[mips] Restrict tail call optimization

[sdardis]

Summary

The tail call optimization was being used without proper consideration of
ABI requirements for saving and restoring the GP. This patch restricts tail
call optimization to functions within the same translation unit.

[sdardis]

Rebase and ping.

[vkalintiris]

LGTM.

lib/Target/Mips/MipsISelLowering.cpp

MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   // caller side but removing it breaks the frame size calculation.
   CCInfo.AllocateStack(ABI.GetCalleeAllocdArgSizeInBytes(CallConv), 1);
 
   CCInfo.AnalyzeCallOperands(Outs, CC_Mips, CLI.getArgs(), Callee.getNode());
 
   // Get a count of how many bytes are to be pushed on the stack.
   unsigned NextStackOffset = CCInfo.getNextStackOffset();
 
-  // Check if it's really possible to do a tail call.
-  if (IsTailCall)
+  // Check if it's really possible to do a tail call. Restrict it to functions
+  // that are part of this compilation unit.
+  bool InternalLinkage = false;
+  if (IsTailCall) {
     IsTailCall = isEligibleForTailCallOptimization(
         CCInfo, NextStackOffset, *MF.getInfo<MipsFunctionInfo>());
-
+     if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
+      InternalLinkage = G->getGlobal()->hasInternalLinkage();
+      IsTailCall &= (InternalLinkage || G->getGlobal()->hasLocalLinkage() ||
+                     G->getGlobal()->hasPrivateLinkage() ||
+                     G->getGlobal()->hasHiddenVisibility() ||
+                     G->getGlobal()->hasProtectedVisibility());
+     }
+  }
   if (!IsTailCall && CLI.CS && CLI.CS->isMustTailCall())
     report_fatal_error("failed to perform tail call elimination on a call "
                        "site marked musttail");
 
   if (IsTailCall)
     ++NumTailCalls;
 
   // Chain is the output chain of the last Load/Store or CopyToReg node.
MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   if (!MemOpChains.empty())
     Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, MemOpChains);
 
   // If the callee is a GlobalAddress/ExternalSymbol node (quite common, every
   // direct call is) turn it into a TargetGlobalAddress/TargetExternalSymbol
   // node so that legalize doesn't hack it.
   bool IsPICCall = (ABI.IsN64() || IsPIC); // true if calls are translated to
                                            // jalr $25
-  bool GlobalOrExternal = false, InternalLinkage = false, IsCallReloc = false;
   SDValue CalleeLo;
   EVT Ty = Callee.getValueType();
+  bool GlobalOrExternal = false, IsCallReloc = false;
 
   if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
     if (IsPICCall) {
       const GlobalValue *Val = G->getGlobal();
       InternalLinkage = Val->hasInternalLinkage();
 
       if (InternalLinkage)
         Callee = getAddrLocal(G, DL, Ty, DAG, ABI.IsN32() || ABI.IsN64());

test/CodeGen/Mips/fastcc.ll

 ; NOODDSPREG-DAG:    lwc1    $f8, 16($[[R0]])
 ; NOODDSPREG-DAG:    lwc1    $f10, 20($[[R0]])
 ; NOODDSPREG-DAG:    lwc1    $f12, 24($[[R0]])
 ; NOODDSPREG-DAG:    lwc1    $f14, 28($[[R0]])
 ; NOODDSPREG-DAG:    lwc1    $f16, 32($[[R0]])
 ; NOODDSPREG-DAG:    lwc1    $f18, 36($[[R0]])
 
 ; NOODDSPREG-DAG:    lwc1    $[[F0:f[0-9]*[02468]]], 40($[[R0]])
-; NOODDSPREG-DAG:    swc1    $[[F0]], 8($sp)
+; NOODDSPREG-DAG:    swc1    $[[F0]], 0($sp)
 
