[LoongArch] Stack realignment support

[wangleiat]

Summary

This patch adds support for stack realignment while adding support for
variable sized objects.

[xen0n]

Some ambitious stack alignments there... VLEN=4096 LoongArch core when? ;-)

llvm/lib/Target/LoongArch/LoongArchFrameLowering.cpp

210–215

Use or rd, rj, zero for LoongArch moves.

[wangleiat]

llvm/lib/Target/LoongArch/LoongArchFrameLowering.cpp

210–215

Use or rd, rj, zero for LoongArch moves.

Thanks. I'll fix it right away.

[wangleiat]

Address xen0n's comments.

[xen0n]

Seems good on a cursory look.

[SixWeining]

LGTM except one nit.

llvm/lib/Target/LoongArch/LoongArchFrameLowering.cpp

190

Inline one-time-used variable.

llvm/lib/Target/LoongArch/LoongArchFrameLowering.cpp

if (hasFP(MF)) {
 
     // Emit ".cfi_def_cfa $fp, LoongArchFI->getVarArgsSaveSize()"
     unsigned CFIIndex = MF.addFrameInst(
         MCCFIInstruction::cfiDefCfa(nullptr, RI->getDwarfRegNum(FPReg, true),
                                     LoongArchFI->getVarArgsSaveSize()));
     BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
         .addCFIIndex(CFIIndex)
         .setMIFlag(MachineInstr::FrameSetup);
+
+    // Realign stack.
+    if (RI->hasStackRealignment(MF)) {
+      Align MaxAlignment = MFI.getMaxAlign();
+      unsigned ShiftAmount = Log2(MaxAlignment);
+      Register VR =
+          MF.getRegInfo().createVirtualRegister(&LoongArch::GPRRegClass);
+      BuildMI(MBB, MBBI, DL,
+              TII->get(STI.is64Bit() ? LoongArch::SRLI_D : LoongArch::SRLI_W),
+              VR)
+          .addReg(SPReg)
+          .addImm(ShiftAmount)
+          .setMIFlag(MachineInstr::FrameSetup);
+      BuildMI(MBB, MBBI, DL,
+              TII->get(STI.is64Bit() ? LoongArch::SLLI_D : LoongArch::SLLI_W),
+              SPReg)
+          .addReg(VR)
+          .addImm(ShiftAmount)
+          .setMIFlag(MachineInstr::FrameSetup);
+      // FP will be used to restore the frame in the epilogue, so we need
+      // another base register BP to record SP after re-alignment. SP will
+      // track the current stack after allocating variable sized objects.
+      if (hasBP(MF)) {
+        // move BP, $sp
+        BuildMI(MBB, MBBI, DL, TII->get(LoongArch::OR),
+                LoongArchABI::getBPReg())
+            .addReg(SPReg)
+            .addReg(LoongArch::R0)
+            .setMIFlag(MachineInstr::FrameSetup);
+      }
+    }
   }
 }
 
 void LoongArchFrameLowering::emitEpilogue(MachineFunction &MF,
                                           MachineBasicBlock &MBB) const {
   const LoongArchRegisterInfo *RI = STI.getRegisterInfo();
   MachineFrameInfo &MFI = MF.getFrameInfo();
   auto *LoongArchFI = MF.getInfo<LoongArchMachineFunctionInfo>();
LoongArchFrameLowering::eliminateCallFramePseudoInstr(
   return MBB.erase(MI);
 }
 
 StackOffset LoongArchFrameLowering::getFrameIndexReference(
     const MachineFunction &MF, int FI, Register &FrameReg) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   const TargetRegisterInfo *RI = MF.getSubtarget().getRegisterInfo();
   auto *LoongArchFI = MF.getInfo<LoongArchMachineFunctionInfo>();
+  uint64_t StackSize = MFI.getStackSize();
 
   // Callee-saved registers should be referenced relative to the stack
   // pointer (positive offset), otherwise use the frame pointer (negative
   // offset).
   const auto &CSI = MFI.getCalleeSavedInfo();
   int MinCSFI = 0;
   int MaxCSFI = -1;
   StackOffset Offset =
       StackOffset::getFixed(MFI.getObjectOffset(FI) - getOffsetOfLocalArea() +
                             MFI.getOffsetAdjustment());
 
   if (CSI.size()) {
     MinCSFI = CSI[0].getFrameIdx();
     MaxCSFI = CSI[CSI.size() - 1].getFrameIdx();
   }
 
