search for: tailjmp

tailcall optimization stop working in jit (lli) in revision 45527. i guess this is because the jit is tailjmping to the wrong function address. the following change would reenable tailcallopt in jit. But i am pretty sure that this is not the correct fix (since i don't really understand what is going on :). maybe evan can comment on this? regards arnold Index: lib/Target/X86/X86CodeEmitter.cpp =...

...tructions: // Assume undefined functions may be outside the Small codespace. bool NeedStub = (Is64BitMode && (TM.getCodeModel() == CodeModel::Large || TM.getSubtarget<X86Subtarget>().isTargetDarwin())) || Opcode == X86::TAILJMPd; emitGlobalAddress(MO.getGlobal(), X86::reloc_pcrel_word, MO.getOffset(), 0, NeedStub); This causes every external call to be emitted as a call to a stub which then jumps to the real function. I understand, thanks to the helpful folks on #llvm, that calls across...

...me undefined functions may be outside the Small > codespace. > bool NeedStub = > (Is64BitMode && > (TM.getCodeModel() == CodeModel::Large || > TM.getSubtarget<X86Subtarget>().isTargetDarwin())) || > Opcode == X86::TAILJMPd; > emitGlobalAddress(MO.getGlobal(), X86::reloc_pcrel_word, > MO.getOffset(), 0, NeedStub); > > This causes every external call to be emitted as a call to a stub > which then jumps to the real function. > I understand, thanks to the helpful folks o...

...ined functions may be outside the Small codespace. >>       bool NeedStub = >>         (Is64BitMode && >>             (TM.getCodeModel() == CodeModel::Large || >>              TM.getSubtarget<X86Subtarget>().isTargetDarwin())) || >>         Opcode == X86::TAILJMPd; >>       emitGlobalAddress(MO.getGlobal(), X86::reloc_pcrel_word, >>                         MO.getOffset(), 0, NeedStub); >> >> This causes every external call to be emitted as a call to a stub >> which then jumps to the real function. >> I understand, thanks...

...call clobbered registers to the eventual CALL MachineInstr? > > The simplest way would probably be to add separate X86ISD opcodes for > each desired set of call-clobbered registers. The problem is the large number of call-like instructions. We would need copies of CALL*, TCRETURN*, and TAILJMP* for each calling convention. I was hoping we could avoid that, and even get rid of the WINCALL instructions. What if InstrEmitter::EmitMachineNode called a target hook to add call-clobbered registers when II.isCall()? We would need some way of communicating the calling convention to the target ho...

Great! Are there other issues I should be aware of? Félix Le 2010-11-25 à 07:43:23, Anton Korobeynikov a écrit : > Hello Felix > >> I have a project in mind that involves using the JIT for a few targets (x86 >> and x86_64 processors on Mac OS, Linux and Windows). However, there is an >> open bug that says LLVM generates incorrect code for Win64. >> Eli's last

Felix, > Great! Are there other issues I should be aware of? There might be some issues wrt tailcalls, but I'd not expect anything else to be major broken. -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University

...lid after the callee saved registers have been restored? (some X86InstrInfo.td foo magic maybe ?) Or do i have to insert code into PEI::saveCalleeSavedRegisters to detect that there is a tail called function that uses a callee saved register and move it to another (EAX). that's how the TAILJMP instruction looks: let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1, noResults = 1 in def TAILJMPr : I<0xFF, MRM4r, (ops GR32:$dst), "jmp {*}$dst # TAIL CALL jmpr", []>; btw. i am currently working with llvm-2.0 codebase. thanks for any help...

On Oct 19, 2010, at 8:00 PM, Jakob Stoklund Olesen wrote: > > > One problem is that calling conventions are handled while building the selection DAG, and the DAG doesn't really know to represent clobbered registers. > > Perhaps X86TargetLowering::LowerCall() could decorate the X86ISD::CALL node with the calling convention somehow? > > Dan, do you have any thoughts on

Hello, I am trying to insert a .byte/.word in the beginning of a specific LLVM IR instruction when it prints out in assembly (the inserted ‘instruction' only appears in assembly, not in LLVM IR), and I am guessing the best way to do that is to insert it in SelectionDAG as it strips down some LLVM IR instructions when it’s lowered. Can I get some guidance on what function I should use to insert