[GlobalISel] introduce MIFlag::NoConvergent

[sameerds]

Summary

Some opcodes in MIR are defined to be convergent by the target by setting
IsConvergent in the corresponding TD file. For example, in AMDGPU, the opcodes
SI_CALL and G_INTRINSIC* are marked as convergent. But this is too conservative,
since calls to functions that do not execute convergent operations should not be
marked convergent. This information is available in LLVM IR.

The new flag MIFlag::NoConvergent now allows the IR translator to mark an
instruction as not performing any convergent operations. It is relevant only on
occurrences of opcodes that are marked convergent. It may be specified on
occurrences of other opcodes, but that's redundant.

[sameerds]

Adding @xbolva00, @RKSimon and @craig.topper since a lot of thought seems to have been involved when the size of MIFlags was bumped to uint32_t, in D118118. Note that we intend to use this new flag MIFlags::NoConvergent only in GlobalISel.

[arsenm]

I have a mostly updated copy of the noconvergent patch, fixing the remaining tests seems permanently stuck as the #3 thing on my todo list. I've also been debating splitting the existing isConvergent MI property into two halves (doesn't really matter for this patch, it's a lot easier to change MI flags than IR ones)

[sameerds]

In D157475#4574293, @arsenm wrote:

I have a mostly updated copy of the noconvergent patch, fixing the remaining tests seems permanently stuck as the #3 thing on my todo list. I've also been debating splitting the existing isConvergent MI property into two halves (doesn't really matter for this patch, it's a lot easier to change MI flags than IR ones)

It's not clear whether you want me to hold this change, or change it in some way.

[yassingh]

llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h

151

Missed doc comment here

llvm/test/CodeGen/MIR/AMDGPU/noconvergent.mir

2

Is this test auto updatable?

[arsenm]

llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h

151

Fix the doc

llvm/test/CodeGen/MIR/AMDGPU/noconvergent.mir

30

Arguably the verifier should reject this

[sameerds]

• rebased
• machine verifier disallows NoConvergent on opcodes that are not isConvergent
• new test for the machine verifier
• added missing comment

[sameerds]

llvm/test/CodeGen/MIR/AMDGPU/noconvergent.mir

30

Yeah, I had thought of silently dropping noconvergent when it is not relevant. But complaining loudly in the machine verifier is a much better idea. Implemented, with test.

[arsenm]

llvm/lib/CodeGen/MachineVerifier.cpp

1821–1822

&&

[sameerds]

• rebased
• fixed formatting near assert
• added clang-format off/on to allow existing code through pre-checks

[sameerds]

• updated the lit test to match the changed error message

llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h

struct CallLoweringInfo {
     /// VReg to hold the hidden sret parameter.
     Register DemoteRegister;
 
     /// The stack index for sret demotion.
     int DemoteStackIndex;
 
     /// Expected type identifier for indirect calls with a CFI check.
     const ConstantInt *CFIType = nullptr;
+
+    /// True if this call results in convergent operations.
+    bool IsConvergent = true;
   };
 
   /// Argument handling is mostly uniform between the four places that
   /// make these decisions: function formal arguments, call
   /// instruction args, call instruction returns and function
   /// returns. However, once a decision has been made on where an
   /// argument should go, exactly what happens can vary slightly. This
   /// class abstracts the differences.

llvm/include/llvm/CodeGen/MachineInstr.h

enum MIFlag {
     IsExact = 1 << 13,       // Instruction supports division is
                              // known to be exact.
     NoFPExcept = 1 << 14,    // Instruction does not raise
                              // floatint-point exceptions.
     NoMerge = 1 << 15,       // Passes that drop source location info
                              // (e.g. branch folding) should skip
                              // this instruction.
     Unpredictable = 1 << 16, // Instruction with unpredictable condition.
+    NoConvergent = 1 << 17,  // Call does not require convergence guarantees.
   };
 
 private:
   const MCInstrDesc *MCID;              // Instruction descriptor.
   MachineBasicBlock *Parent = nullptr;  // Pointer to the owning basic block.
 
   // Operands are allocated by an ArrayRecycler.
   MachineOperand *Operands = nullptr;   // Pointer to the first operand.
public:
   /// Convergent instructions can not be made control-dependent on any
   /// additional values.
   bool isConvergent(QueryType Type = AnyInBundle) const {
     if (isInlineAsm()) {
       unsigned ExtraInfo = getOperand(InlineAsm::MIOp_ExtraInfo).getImm();
       if (ExtraInfo & InlineAsm::Extra_IsConvergent)
         return true;
     }
+    if (getFlag(NoConvergent))
+      return false;
     return hasProperty(MCID::Convergent, Type);
   }
 
   /// Returns true if the specified instruction has a delay slot
   /// which must be filled by the code generator.
   bool hasDelaySlot(QueryType Type = AnyInBundle) const {
     return hasProperty(MCID::DelaySlot, Type);
   }

llvm/lib/CodeGen/GlobalISel/CallLowering.cpp

bool CallLowering::lowerCall(MachineIRBuilder &MIRBuilder, const CallBase &CB,
   CallingConv::ID CallConv = CB.getCallingConv();
   Type *RetTy = CB.getType();
   bool IsVarArg = CB.getFunctionType()->isVarArg();
 
   SmallVector<BaseArgInfo, 4> SplitArgs;
   getReturnInfo(CallConv, RetTy, CB.getAttributes(), SplitArgs, DL);
   Info.CanLowerReturn = canLowerReturn(MF, CallConv, SplitArgs, IsVarArg);
 
+  Info.IsConvergent = CB.isConvergent();
+
   if (!Info.CanLowerReturn) {
     // Callee requires sret demotion.
     insertSRetOutgoingArgument(MIRBuilder, CB, Info);
 
     // The sret demotion isn't compatible with tail-calls, since the sret
     // argument points into the caller's stack frame.
     CanBeTailCalled = false;
   }

llvm/lib/CodeGen/MIRParser/MILexer.h

enum TokenKind {
     kw_cfi_type,
     kw_bbsections,
     kw_bb_id,
     kw_unknown_size,
     kw_unknown_address,
     kw_ir_block_address_taken,
     kw_machine_block_address_taken,
     kw_call_frame_size,
+    kw_noconvergent,
 
     // Metadata types.
     kw_distinct,
 
     // Named metadata keywords
     md_tbaa,
     md_alias_scope,
     md_noalias,

llvm/lib/CodeGen/MIRParser/MILexer.cpp

return StringSwitch<MIToken::TokenKind>(Identifier)
       .Case("bb_id", MIToken::kw_bb_id)
       .Case("unknown-size", MIToken::kw_unknown_size)
       .Case("unknown-address", MIToken::kw_unknown_address)
       .Case("distinct", MIToken::kw_distinct)
       .Case("ir-block-address-taken", MIToken::kw_ir_block_address_taken)
       .Case("machine-block-address-taken",
             MIToken::kw_machine_block_address_taken)
       .Case("call-frame-size", MIToken::kw_call_frame_size)
+      .Case("noconvergent", MIToken::kw_noconvergent)
       .Default(MIToken::Identifier);
 }
 
 static Cursor maybeLexIdentifier(Cursor C, MIToken &Token) {
   if (!isalpha(C.peek()) && C.peek() != '_')
     return std::nullopt;
   auto Range = C;
   while (isIdentifierChar(C.peek()))

llvm/lib/CodeGen/MIRParser/MIParser.cpp

return error(Operands.empty() ? Token.location() : Operands.back().End,
                      printImplicitRegisterFlag(I) + " $" +
                      getRegisterName(TRI, I.getReg()) + "'");
   }
   return false;
 }
 