   %0 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 0), align 4
   %1 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 1), align 4
   %2 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 2), align 4
   %3 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 3), align 4
   %4 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 4), align 4
   %5 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 5), align 4
   %6 = load float, float* getelementptr ([11 x float], [11 x float]* @fa, i32 0, i32 6), align 4

test/CodeGen/Mips/i64arg.ll

 ; RUN: llc -march=mips -relocation-model=pic -mips-tail-calls=1 < %s | FileCheck %s
 
 define void @f1(i64 %ll1, float %f, i64 %ll, i32 %i, float %f2) nounwind {
 entry:
-; CHECK-DAG: lw $[[R2:[0-9]+]], 64($sp)
-; CHECK-DAG: lw $[[R3:[0-9]+]], 68($sp)
+; CHECK-DAG: lw $[[R2:[0-9]+]], 80($sp)
+; CHECK-DAG: lw $[[R3:[0-9]+]], 84($sp)
 ; CHECK-DAG: move $[[R1:[0-9]+]], $5
 ; CHECK-DAG: move $[[R0:[0-9]+]], $4
 ; CHECK-DAG: ori $6, ${{[0-9]+}}, 3855
 ; CHECK-DAG: ori $7, ${{[0-9]+}}, 22136
 ; CHECK-DAG: lw  $25, %call16(ff1)
 ; CHECK: jalr $25
   tail call void @ff1(i32 %i, i64 1085102592623924856) nounwind
 ; CHECK-DAG: lw $25, %call16(ff2)
 ; CHECK-DAG: move $4, $[[R2]]
 ; CHECK-DAG: move $5, $[[R3]]
 ; CHECK: jalr $25
   tail call void @ff2(i64 %ll, double 3.000000e+00) nounwind
   %sub = add nsw i32 %i, -1
 ; CHECK-DAG: lw $25, %call16(ff3)
-; CHECK-DAG: sw $[[R1]], 76($sp)
-; CHECK-DAG: sw $[[R0]], 72($sp)
+; CHECK-DAG: sw $[[R1]], 28($sp)
+; CHECK-DAG: sw $[[R0]], 24($sp)
 ; CHECK-DAG: move $6, $[[R2]]
 ; CHECK-DAG: move $7, $[[R3]]
-; CHECK:     jr $25
+; CHECK:     jalr $25
   tail call void @ff3(i32 %i, i64 %ll, i32 %sub, i64 %ll1) nounwind
   ret void
 }
 
 declare void @ff1(i32, i64)
 
 declare void @ff2(i64, double)
 
 declare void @ff3(i32, i64, i32, i64)

test/CodeGen/Mips/lazy-binding.ll

 declare void @foo2()
 
 ; CHECK-LABEL: foo1:
 ; CHECK: lw $25, %call16(foo2)(${{[0-9]+}})
 ; CHECK: jalr $25
 ; CHECK: lw $25, %call16(foo2)(${{[0-9]+}})
 ; CHECK: jalr $25
 ; CHECK: lw $25, %call16(foo2)(${{[0-9]+}})
-; CHECK: jr $25
+; CHECK: jalr $25
 
 define void @foo1() {
 entry:
   tail call void @foo2()
   tail call void @foo2()
   tail call void @foo2()
   ret void
 }

test/CodeGen/Mips/llvm-ir/call.ll

; R6C:           jalrc $[[TGT]]
 
 
   %1 = call float @extern_float_void()
   %2 = fadd float %1, 1.0
 ; R6C:           jrc $ra
   ret float %2
 }
 
