[CSInfo][MIPS] Call delay slot support in DwarfDebug

[ntesic]

Summary

Added call site info support for call instructions with delay slot.
Search for instructions inside call delay slot, which load value
into parameter forwarding registers.
Return address of the call points to instruction after call delay slot,
which is not the one, immediately after the call instruction.

Tests:
-dbg-call-site-delay-slot-interpretation.mir, dbg-call-site-delay-slot-interpretation-64.mir - confirm creation of call_site_parameter DIE for parameters loaded in call delay slot
-dbg-call-site-low-pc.ll - confirm call return address (DW_AT_low_pc) of call_site DIE (label after call delay slot)

[dstenb]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

805

Nit: s/forwaring/forwarding

811–812

Might there be some target where a parameter value is produced by a series of delay slot instructions? If so, should this search start from the bottom of the delay slot bundle?

[djtodoro]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

811–812

I guess that we are operating with the targets with only one instruction within delay slot. I think we can get rid of the for loop, and add a comment + assert() describing that we expect only one instruction within the bundle.

@dstenb Does that sound OK to you?

[dstenb]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

811–812

That sounds good to me!

[ntesic]

Address comments.
@dstenb @djtodoro Thanks for your suggestions.

[dstenb]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

859–860

Do you know what would happen if you didn't have this guard, and enabled call site information for another target with delay slots?

If it is something that is relatively easy to catch, or perhaps not very serious, I think I'd prefer to have this code target-agnostic (i.e. just MI.hasDelaySlot), and just leave it up to the developer enabling call site information for the target to make sure that there are no delay slot quirks that need to be considered.

1897–1898

Same as above.

[vsk]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

859–860

+ 1

[djtodoro]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

859–860

+1, but I think we should at least make an assertion describing what we currently expect, e.g. bundled call instruction should be mips-call-bundle like:

call {
   delay slot
  }
label

llvm/test/DebugInfo/MIR/Mips/dbg-call-site-delay-slot-interpretation-64bit.mir

94

Please use only one DILocation per scope.
e.g. attach every {!21, !22, !23, !24} dbg! to !21

[ntesic]

The main purpose of the "delay slot guard" is to ensure that call_site DIE won't be created for call with delay slot, because it could have incorrect "return PC" value (DW_AT_low_pc, DW_AT_call_return_pc).
This will happen if the call and its delay slot instruction are not in the expected order.

Result of the wrong instruction order in call-delay slot bundle is incorrect label insertion in DwarfDebug::beginInstruction().
Label will be inserted immediately after call instruction, but it should be inserted after delay slot instruction.
Please, notice label "$tmp2" after call instruction "jal" in the following assembler output.

jal	f1
$tmp2:
	.loc	1 6 11                  # m.c:6:11
	sw	$4, %lo(global)($1)

During creation of call_site DIEs in DwarfDebug::constructCallSiteEntryDIEs(), the incorrectly placed label is used and wrong value of "returnPC" interpreted.
So, the call return address points to the delay slot instruction and it should point to the instruction after.
Please, find value of "DW_AT_low_pc" which points to delay slot instruction "sw $4 ..." in the following dwarfdump and objdump outputs.

0x00000067:     DW_TAG_GNU_call_site
                  DW_AT_abstract_origin	(0x00000079 "f1")
                  DW_AT_low_pc	(0x0000000000000010)

c: 0c 00 00 00                  	jal	0 <foo>
      10: ac 24 00 00                  	sw	$4, 0($1)
      14: 24 02 00 7b                  	addiu	$2, $zero, 123 <foo+0x7b>

This problem for MIPS architecture was caused in LiveDebugValues pass, by insertion of DBG_VALUE after the call instruction, instead of insertion after whole bundle.
Please, find the following output after livedebugvalues pass, clearly incorrect.

JAL @f1 ...{
}
DBG_VALUE $a0
SW renamable $a0

Fortunately, the livedebugvalues problem was fixed by https://reviews.llvm.org/D66888.

