From d85a04b3df05ae8a6bcbc770eab5fc4501155c17 Mon Sep 17 00:00:00 2001 From: Hal Finkel Date: Sat, 13 Apr 2013 08:09:20 +0000 Subject: [PATCH] Spill and restore PPC CR registers using the FP when we have one For functions that need to spill CRs, and have dynamic stack allocations, the value of the SP during the restore is not what it was during the save, and so we need to use the FP in these cases (as for all of the other spills and restores, but the CR restore has a special code path because its reserved slot, like the link register, is specified directly relative to the adjusted SP). llvm-svn: 179457 --- llvm/lib/Target/PowerPC/PPCFrameLowering.cpp | 20 ++++++++++++++------ llvm/test/CodeGen/PowerPC/crsave.ll | 15 ++++++++++++++- 2 files changed, 28 insertions(+), 7 deletions(-) diff --git a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp index 3244b904ee64..8541e1ba5cd8 100644 --- a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp +++ b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp @@ -1141,12 +1141,14 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, // save slot via GPR12 (available in the prolog for 32- and 64-bit). if (Subtarget.isPPC64()) { // 64-bit: SP+8 + bool is31 = needsFP(*MF); + unsigned FP8Reg = is31 ? PPC::X31 : PPC::X1; MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::MFCR8), PPC::X12)); MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::STW8)) .addReg(PPC::X12, getKillRegState(true)) .addImm(8) - .addReg(PPC::X1)); + .addReg(FP8Reg)); } else { // 32-bit: FP-relative. Note that we made sure CR2-CR4 all have // the same frame index in PPCRegisterInfo::hasReservedSpillSlot. @@ -1170,7 +1172,8 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, } static void -restoreCRs(bool isPPC64, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, +restoreCRs(bool isPPC64, bool is31, + bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, const std::vector &CSI, unsigned CSIIndex) { @@ -1182,9 +1185,10 @@ restoreCRs(bool isPPC64, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, if (isPPC64) { // 64-bit: SP+8 + unsigned FP8Reg = is31 ? PPC::X31 : PPC::X1; MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::LWZ8), PPC::X12) .addImm(8) - .addReg(PPC::X1)); + .addReg(FP8Reg)); RestoreOp = PPC::MTCRF8; MoveReg = PPC::X12; } else { @@ -1297,7 +1301,9 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, // least one CR register, restore all spilled CRs together. if ((CR2Spilled || CR3Spilled || CR4Spilled) && !(PPC::CR2 <= Reg && Reg <= PPC::CR4)) { - restoreCRs(Subtarget.isPPC64(), CR2Spilled, CR3Spilled, CR4Spilled, + bool is31 = needsFP(*MF); + restoreCRs(Subtarget.isPPC64(), is31, + CR2Spilled, CR3Spilled, CR4Spilled, MBB, I, CSI, CSIIndex); CR2Spilled = CR3Spilled = CR4Spilled = false; } @@ -1320,9 +1326,11 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, } // If we haven't yet spilled the CRs, do so now. - if (CR2Spilled || CR3Spilled || CR4Spilled) - restoreCRs(Subtarget.isPPC64(), CR2Spilled, CR3Spilled, CR4Spilled, + if (CR2Spilled || CR3Spilled || CR4Spilled) { + bool is31 = needsFP(*MF); + restoreCRs(Subtarget.isPPC64(), is31, CR2Spilled, CR3Spilled, CR4Spilled, MBB, I, CSI, CSIIndex); + } return true; } diff --git a/llvm/test/CodeGen/PowerPC/crsave.ll b/llvm/test/CodeGen/PowerPC/crsave.ll index 3e98dbd254d9..109e595900a6 100644 --- a/llvm/test/CodeGen/PowerPC/crsave.ll +++ b/llvm/test/CodeGen/PowerPC/crsave.ll @@ -1,5 +1,6 @@ ; RUN: llc -O0 -disable-fp-elim -mtriple=powerpc-unknown-linux-gnu < %s | FileCheck %s -check-prefix=PPC32 -; RUN: llc -O0 -disable-fp-elim -mtriple=powerpc64-unknown-linux-gnu < %s | FileCheck %s -check-prefix=PPC64 +; RUN: llc -O0 -disable-fp-elim -mtriple=powerpc64-unknown-linux-gnu < %s | FileCheck %s -check-prefix=PPC64-FP +; RUN: llc -O0 -mtriple=powerpc64-unknown-linux-gnu < %s | FileCheck %s -check-prefix=PPC64 declare void @foo() @@ -23,6 +24,11 @@ entry: ; PPC64: lwz 12, 8(1) ; PPC64-NEXT: mtcrf 32, 12 +; PPC64-FP: mfcr 12 +; PPC64-FP-NEXT: stw 12, 8(31) +; PPC64-FP: lwz 12, 8(31) +; PPC64-FP-NEXT: mtcrf 32, 12 + define i32 @test_cr234() nounwind { entry: %ret = alloca i32, align 4 @@ -47,3 +53,10 @@ entry: ; PPC64-NEXT: mtcrf 16, 12 ; PPC64-NEXT: mtcrf 8, 12 +; PPC64-FP: mfcr 12 +; PPC64-FP-NEXT: stw 12, 8(31) +; PPC64-FP: lwz 12, 8(31) +; PPC64-FP-NEXT: mtcrf 32, 12 +; PPC64-FP-NEXT: mtcrf 16, 12 +; PPC64-FP-NEXT: mtcrf 8, 12 + -- 2.34.1