 bool MIParser::parseInstruction(unsigned &OpCode, unsigned &Flags) {
   // Allow frame and fast math flags for OPCODE
+  // clang-format off
   while (Token.is(MIToken::kw_frame_setup) ||
          Token.is(MIToken::kw_frame_destroy) ||
          Token.is(MIToken::kw_nnan) ||
          Token.is(MIToken::kw_ninf) ||
          Token.is(MIToken::kw_nsz) ||
          Token.is(MIToken::kw_arcp) ||
          Token.is(MIToken::kw_contract) ||
          Token.is(MIToken::kw_afn) ||
          Token.is(MIToken::kw_reassoc) ||
          Token.is(MIToken::kw_nuw) ||
          Token.is(MIToken::kw_nsw) ||
          Token.is(MIToken::kw_exact) ||
          Token.is(MIToken::kw_nofpexcept) ||
+         Token.is(MIToken::kw_noconvergent) ||
          Token.is(MIToken::kw_unpredictable)) {
+    // clang-format on
     // Mine frame and fast math flags
     if (Token.is(MIToken::kw_frame_setup))
       Flags |= MachineInstr::FrameSetup;
     if (Token.is(MIToken::kw_frame_destroy))
       Flags |= MachineInstr::FrameDestroy;
     if (Token.is(MIToken::kw_nnan))
       Flags |= MachineInstr::FmNoNans;
     if (Token.is(MIToken::kw_ninf))
while (Token.is(MIToken::kw_frame_setup) ||
     if (Token.is(MIToken::kw_nsw))
       Flags |= MachineInstr::NoSWrap;
     if (Token.is(MIToken::kw_exact))
       Flags |= MachineInstr::IsExact;
     if (Token.is(MIToken::kw_nofpexcept))
       Flags |= MachineInstr::NoFPExcept;
     if (Token.is(MIToken::kw_unpredictable))
       Flags |= MachineInstr::Unpredictable;
+    if (Token.is(MIToken::kw_noconvergent))
+      Flags |= MachineInstr::NoConvergent;
 
     lex();
   }
   if (Token.isNot(MIToken::Identifier))
     return error("expected a machine instruction");
   StringRef InstrName = Token.stringValue();
   if (PFS.Target.parseInstrName(InstrName, OpCode))
     return error(Twine("unknown machine instruction name '") + InstrName + "'");

llvm/lib/CodeGen/MIRPrinter.cpp

void MIPrinter::print(const MachineInstr &MI) {
   if (MI.getFlag(MachineInstr::IsExact))
     OS << "exact ";
   if (MI.getFlag(MachineInstr::NoFPExcept))
     OS << "nofpexcept ";
   if (MI.getFlag(MachineInstr::NoMerge))
     OS << "nomerge ";
   if (MI.getFlag(MachineInstr::Unpredictable))
     OS << "unpredictable ";
+  if (MI.getFlag(MachineInstr::NoConvergent))
+    OS << "noconvergent ";
 
   OS << TII->getName(MI.getOpcode());
   if (I < E)
     OS << ' ';
 
   bool NeedComma = false;
   for (; I < E; ++I) {
     if (NeedComma)

llvm/lib/CodeGen/MachineVerifier.cpp

 void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
   const MCInstrDesc &MCID = MI->getDesc();
   if (MI->getNumOperands() < MCID.getNumOperands()) {
     report("Too few operands", MI);
     errs() << MCID.getNumOperands() << " operands expected, but "
            << MI->getNumOperands() << " given.\n";
   }
 
+  if (MI->getFlag(MachineInstr::NoConvergent) && !MCID.isConvergent())
+    report("NoConvergent flag expected only on convergent instructions.", MI);
+
   if (MI->isPHI()) {
     if (MF->getProperties().hasProperty(
             MachineFunctionProperties::Property::NoPHIs))
       report("Found PHI instruction with NoPHIs property set", MI);
 
     if (FirstNonPHI)
       report("Found PHI instruction after non-PHI", MI);
   } else if (FirstNonPHI == nullptr)

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

bool AMDGPUCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
 
   // Create a temporarily-floating call instruction so we can add the implicit
   // uses of arg registers.
   unsigned Opc = getCallOpcode(MF, Info.Callee.isReg(), false, Info.CallConv);
 
   auto MIB = MIRBuilder.buildInstrNoInsert(Opc);
   MIB.addDef(TRI->getReturnAddressReg(MF));
 
+  if (!Info.IsConvergent)
+    MIB.setMIFlag(MachineInstr::NoConvergent);
+
   if (!addCallTargetOperands(MIB, MIRBuilder, Info))
     return false;
 
   // Tell the call which registers are clobbered.
   const uint32_t *Mask = TRI->getCallPreservedMask(MF, Info.CallConv);
   MIB.addRegMask(Mask);
 
   SmallVector<CCValAssign, 16> ArgLocs;

llvm/test/CodeGen/AMDGPU/GlobalISel/dereferenceable-declaration.ll

 ; Should have dereferenceable on mem operand
 define i64 @load_deref_declaration_only() {
   ; CHECK-LABEL: name: load_deref_declaration_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @declared_with_ret_deref, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @declared_with_ret_deref, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call ptr @declared_with_ret_deref()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; No dereferenceable on mem operand
 define i64 @load_deref_unknown_decl() {
   ; CHECK-LABEL: name: load_deref_unknown_decl
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @unknown_decl
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @unknown_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @unknown_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call ptr @unknown_decl()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; Should have dereferenceable on mem operand
 define i64 @load_deref_callsite_only() {
   ; CHECK-LABEL: name: load_deref_callsite_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @unknown_decl
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @unknown_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @unknown_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call dereferenceable(8) ptr @unknown_decl()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; Both loads should have effective dereferenceable(8) since the
 ; maximum should be used.
 define i64 @load_deref_maxmimum_callsite_declaration_only() {
   ; CHECK-LABEL: name: load_deref_maxmimum_callsite_declaration_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @declared_with_ret_deref, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @declared_with_ret_deref, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call0)
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV1:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref4
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV1]](p0), @declared_with_ret_deref4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV1]](p0), @declared_with_ret_deref4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV1:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN1:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV1]], 8
   ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN1]](p0) :: (dereferenceable load (s64) from %ir.call1)
   ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s64) = G_ADD [[LOAD]], [[LOAD1]]
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ADD]](s64)

 ; Should have deref_or_nullerenceable on mem operand
 define i64 @load_deref_or_null_declaration_only() {
   ; CHECK-LABEL: name: load_deref_or_null_declaration_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref_or_null
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @declared_with_ret_deref_or_null, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @declared_with_ret_deref_or_null, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call nonnull ptr @declared_with_ret_deref_or_null()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; No deref_or_nullerenceable on mem operand
 define i64 @load_deref_or_null_nonnull_decl() {
   ; CHECK-LABEL: name: load_deref_or_null_nonnull_decl
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @nonnull_decl
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @nonnull_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @nonnull_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call ptr @nonnull_decl()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; Should have deref_or_nullerenceable on mem operand
 define i64 @load_deref_or_null_callsite_only() {
   ; CHECK-LABEL: name: load_deref_or_null_callsite_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @nonnull_decl
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @nonnull_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @nonnull_decl, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call)
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   SI_RETURN implicit $vgpr0, implicit $vgpr1
   %call = call dereferenceable_or_null(8) ptr @nonnull_decl()
   %load = load i64, ptr %call, align 8
   ret i64 %load
 }
 