Do you know what would happen if you didn't have this guard, and enabled call site information for another target with delay slots?

So, if the "call-delay slot" instruction order is not as expected, it will cause wrong "return PC" value.

If it is something that is relatively easy to catch, or perhaps not very serious, I think I'd prefer to have this code target-agnostic (i.e. just MI.hasDelaySlot), and just leave it up to the developer enabling call site information for the target to make sure that there are no delay slot quirks that need to be considered.

We can use assertions in "delaySlotSupported" lambda, in order to ensure expected order of instructions (following example), instead of the target check.

call_instruction {
   delay_slot_instruction
  }

In beginInstruction() guard we should expect that delay slot instruction is immediately after call instruction, before label insertion.
In constructCallSiteEntryDIEs() guard we should expect that both the call instruction and its successor are followed by the same label.

@dstenb @vsk @djtodoro Thanks for your comments and, please,
let me know what do you think about this approach.

[dstenb]

In D78107#2029294, @ntesic wrote:

In beginInstruction() guard we should expect that delay slot instruction is immediately after call instruction, before label insertion.
In constructCallSiteEntryDIEs() guard we should expect that both the call instruction and its successor are followed by the same label.

@dstenb @vsk @djtodoro Thanks for your comments and, please,
let me know what do you think about this approach.

Sorry for the delay! Adding such assertions sounds good to me.

[ntesic]

Address comments.
Change of delay slot guard - use assertions instead of target check.

[ntesic]

Capture "&" instead of "this" in delaySlotSupported lambda.

[dstenb]

Minor comments. Looks good to me otherwise!

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

857

I think it might be worth making this an early exit:

if (!MI.isBundledWithSucc())
  return false;

890

I think this might read a bit as that we expect a delay slot instruction here. Could we perhaps break out the hasDelaySlot() check to here:

if (MI.hasDelaySlot() && !delaySlotSupported(*&MI))
  return;

or perhaps instead rename the delaySlotSupported() function to indicate this (I don't have any better suggestion).

1895

Same as above.

1907

Same as above.

[ntesic]

Address comment - delaySlotSupported change.
Compress test-CHECK-lines.

[ntesic]

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

890

I agree, it seems more intuitive. Thanks for the suggestion.

[ntesic]

@dstenb Thanks for the review. Any additional comments?

[dstenb]

LGTM. Sorry for the delay!

llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

static bool interpretNextInstr(const MachineInstr *CurMI,
   // forwarding registers or we finished the interpretation of all
   // parameters.
   if (CurMI->isCall())
     return false;
 
   if (ForwardedRegWorklist.empty())
     return false;
 
+  // Avoid NOP description.
+  if (CurMI->getNumOperands() == 0)
+    return true;
+
   interpretValues(CurMI, ForwardedRegWorklist, Params);
 
   return true;
 }
 
 /// Try to interpret values loaded into registers that forward parameters
 /// for \p CallMI. Store parameters with interpreted value into \p Params.
 static void collectCallSiteParameters(const MachineInstr *CallMI,
static void collectCallSiteParameters(const MachineInstr *CallMI,
   // the describeLoadedValue()). For those remaining arguments in the working
   // list, for which we do not describe a loaded value by
   // the describeLoadedValue(), we try to generate an entry value expression
   // for their call site value description, if the call is within the entry MBB.
   // TODO: Handle situations when call site parameter value can be described
   // as the entry value within basic blocks other than the first one.
   bool ShouldTryEmitEntryVals = MBB->getIterator() == MF->begin();
 
