not reduce the size of the patch because 
under your plan i would still have to rename 
df_simulate_one_insn_backwards to df_simulate_one_insn.  (actually it 
does reduce the size of the patch if you count characters since the 
names are shorter for this call but that hardly matters.) 

I will do what ever you want.   Also, I guess i need to look at vlads 
code to make sure he was not using the forwards version.  

kenny



More information about the Gcc-patches mailing list