This is the mail archive of the mailing list for the binutils 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: [PATCH] enable fdpic targets/emulations for sh*-*-linux*

On Sun, Oct 04, 2015 at 03:42:20AM -0500, Segher Boessenkool wrote:
> On Sun, Oct 04, 2015 at 12:32:54PM +0900, Oleg Endo wrote:
> > > I think we should focus on the big performance problem that would make
> > > a much much bigger difference: very bad codegen by gcc. Aside from
> > > lack of shrink-wrapping, poor hanling of the PIC register (like the
> > > way x86 used to handle %ebx, as permanently-reserved and unusable)
> > > stands out at something that needs to be fixed.
> > 
> > Shrink wrapping is being done for all backends.  In fact, it has been
> > improved recently.  Of course there could be some SH specific cases
> > which aren't optimized well.  Please open PRs in GCC's bugzilla for
> > issues with the compiler.
> If the prologue sets up a register that is used pretty much everywhere
> (like a PIC register) we cannot shrink-wrap much.  After shrink-wrapping
> you always still have a single prologue (and epilogue).
> Your problem might be different; I can't tell without examples.

Indeed, this is my concern. Shrink-wrapping cannot be effective until
all hard-coded prologue/epilogue logic (and fixed register use for GOT
registers, frame pointers, etc.) is eliminated and replaced by
representing it as proper patterns/constraints that the compiler can
manipulate. Right now we have prologue logic like "if anything in the
function uses the GOT, emit GOT register setup in the prologue".


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