-define void @musttail_call_void_void() {
-; ALL-LABEL: musttail_call_void_void:
-
-; O32:           lw $[[TGT:[0-9]+]], %call16(extern_void_void)($gp)
-
-; N64:           ld $[[TGT:[0-9]+]], %call16(extern_void_void)($gp)
-
-; NOT-R6C:       jr   $[[TGT]]
-; R6C:           jrc  $[[TGT]]
-
-  musttail call void @extern_void_void()
-  ret void
-}
-
-define i32 @musttail_call_i32_void() {
-; ALL-LABEL: musttail_call_i32_void:
-
-; O32:           lw $[[TGT:[0-9]+]], %call16(extern_i32_void)($gp)
-
-; N64:           ld $[[TGT:[0-9]+]], %call16(extern_i32_void)($gp)
-
-; NOT-R6C:       jr   $[[TGT]]
-; R6C:           jrc  $[[TGT]]
-
-  %1 = musttail call i32 @extern_i32_void()
-  ret i32 %1
-}
-
-define float @musttail_call_float_void() {
-; ALL-LABEL: musttail_call_float_void:
-
-; O32:           lw $[[TGT:[0-9]+]], %call16(extern_float_void)($gp)
-
-; N64:           ld $[[TGT:[0-9]+]], %call16(extern_float_void)($gp)
-
-; NOT-R6C:       jr   $[[TGT]]
-; R6C:           jrc  $[[TGT]]
-
-  %1 = musttail call float @extern_float_void()
-  ret float %1
-}
-
 define i32 @indirect_call_void_void(void ()* %addr) {
 ; ALL-LABEL: indirect_call_void_void:
 
 ; ALL:           move $25, $4
 ; NOT-R6C:       jalr $25
 ; R6C:           jalrc $25
 
   call void %addr()

test/CodeGen/Mips/tailcall/tail-call-arguments-clobber.ll

 
 ; This test is more fragile than I'd like. The -NEXT directives enforce an
 ; assumption that any GOT related instructions will not appear between the
 ; loads and stores.
 
 ; O32 case: The last two arguments should appear at 16(sp), 20(sp). The order
 ;           of the loads doesn't matter, but they have to become before the
 ;           stores
-declare i32 @func2(i32, i32, i32, i32, i32, i32)
+define internal i32 @func2(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f) {
+  %1 = add i32 %a, %b
+  %2 = add i32 %1, %c
+  %3 = add i32 %2, %d
+  %4 = add i32 %3, %e
+  %5 = add i32 %4, %f
+  ret i32 %5
+}
 
 define i32 @func1(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f){
 ; MIPS32-LABEL: func1:
 
 ; MIPS32: lw ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS32-NEXT: lw ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS32-NEXT: sw ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS32-NEXT: sw ${{[0-9]+}}, {{[0-9]+}}($sp)
-  %retval = tail call i32 @func1(i32 %a, i32 %f, i32 %c, i32 %d, i32 %e, i32 %b)
+  %retval = tail call i32 @func2(i32 %a, i32 %f, i32 %c, i32 %d, i32 %e, i32 %b)
 
   ret i32 %retval
 }
 
 ; N64, N32 cases: N64 and N32 both pass 8 arguments in registers. The order
 ;           of the loads doesn't matter, but they have to become before the
 ;           stores
 
-declare i64 @func4(i64, i64, i64, i64, i64, i64, i64, i64, i64, i64)
-
+define internal i64 @func4(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e,
+                           i64 %f, i64 %g, i64 %h, i64 %i, i64 %j) {
+  %1 = add i64 %a, %b
+  %2 = add i64 %1, %c
+  %3 = add i64 %2, %d
+  %4 = add i64 %3, %e
+  %5 = add i64 %4, %f
+  %6 = add i64 %1, %g
+  %7 = add i64 %2, %h
+  %8 = add i64 %3, %i
+  %9 = add i64 %4, %j
+  ret i64 %5
+}
 define i64 @func3(i64 %a, i64 %b, i64 %c, i64 %d,
                   i64 %e, i64 %f, i64 %g, i64 %h,
                   i64 %i, i64 %j){
 ; MIPS64-LABEL: func3:
 
