[mips] Enable tail calls by default

Enable tail calls by default for (micro)MIPS(64).

microMIPS is slightly more tricky than doing it for MIPS(R6) or microMIPSR6.
microMIPS has two instruction encodings: 16bit and 32bit along with some
restrictions on the size of the instruction that can fill the delay slot.
For safe tail calls for microMIPS, the delay slot filler attempts to find
a correct size instruction for the delay slot of TAILCALL pseudos.

Reviewers: dsanders, vkalintris

Subscribers: jfb, dsanders, sdardis, llvm-commits

Differential Revision: https://reviews.llvm.org/D21138

llvm-svn: 277708

19 files changed, 173 insertions, 58 deletions

diff --git a/llvm/lib/Target/Mips/MicroMipsInstrInfo.td b/llvm/lib/Target/Mips/MicroMipsInstrInfo.td
index fdf3c5f0e8c..de2f7d49844 100644
--- a/llvm/lib/Target/Mips/MicroMipsInstrInfo.td
+++ b/llvm/lib/Target/Mips/MicroMipsInstrInfo.td
@@ -471,6 +471,18 @@ class JumpRegCMM16<string opstr, RegisterOperand RO> :
   let isIndirectBranch = 1;
 }
 
+let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1, hasDelaySlot = 1,
+    hasExtraSrcRegAllocReq = 1, isCTI = 1, Defs = [AT] in {
+  class TailCall_MM<Instruction JumpInst> :
+    PseudoSE<(outs), (ins calltarget:$target), [], II_J>,
+    PseudoInstExpansion<(JumpInst jmptarget_mm:$target)>;
+
+  class TailCallReg_MM<RegisterOperand RO, Instruction JRInst,
+                       RegisterOperand ResRO = RO> :
+    PseudoSE<(outs), (ins RO:$rs), [(MipsTailCall RO:$rs)], II_JR>,
+    PseudoInstExpansion<(JRInst ResRO:$rs)>;
+}
+
 // Break16 and Sdbbp16
 class BrkSdbbp16MM<string opstr> :
   MicroMipsInst16<(outs), (ins uimm4:$code_),
@@ -969,6 +981,12 @@ let DecoderNamespace = "MicroMips", Predicates = [InMicroMips] in {
   def PREFX_MM : PrefetchIndexed<"prefx">, POOL32F_PREFX_FM_MM<0x15, 0x1A0>;
 }
 
+let AdditionalPredicates = [InMicroMips] in {
+  def TAILCALL_MM : TailCall_MM<J_MM>, ISA_MIPS1_NOT_32R6_64R6;
+  def TAILCALLREG_MM : TailCallReg_MM<GPR32Opnd, JR_MM>,
+                       ISA_MIPS1_NOT_32R6_64R6;
+}
+
 let DecoderNamespace = "MicroMips" in {
   def RDHWR_MM : MMRel, R6MMR6Rel, ReadHardware<GPR32Opnd, HWRegsOpnd>,
                  RDHWR_FM_MM, ISA_MICROMIPS32_NOT_MIPS32R6;
diff --git a/llvm/lib/Target/Mips/Mips32r6InstrInfo.td b/llvm/lib/Target/Mips/Mips32r6InstrInfo.td
index f552f8d37af..be239b0aa60 100644
--- a/llvm/lib/Target/Mips/Mips32r6InstrInfo.td
+++ b/llvm/lib/Target/Mips/Mips32r6InstrInfo.td
@@ -848,6 +848,8 @@ let AdditionalPredicates = [NotInMicroMips] in {
   def SDC2_R6 : SDC2_R6_ENC, SDC2_R6_DESC, ISA_MIPS32R6;
   def SWC2_R6 : SWC2_R6_ENC, SWC2_R6_DESC, ISA_MIPS32R6;
 }
+def TAILCALL_R6 : TailCall<J>, ISA_MIPS32R6;
+def TAILCALLREG_R6 : TailCallReg<GPR32Opnd, JR>, GPR_32, ISA_MIPS32R6;
 
 //===----------------------------------------------------------------------===//
 //
diff --git a/llvm/lib/Target/Mips/Mips64InstrInfo.td b/llvm/lib/Target/Mips/Mips64InstrInfo.td
index 39313e11890..c3d8754416a 100644
--- a/llvm/lib/Target/Mips/Mips64InstrInfo.td
+++ b/llvm/lib/Target/Mips/Mips64InstrInfo.td
@@ -228,22 +228,24 @@ def LL64 : LLBase<"ll", GPR32Opnd>, LW_FM<0x30>, PTR_64,
            ISA_MIPS2_NOT_32R6_64R6;
 def SC64 : SCBase<"sc", GPR32Opnd>, LW_FM<0x38>, PTR_64,
            ISA_MIPS2_NOT_32R6_64R6;
+def JR64   : IndirectBranch<"jr", GPR64Opnd>, MTLO_FM<8>, PTR_64;
 }
 
+def JALR64 : JumpLinkReg<"jalr", GPR64Opnd>, JALR_FM;
+
 /// Jump and Branch Instructions
 let isCodeGenOnly = 1 in {
-  def JR64   : IndirectBranch<"jr", GPR64Opnd>, MTLO_FM<8>;
   def BEQ64  : CBranch<"beq", brtarget, seteq, GPR64Opnd>, BEQ_FM<4>;
   def BNE64  : CBranch<"bne", brtarget, setne, GPR64Opnd>, BEQ_FM<5>;
   def BGEZ64 : CBranchZero<"bgez", brtarget, setge, GPR64Opnd>, BGEZ_FM<1, 1>;
   def BGTZ64 : CBranchZero<"bgtz", brtarget, setgt, GPR64Opnd>, BGEZ_FM<7, 0>;
   def BLEZ64 : CBranchZero<"blez", brtarget, setle, GPR64Opnd>, BGEZ_FM<6, 0>;
   def BLTZ64 : CBranchZero<"bltz", brtarget, setlt, GPR64Opnd>, BGEZ_FM<1, 0>;
-  def JALR64 : JumpLinkReg<"jalr", GPR64Opnd>, JALR_FM;
   def JALR64Pseudo : JumpLinkRegPseudo<GPR64Opnd, JALR, RA, GPR32Opnd>;
-  def TAILCALL64_R : TailCallReg<GPR64Opnd, JR, GPR32Opnd>;
+  def TAILCALLREG64 : TailCallReg<GPR64Opnd, JR64>, GPR_64, ISA_MIPS1_NOT_32R6_64R6;
 }
 
+
 def PseudoReturn64 : PseudoReturnBase<GPR64Opnd>;
 def PseudoIndirectBranch64 : PseudoIndirectBranchBase<GPR64Opnd>;
 
diff --git a/llvm/lib/Target/Mips/Mips64r6InstrInfo.td b/llvm/lib/Target/Mips/Mips64r6InstrInfo.td
index cc01400bb0d..0f366608cba 100644
--- a/llvm/lib/Target/Mips/Mips64r6InstrInfo.td
+++ b/llvm/lib/Target/Mips/Mips64r6InstrInfo.td
@@ -165,6 +165,10 @@ def BLTZC64 : BLTZC_ENC, BLTZC64_DESC, ISA_MIPS64R6, GPR_64;
 def BGEZC64 : BGEZC_ENC, BGEZC64_DESC, ISA_MIPS64R6, GPR_64;
 }
 
+def TAILCALL64_R6 : TailCall<J>, ISA_MIPS64R6;
+def TAILCALLREG64_R6 : TailCallReg<GPR64Opnd, JR64>, GPR_64,
+                       ISA_MIPS64R6;
+
 //===----------------------------------------------------------------------===//
 //
 // Instruction Aliases
diff --git a/llvm/lib/Target/Mips/MipsDelaySlotFiller.cpp b/llvm/lib/Target/Mips/MipsDelaySlotFiller.cpp
index 44bb872554d..c473c6ccd06 100644
--- a/llvm/lib/Target/Mips/MipsDelaySlotFiller.cpp
+++ b/llvm/lib/Target/Mips/MipsDelaySlotFiller.cpp
@@ -543,6 +543,9 @@ Iter Filler::replaceWithCompactBranch(MachineBasicBlock &MBB, Iter Branch,
 
