search for: eliminatecallframepseudoinstr

Hi all, I ran into this while trying to figure out why the X86 getSUBriOpcode/getADDriOpcode functions are duplicated, appearing once in X86RegisterInfo.cpp and once in X86FrameLowering.cpp, The method TargetRegisterInfo::eliminateCallFramePseudoInstr doesn't appear to really belong in this interface. It adds instructions into the MachineFunction given to it, which isn't what TargetRegisterInfo is supposed to do. ISTM that eliminateCallFramePseudoInstr belongs in TargerFrameLowering, since it's being used during prolog/epilog insert...

> ISTM that eliminateCallFramePseudoInstr belongs in > TargerFrameLowering, since it's being used during prolog/epilog > insertion. Moving it there would avoid the code duplication and > possibly other layering problems. > What do you think Go ahead and move. It's s historical artifact why it is inside TRI. -- With be...

Heads up for out-of-tree targets: in r265036, TargetFrameLowering::eliminateCallFramePseudoInstr changed to return an iterator to the next instruction, instead of returning void. If your target was previously doing "MBB.erase(I); return;" your code should now be doing "return MBB.erase(I);". Thanks, Hans

For a Z80 backend, "eliminateCallFramePseudoInstr()" shall adjust the stack pointer in three possible ways, e.g. after a function call, depending on the amount (= adjustment size) *and some other rules*: 1. via one or more target "pop <reg>" instructions (SP increments +2 per instruction), using an unused reg (disregarding...

.... At these point, they have a constant arguments which corresponds to the fixed frame size for argument passing. But the size of the stack frame isn't finalized until frame layout has been completed, these are not translated into stack dec / inc until very late (see X86RegisterInfo.cpp::eliminateCallFramePseudoInstr). May I ask why are you adding a command line option to control stack frame size? If you must do it, it should be separate from the general mechanism. You can add a couple of special instructions that are "flagged" to the CALLSEQ nodes that will be emitted as stack inc / dec code...

Hello, I am trying to add an instruction before each function call to add/ subtract the stack pointer by a value specified at the command line. I wonder if I can do that during lowering. For example, in X86TargetLowering::LowerCALL. I appreciate it if you give me some hints how and where I can do that. Thank you, Babak

...Chain = DAG.getNode(ISD::CALLSEQ_END, NodeTys, &Ops[0], Ops.size()); InFlag = Chain.getValue(1); The third operand is the number of bytes the callee pops of the stack on return (on x86). This gets lowered to a ADJCALLSTACKUP pseudo machineinstruction. Later when X86RegisterInfo::eliminateCallFramePseudoInstr is called and framepointerelimination is enabled the following code gets called: ... else if (I->getOpcode() == X86::ADJCALLSTACKUP) { // If we are performing frame pointer elimination and if the callee pops // something off the stack pointer, add it back. We do this until we...

Comments: CheckDAGForTailCallsAndFixThem - 1. for (SelectionDAG::allnodes_iterator BE = DAG.allnodes_begin(), + BI = prior(DAG.allnodes_end()); BI != BE; BI--) { Please use pre-decrement instead of post-decrement. 2. The function is slower than it should be. You are scanning all the nodes in the DAG twice. You should just examine DAG.getRoot() to make determine whether it's a

..._END, NodeTys, &Ops[0], Ops.size > ()); > InFlag = Chain.getValue(1); > > The third operand is the number of bytes the callee pops of the > stack on return (on x86). This gets lowered to a ADJCALLSTACKUP > pseudo machineinstruction. > > Later when X86RegisterInfo::eliminateCallFramePseudoInstr is called > and framepointerelimination is enabled the following code gets called: > ... > else if (I->getOpcode() == X86::ADJCALLSTACKUP) { > // If we are performing frame pointer elimination and if the > callee pops > // something off the stack pointer, add it ba...

....size >> ()); >> InFlag = Chain.getValue(1); >> >> The third operand is the number of bytes the callee pops of the >> stack on return (on x86). This gets lowered to a ADJCALLSTACKUP >> pseudo machineinstruction. >> >> Later when X86RegisterInfo::eliminateCallFramePseudoInstr is called >> and framepointerelimination is enabled the following code gets >> called: >> ... >> else if (I->getOpcode() == X86::ADJCALLSTACKUP) { >> // If we are performing frame pointer elimination and if the >> callee pops >> // somethi...

...ed to track the SP adjustment for frame index elimination. if (TFI->canSimplifyCallFramePseudos(Fn)) => TFI->eliminateCallFramePseudoInstr(Fn, *I->getParent(), I); Perhaps there is a bug in canSimplifyCallFramePseudos? Evan On Sep 26, 2013, at 12:00 PM, Krzysztof Parzyszek <kparzysz at codeaurora.org> wrote: > Consider this example: > > --- ex.ll --- > declare void @bar() > > ; Function Attrs: nounwind...

.../ the target doesn't indicate otherwise, remove the call frame pseudos > // here. The sub/add sp instruction pairs are still inserted, but we don't > // need to track the SP adjustment for frame index elimination. > if (TFI->canSimplifyCallFramePseudos(Fn)) > => TFI->eliminateCallFramePseudoInstr(Fn, *I->getParent(), I); > > Perhaps there is a bug in canSimplifyCallFramePseudos? > > Evan > > On Sep 26, 2013, at 12:00 PM, Krzysztof Parzyszek > <kparzysz at codeaurora.org <mailto:kparzysz at codeaurora.org>> wrote: > >> Consider this example: >&...

Consider this example: --- ex.ll --- declare void @bar() ; Function Attrs: nounwind optsize define void @main() { entry: %hin = alloca [256 x i32], align 4 %xin = alloca [256 x i32], align 4 call void @bar() ret void } ------------- Freshly built llc: llc -O2 -march=x86 < ex.ll -print-before-all # *** IR Dump Before Prologue/Epilogue Insertion & Frame Finalization ***: #