[WebAssembly] Fix epilogue insertion for indirect tail calls

[tlively]

Summary

Previously epilogues were incorrectly inserted after indirect tail calls because
they did not have the isTerminator property. Add that property and test that
they get correct epilogues. To be safe, also add other properties that were
defined for direct tail calls.

[aheejin]

Someone uploaded a CL fixing this a while ago (D142615), and not sure why they didn't land it. Anyway LGTM.

[tlively]

In D146569#4211332, @aheejin wrote:

Someone uploaded a CL fixing this a while ago (D142615), and not sure why they didn't land it. Anyway LGTM.

Oh bummer, it would have saved me some time if that had landed. Oh well, I'll land this one.

llvm/lib/Target/WebAssembly/WebAssemblyInstrCall.td

 
 let isReturn = 1, isTerminator = 1, hasCtrlDep = 1, isBarrier = 1 in
 defm RET_CALL :
   I<(outs), (ins function32_op:$callee, variable_ops),
     (outs), (ins function32_op:$callee), [],
     "return_call    \t$callee", "return_call\t$callee", 0x12>,
   Requires<[HasTailCall]>;
 
-let isReturn = 1 in
+let isReturn = 1, isTerminator = 1, hasCtrlDep = 1, isBarrier = 1 in
 defm RET_CALL_INDIRECT :
   I<(outs), (ins TypeIndex:$type, table32_op:$table, variable_ops),
     (outs), (ins TypeIndex:$type, table32_op:$table), [],
     "return_call_indirect\t", "return_call_indirect\t$type, $table",
     0x13>,
   Requires<[HasTailCall]>;
 
 } // Uses = [SP32,SP64], isCall = 1

llvm/test/CodeGen/WebAssembly/tailcall.ll

 ; FAST-NEXT:    global.set __stack_pointer, $pop4
 ; FAST-NEXT:    return $1
   %a = alloca [64 x i32]
   %i = ptrtoint ptr %a to i32
   %v = tail call i32 @stack_arg_cast(i32 %i)
   ret i32 %v
 }
 
+; Checks that epilogues are inserted after return calls.
+define i32 @direct_epilogue() {
+; CHECK-LABEL: direct_epilogue:
+; CHECK:         .functype direct_epilogue () -> (i32)
+; CHECK-NEXT:  # %bb.0:
+; CHECK-NEXT:    global.get $push0=, __stack_pointer
+; CHECK-NEXT:    i32.const $push1=, 256
+; CHECK-NEXT:    i32.sub $push5=, $pop0, $pop1
+; CHECK-NEXT:    local.tee $push4=, $0=, $pop5
+; CHECK-NEXT:    global.set __stack_pointer, $pop4
+; CHECK-NEXT:    i32.const $push2=, 256
+; CHECK-NEXT:    i32.add $push3=, $0, $pop2
+; CHECK-NEXT:    global.set __stack_pointer, $pop3
+; CHECK-NEXT:    return_call direct_epilogue
+  %a = alloca [64 x i32]
+  %v = musttail call i32 @direct_epilogue()
+  ret i32 %v
+}
+
+define i32 @indirect_epilogue(ptr %p) {
+; CHECK-LABEL: indirect_epilogue:
+; CHECK:         .functype indirect_epilogue (i32) -> (i32)
+; CHECK-NEXT:  # %bb.0:
+; CHECK-NEXT:    global.get $push0=, __stack_pointer
+; CHECK-NEXT:    i32.const $push1=, 256
+; CHECK-NEXT:    i32.sub $push5=, $pop0, $pop1
+; CHECK-NEXT:    local.tee $push4=, $1=, $pop5
+; CHECK-NEXT:    global.set __stack_pointer, $pop4
+; CHECK-NEXT:    i32.const $push2=, 256
+; CHECK-NEXT:    i32.add $push3=, $1, $pop2
+; CHECK-NEXT:    global.set __stack_pointer, $pop3
+; CHECK-NEXT:    return_call_indirect , $0, $0
+  %a = alloca [64 x i32]
+  %v = musttail call i32 %p(ptr %p)
+  ret i32 %v
+}
+
 ; Check that the signatures generated for external indirectly
 ; return-called functions include the proper return types
 
 ; YAML-LABEL: - Index:           8
 ; YAML-NEXT:    ParamTypes:
 ; YAML-NEXT:      - I32
 ; YAML-NEXT:      - F32
 ; YAML-NEXT:      - I64