This is the mail archive of the 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: [lto] make builtins go

I don't think my suggestion prevents optimizing (2).  :-)   In the (2)
case __builtin_sin will be serialized out, read in by LTO, and optimize.

That makes sense. I didn't know that we transform sin to "__builtin_sin" even at "-O0".

We do not generally convert calls to sin() to calls to __builtin_sin(), but if the middle-end introduces new calls to sin it generally uses the builtin variant. We would need to start doing so instead, to make Paolos suggestion work - which I think would be a sensible thing to do to address this problem.

What I meant was only happening at the boundary between LTO-write and LTO-read: we can skip optimization on "sin" at LTO time, and not lose anything for translation units not compiled with -fno-builtin-xyz, if we serialize every "sin" as "__builtin_sin".

But yes, it would make sense to do this translation much earlier than at serialization time, possibly even when building GENERIC. Whether to do that in mainline and import it into LTO, or the other way round, I don't know; but if you guys are thinking of doing it in LTO only, well, for sure it's a more compelling change than adopting a new representation for CALL_EXPRs. :-)


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