 ; Both loads should have effective deref_or_nullerenceable(8) since the
 ; maximum should be used.
 define i64 @load_deref_or_null_maxmimum_callsite_declaration_only() {
   ; CHECK-LABEL: name: load_deref_or_null_maxmimum_callsite_declaration_only
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref_or_null
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @declared_with_ret_deref_or_null, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @declared_with_ret_deref_or_null, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN]](p0) :: (dereferenceable load (s64) from %ir.call0)
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV1:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @declared_with_ret_deref_or_null4
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV1]](p0), @declared_with_ret_deref_or_null4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV1]](p0), @declared_with_ret_deref_or_null4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV1:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN1:%[0-9]+]]:_(p0) = G_ASSERT_ALIGN [[MV1]], 8
   ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[ASSERT_ALIGN1]](p0) :: (dereferenceable load (s64) from %ir.call1)
   ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s64) = G_ADD [[LOAD]], [[LOAD1]]
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ADD]](s64)

llvm/test/CodeGen/AMDGPU/GlobalISel/dropped_debug_info_assert.ll

define amdgpu_kernel void @call_debug_loc() {
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc, debug-location !6
   ; CHECK-NEXT:   [[COPY10:%[0-9]+]]:sreg_64 = COPY [[COPY8]], debug-location !6
   ; CHECK-NEXT:   [[COPY11:%[0-9]+]]:sreg_64 = COPY [[COPY7]], debug-location !6
   ; CHECK-NEXT:   [[COPY12:%[0-9]+]]:sreg_64 = COPY [[COPY6]], debug-location !6
   ; CHECK-NEXT:   [[COPY13:%[0-9]+]]:sreg_32 = COPY [[COPY5]], debug-location !6
   ; CHECK-NEXT:   [[COPY14:%[0-9]+]]:sreg_32 = COPY [[COPY4]], debug-location !6
   ; CHECK-NEXT:   [[COPY15:%[0-9]+]]:sreg_32 = COPY [[COPY3]], debug-location !6
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:sreg_32 = IMPLICIT_DEF debug-location !6
-  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 10
-  ; CHECK-NEXT:   [[COPY16:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
+  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 10, debug-location !6
+  ; CHECK-NEXT:   [[COPY16:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]], debug-location !6
   ; CHECK-NEXT:   [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 [[COPY16]], [[COPY1]], implicit $exec, debug-location !6
-  ; CHECK-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 20
-  ; CHECK-NEXT:   [[COPY17:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]]
+  ; CHECK-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 20, debug-location !6
+  ; CHECK-NEXT:   [[COPY17:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_1]], debug-location !6
   ; CHECK-NEXT:   [[V_LSHLREV_B32_e64_1:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 [[COPY17]], [[COPY]], implicit $exec, debug-location !6
   ; CHECK-NEXT:   [[V_OR3_B32_e64_:%[0-9]+]]:vgpr_32 = V_OR3_B32_e64 [[COPY2]], [[V_LSHLREV_B32_e64_]], [[V_LSHLREV_B32_e64_1]], implicit $exec, debug-location !6
   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3, debug-location !6
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY18]], debug-location !6
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]], debug-location !6
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]], debug-location !6
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY9]], debug-location !6
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]], debug-location !6
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]], debug-location !6
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]], debug-location !6
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]], debug-location !6
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]], debug-location !6
   ; CHECK-NEXT:   $vgpr31 = COPY [[V_OR3_B32_e64_]], debug-location !6
   ; CHECK-NEXT:   [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:sreg_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def $scc, debug-location !6
   ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[SI_PC_ADD_REL_OFFSET]], 0, 0, debug-location !6 :: (dereferenceable invariant load (p0) from got, addrspace 4)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = SI_CALL [[S_LOAD_DWORDX2_IMM]], @callee, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, debug-location !6
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent SI_CALL [[S_LOAD_DWORDX2_IMM]], @callee, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, debug-location !6
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc, debug-location !6
   ; CHECK-NEXT:   S_ENDPGM 0
 entry:
   call void @callee(), !dbg !6
   ret void
 }
 
 define void @returnaddress_debug_loc(ptr addrspace(1) %ptr) {
   ; CHECK-LABEL: name: returnaddress_debug_loc
   ; CHECK: bb.1.entry:
   ; CHECK-NEXT:   liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr30_sgpr31, debug-location !6
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
   ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1
-  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:vreg_64 = COPY [[COPY]]
+  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:vreg_64 = COPY [[COPY]], debug-location !6
   ; CHECK-NEXT:   GLOBAL_STORE_DWORDX2 [[REG_SEQUENCE]], [[COPY3]], 0, 0, implicit $exec :: (store (p0) into %ir.ptr, addrspace 1)
   ; CHECK-NEXT:   SI_RETURN
 entry:
   %returnaddr = call ptr @llvm.returnaddress(i32 0), !dbg !6
   store ptr %returnaddr, ptr addrspace(1) %ptr, align 8
   ret void
 }
 

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-assert-align.ll

define void @call_result_align_1() {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @returns_ptr, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @returns_ptr, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   G_STORE [[C]](s8), [[MV]](p1) :: (store (s8) into %ir.ptr, addrspace 1)
   ; CHECK-NEXT:   SI_RETURN
   %ptr = call align 1 ptr addrspace(1) @returns_ptr()
   store i8 0, ptr addrspace(1) %ptr, align 1
define void @call_result_align_8() {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @returns_ptr, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @returns_ptr, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p1) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   G_STORE [[C]](s8), [[ASSERT_ALIGN]](p1) :: (store (s8) into %ir.ptr, align 8, addrspace 1)
   ; CHECK-NEXT:   SI_RETURN
   %ptr = call align 8 ptr addrspace(1) @returns_ptr()
define void @declaration_result_align_8() {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @returns_ptr_align8, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @returns_ptr_align8, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[ASSERT_ALIGN:%[0-9]+]]:_(p1) = G_ASSERT_ALIGN [[MV]], 8
   ; CHECK-NEXT:   G_STORE [[C]](s8), [[ASSERT_ALIGN]](p1) :: (store (s8) into %ir.ptr, align 8, addrspace 1)
   ; CHECK-NEXT:   SI_RETURN
   %ptr = call ptr addrspace(1) @returns_ptr_align8()

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-abi-attribute-hints.ll

define amdgpu_kernel void @kernel_call_no_workitem_ids() {
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY7]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY8]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY10]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY11]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY12]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
   ret void
 }
 
 define amdgpu_kernel void @kernel_call_no_workgroup_ids() {
   ; CHECK-LABEL: name: kernel_call_no_workgroup_ids
define amdgpu_kernel void @kernel_call_no_workgroup_ids() {
   ; CHECK-NEXT:   [[COPY14:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY14]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY7]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY8]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY10]](s64)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define amdgpu_kernel void @kernel_call_no_other_sgprs() {
   ; CHECK-LABEL: name: kernel_call_no_other_sgprs
define amdgpu_kernel void @kernel_call_no_other_sgprs() {
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY7]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
   ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY8]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define void @func_call_no_workitem_ids() {
   ; CHECK-LABEL: name: func_call_no_workitem_ids
define void @func_call_no_workitem_ids() {
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY8]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY15]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
   ret void
 }
 
 define void @func_call_no_workgroup_ids() {
   ; CHECK-LABEL: name: func_call_no_workgroup_ids
define void @func_call_no_workgroup_ids() {
   ; CHECK-NEXT:   [[COPY12:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY12]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY6]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY7]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY8]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY9]](s64)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY10]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY11]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define void @func_call_no_other_sgprs() {
   ; CHECK-LABEL: name: func_call_no_other_sgprs
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $sgpr15, $vgpr31, $sgpr8_sgpr9
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:_(p4) = COPY [[COPY2]]
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY6]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY3]](p4)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY4]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY5]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-implicit-args.ll

define amdgpu_kernel void @test_call_external_void_func_i32([17 x i8]) #0 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_call_external_void_func_i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
define amdgpu_kernel void @test_call_external_void_func_i32([17 x i8]) #0 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define void @test_func_call_external_void_func_i32() #0 {
   ; GFX900-LABEL: name: test_func_call_external_void_func_i32
define void @test_func_call_external_void_func_i32() #0 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   SI_RETURN
+  ;
   ; GFX908-LABEL: name: test_func_call_external_void_func_i32
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr12, $sgpr13, $sgpr14, $sgpr15, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr13
define void @test_func_call_external_void_func_i32() #0 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   SI_RETURN
   call void @external_void_func_i32(i32 99)
   ret void
 }
 
 ; Explicit argument is split between registers ad the stack due to v31
 ; being used for workitem IDs.
define amdgpu_kernel void @test_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_call_external_void_func_v32i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
define amdgpu_kernel void @test_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_v32i32(<32 x i32> zeroinitializer)
   ret void
 }
 
 define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900-LABEL: name: test_func_call_external_void_func_v32i32
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY27]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[COPY28]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY29]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY30]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY31]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY32]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[COPY33]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[COPY34]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX900-NEXT:   SI_RETURN
+  ;
   ; GFX908-LABEL: name: test_func_call_external_void_func_v32i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908-NEXT:   liveins: $sgpr12, $sgpr13, $sgpr14, $sgpr15, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8, $vgpr9, $vgpr10, $vgpr11, $vgpr12, $vgpr13, $vgpr14, $vgpr15, $vgpr16, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr13
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY27]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[COPY28]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY29]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY30]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY31]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY32]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[COPY33]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[COPY34]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX908-NEXT:   SI_RETURN
   call void @external_void_func_v32i32(<32 x i32> zeroinitializer)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX900-LABEL: name: test_only_workitem_id_x
define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[COPY15]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_x
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[COPY15]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX900-LABEL: name: test_only_workitem_id_y
define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_y
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr1, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX900-LABEL: name: test_only_workitem_id_z
define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_z
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY9]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY12]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX900-LABEL: name: test_only_workitem_id_xy
define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_xy
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr15
define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX900-LABEL: name: test_only_workitem_id_yz
define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_yz
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr15
define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX900-LABEL: name: test_only_workitem_id_xz
define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX900-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX900-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX900-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX900-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX900-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX900-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX900-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX900-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX900-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX900-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900-NEXT:   S_ENDPGM 0
+  ;
   ; GFX908-LABEL: name: test_only_workitem_id_xz
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908-NEXT:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908-NEXT: {{  $}}
   ; GFX908-NEXT:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr15
define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX908-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX908-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX908-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX908-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX908-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX908-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GFX908-NEXT:   $vgpr31 = COPY [[OR]](s32)
-  ; GFX908-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GFX908-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GFX908-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 declare i32 @llvm.amdgcn.workitem.id.x() #1
 declare i32 @llvm.amdgcn.workitem.id.y() #1

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll

 
 define amdgpu_gfx void @test_gfx_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_void()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK-NEXT:   liveins: $vgpr0
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
   ; CHECK-NEXT:   $vgpr0 = COPY [[C]](s32)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm_inreg
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK-NEXT:   liveins: $sgpr4
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
   ; CHECK-NEXT:   $sgpr4 = COPY [[C]](s32)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_struct_i8_i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (invariant load (p1) from `ptr addrspace(4) undef`, addrspace 4)
   ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK-NEXT:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK-NEXT:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
   ; CHECK-NEXT:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK-NEXT:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK-NEXT:   $vgpr0 = COPY [[ANYEXT1]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[LOAD2]](s32)
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load { i8, i32 }, ptr addrspace(1) %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #0 {
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
   ; CHECK-NEXT:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK-NEXT:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK-NEXT:   $sgpr4 = COPY [[ANYEXT1]](s32)
   ; CHECK-NEXT:   $sgpr5 = COPY [[LOAD2]](s32)
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load { i8, i32 }, ptr addrspace(1) %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32_inreg({ i8, i32 } inreg %val)
   ret void
 }
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll

define amdgpu_kernel void @test_call_external_i32_func_i32_imm(ptr addrspace(1) %out) #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i32_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](s32), [[LOAD]](p1) :: (volatile store (s32) into %ir.out.load, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i32 @external_i32_func_i32(i32 42)
   store volatile i32 %val, ptr addrspace(1) %out
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_func_i32_imm(ptr addrspace(1) %out) #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_func_i32_imm
   ; GCN: bb.1 (%ir-block.0):
   ; GCN-NEXT:   liveins: $vgpr0, $vgpr1
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
   ; GCN-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_i32
   ; GCN-NEXT:   $vgpr0 = COPY [[C]](s32)
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_i32_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY3]](s32), [[MV]](p1) :: (volatile store (s32) into %ir.out, addrspace 1)
   ; GCN-NEXT:   SI_RETURN
   %val = call amdgpu_gfx i32 @external_gfx_i32_func_i32(i32 42)
   store volatile i32 %val, ptr addrspace(1) %out
   ret void
 }
define amdgpu_kernel void @test_call_external_i1_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY21]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC]](s1), [[DEF]](p1) :: (volatile store (s1) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i1 @external_i1_func_void()
   store volatile i1 %val, ptr addrspace(1) undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i1_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i1_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i1_func_void
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i1_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_i1_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY1]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC]](s1), [[DEF]](p1) :: (volatile store (s1) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   SI_RETURN
   %val = call amdgpu_gfx i1 @external_gfx_i1_func_void()
   store volatile i1 %val, ptr addrspace(1) undef
   ret void
define amdgpu_kernel void @test_call_external_i1_zeroext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i1_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_ZEXT:%[0-9]+]]:_(s32) = G_ASSERT_ZEXT [[COPY21]], 1
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[ASSERT_ZEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC]](s1)
   ; GCN-NEXT:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i1 @external_i1_zeroext_func_void()
define amdgpu_kernel void @test_call_external_i1_signext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i1_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_SEXT:%[0-9]+]]:_(s32) = G_ASSERT_SEXT [[COPY21]], 1
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[ASSERT_SEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC]](s1)
   ; GCN-NEXT:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i1 @external_i1_signext_func_void()
define amdgpu_kernel void @test_call_external_i8_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i8_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY21]](s32)
   ; GCN-NEXT:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC1]](s8), [[DEF]](p1) :: (volatile store (s8) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i8 @external_i8_func_void()
   store volatile i8 %val, ptr addrspace(1) undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i8_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i8_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i8_func_void
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i8_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_i8_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
   ; GCN-NEXT:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC1]](s8), [[DEF]](p1) :: (volatile store (s8) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   SI_RETURN
   %val = call amdgpu_gfx i8 @external_gfx_i8_func_void()
   store volatile i8 %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_i8_zeroext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i8_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_ZEXT:%[0-9]+]]:_(s32) = G_ASSERT_ZEXT [[COPY21]], 8
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[ASSERT_ZEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC]](s8)
   ; GCN-NEXT:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i8 @external_i8_zeroext_func_void()
define amdgpu_kernel void @test_call_external_i8_signext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i8_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_SEXT:%[0-9]+]]:_(s32) = G_ASSERT_SEXT [[COPY21]], 8
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[ASSERT_SEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC]](s8)
   ; GCN-NEXT:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i8 @external_i8_signext_func_void()
define amdgpu_kernel void @test_call_external_i16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY21]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC]](s16), [[DEF]](p1) :: (volatile store (s16) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i16 @external_i16_func_void()
   store volatile i16 %val, ptr addrspace(1) undef
   ret void
define amdgpu_kernel void @test_call_external_i16_zeroext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i16_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_ZEXT:%[0-9]+]]:_(s32) = G_ASSERT_ZEXT [[COPY21]], 16
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[ASSERT_ZEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC]](s16)
   ; GCN-NEXT:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i16 @external_i16_zeroext_func_void()
define amdgpu_kernel void @test_call_external_i16_signext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i16_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[ASSERT_SEXT:%[0-9]+]]:_(s32) = G_ASSERT_SEXT [[COPY21]], 16
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[ASSERT_SEXT]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC]](s16)
   ; GCN-NEXT:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i16 @external_i16_signext_func_void()
define amdgpu_kernel void @test_call_external_i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i32 @external_i32_func_void()
   store volatile i32 %val, ptr addrspace(1) undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_void
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_i32_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY1]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   SI_RETURN
   %val = call amdgpu_gfx i32 @external_gfx_i32_func_void()
   store volatile i32 %val, ptr addrspace(1) undef
   ret void
 }
define amdgpu_kernel void @test_call_external_i48_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i48_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC]](s48), [[DEF]](p1) :: (volatile store (s48) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i48 @external_i48_func_void()
define amdgpu_kernel void @test_call_external_i48_zeroext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i48_zeroext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[TRUNC]](s48)
   ; GCN-NEXT:   G_STORE [[ZEXT]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i48_signext_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i48_signext_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[TRUNC]](s48)
   ; GCN-NEXT:   G_STORE [[SEXT]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i64_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call i64 @external_i64_func_void()
   store volatile i64 %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_p1_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_p1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_p1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[MV]](p1), [[DEF]](p1) :: (volatile store (p1) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call ptr addrspace(1) @external_p1_func_void()
   store volatile ptr addrspace(1) %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_v2p1_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2p1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2p1_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   [[MV1:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY23]](s32), [[COPY24]](s32)
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p1>) = G_BUILD_VECTOR [[MV]](p1), [[MV1]](p1)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
define amdgpu_kernel void @test_call_external_p3_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_p3_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_p3_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(p3) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](p3), [[DEF]](p3) :: (volatile store (p3) into `ptr addrspace(3) undef`, addrspace 3)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call ptr addrspace(3) @external_p3_func_void()
   store volatile ptr addrspace(3) %val, ptr addrspace(3) undef
   ret void
 }
define amdgpu_kernel void @test_call_external_v2p3_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2p3_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2p3_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(p3) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(p3) = COPY $vgpr1
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p3>) = G_BUILD_VECTOR [[COPY21]](p3), [[COPY22]](p3)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<2 x p3>), [[DEF]](p3) :: (volatile store (<2 x p3>) into `ptr addrspace(3) undef`, addrspace 3)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <2 x ptr addrspace(3)> @external_v2p3_func_void()
   store volatile <2 x ptr addrspace(3)> %val, ptr addrspace(3) undef
define amdgpu_kernel void @test_call_external_f16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY21]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[TRUNC]](s16), [[DEF]](p1) :: (volatile store (s16) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call half @external_f16_func_void()
   store volatile half %val, ptr addrspace(1) undef
   ret void
define amdgpu_kernel void @test_call_external_f32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call float @external_f32_func_void()
   store volatile float %val, ptr addrspace(1) undef
   ret void
 }
define amdgpu_kernel void @test_call_external_f64_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_f64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call double @external_f64_func_void()
   store volatile double %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_v2f64_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2f64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2f64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY23]](s32), [[COPY24]](s32)
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
define amdgpu_kernel void @test_call_external_v2i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<2 x s32>), [[DEF]](p1) :: (volatile store (<2 x s32>) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <2 x i32> @external_v2i32_func_void()
   store volatile <2 x i32> %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_v3i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v3i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s32>), [[DEF]](p1) :: (volatile store (<3 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <3 x i32> @external_v3i32_func_void()
define amdgpu_kernel void @test_call_external_v4i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v4i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<4 x s32>), [[DEF]](p1) :: (volatile store (<4 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v5i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v5i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v5i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<5 x s32>), [[DEF]](p1) :: (volatile store (<5 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
define amdgpu_kernel void @test_call_external_v8i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v8i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v8i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN-NEXT:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN-NEXT:   [[COPY27:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN-NEXT:   [[COPY28:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v16i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v16i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v16i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN-NEXT:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN-NEXT:   [[COPY27:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN-NEXT:   [[COPY28:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v32i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v32i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15, implicit-def $vgpr16, implicit-def $vgpr17, implicit-def $vgpr18, implicit-def $vgpr19, implicit-def $vgpr20, implicit-def $vgpr21, implicit-def $vgpr22, implicit-def $vgpr23, implicit-def $vgpr24, implicit-def $vgpr25, implicit-def $vgpr26, implicit-def $vgpr27, implicit-def $vgpr28, implicit-def $vgpr29, implicit-def $vgpr30, implicit-def $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v32i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15, implicit-def $vgpr16, implicit-def $vgpr17, implicit-def $vgpr18, implicit-def $vgpr19, implicit-def $vgpr20, implicit-def $vgpr21, implicit-def $vgpr22, implicit-def $vgpr23, implicit-def $vgpr24, implicit-def $vgpr25, implicit-def $vgpr26, implicit-def $vgpr27, implicit-def $vgpr28, implicit-def $vgpr29, implicit-def $vgpr30, implicit-def $vgpr31
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN-NEXT:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN-NEXT:   [[COPY27:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN-NEXT:   [[COPY28:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v2i16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](<2 x s16>), [[DEF]](p1) :: (volatile store (<2 x s16>) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <2 x i16> @external_v2i16_func_void()
   store volatile <2 x i16> %val, ptr addrspace(1) undef
   ret void
 }
define amdgpu_kernel void @test_call_external_v3i16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v3i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
   ; GCN-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v4i16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v4i16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](p1) :: (volatile store (<4 x s16>) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <4 x i16> @external_v4i16_func_void()
   store volatile <4 x i16> %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_v2f16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v2f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](<2 x s16>), [[DEF]](p1) :: (volatile store (<2 x s16>) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <2 x half> @external_v2f16_func_void()
   store volatile <2 x half> %val, ptr addrspace(1) undef
   ret void
 }
define amdgpu_kernel void @test_call_external_v3f16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v3f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
   ; GCN-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v4f16_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v4f16_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](p1) :: (volatile store (<4 x s16>) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <4 x half> @external_v4f16_func_void()
   store volatile <4 x half> %val, ptr addrspace(1) undef
define amdgpu_kernel void @test_call_external_v3f32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v3f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s32>), [[DEF]](p1) :: (volatile store (<3 x s32>) into `ptr addrspace(1) undef`, align 16, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <3 x float> @external_v3f32_func_void()
define amdgpu_kernel void @test_call_external_v5f32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v5f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v5f32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN-NEXT:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<5 x s32>), [[DEF]](p1) :: (volatile store (<5 x s32>) into `ptr addrspace(1) undef`, align 32, addrspace 1)
define amdgpu_kernel void @test_call_external_i32_i64_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_i64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i32_i64_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY22]](s32), [[COPY23]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call { i32, i64 } @external_i32_i64_func_void()
   %val.0 = extractvalue { i32, i64 } %val, 0
   %val.1 = extractvalue { i32, i64 } %val, 1
   store volatile i32 %val.0, ptr addrspace(1) undef
   store volatile i64 %val.1, ptr addrspace(1) undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_i64_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_i64_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_i64_func_void
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_i64_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_i32_i64_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY2]](s32), [[COPY3]](s32)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY1]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   SI_RETURN
define amdgpu_kernel void @test_call_external_a2i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_a2i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_a2i32_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   G_STORE [[COPY21]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   G_STORE [[COPY22]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call [2 x i32] @external_a2i32_func_void()
   %val.0 = extractvalue [2 x i32] %val, 0
define amdgpu_kernel void @test_call_external_a5i8_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_a5i8_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_a5i8_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY21]](s32)
   ; GCN-NEXT:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY22]](s32)
   ; GCN-NEXT:   [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC2]](s16)
   ; GCN-NEXT:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY23]](s32)
define amdgpu_kernel void @test_call_external_v32i32_i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v32i32_i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v32i32_i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[LOAD:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<32 x s32>) from %stack.0, addrspace 5)
   ; GCN-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
   ; GCN-NEXT:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C3]](s32)
   ; GCN-NEXT:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s32) from %stack.0, align 128, addrspace 5)
   ; GCN-NEXT:   G_STORE [[LOAD]](<32 x s32>), [[DEF]](p1) :: (volatile store (<32 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   G_STORE [[LOAD1]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i32_v32i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_v32i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_i32_v32i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[FRAME_INDEX]](p5) :: (load (s32) from %stack.0, align 128, addrspace 5)
   ; GCN-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
   ; GCN-NEXT:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C3]](s32)
   ; GCN-NEXT:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[PTR_ADD1]](p5) :: (load (<32 x s32>) from %stack.0, addrspace 5)
   ; GCN-NEXT:   G_STORE [[LOAD]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   G_STORE [[LOAD1]](<32 x s32>), [[DEF]](p1) :: (volatile store (<32 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v33i32_func_void() #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v33i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v33i32_func_void, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[LOAD:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<33 x s32>) from %stack.0, align 256, addrspace 5)
   ; GCN-NEXT:   G_STORE [[LOAD]](<33 x s32>), [[DEF]](p1) :: (volatile store (<33 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <33 x i32> @external_v33i32_func_void()
   store volatile <33 x i32> %val, ptr addrspace(1) undef, align 8
   ret void
 }
define amdgpu_kernel void @test_call_external_v33i32_func_v33i32_i32(ptr addrspace(1) %p, i32 %idx) #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v33i32_func_v33i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_v33i32_func_v33i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[LOAD2:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<33 x s32>) from %stack.0, align 256, addrspace 5)
   ; GCN-NEXT:   G_STORE [[LOAD2]](<33 x s32>), [[DEF]](p1) :: (volatile store (<33 x s32>) into `ptr addrspace(1) undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <33 x i32> @external_v33i32_func_v33i32_i32(ptr addrspace(1) %p, i32 %idx)
   store volatile <33 x i32> %val, ptr addrspace(1) undef, align 8
   ret void
 }
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-sret.ll

define amdgpu_kernel void @test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32(i32) #0 {
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; GCN-NEXT:   [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX1]], [[C2]](s32)
   ; GCN-NEXT:   [[LOAD:%[0-9]+]]:_(s8) = G_LOAD [[FRAME_INDEX1]](p5) :: (dereferenceable load (s8) from %ir.out.val, addrspace 5)
   ; GCN-NEXT:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (dereferenceable load (s32) from %ir.out.gep1, addrspace 5)
   ; GCN-NEXT:   G_STORE [[LOAD]](s8), [[DEF]](p1) :: (volatile store (s8) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   G_STORE [[LOAD1]](s32), [[DEF]](p1) :: (volatile store (s32) into `ptr addrspace(1) undef`, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %in.val = alloca { i8, i32 }, align 4, addrspace(5)

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll

define amdgpu_kernel void @test_call_external_void_func_void() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_void()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_void()
   ret void
 }
 
 define void @test_func_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_func_call_external_void_func_void
define void @test_func_call_external_void_func_void() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @external_void_func_void()
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_empty_struct() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_empty_struct
define amdgpu_kernel void @test_call_external_void_func_empty_struct() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_empty_struct, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_empty_struct, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_empty_struct({} zeroinitializer, i32 23)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_empty_array() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_empty_array
define amdgpu_kernel void @test_call_external_void_func_empty_array() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_empty_array, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_empty_array, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_empty_array([0 x i8] zeroinitializer, i32 23)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i1_imm
define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i1, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_i1(i1 true)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i1_signext
define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i1_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i1, ptr addrspace(1) undef
   call void @external_void_func_i1_signext(i1 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i1_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i1, ptr addrspace(1) undef
   call void @external_void_func_i1_zeroext(i1 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i8, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_i8(i8 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i8_signext
define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i8_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i8, ptr addrspace(1) undef
   call void @external_void_func_i8_signext(i8 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i8_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i8, ptr addrspace(1) undef
   call void @external_void_func_i8_zeroext(i8 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_i16(i16 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i16_signext
define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i16_signext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i16, ptr addrspace(1) undef
   call void @external_void_func_i16_signext(i16 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i16_zeroext, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i16, ptr addrspace(1) undef
   call void @external_void_func_i16_zeroext(i16 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK-NEXT:   liveins: $vgpr0
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
   ; CHECK-NEXT:   $vgpr0 = COPY [[C]](s32)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm_inreg
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK-NEXT:   liveins: $sgpr4
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
   ; CHECK-NEXT:   $sgpr4 = COPY [[C]](s32)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i64_imm
define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_i64(i64 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i64
define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <2 x i64>, ptr addrspace(1) null
   call void @external_void_func_v2i64(<2 x i64> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v2i64(<2 x i64> <i64 8589934593, i64 17179869187>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i48
define amdgpu_kernel void @test_call_external_void_func_i48(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i48, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i48, ptr addrspace(1) undef
   call void @external_void_func_i48(i48 %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48_signext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i48_signext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48_signext, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i48_signext, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i48, ptr addrspace(1) undef
   call void @external_void_func_i48_signext(i48 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i48_zeroext(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48_zeroext, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_i48_zeroext, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %var = load volatile i48, ptr addrspace(1) undef
   call void @external_void_func_i48_zeroext(i48 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_p0_imm(ptr %arg) #0 {
define amdgpu_kernel void @test_call_external_void_func_p0_imm(ptr %arg) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_p0, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_p0, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_p0(ptr %arg)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2p0() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2p0
define amdgpu_kernel void @test_call_external_void_func_v2p0() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2p0, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2p0, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <2 x ptr>, ptr addrspace(1) null
   call void @external_void_func_v2p0(<2 x ptr> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 {
define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %load = load <2 x i64>, ptr addrspace(1) null
   %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 undef>, <3 x i32> <i32 0, i32 1, i32 2>
 