 // For given opcode returns opcode of corresponding instruction with short
 // delay slot.
+// For the pseudo TAILCALL*_MM instrunctions return the short delay slot
+// form. Unfortunately, TAILCALL<->b16 is denied as b16 has a limited range
+// that is too short to make use of for tail calls.
 static int getEquivalentCallShort(int Opcode) {
   switch (Opcode) {
   case Mips::BGEZAL:
@@ -555,6 +558,10 @@ static int getEquivalentCallShort(int Opcode) {
     return Mips::JALRS_MM;
   case Mips::JALR16_MM:
     return Mips::JALRS16_MM;
+  case Mips::TAILCALL_MM:
+    llvm_unreachable("Attempting to shorten the TAILCALL_MM pseudo!");
+  case Mips::TAILCALLREG_MM:
+    return Mips::JR16_MM;
   default:
     llvm_unreachable("Unexpected call instruction for microMIPS.");
   }
@@ -606,6 +613,12 @@ bool Filler::runOnMachineBasicBlock(MachineBasicBlock &MBB) {
             DSI->isCall()) {
           // If instruction in delay slot is 16b change opcode to
           // corresponding instruction with short delay slot.
+
+          // TODO: Implement an instruction mapping table of 16bit opcodes to
+          // 32bit opcodes so that an instruction can be expanded. This would
+          // save 16 bits as a TAILCALL_MM pseudo requires a fullsized nop.
+          // TODO: Permit b16 when branching backwards to the the same function
+          // if it is in range.
           DSI->setDesc(TII->get(getEquivalentCallShort(DSI->getOpcode())));
         }
         continue;
@@ -692,9 +705,14 @@ bool Filler::searchRange(MachineBasicBlock &MBB, IterTy Begin, IterTy End,
     bool InMicroMipsMode = STI.inMicroMipsMode();
     const MipsInstrInfo *TII = STI.getInstrInfo();
     unsigned Opcode = (*Slot).getOpcode();
-    if (InMicroMipsMode && TII->getInstSizeInBytes(*CurrI) == 2 &&
+    // This is complicated by the tail call optimization. For non-PIC code
+    // there is only a 32bit sized unconditional branch which can be assumed
+    // to be able to reach the target. b16 only has a range of +/- 1 KB.
+    // It's entirely possible that the target function is reachable with b16
+    // but we don't have enough information to make that decision.
+     if (InMicroMipsMode && TII->getInstSizeInBytes(*CurrI) == 2 &&
         (Opcode == Mips::JR || Opcode == Mips::PseudoIndirectBranch ||
-         Opcode == Mips::PseudoReturn))
+         Opcode == Mips::PseudoReturn || Opcode == Mips::TAILCALL))
       continue;
 
     Filler = CurrI;
diff --git a/llvm/lib/Target/Mips/MipsInstrInfo.cpp b/llvm/lib/Target/Mips/MipsInstrInfo.cpp
index 1f38ca6447b..c317ecb91b2 100644
--- a/llvm/lib/Target/Mips/MipsInstrInfo.cpp
+++ b/llvm/lib/Target/Mips/MipsInstrInfo.cpp
@@ -282,6 +282,7 @@ unsigned MipsInstrInfo::getEquivalentCompactForm(
     case Mips::JR:
     case Mips::PseudoReturn:
     case Mips::PseudoIndirectBranch:
+    case Mips::TAILCALLREG_MM:
       canUseShortMicroMipsCTI = true;
       break;
     }
@@ -362,6 +363,8 @@ unsigned MipsInstrInfo::getEquivalentCompactForm(
     case Mips::JR:
     case Mips::PseudoReturn:
     case Mips::PseudoIndirectBranch:
+    case Mips::TAILCALLREG_MM:
+    case Mips::TAILCALLREG_R6:
       if (canUseShortMicroMipsCTI)
         return Mips::JRC16_MM;
       return Mips::JIC;
@@ -370,6 +373,7 @@ unsigned MipsInstrInfo::getEquivalentCompactForm(
     case Mips::JR64:
     case Mips::PseudoReturn64:
     case Mips::PseudoIndirectBranch64:
+    case Mips::TAILCALLREG64_R6:
       return Mips::JIC64;
     case Mips::JALR64Pseudo:
       return Mips::JIALC64;
diff --git a/llvm/lib/Target/Mips/MipsInstrInfo.td b/llvm/lib/Target/Mips/MipsInstrInfo.td
index 43e4818e264..ff8a5fba732 100644
--- a/llvm/lib/Target/Mips/MipsInstrInfo.td
+++ b/llvm/lib/Target/Mips/MipsInstrInfo.td
@@ -1918,8 +1918,11 @@ def BLTZAL : MMRel, BGEZAL_FT<"bltzal", brtarget, GPR32Opnd>, BGEZAL_FM<0x10>,
 def BLTZALL : MMRel, BGEZAL_FT<"bltzall", brtarget, GPR32Opnd, 0>,
               BGEZAL_FM<0x12>, ISA_MIPS2_NOT_32R6_64R6;
 def BAL_BR : BAL_BR_Pseudo<BGEZAL>;
-def TAILCALL : TailCall<J>;
-def TAILCALL_R : TailCallReg<GPR32Opnd, JR>;
+
+let AdditionalPredicates = [NotInMicroMips] in {
+def TAILCALL : TailCall<J>, ISA_MIPS1_NOT_32R6_64R6;
+def TAILCALLREG : TailCallReg<GPR32Opnd, JR>, ISA_MIPS1_NOT_32R6_64R6, GPR_32;
+}
 
 // Indirect branches are matched as PseudoIndirectBranch/PseudoIndirectBranch64
 // then are expanded to JR, JR64, JALR, or JALR64 depending on the ISA.
diff --git a/llvm/lib/Target/Mips/MipsSEISelLowering.cpp b/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
index 80c000d5746..73c461ba9f0 100644
--- a/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
@@ -27,8 +27,8 @@ using namespace llvm;
 #define DEBUG_TYPE "mips-isel"
 
 static cl::opt<bool>
-EnableMipsTailCalls("enable-mips-tail-calls", cl::Hidden,
-                    cl::desc("MIPS: Enable tail calls."), cl::init(false));
+UseMipsTailCalls("mips-tail-calls", cl::Hidden,
+                    cl::desc("MIPS: permit tail calls."), cl::init(true));
 
 static cl::opt<bool> NoDPLoadStore("mno-ldc1-sdc1", cl::init(false),
                                    cl::desc("Expand double precision loads and "
@@ -1178,7 +1178,7 @@ MipsSETargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
 bool MipsSETargetLowering::isEligibleForTailCallOptimization(
     const CCState &CCInfo, unsigned NextStackOffset,
     const MipsFunctionInfo &FI) const {
-  if (!EnableMipsTailCalls)
+  if (!UseMipsTailCalls)
     return false;
 
   // Exception has to be cleared with eret.
diff --git a/llvm/test/CodeGen/Mips/biggot.ll b/llvm/test/CodeGen/Mips/biggot.ll
index aafaf55ae18..94940548ead 100644
--- a/llvm/test/CodeGen/Mips/biggot.ll
+++ b/llvm/test/CodeGen/Mips/biggot.ll
@@ -1,11 +1,12 @@
 ; RUN: llc -march=mipsel -mxgot -relocation-model=pic < %s | FileCheck %s -check-prefix=O32
-; RUN: llc -march=mips64el -mcpu=mips64r2 -mattr=+n64 -mxgot -relocation-model=pic < %s | \
+; RUN: llc -march=mips64el -mcpu=mips64r2 -mxgot -relocation-model=pic < %s | \
 ; RUN: FileCheck %s -check-prefix=N64
 
 @v0 = external global i32
 
 define void @foo1() nounwind {
 entry:
+; O32-LABEL: foo1:
 ; O32: lui $[[R0:[0-9]+]], %got_hi(v0)
 ; O32: addu  $[[R1:[0-9]+]], $[[R0]], ${{[a-z0-9]+}}
 ; O32: lw  ${{[0-9]+}}, %got_lo(v0)($[[R1]])
@@ -13,11 +14,12 @@ entry:
 ; O32: addu  $[[R3:[0-9]+]], $[[R2]], ${{[a-z0-9]+}}
 ; O32: lw  ${{[0-9]+}}, %call_lo(foo0)($[[R3]])
 
+; N64-LABEL: foo1:
 ; N64: lui $[[R0:[0-9]+]], %got_hi(v0)
 ; N64: daddu  $[[R1:[0-9]+]], $[[R0]], ${{[a-z0-9]+}}
-; N64: ld  ${{[0-9]+}}, %got_lo(v0)($[[R1]])
 ; N64: lui $[[R2:[0-9]+]], %call_hi(foo0)
 ; N64: daddu  $[[R3:[0-9]+]], $[[R2]], ${{[a-z0-9]+}}
+; N64: ld  ${{[0-9]+}}, %got_lo(v0)($[[R1]])
 ; N64: ld  ${{[0-9]+}}, %call_lo(foo0)($[[R3]])
 
   %0 = load i32, i32* @v0, align 4
diff --git a/llvm/test/CodeGen/Mips/brdelayslot.ll b/llvm/test/CodeGen/Mips/brdelayslot.ll
index 3a5d4dcded2..136cc6534aa 100644
--- a/llvm/test/CodeGen/Mips/brdelayslot.ll
+++ b/llvm/test/CodeGen/Mips/brdelayslot.ll
@@ -22,7 +22,7 @@ entry:
 ; None: nop
 ; None: .end
 
-  tail call void @foo2(i32 3) nounwind
+  call void @foo2(i32 3) nounwind
   ret void
 }
 
@@ -37,7 +37,7 @@ entry:
 ; Default:     cvt.d.w
 
   %conv = sitofp i32 %a to double
-  tail call void @foo4(double %conv) nounwind
+  call void @foo4(double %conv) nounwind
   ret void
 }
 
@@ -83,7 +83,7 @@ if.end:
 
 define void @foo6(float %a0, double %a1) nounwind {
 entry:
-  tail call void @foo7(double %a1, float %a0) nounwind
+  call void @foo7(double %a1, float %a0) nounwind
   ret void
 }
 
@@ -101,7 +101,7 @@ define i32 @foo8(i32 %a) nounwind {
 entry:
   store i32 %a, i32* @g1, align 4
   %0 = load void ()*, void ()** @foo9, align 4
-  tail call void %0() nounwind
+  call void %0() nounwind
   %1 = load i32, i32* @g1, align 4
   %add = add nsw i32 %1, %a
   ret i32 %add
diff --git a/llvm/test/CodeGen/Mips/cconv/reserved-space.ll b/llvm/test/CodeGen/Mips/cconv/reserved-space.ll
index a17377ba0e5..0a922345ddf 100644
--- a/llvm/test/CodeGen/Mips/cconv/reserved-space.ll
+++ b/llvm/test/CodeGen/Mips/cconv/reserved-space.ll
@@ -17,7 +17,7 @@ declare void @foo() nounwind;
 
 define void @reserved_space() nounwind {
 entry:
-        tail call void @foo()
+        call void @foo()
         ret void
 }
 
diff --git a/llvm/test/CodeGen/Mips/fastcc.ll b/llvm/test/CodeGen/Mips/fastcc.ll
index 4b6191f9ae8..bf082f630de 100644
--- a/llvm/test/CodeGen/Mips/fastcc.ll
+++ b/llvm/test/CodeGen/Mips/fastcc.ll
@@ -290,7 +290,7 @@ entry:
 ; NOODDSPREG-DAG:    lwc1    $f18, 36($[[R0]])
 
 ; NOODDSPREG-DAG:    lwc1    $[[F0:f[0-9]*[02468]]], 40($[[R0]])
-; NOODDSPREG-DAG:    swc1    $[[F0]], 0($sp)
+; NOODDSPREG-DAG:    swc1    $[[F0]], 8($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
diff --git a/llvm/test/CodeGen/Mips/gpreg-lazy-binding.ll b/llvm/test/CodeGen/Mips/gpreg-lazy-binding.ll
index 98849ded707..0330d7c36ef 100644
--- a/llvm/test/CodeGen/Mips/gpreg-lazy-binding.ll
+++ b/llvm/test/CodeGen/Mips/gpreg-lazy-binding.ll
@@ -6,7 +6,7 @@
 ; CHECK:     jalr  $25
 ; CHECK:     nop
 ; CHECK-NOT: move  $gp
-; CHECK:     jalr  $25
+; CHECK:     jr    $25
 
 define void @f0() nounwind {
 entry:
diff --git a/llvm/test/CodeGen/Mips/i64arg.ll b/llvm/test/CodeGen/Mips/i64arg.ll
index 22a0c1f51f7..036b88e4c77 100644
--- a/llvm/test/CodeGen/Mips/i64arg.ll
+++ b/llvm/test/CodeGen/Mips/i64arg.ll
@@ -2,14 +2,14 @@
 
 define void @f1(i64 %ll1, float %f, i64 %ll, i32 %i, float %f2) nounwind {
 entry:
-; CHECK-DAG: lw $[[R2:[0-9]+]], 80($sp)
-; CHECK-DAG: lw $[[R3:[0-9]+]], 84($sp)
+; CHECK-DAG: lw $[[R2:[0-9]+]], 64($sp)
+; CHECK-DAG: lw $[[R3:[0-9]+]], 68($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
+; CHECK: jalr $25
   tail call void @ff1(i32 %i, i64 1085102592623924856) nounwind
 ; CHECK-DAG: lw $25, %call16(ff2)
 ; CHECK-DAG: move $4, $[[R2]]
@@ -18,11 +18,11 @@ entry:
   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 $[[R1]], 28($sp)
-; CHECK-DAG: sw $[[R0]], 24($sp)
+; CHECK-DAG: sw $[[R1]], 76($sp)
+; CHECK-DAG: sw $[[R0]], 72($sp)
 ; CHECK-DAG: move $6, $[[R2]]
 ; CHECK-DAG: move $7, $[[R3]]
-; CHECK:     jalr $25
+; CHECK:     jr $25
   tail call void @ff3(i32 %i, i64 %ll, i32 %sub, i64 %ll1) nounwind
   ret void
 }
diff --git a/llvm/test/CodeGen/Mips/indirectcall.ll b/llvm/test/CodeGen/Mips/indirectcall.ll
index ac565d64667..97c29d09dc9 100644
--- a/llvm/test/CodeGen/Mips/indirectcall.ll
+++ b/llvm/test/CodeGen/Mips/indirectcall.ll
@@ -2,7 +2,7 @@
 