+  // Search for a loading value in forwarding registers inside call delay slot.
+  if (CallMI->hasDelaySlot()) {
+    auto Suc = std::next(CallMI->getIterator());
+    // Only one-instruction delay slot is supported.
+    auto BundleEnd = llvm::getBundleEnd(CallMI->getIterator());
+    assert(std::next(Suc) == BundleEnd &&
+           "More than one instruction in call delay slot");
+    // Try to interpret value loaded by instruction.
+    if (!interpretNextInstr(&*Suc, ForwardedRegWorklist, Params))
+      return;
+  }
+
   // Search for a loading value in forwarding registers.
   for (; I != MBB->rend(); ++I) {
     // Try to interpret values loaded by instruction.
     if (!interpretNextInstr(&*I, ForwardedRegWorklist, Params))
       return;
   }
 
   // Emit the call site parameter's value as an entry value.
void DwarfDebug::constructCallSiteEntryDIEs(const DISubprogram &SP,
   // for both tail and non-tail calls. Don't use DW_AT_call_all_source_calls
   // because one of its requirements is not met: call site entries for
   // optimized-out calls are elided.
   CU.addFlag(ScopeDIE, CU.getDwarf5OrGNUAttr(dwarf::DW_AT_call_all_calls));
 
   const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
   assert(TII && "TargetInstrInfo not found: cannot label tail calls");
 
+  // Delay slot support check.
+  auto delaySlotSupported = [&](const MachineInstr &MI) {
+    if (!MI.isBundledWithSucc())
+      return false;
+    auto Suc = std::next(MI.getIterator());
+    auto CallInstrBundle = getBundleStart(MI.getIterator());
+    auto DelaySlotBundle = getBundleStart(Suc);
+    // Ensure that label after call is following delay slot instruction.
+    // Ex. CALL_INSTRUCTION {
+    //       DELAY_SLOT_INSTRUCTION }
+    //      LABEL_AFTER_CALL
+    assert(getLabelAfterInsn(&*CallInstrBundle) ==
+               getLabelAfterInsn(&*DelaySlotBundle) &&
+           "Call and its successor instruction don't have same label after.");
+    return true;
+  };
+
   // Emit call site entries for each call or tail call in the function.
   for (const MachineBasicBlock &MBB : MF) {
     for (const MachineInstr &MI : MBB.instrs()) {
       // Bundles with call in them will pass the isCall() test below but do not
       // have callee operand information so skip them here. Iterator will
       // eventually reach the call MI.
       if (MI.isBundle())
         continue;
 
       // Skip instructions which aren't calls. Both calls and tail-calling jump
       // instructions (e.g TAILJMPd64) are classified correctly here.
       if (!MI.isCandidateForCallSiteEntry())
         continue;
 
       // Skip instructions marked as frame setup, as they are not interesting to
       // the user.
       if (MI.getFlag(MachineInstr::FrameSetup))
         continue;
 
-      // TODO: Add support for targets with delay slots (see: beginInstruction).
-      if (MI.hasDelaySlot())
+      // Check if delay slot support is enabled.
+      if (MI.hasDelaySlot() && !delaySlotSupported(*&MI))
         return;
 
       // If this is a direct call, find the callee's subprogram.
       // In the case of an indirect call find the register that holds
       // the callee.
       const MachineOperand &CalleeOp = MI.getOperand(0);
       if (!CalleeOp.isGlobal() && !CalleeOp.isReg())
         continue;

 
 // Process beginning of an instruction.
 void DwarfDebug::beginInstruction(const MachineInstr *MI) {
   const MachineFunction &MF = *MI->getMF();
   const auto *SP = MF.getFunction().getSubprogram();
   bool NoDebug =
       !SP || SP->getUnit()->getEmissionKind() == DICompileUnit::NoDebug;
 
