[llvm] r271066 - Start using shouldAssumeDSOLocal on ARM.

[Steven Wu via llvm-commits]

Hi Rafael

I run into a problem for this commit. See comment below.

> On May 27, 2016, at 3:41 PM, Rafael Espindola via llvm-commits <llvm-commits at lists.llvm.org> wrote:
> 
> Author: rafael
> Date: Fri May 27 17:41:51 2016
> New Revision: 271066
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=271066&view=rev
> Log:
> Start using shouldAssumeDSOLocal on ARM.
> 
> Given where this is used it should be a nop.
> 
> Modified:
>    llvm/trunk/lib/Target/ARM/ARMSubtarget.cpp
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMSubtarget.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMSubtarget.cpp?rev=271066&r1=271065&r2=271066&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMSubtarget.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/ARMSubtarget.cpp Fri May 27 17:41:51 2016
> @@ -22,6 +22,7 @@
> #include "Thumb1FrameLowering.h"
> #include "Thumb1InstrInfo.h"
> #include "Thumb2InstrInfo.h"
> +#include "llvm/CodeGen/Analysis.h"
> #include "llvm/CodeGen/MachineRegisterInfo.h"
> #include "llvm/IR/Attributes.h"
> #include "llvm/IR/Function.h"
> @@ -273,40 +274,19 @@ bool ARMSubtarget::isAAPCS16_ABI() const
>   return TM.TargetABI == ARMBaseTargetMachine::ARM_ABI_AAPCS16;
> }
> 
> -
> -/// GVIsIndirectSymbol - true if the GV will be accessed via an indirect symbol.
> +/// true if the GV will be accessed via an indirect symbol.
> bool
> ARMSubtarget::GVIsIndirectSymbol(const GlobalValue *GV,
>                                  Reloc::Model RelocM) const {
> -  if (RelocM == Reloc::Static)
> -    return false;
> -
> -  bool isDef = GV->isStrongDefinitionForLinker();
> +  if (!shouldAssumeDSOLocal(RelocM, TargetTriple, *GV->getParent(), GV))
> +    return true;
> 
> -  if (!isTargetMachO()) {
> -    // Extra load is needed for all externally visible.
> -    if (GV->hasLocalLinkage() || GV->hasHiddenVisibility())
> -      return false;
> +  // 32 bit macho has no relocation for a-b if a is undefined, even if b is in
> +  // the section that is being relocated. This means we have to use o load even
> +  // for GVs that are known to be local to the dso.
> +  if (isTargetDarwin() && RelocM == Reloc::PIC_ &&
> +      (GV->isDeclarationForLinker() || GV->hasCommonLinkage()))
Is there any specific reason using isTargetDarwin() instead of isTargetMachO()? This seems like a general limitation for MachO which is not specific to Darwin. When I compiling embedded architecture for MachO (armv7em), my target is not Darwin but it is still MachO (default triple is thumbv7em-apple-unknown-macho). I will get error like:
fatal error: error in backend: symbol '*' can not be undefined in a subtraction expression

Thanks

Steven

>     return true;
> -  } else {
> -    // If this is a strong reference to a definition, it is definitely not
> -    // through a stub.
> -    if (isDef)
> -      return false;
> -
> -    // Unless we have a symbol with hidden visibility, we have to go through a
> -    // normal $non_lazy_ptr stub because this symbol might be resolved late.
> -    if (!GV->hasHiddenVisibility())  // Non-hidden $non_lazy_ptr reference.
> -      return true;
> -
> -    if (RelocM == Reloc::PIC_) {
> -      // If symbol visibility is hidden, we have a stub for common symbol
> -      // references and external declarations.
> -      if (GV->isDeclarationForLinker() || GV->hasCommonLinkage())
> -        // Hidden $non_lazy_ptr reference.
> -        return true;
> -    }
> -  }
> 
>   return false;
> }
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits