[MIPS][ISel] Fix musttail

llvm/lib/Target/Mips/Mips16ISelLowering.cpp

@@ -247,8 +247,8 @@ Mips16TargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
 }
 
 bool Mips16TargetLowering::isEligibleForTailCallOptimization(
-    const CCState &CCInfo, unsigned NextStackOffset,
-    const MipsFunctionInfo &FI) const {
+    const CCState &CCInfo, unsigned NextStackOffset, const MipsFunctionInfo &FI,
+    bool IsMustTail) const {
   // No tail call optimization for mips16.
   return false;
 }

llvm/lib/Target/Mips/Mips16ISelLowering.h

@@ -33,7 +33,7 @@ namespace llvm {
   private:
     bool isEligibleForTailCallOptimization(
         const CCState &CCInfo, unsigned NextStackOffset,
-        const MipsFunctionInfo &FI) const override;
+        const MipsFunctionInfo &FI, bool IsMustTail) const override;
 
     void setMips16HardFloatLibCalls();
 

llvm/lib/Target/Mips/MipsISelLowering.cpp

@@ -3404,21 +3404,26 @@ MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
     CSInfo = MachineFunction::CallSiteInfo(*CB);
 
-  // Check if it's really possible to do a tail call. Restrict it to functions
-  // that are part of this compilation unit.
+  // Check if it's really possible to do a tail call.
+  // For non-musttail calls, restrict to functions that won't require $gp
+  // restoration. In PIC mode, calling external functions via tail call can
+  // cause issues with $gp register handling (see D24763).
   bool InternalLinkage = false;
+  bool IsMustTail = CLI.CB && CLI.CB->isMustTailCall();
   if (IsTailCall) {
     IsTailCall = isEligibleForTailCallOptimization(
-        CCInfo, StackSize, *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());
-     }
+        CCInfo, StackSize, *MF.getInfo<MipsFunctionInfo>(), IsMustTail);
+    // For non-musttail calls, restrict to local or non-interposable functions
+    if (IsTailCall && !IsMustTail) {
+      if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
+        InternalLinkage = G->getGlobal()->hasInternalLinkage();
+        IsTailCall &= (G->getGlobal()->hasLocalLinkage() ||
+                       G->getGlobal()->hasHiddenVisibility() ||
+                       G->getGlobal()->hasProtectedVisibility());
+      }
+    }
   }
-  if (!IsTailCall && CLI.CB && CLI.CB->isMustTailCall())
+  if (!IsTailCall && IsMustTail)
     report_fatal_error("failed to perform tail call elimination on a call "
                        "site marked musttail");
 

llvm/lib/Target/Mips/MipsISelLowering.h

@@ -595,10 +595,9 @@ class TargetRegisterClass;
 
     /// isEligibleForTailCallOptimization - Check whether the call is eligible
     /// for tail call optimization.
-    virtual bool
-    isEligibleForTailCallOptimization(const CCState &CCInfo,
-                                      unsigned NextStackOffset,
-                                      const MipsFunctionInfo &FI) const = 0;
+    virtual bool isEligibleForTailCallOptimization(
+        const CCState &CCInfo, unsigned NextStackOffset,
+        const MipsFunctionInfo &FI, bool IsMustTail) const = 0;
 
     /// copyByValArg - Copy argument registers which were used to pass a byval
     /// argument to the stack. Create a stack frame object for the byval

llvm/lib/Target/Mips/MipsSEISelLowering.cpp

@@ -1179,9 +1179,9 @@ MipsSETargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
 }
 
 bool MipsSETargetLowering::isEligibleForTailCallOptimization(
-    const CCState &CCInfo, unsigned NextStackOffset,
-    const MipsFunctionInfo &FI) const {
-  if (!UseMipsTailCalls)
+    const CCState &CCInfo, unsigned NextStackOffset, const MipsFunctionInfo &FI,
+    bool IsMustTail) const {
+  if (!UseMipsTailCalls && !IsMustTail)
     return false;
 
   // Exception has to be cleared with eret.

llvm/lib/Target/Mips/MipsSEISelLowering.h

@@ -65,7 +65,7 @@ class TargetRegisterClass;
   private:
     bool isEligibleForTailCallOptimization(
         const CCState &CCInfo, unsigned NextStackOffset,
-        const MipsFunctionInfo &FI) const override;
+        const MipsFunctionInfo &FI, bool IsMustTail) const override;
 
     void
     getOpndList(SmallVectorImpl<SDValue> &Ops,

llvm/test/CodeGen/Mips/musttail.ll

@@ -0,0 +1,97 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=mips-unknown-linux-gnu < %s | FileCheck %s --check-prefix=MIPS32
+; RUN: llc -mtriple=mips64-unknown-linux-gnu < %s | FileCheck %s --check-prefix=MIPS64
+
+; Test musttail support for MIPS
+
+define dso_local i32 @callee_args(i32 %a, i32 %b, i32 %c) {
+  ret i32 %a;
+}
+
+define i32 @test_musttail_args(i32 %x, i32 %y, i32 %z) {
+; MIPS32-LABEL: test_musttail_args:
+; MIPS32:       # %bb.0:
+; MIPS32-NEXT:    j callee_args
+; MIPS32-NEXT:    nop
+;
+; MIPS64-LABEL: test_musttail_args:
+; MIPS64:       # %bb.0:
+; MIPS64-NEXT:    j callee_args
+; MIPS64-NEXT:    nop
+  %ret = musttail call i32 @callee_args(i32 %x, i32 %y, i32 %z)
+  ret i32 %ret
+}
+
+; Test musttail with many arguments that spill to stack (involves memory)
+; MIPS O32 ABI: first 4 args in $a0-$a3, rest on stack
+declare i32 @many_args_callee(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h)
+
+define i32 @test_musttail_many_args(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h) {
+; MIPS32-LABEL: test_musttail_many_args:
+; MIPS32:       # %bb.0:
+; MIPS32-NEXT:    lw $1, 24($sp)
+; MIPS32-NEXT:    lw $2, 20($sp)
+; MIPS32-NEXT:    lw $3, 16($sp)
+; MIPS32-NEXT:    sw $3, 16($sp)
+; MIPS32-NEXT:    sw $2, 20($sp)
+; MIPS32-NEXT:    sw $1, 24($sp)
+; MIPS32-NEXT:    lw $1, 28($sp)
+; MIPS32-NEXT:    j many_args_callee
+; MIPS32-NEXT:    sw $1, 28($sp)
+;
+; MIPS64-LABEL: test_musttail_many_args:
+; MIPS64:       # %bb.0:
+; MIPS64-NEXT:    j many_args_callee
+; MIPS64-NEXT:    nop
+  %ret = musttail call i32 @many_args_callee(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h)
+  ret i32 %ret
+}
+
+; Test musttail with large struct passed by value (involves memory)
+%struct.large = type { i32, i32, i32, i32, i32, i32, i32, i32 }
+
+declare i32 @callee_with_struct(%struct.large %s, i32 %x)
+
+define i32 @test_musttail_struct(%struct.large %s, i32 %x) {
+; MIPS32-LABEL: test_musttail_struct:
+; MIPS32:       # %bb.0:
+; MIPS32-NEXT:    lw $1, 28($sp)
+; MIPS32-NEXT:    lw $2, 24($sp)
+; MIPS32-NEXT:    lw $3, 20($sp)
+; MIPS32-NEXT:    lw $8, 16($sp)
+; MIPS32-NEXT:    sw $8, 16($sp)
+; MIPS32-NEXT:    sw $3, 20($sp)
+; MIPS32-NEXT:    sw $2, 24($sp)
+; MIPS32-NEXT:    sw $1, 28($sp)
+; MIPS32-NEXT:    lw $1, 32($sp)
+; MIPS32-NEXT:    j callee_with_struct
+; MIPS32-NEXT:    sw $1, 32($sp)
+;
+; MIPS64-LABEL: test_musttail_struct:
+; MIPS64:       # %bb.0:
+; MIPS64-NEXT:    ld $1, 0($sp)
+; MIPS64-NEXT:    j callee_with_struct
+; MIPS64-NEXT:    sd $1, 0($sp)
+  %ret = musttail call i32 @callee_with_struct(%struct.large %s, i32 %x)
+  ret i32 %ret
+}
+
+; Test musttail with mixed int and float arguments that use stack
+declare float @mixed_args_callee(i32 %a, float %b, i32 %c, float %d, i32 %e, float %f)
+
+define float @test_musttail_mixed_args(i32 %a, float %b, i32 %c, float %d, i32 %e, float %f) {
+; MIPS32-LABEL: test_musttail_mixed_args:
+; MIPS32:       # %bb.0:
+; MIPS32-NEXT:    lw $1, 16($sp)
+; MIPS32-NEXT:    sw $1, 16($sp)
+; MIPS32-NEXT:    lwc1 $f0, 20($sp)
+; MIPS32-NEXT:    j mixed_args_callee
+; MIPS32-NEXT:    swc1 $f0, 20($sp)
+;
+; MIPS64-LABEL: test_musttail_mixed_args:
+; MIPS64:       # %bb.0:
+; MIPS64-NEXT:    j mixed_args_callee
+; MIPS64-NEXT:    nop
+  %ret = musttail call float @mixed_args_callee(i32 %a, float %b, i32 %c, float %d, i32 %e, float %f)
+  ret float %ret
+}