+  // Delay slot support check.
+  auto delaySlotSupported = [](const MachineInstr &MI) {
+    if (!MI.isBundledWithSucc())
+      return false;
+    auto Suc = std::next(MI.getIterator());
+    // Ensure that delay slot instruction is successor of the call instruction.
+    // Ex. CALL_INSTRUCTION {
+    //        DELAY_SLOT_INSTRUCTION }
+    assert(Suc->isBundledWithPred() &&
+           "Call bundle instructions are out of order");
+    return true;
+  };
+
   // When describing calls, we need a label for the call instruction.
-  // TODO: Add support for targets with delay slots.
   if (!NoDebug && SP->areAllCallsDescribed() &&
       MI->isCandidateForCallSiteEntry(MachineInstr::AnyInBundle) &&
-      !MI->hasDelaySlot()) {
+      (!MI->hasDelaySlot() || delaySlotSupported(*MI))) {
     const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
     bool IsTail = TII->isTailCall(*MI);
     // For tail calls, we need the address of the branch instruction for
     // DW_AT_call_pc.
     if (IsTail)
       requestLabelBeforeInsn(MI);
     // For non-tail calls, we need the return address for the call for
     // DW_AT_call_return_pc. Under GDB tuning, this information is needed for

llvm/test/DebugInfo/MIR/Mips/dbg-call-site-delay-slot-interpretation-64bit.mir

+## Test mips64:
+# RUN: llc -emit-call-site-info -start-after=machineverifier -filetype=obj -mtriple=mips64-linux-gnu %s -o -| llvm-dwarfdump -| FileCheck %s
+## Test mips64el:
+# RUN: llc -emit-call-site-info -start-after=machineverifier -filetype=obj -mtriple=mips64el-linux-gnu %s -o -| llvm-dwarfdump -| FileCheck %s
+
+## Built from source:
+## extern int __attribute__((noinline)) sum(int a, int b);
+## void __attribute__((noinline)) set(int *adr, int val) {
+##   val++;
+##   *adr = val + sum(val, val);
+## }
+## Using command:
+## clang -g -O2 -target mips64-linux-gnu m.c -c -mllvm -stop-before=machineverifier
+## Check that call site interpretation analysis can interpret calls with delay slot and
+## parameters set outside and inside of the call delay slot.
+
+## Test mips64:
+# CHECK:        DW_TAG_GNU_call_site
+# CHECK-NEXT:     DW_AT_abstract_origin {{.*}} "sum"
+# CHECK-NEXT:     DW_AT_low_pc
+# CHECK-EMPTY:
+## Parameter forwarding register A1_64 is set in call delay slot.
+# CHECK-NEXT:     DW_TAG_GNU_call_site_parameter
+# CHECK-NEXT:       DW_AT_location      (DW_OP_reg5 A1_64)
+# CHECK-NEXT:       DW_AT_GNU_call_site_value   (DW_OP_breg17 S1_64+0)
+# CHECK-EMPTY:
+# CHECK-NEXT:     DW_TAG_GNU_call_site_parameter
+# CHECK-NEXT:       DW_AT_location      (DW_OP_reg4 A0_64)
+# CHECK-NEXT:       DW_AT_GNU_call_site_value   (DW_OP_breg17 S1_64+0)
+
+--- |
+  ; ModuleID = 'm.ll'
+  source_filename = "m.c"
+  target datalayout = "E-m:e-i8:8:32-i16:16:32-i64:64-n32:64-S128"
+  target triple = "mips64-unknown-linux-gnu"
+  ; Function Attrs: noinline nounwind
+  define void @set(i32* nocapture %adr, i32 signext %val) local_unnamed_addr !dbg !13 {
+  entry:
+    call void @llvm.dbg.value(metadata i32* %adr, metadata !18, metadata !DIExpression()), !dbg !20
+    call void @llvm.dbg.value(metadata i32 %val, metadata !19, metadata !DIExpression()), !dbg !20
+    %inc = add nsw i32 %val, 1, !dbg !20
+    call void @llvm.dbg.value(metadata i32 %inc, metadata !19, metadata !DIExpression()), !dbg !20
+    %call = tail call signext i32 @sum(i32 signext %inc, i32 signext %inc), !dbg !20
+    %add = add nsw i32 %call, %inc, !dbg !20
+    store i32 %add, i32* %adr, align 4, !dbg !20
+    ret void
+  }
+
+  declare !dbg !4 signext i32 @sum(i32 signext, i32 signext) local_unnamed_addr
+
+  ; Function Attrs: nounwind readnone speculatable willreturn
+  declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+  !llvm.dbg.cu = !{!0}
+  !llvm.module.flags = !{!8, !9, !10, !11}
+  !llvm.ident = !{!12}
+
+  !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, splitDebugInlining: false, nameTableKind: None)
+  !1 = !DIFile(filename: "m.c", directory: "/dir")
+  !2 = !{}
+  !3 = !{!4}
+  !4 = !DISubprogram(name: "sum", scope: !1, file: !1, line: 1, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+  !5 = !DISubroutineType(types: !6)
+  !6 = !{!7, !7, !7}
+  !7 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+  !8 = !{i32 7, !"Dwarf Version", i32 4}
+  !9 = !{i32 2, !"Debug Info Version", i32 3}
+  !10 = !{i32 1, !"wchar_size", i32 4}
+  !11 = !{i32 7, !"PIC Level", i32 1}
+  !12 = !{!"clang version 11.0.0"}
+  !13 = distinct !DISubprogram(name: "set", scope: !1, file: !1, line: 2, type: !14, scopeLine: 2, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !17)
+  !14 = !DISubroutineType(types: !15)
+  !15 = !{null, !16, !7}
+  !16 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !7, size: 64)
+  !17 = !{!18, !19}
+  !18 = !DILocalVariable(name: "adr", arg: 1, scope: !13, file: !1, line: 2, type: !16)
+  !19 = !DILocalVariable(name: "val", arg: 2, scope: !13, file: !1, line: 2, type: !7)
+  !20 = !DILocation(line: 0, scope: !13)
+
+...
+---
+name:            set
+alignment:       8
+stack:
+  - { id: 0, name: '', type: spill-slot, offset: -8, size: 8, alignment: 8,
+      stack-id: default, callee-saved-register: '$ra_64', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+  - { id: 1, name: '', type: spill-slot, offset: -16, size: 8, alignment: 8,
+      stack-id: default, callee-saved-register: '$s1_64', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+  - { id: 2, name: '', type: spill-slot, offset: -24, size: 8, alignment: 8,
+      stack-id: default, callee-saved-register: '$s0_64', callee-saved-restored: true,
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+callSites:
+  - { bb: 0, offset: 17, fwdArgRegs:
+      - { arg: 0, reg: '$a0_64' }
+      - { arg: 1, reg: '$a1_64' } }
+body:             |
+  bb.0.entry:
+    DBG_VALUE $a0_64, $noreg, !18, !DIExpression(), debug-location !20
+    DBG_VALUE $a1_64, $noreg, !19, !DIExpression(), debug-location !20
+    DBG_VALUE $a1_64, $noreg, !19, !DIExpression(), debug-location !20
+    $sp_64 = DADDiu $sp_64, -32
+    CFI_INSTRUCTION def_cfa_offset 32
+    SD killed $ra_64, $sp_64, 24 :: (store 8 into %stack.0)
+    SD killed $s1_64, $sp_64, 16 :: (store 8 into %stack.1)
+    SD killed $s0_64, $sp_64, 8 :: (store 8 into %stack.2)
+    CFI_INSTRUCTION offset $ra_64, -8
+    CFI_INSTRUCTION offset $s1_64, -16
+    CFI_INSTRUCTION offset $s0_64, -24
+    $s0_64 = OR64 $a0_64, $zero_64
+    DBG_VALUE $a1, $noreg, !19, !DIExpression(), debug-location !20
+    DBG_VALUE $s0_64, $noreg, !18, !DIExpression(), debug-location !20
+    renamable $s1 = ADDiu renamable $a1, 1, implicit killed $a1_64, implicit-def $s1_64, debug-location !20
+    DBG_VALUE $s1, $noreg, !19, !DIExpression(), debug-location !20
+    $a0_64 = OR64 $s1_64, $zero_64, debug-location !20
+    JAL @sum, csr_n64, implicit-def dead $ra, implicit $a0_64, implicit $a1_64, implicit-def $sp, implicit-def $v0, debug-location !20 {
+      $a1_64 = OR64 $s1_64, $zero_64, debug-location !20
+    }
+    renamable $at = nsw ADDu killed renamable $v0, renamable $s1, implicit killed $s1_64, debug-location !20
+    SW killed renamable $at, killed renamable $s0_64, 0
+    $s0_64 = LD $sp_64, 8
+    $s1_64 = LD $sp_64, 16
+    $ra_64 = LD $sp_64, 24
+    PseudoReturn64 undef $ra_64 {
+      $sp_64 = DADDiu $sp_64, 32
+    }
+
+...

llvm/test/DebugInfo/MIR/Mips/dbg-call-site-delay-slot-interpretation.mir

+## Test mips32:
+# RUN: llc -emit-call-site-info -start-after=machineverifier -filetype=obj -mtriple=mips-linux-gnu %s -o -| llvm-dwarfdump -| FileCheck %s
+## Test mipsel:
+# RUN: llc -emit-call-site-info -start-after=machineverifier -filetype=obj -mtriple=mipsel-linux-gnu %s -o -| llvm-dwarfdump -| FileCheck %s
+
+## Built from source:
+## extern int __attribute__((noinline)) sum(int a, int b);
+## void __attribute__((noinline)) set(int *adr, int val) {
+##   val++;
+##   *adr = val + sum(val, val);
+## }
+## Using command:
+## clang -g -O2 -target mips-linux-gnu m.c -c -mllvm -stop-before=machineverifier
+## Check that call site interpretation analysis can interpret calls with delay slot and
+## parameters set outside and inside of the call delay slot.
+
+## Test mips32:
+# CHECK:        DW_TAG_GNU_call_site
+# CHECK-NEXT:     DW_AT_abstract_origin {{.*}} "sum"
+# CHECK-NEXT:     DW_AT_low_pc
+# CHECK-EMPTY:
+## Parameter forwarding register A1_64 is set in call delay slot.
+# CHECK-NEXT:     DW_TAG_GNU_call_site_parameter
+# CHECK-NEXT:       DW_AT_location      (DW_OP_reg5 A1_64)
+# CHECK-NEXT:       DW_AT_GNU_call_site_value   (DW_OP_breg17 S1_64+0)
+# CHECK-EMPTY:
+# CHECK-NEXT:     DW_TAG_GNU_call_site_parameter
+# CHECK-NEXT:       DW_AT_location      (DW_OP_reg4 A0_64)
+# CHECK-NEXT:       DW_AT_GNU_call_site_value   (DW_OP_breg17 S1_64+0)
+
+--- |
+  ; ModuleID = 'm.ll'
+  source_filename = "m.c"
+  target datalayout = "E-m:m-p:32:32-i8:8:32-i16:16:32-i64:64-n32-S64"
+  target triple = "mips-unknown-linux-gnu"
+  ; Function Attrs: noinline nounwind
+  define dso_local void @set(i32* nocapture %adr, i32 signext %val) local_unnamed_addr !dbg !12 {
+  entry:
+    call void @llvm.dbg.value(metadata i32* %adr, metadata !17, metadata !DIExpression()), !dbg !19
+    call void @llvm.dbg.value(metadata i32 %val, metadata !18, metadata !DIExpression()), !dbg !19
+    %inc = add nsw i32 %val, 1, !dbg !19
+    call void @llvm.dbg.value(metadata i32 %inc, metadata !18, metadata !DIExpression()), !dbg !19
+    %call = tail call i32 @sum(i32 signext %inc, i32 signext %inc), !dbg !19
+    %add = add nsw i32 %call, %inc, !dbg !19
+    store i32 %add, i32* %adr, align 4, !dbg !19
+    ret void
+  }
+  declare !dbg !4 dso_local i32 @sum(i32 signext, i32 signext) local_unnamed_addr
+  ; Function Attrs: nounwind readnone speculatable willreturn
+  declare void @llvm.dbg.value(metadata, metadata, metadata)
+  
+  !llvm.dbg.cu = !{!0}
+  !llvm.module.flags = !{!8, !9, !10}
+  !llvm.ident = !{!11}
+  
+  !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, splitDebugInlining: false, nameTableKind: None)
+  !1 = !DIFile(filename: "m.c", directory: "/dir")
+  !2 = !{}
+  !3 = !{!4}
+  !4 = !DISubprogram(name: "sum", scope: !1, file: !1, line: 1, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+  !5 = !DISubroutineType(types: !6)
+  !6 = !{!7, !7, !7}
+  !7 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+  !8 = !{i32 7, !"Dwarf Version", i32 4}