 define void @foo0(void (i32)* nocapture %f1) nounwind {
 entry:
-; CHECK: jalr $25
+; CHECK: jr $25
   tail call void %f1(i32 13) nounwind
   ret void
 }
diff --git a/llvm/test/CodeGen/Mips/lazy-binding.ll b/llvm/test/CodeGen/Mips/lazy-binding.ll
index 87040cc2b3a..2e2f35064c4 100644
--- a/llvm/test/CodeGen/Mips/lazy-binding.ll
+++ b/llvm/test/CodeGen/Mips/lazy-binding.ll
@@ -30,7 +30,7 @@ declare void @foo2()
 ; CHECK: lw $25, %call16(foo2)(${{[0-9]+}})
 ; CHECK: jalr $25
 ; CHECK: lw $25, %call16(foo2)(${{[0-9]+}})
-; CHECK: jalr $25
+; CHECK: jr $25
 
 define void @foo1() {
 entry:
diff --git a/llvm/test/CodeGen/Mips/llvm-ir/call.ll b/llvm/test/CodeGen/Mips/llvm-ir/call.ll
index 0d524d43960..1d0c76877dd 100644
--- a/llvm/test/CodeGen/Mips/llvm-ir/call.ll
+++ b/llvm/test/CodeGen/Mips/llvm-ir/call.ll
@@ -1,18 +1,29 @@
 ; Test the 'call' instruction and the tailcall variant.
 
-; FIXME: We should remove the need for -enable-mips-tail-calls
-; RUN: llc -march=mips   -mcpu=mips32   -relocation-model=pic -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
-; RUN: llc -march=mips   -mcpu=mips32r2 -relocation-model=pic -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
-; RUN: llc -march=mips   -mcpu=mips32r3 -relocation-model=pic -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
-; RUN: llc -march=mips   -mcpu=mips32r5 -relocation-model=pic -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
-; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -disable-mips-delay-filler -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,R6C
-; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -mattr=+fp64,+nooddspreg -disable-mips-delay-filler -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,O32,R6C
-; RUN: llc -march=mips64 -mcpu=mips4    -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
-; RUN: llc -march=mips64 -mcpu=mips64   -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
-; RUN: llc -march=mips64 -mcpu=mips64r2 -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
-; RUN: llc -march=mips64 -mcpu=mips64r3 -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
-; RUN: llc -march=mips64 -mcpu=mips64r5 -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
-; RUN: llc -march=mips64 -mcpu=mips64r6 -disable-mips-delay-filler -enable-mips-tail-calls < %s | FileCheck %s -check-prefixes=ALL,N64,R6C
+; RUN: llc -march=mips   -mcpu=mips32   -relocation-model=pic  < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r2 -relocation-model=pic  < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r3 -relocation-model=pic  < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r5 -relocation-model=pic  < %s | FileCheck %s -check-prefixes=ALL,O32,NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -disable-mips-delay-filler  < %s | FileCheck %s -check-prefixes=ALL,O32,R6C
+; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -mattr=+fp64,+nooddspreg -disable-mips-delay-filler  < %s | FileCheck %s -check-prefixes=ALL,O32,R6C
+; RUN: llc -march=mips64 -mcpu=mips4     < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64    < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r2  < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r3  < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r5  < %s | FileCheck %s -check-prefixes=ALL,N64,NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r6 -disable-mips-delay-filler  < %s | FileCheck %s -check-prefixes=ALL,N64,R6C
+; RUN: llc -march=mips   -mcpu=mips32   -relocation-model=pic  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r2 -relocation-model=pic  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r3 -relocation-model=pic  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r5 -relocation-model=pic  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=NOT-R6C
+; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -disable-mips-delay-filler  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=R6C
+; RUN: llc -march=mips   -mcpu=mips32r6 -relocation-model=pic -mattr=+fp64,+nooddspreg -disable-mips-delay-filler  < %s | FileCheck %s -check-prefix=ALL -check-prefix=O32 -check-prefix=R6C
+; RUN: llc -march=mips64 -mcpu=mips4     < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64    < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r2  < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r3  < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r5  < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=NOT-R6C
+; RUN: llc -march=mips64 -mcpu=mips64r6 -disable-mips-delay-filler  < %s | FileCheck %s -check-prefix=ALL -check-prefix=N64 -check-prefix=R6C
 
 declare void @extern_void_void()
 declare i32 @extern_i32_void()
@@ -76,7 +87,8 @@ define void @musttail_call_void_void() {
 