 ; MIPS64: ld ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS64-NEXT: ld ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS64-NEXT: sd ${{[0-9]+}}, {{[0-9]+}}($sp)
 ; MIPS64-NEXT: sd ${{[0-9]+}}, {{[0-9]+}}($sp)
   %retval = tail call i64 @func4(i64 %a, i64 %j, i64 %c, i64 %d, i64 %e, i64 %f, i64 %g, i64 %h, i64 %i, i64 %b)
 
   ret i64 %retval
 }
-
-

test/CodeGen/Mips/tailcall/tailcall-wrong-isa.ll

 ; RUN: llc -filetype=obj -march=mipsel -relocation-model=pic -mcpu=mips32r6 -mips-tail-calls=1 < %s -o - \
 ; RUN:   | llvm-objdump  -d - | FileCheck %s  -check-prefix=PIC32R6
 ; RUN: llc -filetype=obj -march=mipsel -relocation-model=static -mcpu=mips32r6 -mips-tail-calls=1 < %s -o - \
 ; RUN:   | llvm-objdump  -d - | FileCheck %s -check-prefix=STATIC32R6
 
 ; RUN: llc -filetype=obj -march=mips64el -mcpu=mips64r6 -mips-tail-calls=1 < %s -o - \
 ; RUN:   | llvm-objdump  -d - | FileCheck %s -check-prefix=N64R6
 
-declare i8 @f2(i8)
+define internal i8 @f2(i8) {
+  ret i8 4
+}
 
 define i8 @f1(i8 signext %i) nounwind {
   %a = tail call i8 @f2(i8 %i)
   ret i8 %a
 }
 
-; PIC32:      {{[0-9]}}: 08 00 20 03  jr $25
-; STATIC32:   {{[0-9]}}: 00 00 00 08  j 0
+; ALL:        f1:
+; PIC32:      {{[0-9a-z]}}: 08 00 20 03  jr $25
+; STATIC32:   {{[0-9a-z]}}: 00 00 00 08  j 0
 
 ; N64:        {{[0-9a-z]+}}: 08 00 20 03  jr $25
 
 ; PIC32MM:    {{[0-9a-z]+}}: b9 45 jrc $25
-; STATIC32MM: {{[0-9]}}: 00 d4 00 00 j 0
+; STATIC32MM: {{[0-9a-z]}}: 00 d4 00 00 j 0
 
-; PIC32R6:    {{[0-9]}}: 00 00 19 d8  jrc $25
-; STATIC32R6: {{[0-9]}}: 00 00 00 08  j 0
+; PIC32R6:    {{[0-9a-z]}}: 00 00 19 d8  jrc $25
+; STATIC32R6: {{[0-9a-z]}}: 00 00 00 08  j 0
 
-; N64R6:      {{[0-9a-z]+}}: 09 00 20 03  jr $25
+; N64R6:      {{[0-9a-z]+}}: 00 00 19 d8  jrc $25
 

test/CodeGen/Mips/tailcall/tailcall.ll

 ; RUN: llc -march=mipsel -relocation-model=pic  \
-; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=PIC32
+; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,PIC32
 ; RUN: llc -march=mipsel -relocation-model=static  \
-; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=STATIC32
+; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,STATIC32
 ; RUN: llc -march=mips64el -mcpu=mips64r2  \
-; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=N64
+; RUN:     -verify-machineinstrs -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,N64
 ; RUN: llc -march=mipsel -mattr=mips16 -relocation-model=pic \
 ; RUN:      -verify-machineinstrs -mips-tail-calls=1 < %s | \
-; RUN:     FileCheck %s -check-prefix=PIC16
+; RUN:     FileCheck %s -check-prefixes=ALL,PIC16
 
 ; RUN: llc -march=mipsel -relocation-model=pic -mattr=+micromips -mips-tail-calls=1 < %s | \
-; RUN:     FileCheck %s -check-prefix=PIC32
+; RUN:     FileCheck %s -check-prefixes=ALL,PIC32MM
 ; RUN: llc -march=mipsel -relocation-model=static -mattr=+micromips \
-; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=STATIC32
+; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,STATIC32
 