   call void @external_void_func_v3i64(<3 x i64> %val)
   ret void
 }
define amdgpu_kernel void @test_call_external_void_func_v4i64() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %load = load <2 x i64>, ptr addrspace(1) null
   %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 17179869187>, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   call void @external_void_func_v4i64(<4 x i64> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_f16_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_f16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_f16(half 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_f32_imm
define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_f32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_f32(float 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2f32_imm
define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v2f32(<2 x float> <float 1.0, float 2.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3f32_imm
define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v3f32(<3 x float> <float 1.0, float 2.0, float 4.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5f32_imm
define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v5f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v5f32(<5 x float> <float 1.0, float 2.0, float 4.0, float -1.0, float 0.5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_f64_imm
define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_f64(double 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2f64_imm
define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v2f64(<2 x double> <double 2.0, double 4.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3f64_imm
define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3f64, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v3f64(<3 x double> <double 2.0, double 4.0, double 8.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i16
define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <2 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v2i16(<2 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <3 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v3i16(<3 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3f16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <3 x half>, ptr addrspace(1) undef
   call void @external_void_func_v3f16(<3 x half> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <4 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v4i16(<4 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v4i16(<4 x i16> <i16 1, i16 2, i16 3, i16 4>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5i16
define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v5i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <5 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v5i16(<5 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v7i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v7i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <7 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v7i16(<7 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v63i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v63i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <63 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v63i16(<63 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v65i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v65i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <65 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v65i16(<65 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v66i16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v66i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v66i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v66i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <66 x i16>, ptr addrspace(1) undef
   call void @external_void_func_v66i16(<66 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2f16, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <2 x half>, ptr addrspace(1) undef
   call void @external_void_func_v2f16(<2 x half> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <2 x i32>, ptr addrspace(1) undef
   call void @external_void_func_v2i32(<2 x i32> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v2i32(<2 x i32> <i32 1, i32 2>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3i32_imm
define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v3i32(<3 x i32> <i32 3, i32 4, i32 5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3i32_i32
define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v3i32_i32(<3 x i32> <i32 3, i32 4, i32 5>, i32 6)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v4i32
define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = load <4 x i32>, ptr addrspace(1) undef
   call void @external_void_func_v4i32(<4 x i32> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v4i32(<4 x i32> <i32 1, i32 2, i32 3, i32 4>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5i32_imm
define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v5i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v5i32(<5 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v8i32
define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <8 x i32>, ptr addrspace(1) %ptr
   call void @external_void_func_v8i32(<8 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v8i32_imm() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void @external_void_func_v8i32(<8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v16i32
define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v16i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v16i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <16 x i32>, ptr addrspace(1) %ptr
   call void @external_void_func_v16i32(<16 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v32i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <32 x i32>, ptr addrspace(1) %ptr
   call void @external_void_func_v32i32(<32 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v32i32_i32(i32) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF2]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val0 = load <32 x i32>, ptr addrspace(1) %ptr0
   %val1 = load i32, ptr addrspace(1) undef
   call void @external_void_func_v32i32_i32(<32 x i32> %val0, i32 %val1)
   ret void
 }
define amdgpu_kernel void @test_call_external_void_func_v32i32_i8_i8_i16() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF2]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i8_i8_i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i8_i8_i16, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 16, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val0 = load <32 x i32>, ptr addrspace(1) %ptr0
   %val1 = load i8, ptr addrspace(1) undef
   %val2 = load i8, ptr addrspace(1) undef
   %val3 = load i16, ptr addrspace(1) undef
   call void @external_void_func_v32i32_i8_i8_i16(<32 x i32> %val0, i8 %val1, i8 %val2, i16 %val3)
define amdgpu_kernel void @test_call_external_void_func_v32i32_p3_p5() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF2]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_p3_p5, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v32i32_p3_p5, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val0 = load <32 x i32>, ptr addrspace(1) %ptr0
   %val1 = load ptr addrspace(3), ptr addrspace(1) undef
   %val2 = load ptr addrspace(5), ptr addrspace(1) undef
   call void @external_void_func_v32i32_p3_p5(<32 x i32> %val0, ptr addrspace(3) %val1, ptr addrspace(5) %val2)
   ret void
define amdgpu_kernel void @test_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_struct_i8_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_struct_i8_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load { i8, i32 }, ptr addrspace(1) %ptr0
   call void @external_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
   ; CHECK-NEXT:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK-NEXT:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK-NEXT:   $vgpr0 = COPY [[ANYEXT1]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[LOAD2]](s32)
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_si_gfx, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load { i8, i32 }, ptr addrspace(1) %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #0 {
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
   ; CHECK-NEXT:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK-NEXT:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK-NEXT:   $sgpr4 = COPY [[ANYEXT1]](s32)
   ; CHECK-NEXT:   $sgpr5 = COPY [[LOAD2]](s32)
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   %ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load { i8, i32 }, ptr addrspace(1) %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32_inreg({ i8, i32 } inreg %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_byval_struct_i8_i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_byval_struct_i8_i32, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_byval_struct_i8_i32, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %val = alloca { i8, i32 }, align 4, addrspace(5)
   %gep0 = getelementptr inbounds { i8, i32 }, ptr addrspace(5) %val, i32 0, i32 0
   %gep1 = getelementptr inbounds { i8, i32 }, ptr addrspace(5) %val, i32 0, i32 1
   store i8 3, ptr addrspace(5) %gep0
   store i32 8, ptr addrspace(5) %gep1
   call void @external_void_func_byval_struct_i8_i32(ptr addrspace(5) byval({ i8, i32 }) %val)
define void @call_byval_3ai32_byval_i8_align32(ptr addrspace(5) %incoming0, ptr addrspace(5) align 32 %incoming1) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY13]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY17]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY18]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY19]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @void_func_byval_a3i32_byval_i8_align32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @void_func_byval_a3i32_byval_i8_align32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @void_func_byval_a3i32_byval_i8_align32(ptr addrspace(5) byval([3 x i32]) %incoming0, ptr addrspace(5) align 32 %incoming1, i32 999)
   ret void
 }
 
 declare void @void_func_byval_a4i64_align4(ptr addrspace(5) byval([4 x i64]) align 4 %arg0) #0
 
define void @call_byval_a4i64_align4_higher_source_align(ptr addrspace(5) align 256 %incoming_high_align) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY12]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY17]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY18]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @void_func_byval_a4i64_align4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @void_func_byval_a4i64_align4, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 32, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call void @void_func_byval_a4i64_align4(ptr addrspace(5) byval([4 x i64]) align 4 %incoming_high_align)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i8() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i8
define amdgpu_kernel void @test_call_external_void_func_v2i8() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v2i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <2 x i8>, ptr addrspace(1) %ptr
   call void @external_void_func_v2i8(<2 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v3i8() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v3i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <3 x i8>, ptr addrspace(1) %ptr
   call void @external_void_func_v3i8(<3 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v4i8() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v4i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <4 x i8>, ptr addrspace(1) %ptr
   call void @external_void_func_v4i8(<4 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v8i8() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v8i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <8 x i8>, ptr addrspace(1) %ptr
   call void @external_void_func_v8i8(<8 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v16i8() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF1]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v16i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_v16i8, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   %ptr = load ptr addrspace(1), ptr addrspace(4) undef
   %val = load <16 x i8>, ptr addrspace(1) %ptr
   call void @external_void_func_v16i8(<16 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val, double %tmp) #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @stack_passed_f64_arg, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @stack_passed_f64_arg, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
 entry:
   call void @stack_passed_f64_arg(<32 x i32> %val, double %tmp)
   ret void
 }
 