+  !9 = !{i32 2, !"Debug Info Version", i32 3}
+  !10 = !{i32 1, !"wchar_size", i32 4}
+  !11 = !{!"clang version 11.0.0"}
+  !12 = distinct !DISubprogram(name: "set", scope: !1, file: !1, line: 2, type: !13, scopeLine: 2, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !16)
+  !13 = !DISubroutineType(types: !14)
+  !14 = !{null, !15, !7}
+  !15 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !7, size: 32)
+  !16 = !{!17, !18}
+  !17 = !DILocalVariable(name: "adr", arg: 1, scope: !12, file: !1, line: 2, type: !15)
+  !18 = !DILocalVariable(name: "val", arg: 2, scope: !12, file: !1, line: 2, type: !7)
+  !19 = !DILocation(line: 0, scope: !12)
+
+...
+---
+name:            set
+alignment:       4
+stack:
+  - { id: 0, name: '', type: spill-slot, offset: -4, size: 4, alignment: 4, 
+      stack-id: default, callee-saved-register: '$ra', callee-saved-restored: true, 
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+  - { id: 1, name: '', type: spill-slot, offset: -8, size: 4, alignment: 4, 
+      stack-id: default, callee-saved-register: '$s1', callee-saved-restored: true, 
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+  - { id: 2, name: '', type: spill-slot, offset: -12, size: 4, alignment: 4, 
+      stack-id: default, callee-saved-register: '$s0', callee-saved-restored: true, 
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+callSites:
+  - { bb: 0, offset: 16, fwdArgRegs: 
+      - { arg: 0, reg: '$a0' }
+      - { arg: 1, reg: '$a1' } }
+body:             |
+  bb.0.entry:
+    DBG_VALUE $a0, $noreg, !17, !DIExpression(), debug-location !19
+    DBG_VALUE $a1, $noreg, !18, !DIExpression(), debug-location !19
+    DBG_VALUE $a1, $noreg, !18, !DIExpression(), debug-location !19
+    $sp = ADDiu $sp, -32
+    CFI_INSTRUCTION def_cfa_offset 32
+    SW killed $ra, $sp, 28 :: (store 4 into %stack.0)
+    SW killed $s1, $sp, 24 :: (store 4 into %stack.1)
+    SW killed $s0, $sp, 20 :: (store 4 into %stack.2)
+    CFI_INSTRUCTION offset $ra_64, -4
+    CFI_INSTRUCTION offset $s1_64, -8
+    CFI_INSTRUCTION offset $s0_64, -12
+    $s0 = OR $a0, $zero
+    DBG_VALUE $s0, $noreg, !17, !DIExpression(), debug-location !19
+    renamable $s1 = nsw ADDiu killed renamable $a1, 1, debug-location !19
+    DBG_VALUE $s1, $noreg, !18, !DIExpression(), debug-location !19
+    $a0 = OR $s1, $zero, debug-location !19
+    JAL @sum, csr_o32, implicit-def dead $ra, implicit $a0, implicit $a1, implicit-def $sp, implicit-def $v0, debug-location !19 {
+      $a1 = OR $s1, $zero, debug-location !19
+    }
+    renamable $at = nsw ADDu killed renamable $v0, killed renamable $s1, debug-location !19
+    SW killed renamable $at, killed renamable $s0, 0, debug-location !19 :: (store 4 into %ir.adr)
+    $s0 = LW $sp, 20
+    DBG_VALUE $a0, $noreg, !17, !DIExpression(DW_OP_LLVM_entry_value, 1), debug-location !19
+    $s1 = LW $sp, 24
+    $ra = LW $sp, 28
+    PseudoReturn undef $ra {
+      $sp = ADDiu $sp, 32
+    }
+
+...
+

