This is the mail archive of the gcc-patches@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: Fix for df problems, and and an ifcvt improvement


On 04/22/2010 12:25 PM, Paolo Bonzini wrote:

> I fixed it to at least make sense, but I didn't recognize the
> possibility to ignore clobbers (note that for
> df_simulate_one_insn_forwards it won't usually matter to ignore partial
> and conditional sets, because those should already be live).

This isn't always true.  While working on regrename a while ago I was
reminded that there are sequences like
 (set (zero_extract of upper 16 bits) (something))
 (set (zero_extract of lower 16 bits) (something else))

The same happens (more frequently actually) with cond_exec:
 (cond_exec (condition A) (set (reg X) (Y)))
 (cond_exec (reverse-condition A) (set (reg X) (Z)))

> Patch is okay, thanks.

Thanks!


Bernd


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