 ; RUN: llc -march=mipsel -relocation-model=pic -mcpu=mips32r6 -mips-tail-calls=1 < %s | \
-; RUN:     FileCheck %s -check-prefix=PIC32
+; RUN:     FileCheck %s -check-prefixes=ALL,PIC32R6
 ; RUN: llc -march=mipsel -relocation-model=static -mcpu=mips32r6 \
-; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=STATIC32
+; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,STATIC32
 ; RUN: llc -march=mips64el -mcpu=mips64r6  \
-; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=N64
+; RUN:     -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,N64R6
 
 ; RUN: llc -march=mipsel -relocation-model=pic -mcpu=mips32r6 -mattr=+micromips \
-; RUN:      -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=PIC32
+; RUN:      -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,PIC32MM
 ; RUN: llc -march=mipsel -relocation-model=static -mcpu=mips32r6 \
-; RUN:     -mattr=+micromips -mips-tail-calls=1 < %s | FileCheck %s -check-prefix=STATIC32
+; RUN:     -mattr=+micromips -mips-tail-calls=1 < %s | FileCheck %s -check-prefixes=ALL,STATIC32
 ; RUN: llc -march=mips64el -mcpu=mips64r6 -mattr=+micromips -mips-tail-calls=1 < %s \
-; RUN:      | FileCheck %s -check-prefix=N64
+; RUN:      | FileCheck %s -check-prefixes=ALL,N64
 
 @g0 = common global i32 0, align 4
 @g1 = common global i32 0, align 4
 @g2 = common global i32 0, align 4
 @g3 = common global i32 0, align 4
 @g4 = common global i32 0, align 4
 @g5 = common global i32 0, align 4
 @g6 = common global i32 0, align 4
 @g7 = common global i32 0, align 4
 @g8 = common global i32 0, align 4
 @g9 = common global i32 0, align 4
 
 define i32 @caller1(i32 %a0) nounwind {
 entry:
-; PIC32-NOT: jalr
-; STATIC32-NOT: jal
-; N64-NOT: jalr
+; ALL-LABEL: caller1:
+; PIC32: jalr $25
+; PIC32MM: jalr $25
+; PIC32R6: jalr $25
+; STATIC32: jal
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee1(i32 1, i32 1, i32 1, i32 %a0) nounwind
   ret i32 %call
 }
 
 declare i32 @callee1(i32, i32, i32, i32)
 
 define i32 @caller2(i32 %a0, i32 %a1, i32 %a2, i32 %a3) nounwind {
 entry:
-; PIC32: jalr
+; ALL-LABEL: caller2
+; PIC32: jalr $25
+; PIC32MM: jalr $25
+; PIC32R6: jalr $25
 ; STATIC32: jal
-; N64-NOT: jalr
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee2(i32 1, i32 %a0, i32 %a1, i32 %a2, i32 %a3) nounwind
   ret i32 %call
 }
 
 declare i32 @callee2(i32, i32, i32, i32, i32)
 
 define i32 @caller3(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4) nounwind {
 entry:
-; PIC32: jalr
+; ALL-LABEL: caller3:
+; PIC32: jalr $25
+; PIC32R6: jalr $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; N64-NOT: jalr
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee3(i32 1, i32 1, i32 1, i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4) nounwind
   ret i32 %call
 }
 
 declare i32 @callee3(i32, i32, i32, i32, i32, i32, i32, i32)
 
 define i32 @caller4(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6, i32 %a7) nounwind {
 entry:
-; PIC32: jalr
+; ALL-LABEL: caller4:
+; PIC32: jalr $25
+; PIC32R6: jalr $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; N64: jalr
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee4(i32 1, i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6, i32 %a7) nounwind
   ret i32 %call
 }
 
 declare i32 @callee4(i32, i32, i32, i32, i32, i32, i32, i32, i32)
 
 define i32 @caller5() nounwind readonly {
 entry:
-; PIC32: .ent caller5
-; PIC32-NOT: jalr $25
-; PIC32: .end caller5
-; STATIC32: .ent caller5
-; STATIC32-NOT: jal
-; STATIC32: .end caller5
-; N64: .ent caller5
-; N64-NOT: jalr $25
-; N64: .end caller5
-; PIC16: .ent caller5
+; ALL-LABEL: caller5:
+; PIC32: jr $25
+; PIC32R6: jr $25
+; PIC32MM: jr
+; STATIC32: j
+; N64: jr $25
+; N64R6: jr $25
 ; PIC16: jalrc
-; PIC16: .end caller5
 