-  if ((FI >= MinCSFI && FI <= MaxCSFI) || !hasFP(MF)) {
+  if (FI >= MinCSFI && FI <= MaxCSFI) {
     FrameReg = LoongArch::R3;
-    Offset += StackOffset::getFixed(MFI.getStackSize());
+    Offset += StackOffset::getFixed(StackSize);
+  } else if (RI->hasStackRealignment(MF) && !MFI.isFixedObjectIndex(FI)) {
+    // If the stack was realigned, the frame pointer is set in order to allow
+    // SP to be restored, so we need another base register to record the stack
+    // after realignment.
+    FrameReg = hasBP(MF) ? LoongArchABI::getBPReg() : LoongArch::R3;
+    Offset += StackOffset::getFixed(StackSize);
   } else {
     FrameReg = RI->getFrameRegister(MF);
+    if (hasFP(MF))
       Offset += StackOffset::getFixed(LoongArchFI->getVarArgsSaveSize());
+    else
+      Offset += StackOffset::getFixed(StackSize);
   }
 
   return Offset;
 }

llvm/test/CodeGen/LoongArch/stack-realignment-with-variable-sized-objects.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch32 --verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefix=LA32
+; RUN: llc --mtriple=loongarch64 --verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefix=LA64
+
+declare void @callee(i8*, i32*)
+
+define void @caller(i32 %n) {
+; LA32-LABEL: caller:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -64
+; LA32-NEXT:    .cfi_def_cfa_offset 64
+; LA32-NEXT:    st.w $ra, $sp, 60 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 56 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $s8, $sp, 52 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    .cfi_offset 31, -12
+; LA32-NEXT:    addi.w $fp, $sp, 64
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a1, $sp, 6
+; LA32-NEXT:    slli.w $sp, $a1, 6
+; LA32-NEXT:    move $s8, $sp
+; LA32-NEXT:    addi.w $a0, $a0, 15
+; LA32-NEXT:    addi.w $a1, $zero, -16
+; LA32-NEXT:    and $a0, $a0, $a1
+; LA32-NEXT:    sub.w $a0, $sp, $a0
+; LA32-NEXT:    move $sp, $a0
+; LA32-NEXT:    addi.w $a1, $s8, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -64
+; LA32-NEXT:    ld.w $s8, $sp, 52 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $fp, $sp, 56 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 60 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 64
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -64
+; LA64-NEXT:    .cfi_def_cfa_offset 64
+; LA64-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $s8, $sp, 40 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    .cfi_offset 31, -24
+; LA64-NEXT:    addi.d $fp, $sp, 64
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a1, $sp, 6
+; LA64-NEXT:    slli.d $sp, $a1, 6
+; LA64-NEXT:    move $s8, $sp
+; LA64-NEXT:    addi.w $a1, $zero, -16
+; LA64-NEXT:    lu32i.d $a1, 1
+; LA64-NEXT:    bstrpick.d $a0, $a0, 31, 0
+; LA64-NEXT:    addi.d $a0, $a0, 15
+; LA64-NEXT:    and $a0, $a0, $a1
+; LA64-NEXT:    sub.d $a0, $sp, $a0
+; LA64-NEXT:    move $sp, $a0
+; LA64-NEXT:    addi.d $a1, $s8, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -64
+; LA64-NEXT:    ld.d $s8, $sp, 40 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 64
+; LA64-NEXT:    ret
+  %1 = alloca i8, i32 %n
+  %2 = alloca i32, align 64
+  call void @callee(i8* %1, i32 *%2)
+  ret void
+}

llvm/test/CodeGen/LoongArch/stack-realignment.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch32 --verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefix=LA32
+; RUN: llc --mtriple=loongarch64 --verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefix=LA64
+
+declare void @callee(i8*)
+
+define void @caller32() {
+; LA32-LABEL: caller32:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -32
+; LA32-NEXT:    .cfi_def_cfa_offset 32
+; LA32-NEXT:    st.w $ra, $sp, 28 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 24 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 32
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 5
+; LA32-NEXT:    slli.w $sp, $a0, 5
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -32
+; LA32-NEXT:    ld.w $fp, $sp, 24 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 28 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 32
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller32:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -32
+; LA64-NEXT:    .cfi_def_cfa_offset 32
+; LA64-NEXT:    st.d $ra, $sp, 24 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 16 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 32
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 5
+; LA64-NEXT:    slli.d $sp, $a0, 5
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -32
+; LA64-NEXT:    ld.d $fp, $sp, 16 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 24 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 32
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 32
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign32() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign32:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign32:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 32
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller64() {
+; LA32-LABEL: caller64:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -64
+; LA32-NEXT:    .cfi_def_cfa_offset 64
+; LA32-NEXT:    st.w $ra, $sp, 60 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 56 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 64
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 6
+; LA32-NEXT:    slli.w $sp, $a0, 6
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -64
+; LA32-NEXT:    ld.w $fp, $sp, 56 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 60 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 64
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller64:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -64
+; LA64-NEXT:    .cfi_def_cfa_offset 64
+; LA64-NEXT:    st.d $ra, $sp, 56 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 48 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 64
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 6
+; LA64-NEXT:    slli.d $sp, $a0, 6
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -64
+; LA64-NEXT:    ld.d $fp, $sp, 48 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 56 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 64
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 64
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign64() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign64:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign64:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 64
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller128() {
+; LA32-LABEL: caller128:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -128
+; LA32-NEXT:    .cfi_def_cfa_offset 128
+; LA32-NEXT:    st.w $ra, $sp, 124 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 120 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 128
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 7
+; LA32-NEXT:    slli.w $sp, $a0, 7
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -128
+; LA32-NEXT:    ld.w $fp, $sp, 120 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 124 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 128
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller128:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -128
+; LA64-NEXT:    .cfi_def_cfa_offset 128
+; LA64-NEXT:    st.d $ra, $sp, 120 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 112 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 128
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 7
+; LA64-NEXT:    slli.d $sp, $a0, 7
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -128
+; LA64-NEXT:    ld.d $fp, $sp, 112 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 120 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 128
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 128
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign128() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign128:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign128:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 128
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller256() {
+; LA32-LABEL: caller256:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -256
+; LA32-NEXT:    .cfi_def_cfa_offset 256
+; LA32-NEXT:    st.w $ra, $sp, 252 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 248 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 256
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 8
+; LA32-NEXT:    slli.w $sp, $a0, 8
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -256
+; LA32-NEXT:    ld.w $fp, $sp, 248 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 252 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 256
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller256:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -256
+; LA64-NEXT:    .cfi_def_cfa_offset 256
+; LA64-NEXT:    st.d $ra, $sp, 248 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 240 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 256
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 8
+; LA64-NEXT:    slli.d $sp, $a0, 8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -256
+; LA64-NEXT:    ld.d $fp, $sp, 240 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 248 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 256
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 256
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign256() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign256:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign256:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 256
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller512() {
+; LA32-LABEL: caller512:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -512
+; LA32-NEXT:    .cfi_def_cfa_offset 512
+; LA32-NEXT:    st.w $ra, $sp, 508 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 504 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 512
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 9
+; LA32-NEXT:    slli.w $sp, $a0, 9
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -512
+; LA32-NEXT:    ld.w $fp, $sp, 504 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 508 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 512
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller512:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -512
+; LA64-NEXT:    .cfi_def_cfa_offset 512