llvm/test/DebugInfo/Mips/dbg-call-site-low-pc.ll

+;; Test mips32:
+; RUN: llc -emit-call-site-info %s -mtriple=mips -filetype=obj -o -| llvm-dwarfdump -| FileCheck %s
+;; Test mipsel:
+; RUN: llc -emit-call-site-info %s -mtriple=mipsel -filetype=obj -o -| llvm-dwarfdump -| FileCheck %s
+;; Test mips64:
+; RUN: llc -emit-call-site-info %s -mtriple=mips64 -filetype=obj -o -| llvm-dwarfdump -| FileCheck %s
+;; Test mips64el:
+; RUN: llc -emit-call-site-info %s -mtriple=mips64el -filetype=obj -o -| llvm-dwarfdump -| FileCheck %s
+
+;; Source:
+;; __attribute__((noinline))
+;; extern void f1(int a);
+;; __attribute__((noinline))
+;; int main(){
+;;   int x = 10;
+;;   f1(x);
+;;   return ++x;
+;; }
+;; Command: clang -g -O2 -target mips-linux-gnu -S -emit-llvm m.c -c
+;; Confirm that DW_AT_low_pc (call return address) points to instruction after call delay slot.
+
+;; Test mips, mipsel, mips64, mips64el:
+; CHECK:        DW_TAG_GNU_call_site
+; CHECK-NEXT:     DW_AT_abstract_origin {{.*}} "f1"
+; CHECK-NEXT:     DW_AT_low_pc (0x0000000000000010)
+
+; ModuleID = 'm.c'
+source_filename = "m.c"
+target datalayout = "E-m:m-p:32:32-i8:8:32-i16:16:32-i64:64-n32-S64"
+target triple = "mips-unknown-linux-gnu"
+
+; Function Attrs: noinline nounwind
+define dso_local i32 @main() local_unnamed_addr !dbg !12 {
+entry:
+  call void @llvm.dbg.value(metadata i32 10, metadata !16, metadata !DIExpression()), !dbg !17
+  tail call void @f1(i32 signext 10), !dbg !17
+  call void @llvm.dbg.value(metadata i32 11, metadata !16, metadata !DIExpression()), !dbg !17
+  ret i32 11, !dbg !17
+}
+
+declare !dbg !4 dso_local void @f1(i32 signext) local_unnamed_addr
+
+; Function Attrs: nounwind readnone speculatable willreturn
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!8, !9, !10}
+!llvm.ident = !{!11}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "m.c", directory: "/dir")
+!2 = !{}
+!3 = !{!4}
+!4 = !DISubprogram(name: "f1", scope: !1, file: !1, line: 2, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!5 = !DISubroutineType(types: !6)
+!6 = !{null, !7}
+!7 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!8 = !{i32 7, !"Dwarf Version", i32 4}
+!9 = !{i32 2, !"Debug Info Version", i32 3}
+!10 = !{i32 1, !"wchar_size", i32 4}
+!11 = !{!"clang version 11.0.0"}
+!12 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 5, type: !13, scopeLine: 5, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !15)
+!13 = !DISubroutineType(types: !14)
+!14 = !{!7}
+!15 = !{!16}
+!16 = !DILocalVariable(name: "x", scope: !12, file: !1, line: 6, type: !7)
+!17 = !DILocation(line: 0, scope: !12)