   %0 = load i32, i32* @g0, align 4
   %1 = load i32, i32* @g1, align 4
   %2 = load i32, i32* @g2, align 4
   %3 = load i32, i32* @g3, align 4
   %4 = load i32, i32* @g4, align 4
   %5 = load i32, i32* @g5, align 4
   %6 = load i32, i32* @g6, align 4
entry:
   %add8 = add nsw i32 %add7, %a9
   ret i32 %add8
 }
 
 declare i32 @callee8(i32, ...)
 
 define i32 @caller8_0() nounwind {
 entry:
-; PIC32: .ent caller8_0
-; PIC32: jr
-; PIC32: .end caller8_0
-; STATIC32: .ent caller8_0
+; ALL-LABEL: caller8_0:
+; PIC32: jr $25
+; PIC32R6: jrc $25
+; PIC32MM: jrc
 ; STATIC32: j
-; STATIC32: .end caller8_0
-; N64: .ent caller8_0
-; N64-NOT: jalr $25
-; N64: .end caller8_0
-; PIC16: .ent caller8_0
+; N64: jr $25
+; N64R6: jrc $25
 ; PIC16: jalrc
-; PIC16: .end caller8_0
 
   %call = tail call fastcc i32 @caller8_1()
   ret i32 %call
 }
 
 define internal fastcc i32 @caller8_1() nounwind noinline {
 entry:
-; PIC32: .ent caller8_1
-; PIC32: jalr
-; PIC32: .end caller8_1
-; STATIC32: .ent caller8_1
+; ALL-LABEL: caller8_1:
+; PIC32: jalr $25
+; PIC32R6: jalr $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; STATIC32: .end caller8_1
-; N64: .ent caller8_1
-; N64-NOT: jalr $25
-; N64: .end caller8_1
-; PIC16: .ent caller8_1
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
-; PIC16: .end caller8_1
 
   %call = tail call i32 (i32, ...) @callee8(i32 2, i32 1) nounwind
   ret i32 %call
 }
 
 %struct.S = type { [2 x i32] }
 
 @gs1 = external global %struct.S
 
 declare i32 @callee9(%struct.S* byval)
 
 define i32 @caller9_0() nounwind {
 entry:
-; PIC32: .ent caller9_0
-; PIC32: jr
-; PIC32: .end caller9_0
-; STATIC32: .ent caller9_0
+; ALL-LABEL: caller9_0:
+; PIC32: jr $25
+; PIC32R6: jrc $25
+; PIC32MM: jrc
 ; STATIC32: j
-; STATIC32: .end caller9_0
-; N64: .ent caller9_0
-; N64-NOT: jalr $25
-; N64: .end caller9_0
-; PIC16: .ent caller9_0
+; N64: jr $25
+; N64R6: jrc $25
 ; PIC16: jalrc
-; PIC16: .end caller9_0
   %call = tail call fastcc i32 @caller9_1()
   ret i32 %call
 }
 
 define internal fastcc i32 @caller9_1() nounwind noinline {
 entry:
-; PIC32: .ent caller9_1
-; PIC32: jalr
-; PIC32: .end caller9_1
-; STATIC32: .ent caller9_1
+; ALL-LABEL: caller9_1:
+; PIC32: jalr $25
+; PIC32R6: jalrc $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; STATIC32: .end caller9_1
-; N64: .ent caller9_1
-; N64: jalr
-; N64: .end caller9_1
-; PIC16: .ent caller9_1
+; N64: jalr $25
+; N64R6: jalrc $25
 ; PIC16: jalrc
-; PIC16: .end caller9_1
 