 ; N64:           ld $[[TGT:[0-9]+]], %call16(extern_void_void)($gp)
 
-; ALL:           jr $[[TGT]]
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   musttail call void @extern_void_void()
   ret void
@@ -89,7 +101,8 @@ define i32 @musttail_call_i32_void() {
 
 ; N64:           ld $[[TGT:[0-9]+]], %call16(extern_i32_void)($gp)
 
-; ALL:           jr $[[TGT]]
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   %1 = musttail call i32 @extern_i32_void()
   ret i32 %1
@@ -102,7 +115,8 @@ define float @musttail_call_float_void() {
 
 ; N64:           ld $[[TGT:[0-9]+]], %call16(extern_float_void)($gp)
 
-; ALL:           jr $[[TGT]]
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   %1 = musttail call float @extern_float_void()
   ret float %1
@@ -154,7 +168,8 @@ define void @tail_indirect_call_void_void(void ()* %addr) {
 ; ALL-LABEL: tail_indirect_call_void_void:
 
 ; ALL:           move $25, $4
-; ALL:           jr $25
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   tail call void %addr()
   ret void
@@ -164,7 +179,8 @@ define i32 @tail_indirect_call_i32_void(i32 ()* %addr) {
 ; ALL-LABEL: tail_indirect_call_i32_void:
 
 ; ALL:           move $25, $4
-; ALL:           jr $25
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   %1 = tail call i32 %addr()
   ret i32 %1
@@ -174,7 +190,8 @@ define float @tail_indirect_call_float_void(float ()* %addr) {
 ; ALL-LABEL: tail_indirect_call_float_void:
 
 ; ALL:           move $25, $4
-; ALL:           jr $25
+; NOT-R6C:       jr   $[[TGT]]
+; R6C:           jrc  $[[TGT]]
 
   %1 = tail call float %addr()
   ret float %1
@@ -188,7 +205,8 @@ define hidden void @thunk_undef_double(i32 %this, double %volume) unnamed_addr a
 ; ALL-LABEL: thunk_undef_double:
 ; O32: # implicit-def: %A2
 ; O32: # implicit-def: %A3
-; ALL:        jr $25
+; NOT-R6C:    jr   $[[TGT]]
+; R6C:        jrc  $[[TGT]]
 
   tail call void @undef_double(i32 undef, double undef) #8
   ret void
diff --git a/llvm/test/CodeGen/Mips/nacl-branch-delay.ll b/llvm/test/CodeGen/Mips/nacl-branch-delay.ll
index 2927f39a416..b0d7e9c85d9 100644
--- a/llvm/test/CodeGen/Mips/nacl-branch-delay.ll
+++ b/llvm/test/CodeGen/Mips/nacl-branch-delay.ll
@@ -44,7 +44,7 @@ define void @test1() {
 
 define void @test2() {
   store i32 1, i32* @x, align 4
-  tail call void @f2()
+  call void @f2()
   ret void
 
 
diff --git a/llvm/test/CodeGen/Mips/tailcall.ll b/llvm/test/CodeGen/Mips/tailcall.ll
index 61f8e508f40..1db780c5565 100644
--- a/llvm/test/CodeGen/Mips/tailcall.ll
+++ b/llvm/test/CodeGen/Mips/tailcall.ll
@@ -1,13 +1,32 @@
-; RUN: llc -march=mipsel -relocation-model=pic -enable-mips-tail-calls \
+; RUN: llc -march=mipsel -relocation-model=pic  \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s -check-prefix=PIC32
-; RUN: llc -march=mipsel -relocation-model=static -enable-mips-tail-calls \
+; RUN: llc -march=mipsel -relocation-model=static  \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s -check-prefix=STATIC32
-; RUN: llc -march=mips64el -mcpu=mips64r2 -enable-mips-tail-calls \
+; RUN: llc -march=mips64el -mcpu=mips64r2  \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s -check-prefix=N64
 ; RUN: llc -march=mipsel -mattr=mips16 -relocation-model=pic \
-; RUN:     -enable-mips-tail-calls -verify-machineinstrs < %s | \
+; RUN:      -verify-machineinstrs < %s | \
 ; RUN:     FileCheck %s -check-prefix=PIC16
 