+; LA64-NEXT:    st.d $ra, $sp, 504 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 496 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 512
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 9
+; LA64-NEXT:    slli.d $sp, $a0, 9
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -512
+; LA64-NEXT:    ld.d $fp, $sp, 496 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 504 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 512
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 512
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign512() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign512:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign512:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 512
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller1024() {
+; LA32-LABEL: caller1024:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -1024
+; LA32-NEXT:    .cfi_def_cfa_offset 1024
+; LA32-NEXT:    st.w $ra, $sp, 1020 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 1016 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 1024
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 10
+; LA32-NEXT:    slli.w $sp, $a0, 10
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -1024
+; LA32-NEXT:    ld.w $fp, $sp, 1016 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 1020 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 1024
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller1024:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -1024
+; LA64-NEXT:    .cfi_def_cfa_offset 1024
+; LA64-NEXT:    st.d $ra, $sp, 1016 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 1008 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 1024
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 10
+; LA64-NEXT:    slli.d $sp, $a0, 10
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -1024
+; LA64-NEXT:    ld.d $fp, $sp, 1008 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 1016 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 1024
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 1024
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign1024() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign1024:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign1024:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 1024
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller2048() {
+; LA32-LABEL: caller2048:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -2048
+; LA32-NEXT:    .cfi_def_cfa_offset 2048
+; LA32-NEXT:    st.w $ra, $sp, 2044 # 4-byte Folded Spill
+; LA32-NEXT:    st.w $fp, $sp, 2040 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    addi.w $fp, $sp, 2032
+; LA32-NEXT:    addi.w $fp, $fp, 16
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 11
+; LA32-NEXT:    slli.w $sp, $a0, 11
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    addi.w $sp, $fp, -2048
+; LA32-NEXT:    ld.w $fp, $sp, 2040 # 4-byte Folded Reload
+; LA32-NEXT:    ld.w $ra, $sp, 2044 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 2032
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller2048:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -2048
+; LA64-NEXT:    .cfi_def_cfa_offset 2048
+; LA64-NEXT:    st.d $ra, $sp, 2040 # 8-byte Folded Spill
+; LA64-NEXT:    st.d $fp, $sp, 2032 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    addi.d $fp, $sp, 2032
+; LA64-NEXT:    addi.d $fp, $fp, 16
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 11
+; LA64-NEXT:    slli.d $sp, $a0, 11
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    addi.d $sp, $fp, -2048
+; LA64-NEXT:    ld.d $fp, $sp, 2032 # 8-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 2040 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 2032
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 2048
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign2048() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign2048:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign2048:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 2048
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller4096() {
+; LA32-LABEL: caller4096:
+; LA32:       # %bb.0:
+; LA32-NEXT:    lu12i.w $a0, 1
+; LA32-NEXT:    sub.w $sp, $sp, $a0
+; LA32-NEXT:    .cfi_def_cfa_offset 4096
+; LA32-NEXT:    ori $a0, $zero, 4092
+; LA32-NEXT:    add.w $a0, $sp, $a0
+; LA32-NEXT:    st.w $ra, $a0, 0 # 4-byte Folded Spill
+; LA32-NEXT:    ori $a0, $zero, 4088
+; LA32-NEXT:    add.w $a0, $sp, $a0
+; LA32-NEXT:    st.w $fp, $a0, 0 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    .cfi_offset 22, -8
+; LA32-NEXT:    lu12i.w $a0, 1
+; LA32-NEXT:    add.w $fp, $sp, $a0
+; LA32-NEXT:    .cfi_def_cfa 22, 0
+; LA32-NEXT:    srli.w $a0, $sp, 12
+; LA32-NEXT:    slli.w $sp, $a0, 12
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    lu12i.w $a0, 1
+; LA32-NEXT:    sub.w $sp, $fp, $a0
+; LA32-NEXT:    ori $a0, $zero, 4088
+; LA32-NEXT:    add.w $a0, $sp, $a0
+; LA32-NEXT:    ld.w $fp, $a0, 0 # 4-byte Folded Reload
+; LA32-NEXT:    ori $a0, $zero, 4092
+; LA32-NEXT:    add.w $a0, $sp, $a0
+; LA32-NEXT:    ld.w $ra, $a0, 0 # 4-byte Folded Reload
+; LA32-NEXT:    lu12i.w $a0, 1
+; LA32-NEXT:    add.w $sp, $sp, $a0
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller4096:
+; LA64:       # %bb.0:
+; LA64-NEXT:    lu12i.w $a0, 1
+; LA64-NEXT:    sub.d $sp, $sp, $a0
+; LA64-NEXT:    .cfi_def_cfa_offset 4096
+; LA64-NEXT:    ori $a0, $zero, 4088
+; LA64-NEXT:    add.d $a0, $sp, $a0
+; LA64-NEXT:    st.d $ra, $a0, 0 # 8-byte Folded Spill
+; LA64-NEXT:    ori $a0, $zero, 4080
+; LA64-NEXT:    add.d $a0, $sp, $a0
+; LA64-NEXT:    st.d $fp, $a0, 0 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    .cfi_offset 22, -16
+; LA64-NEXT:    lu12i.w $a0, 1
+; LA64-NEXT:    add.d $fp, $sp, $a0
+; LA64-NEXT:    .cfi_def_cfa 22, 0
+; LA64-NEXT:    srli.d $a0, $sp, 12
+; LA64-NEXT:    slli.d $sp, $a0, 12
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    lu12i.w $a0, 1
+; LA64-NEXT:    sub.d $sp, $fp, $a0
+; LA64-NEXT:    ori $a0, $zero, 4080
+; LA64-NEXT:    add.d $a0, $sp, $a0
+; LA64-NEXT:    ld.d $fp, $a0, 0 # 8-byte Folded Reload
+; LA64-NEXT:    ori $a0, $zero, 4088
+; LA64-NEXT:    add.d $a0, $sp, $a0
+; LA64-NEXT:    ld.d $ra, $a0, 0 # 8-byte Folded Reload
+; LA64-NEXT:    lu12i.w $a0, 1
+; LA64-NEXT:    add.d $sp, $sp, $a0
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 4096
+  call void @callee(i8* %1)
+  ret void
+}
+
+define void @caller_no_realign4096() "no-realign-stack" {
+; LA32-LABEL: caller_no_realign4096:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    .cfi_def_cfa_offset 16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    .cfi_offset 1, -4
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(callee)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: caller_no_realign4096:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    .cfi_def_cfa_offset 16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    .cfi_offset 1, -8
+; LA64-NEXT:    addi.d $a0, $sp, 0
+; LA64-NEXT:    bl %plt(callee)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %1 = alloca i8, align 4096
+  call void @callee(i8* %1)
+  ret void
+}