   %call = tail call i32 @callee9(%struct.S* byval @gs1) nounwind
   ret i32 %call
 }
 
 declare i32 @callee10(i32, i32, i32, i32, i32, i32, i32, i32, i32)
 
 define i32 @caller10(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6, i32 %a7, i32 %a8) nounwind {
 entry:
-; PIC32: .ent caller10
-; PIC32-NOT: jalr $25
-; STATIC32: .ent caller10
-; STATIC32-NOT: jal
-; N64: .ent caller10
-; N64-NOT: jalr $25
-; PIC16: .ent caller10
+; ALL-LABEL: caller10:
+; PIC32: jalr $25
+; PIC32R6: jalr $25
+; PIC32MM: jalr $25
+; STATIC32: jal
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee10(i32 %a8, i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6, i32 %a7) nounwind
   ret i32 %call
 }
 
 declare i32 @callee11(%struct.S* byval)
 
 define i32 @caller11() nounwind noinline {
 entry:
-; PIC32: .ent caller11
-; PIC32: jalr
-; STATIC32: .ent caller11
+; ALL-LABEL: caller11:
+; PIC32: jalr $25
+; PIC32R6: jalrc $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; N64: .ent caller11
-; N64: jalr
-; PIC16: .ent caller11
+; N64: jalr $25
+; N64R6: jalrc $25
 ; PIC16: jalrc
 
   %call = tail call i32 @callee11(%struct.S* byval @gs1) nounwind
   ret i32 %call
 }
 
 declare i32 @callee12()
 
 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
 
 define i32 @caller12(%struct.S* nocapture byval %a0) nounwind {
 entry:
-; PIC32: .ent caller12
-; PIC32: jalr
-; STATIC32: .ent caller12
+; ALL-LABEL: caller12:
+; PIC32: jalr $25
+; PIC32R6: jalrc $25
+; PIC32MM: jalr $25
 ; STATIC32: jal
-; N64: .ent caller12
-; N64: jalr
-; PIC16: .ent caller12
+; N64: jalr $25
+; N64R6: jalrc $25
 ; PIC16: jalrc
 
   %0 = bitcast %struct.S* %a0 to i8*
   tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* bitcast (%struct.S* @gs1 to i8*), i8* %0, i32 8, i32 4, i1 false)
   %call = tail call i32 @callee12() nounwind
   ret i32 %call
 }
 
 declare i32 @callee13(i32, ...)
 
 define i32 @caller13() nounwind {
 entry:
-; PIC32: .ent caller13
-; PIC32-NOT: jalr
-; STATIC32: .ent caller13
-; STATIC32-NOT: jal
-; N64: .ent caller13
-; N64-NOT: jalr $25
-; PIC16: .ent caller13
+; ALL-LABEL: caller13
+; PIC32: jalr $25
+; PIC32R6: jalr $25
+; PIC32MM: jalr $25
+; STATIC32: jal
+; N64: jalr $25
+; N64R6: jalr $25
 ; PIC16: jalrc
 
   %call = tail call i32 (i32, ...) @callee13(i32 1, i32 2) nounwind
   ret i32 %call
 }
 
 ; Check that there is a chain edge between the load and store nodes.
 ;
-; PIC32-LABEL: caller14:
-; PIC32: lw ${{[0-9]+}}, 16($sp)
+; ALL-LABEL: caller14:
+; PIC32: lw ${{[0-9]+}}, 48($sp)
 ; PIC32: sw $4, 16($sp)
 
+; PIC32MM: lw ${{[0-9]+}}, 48($sp)
+; PIC32MM: sw16 $4, 16(${{[0-9]+}})
+
 define void @caller14(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
 entry:
   tail call void @callee14(i32 %e, i32 %b, i32 %c, i32 %d, i32 %a)
   ret void
 }
 
 declare void @callee14(i32, i32, i32, i32, i32)