+; RUN: llc -march=mipsel -relocation-model=pic -mattr=+micromips < %s | \
+; RUN:     FileCheck %s -check-prefix=PIC32
+; RUN: llc -march=mipsel -relocation-model=static -mattr=+micromips \
+; RUN:     < %s | FileCheck %s -check-prefix=STATIC32
+
+; RUN: llc -march=mipsel -relocation-model=pic -mcpu=mips32r6 < %s | \
+; RUN:     FileCheck %s -check-prefix=PIC32
+; RUN: llc -march=mipsel -relocation-model=static -mcpu=mips32r6 \
+; RUN:     < %s | FileCheck %s -check-prefix=STATIC32
+; RUN: llc -march=mips64el -mcpu=mips64r6  \
+; RUN:     < %s | FileCheck %s -check-prefix=N64
+
+; RUN: llc -march=mipsel -relocation-model=pic -mcpu=mips32r6 -mattr=+micromips \
+; RUN:      < %s | FileCheck %s -check-prefix=PIC32
+; RUN: llc -march=mipsel -relocation-model=static -mcpu=mips32r6 \
+; RUN:     -mattr=+micromips < %s | FileCheck %s -check-prefix=STATIC32
+; RUN: llc -march=mips64el -mcpu=mips64r6 -mattr=+micromips < %s \
+; RUN:      | FileCheck %s -check-prefix=N64
+
 @g0 = common global i32 0, align 4
 @g1 = common global i32 0, align 4
 @g2 = common global i32 0, align 4
@@ -74,13 +93,13 @@ declare i32 @callee4(i32, i32, i32, i32, i32, i32, i32, i32, i32)
 define i32 @caller5() nounwind readonly {
 entry:
 ; PIC32: .ent caller5
-; PIC32-NOT: jalr
+; PIC32-NOT: jalr $25
 ; PIC32: .end caller5
 ; STATIC32: .ent caller5
 ; STATIC32-NOT: jal
 ; STATIC32: .end caller5
 ; N64: .ent caller5
-; N64-NOT: jalr
+; N64-NOT: jalr $25
 ; N64: .end caller5
 ; PIC16: .ent caller5
 ; PIC16: jalrc
@@ -118,6 +137,19 @@ declare i32 @callee8(i32, ...)
 
 define i32 @caller8_0() nounwind {
 entry:
+; PIC32: .ent caller8_0
+; PIC32: jr
+; PIC32: .end caller8_0
+; STATIC32: .ent caller8_0
+; STATIC32: j
+; STATIC32: .end caller8_0
+; N64: .ent caller8_0
+; N64-NOT: jalr $25
+; N64: .end caller8_0
+; PIC16: .ent caller8_0
+; PIC16: jalrc
+; PIC16: .end caller8_0
+
   %call = tail call fastcc i32 @caller8_1()
   ret i32 %call
 }
@@ -131,7 +163,7 @@ entry:
 ; STATIC32: jal
 ; STATIC32: .end caller8_1
 ; N64: .ent caller8_1
-; N64-NOT: jalr
+; N64-NOT: jalr $25
 ; N64: .end caller8_1
 ; PIC16: .ent caller8_1
 ; PIC16: jalrc
@@ -149,6 +181,18 @@ 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
+; STATIC32: j
+; STATIC32: .end caller9_0
+; N64: .ent caller9_0
+; N64-NOT: jalr $25
+; N64: .end caller9_0
+; PIC16: .ent caller9_0
+; PIC16: jalrc
+; PIC16: .end caller9_0
   %call = tail call fastcc i32 @caller9_1()
   ret i32 %call
 }
@@ -177,11 +221,11 @@ 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
+; PIC32-NOT: jalr $25
 ; STATIC32: .ent caller10
 ; STATIC32-NOT: jal
 ; N64: .ent caller10
-; N64-NOT: jalr
+; N64-NOT: jalr $25
 ; PIC16: .ent caller10
 ; PIC16: jalrc
 
@@ -236,7 +280,7 @@ entry:
 ; STATIC32: .ent caller13
 ; STATIC32-NOT: jal
 ; N64: .ent caller13
-; N64-NOT: jalr
+; N64-NOT: jalr $25
 ; PIC16: .ent caller13
 ; PIC16: jalrc