 define void @stack_12xv3i32() #0 {
define void @stack_12xv3i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_12xv3i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_12xv3i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 20, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
 entry:
   call void @external_void_func_12xv3i32(
       <3 x i32> <i32 0, i32 0, i32 0>,
       <3 x i32> <i32 1, i32 1, i32 1>,
       <3 x i32> <i32 2, i32 2, i32 2>,
       <3 x i32> <i32 3, i32 3, i32 3>,
define void @stack_12xv3f32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_12xv3f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_12xv3f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 20, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
 entry:
   call void @external_void_func_12xv3f32(
       <3 x float> <float 0.0, float 0.0, float 0.0>,
       <3 x float> <float 1.0, float 1.0, float 1.0>,
       <3 x float> <float 2.0, float 2.0, float 2.0>,
       <3 x float> <float 3.0, float 3.0, float 3.0>,
define void @stack_8xv5i32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_8xv5i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_8xv5i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
 entry:
   call void @external_void_func_8xv5i32(
       <5 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0>,
       <5 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1>,
       <5 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2>,
       <5 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3>,
define void @stack_8xv5f32() #0 {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_8xv5f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_void_func_8xv5f32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
 entry:
   call void @external_void_func_8xv5f32(
       <5 x float> <float 0.0, float 0.0, float 0.0, float 0.0, float 0.0>,
       <5 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>,
       <5 x float> <float 2.0, float 2.0, float 2.0, float 2.0, float 2.0>,
       <5 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>,
define amdgpu_ps void @amdgpu_ps_call_default_cc() {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[COPY1]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[DEF1]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[DEF2]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY2]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY3]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[COPY4]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[COPY5]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[C]](p0), 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[C]](p0), 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
 main_body:
   call void null()
   ret void
 }
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-indirect-call.ll

define amdgpu_kernel void @test_indirect_call_sgpr_ptr(ptr %fptr) {
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK-NEXT:   $sgpr15 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[LOAD]](p0), 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[LOAD]](p0), 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   S_ENDPGM 0
   call void %fptr()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_indirect_call_sgpr_ptr(ptr %fptr) {
   ; CHECK-LABEL: name: test_gfx_indirect_call_sgpr_ptr
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $vgpr0, $vgpr1
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; CHECK-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[MV]](p0), 0, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[MV]](p0), 0, csr_amdgpu_si_gfx, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK-NEXT:   SI_RETURN
   call amdgpu_gfx void %fptr()
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-sibling-call.ll

define amdgpu_kernel void @kernel_call_i32_fastcc_i32_i32_unused_result(i32 %a, i32 %b, i32 %c) #1 {
   ; GCN-NEXT:   [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[LOAD]](<2 x s32>), [[C]](s32)
   ; GCN-NEXT:   [[EVEC1:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[LOAD]](<2 x s32>), [[C1]](s32)
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
   ; GCN-NEXT:   $vgpr0 = COPY [[EVEC]](s32)
   ; GCN-NEXT:   $vgpr1 = COPY [[EVEC1]](s32)
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   S_ENDPGM 0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_i32(i32 %a, i32 %b)
   ret void
 }
 
define fastcc i32 @sibling_call_i32_fastcc_i32_byval_i32_byval_parent(i32 %a, ptr addrspace(5) byval(i32) %b.byval, i32 %c) #1 {
   ; GCN-NEXT:   [[AMDGPU_WAVE_ADDRESS:%[0-9]+]]:_(p5) = G_AMDGPU_WAVE_ADDRESS $sgpr32
   ; GCN-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GCN-NEXT:   [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[AMDGPU_WAVE_ADDRESS]], [[C]](s32)
   ; GCN-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; GCN-NEXT:   G_MEMCPY [[PTR_ADD]](p5), [[COPY1]](p5), [[C1]](s32), 0 :: (dereferenceable store (s32) into stack, addrspace 5), (dereferenceable load (s32) from %ir.b.byval, addrspace 5)
   ; GCN-NEXT:   $vgpr0 = COPY [[COPY]](s32)
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_byval_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @i32_fastcc_i32_byval_i32, csr_amdgpu, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GCN-NEXT:   $vgpr0 = COPY [[COPY4]](s32)
   ; GCN-NEXT:   SI_RETURN implicit $vgpr0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_byval_i32(i32 %a, ptr addrspace(5) byval(i32) %b.byval)
   ret i32 %ret
 }
define fastcc i32 @no_sibling_call_callee_more_stack_space(i32 %a, i32 %b) #1 {
   ; GCN-NEXT:   $vgpr25 = COPY [[C]](s32)
   ; GCN-NEXT:   $vgpr26 = COPY [[C]](s32)
   ; GCN-NEXT:   $vgpr27 = COPY [[C]](s32)
   ; GCN-NEXT:   $vgpr28 = COPY [[C]](s32)
   ; GCN-NEXT:   $vgpr29 = COPY [[C]](s32)
   ; GCN-NEXT:   $vgpr30 = COPY [[C]](s32)
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32_a32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32_a32i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; GCN-NEXT:   $vgpr0 = COPY [[COPY3]](s32)
   ; GCN-NEXT:   SI_RETURN implicit $vgpr0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_i32_a32i32(i32 %a, i32 %b, [32 x i32] zeroinitializer)
   ret i32 %ret
 }
 
 ; Have another non-tail in the function
 define fastcc i32 @sibling_call_i32_fastcc_i32_i32_other_call(i32 %a, i32 %b, i32 %c) #1 {
   ; GCN-LABEL: name: sibling_call_i32_fastcc_i32_i32_other_call
   ; GCN: bb.1.entry:
   ; GCN-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
   ; GCN-NEXT:   $vgpr0 = COPY [[COPY]](s32)
   ; GCN-NEXT:   $vgpr1 = COPY [[COPY1]](s32)
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN-NEXT:   $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN-NEXT:   [[GV1:%[0-9]+]]:ccr_sgpr_64(p0) = G_GLOBAL_VALUE @sibling_call_i32_fastcc_i32_i32
   ; GCN-NEXT:   $vgpr0 = COPY [[COPY]](s32)
   ; GCN-NEXT:   $vgpr1 = COPY [[COPY1]](s32)
   ; GCN-NEXT:   $vgpr2 = COPY [[COPY4]](s32)
   ; GCN-NEXT:   [[COPY5:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY5]](<4 x s32>)

llvm/test/CodeGen/MIR/AMDGPU/noconvergent-invalid.mir

+# RUN: not --crash llc -mtriple=amdgcn-amd-amdhsa -run-pass=none -verify-machineinstrs %s -o /dev/null 2>&1 | FileCheck %s
+
+---
+name:            noconvergent_invalid
+registers:
+body: |
+  bb.0:
+    ; CHECK: NoConvergent flag expected only on convergent instructions.
+    %2:sreg_32 = IMPLICIT_DEF
+    %3:sreg_32 = noconvergent S_ADD_I32 %2, 1, implicit-def $scc
+...

llvm/test/CodeGen/MIR/AMDGPU/noconvergent.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn -run-pass none -o - %s | FileCheck %s
+
+---
+# Completely invalid code, but it checks that noconvergent round-trips properly.
+name:            use_noconvergent
+registers:
+body: |
+  bb.0:
+    ; CHECK-LABEL: name: use_noconvergent
+    ; CHECK: [[DEF:%[0-9]+]]:sreg_64 = IMPLICIT_DEF
+    ; CHECK-NEXT: [[SI_CALL:%[0-9]+]]:sreg_64 = SI_CALL [[DEF]], 1, csr_amdgpu
+    ; CHECK-NEXT: [[SI_CALL]]:sreg_64 = noconvergent SI_CALL [[DEF]], 2, csr_amdgpu
+    %0:sreg_64 = IMPLICIT_DEF
+    %1:sreg_64 = SI_CALL %0, 1, csr_amdgpu
+    %1:sreg_64 = noconvergent SI_CALL %0, 2, csr_amdgpu
+...