[WebAssembly] Properly align fp128 arguments in outgoing varargs arguments

[sunfish]

Summary

For outgoing varargs arguments, it's necessary to check the OrigAlign field of the corresponding OutputArg entry to determine argument alignment, rather than just computing an alignment from the argument value type. This is because types like fp128 are split into multiple argument values, with narrower types that don't reflect the ABI alignment of the full fp128.

This fixes the printf("printfL: %4.*Lf\n", 2, lval); testcase.

[jgravelle-google]

Nice, lgtm

lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

677

Is it this diff or is the spacing off here?

[sunfish]

lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

677

I was somehow in 4-space indent mode there. Fixed.

lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

WebAssemblyTargetLowering::LowerCall(CallLoweringInfo &CLI,
 
   // Analyze operands of the call, assigning locations to each operand.
   SmallVector<CCValAssign, 16> ArgLocs;
   CCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
 
   if (IsVarArg) {
     // Outgoing non-fixed arguments are placed in a buffer. First
     // compute their offsets and the total amount of buffer space needed.
-    for (SDValue Arg :
-         make_range(OutVals.begin() + NumFixedArgs, OutVals.end())) {
+    for (unsigned I = NumFixedArgs; I < Outs.size(); ++I) {
+        const ISD::OutputArg &Out = Outs[I];
+        SDValue &Arg = OutVals[I];
       EVT VT = Arg.getValueType();
       assert(VT != MVT::iPTR && "Legalized args should be concrete");
       Type *Ty = VT.getTypeForEVT(*DAG.getContext());
-      unsigned Offset = CCInfo.AllocateStack(Layout.getTypeAllocSize(Ty),
-                                             Layout.getABITypeAlignment(Ty));
+      unsigned Align = std::max(Out.Flags.getOrigAlign(),
+      unsigned Offset = CCInfo.AllocateStack(Layout.getTypeAllocSize(Ty),
+                                             Align);
       CCInfo.addLoc(CCValAssign::getMem(ArgLocs.size(), VT.getSimpleVT(),
                                         Offset, VT.getSimpleVT(),
                                         CCValAssign::Full));
     }
   }
 
   unsigned NumBytes = CCInfo.getAlignedCallFrameSize();
 

test/CodeGen/WebAssembly/varargs.ll

 ; Test a definition a varargs function with a non-legal fixed argument.
 
 ; CHECK-LABEL: nonlegal_fixed:
 ; CHECK-NEXT: .functype nonlegal_fixed (i64, i64, i32) -> (){{$}}
 define void @nonlegal_fixed(fp128 %x, ...) nounwind {
   ret void
 }
 
+; Test that an fp128 argument is properly aligned and allocated
+; within a vararg buffer.
+
+; CHECK-LABEL: call_fp128_alignment:
+; CHECK:      global.get      $push7=, __stack_pointer@GLOBAL
+; CHECK-NEXT: i32.const       $push8=, 32
+; CHECK-NEXT: i32.sub         $push12=, $pop7, $pop8
+; CHECK-NEXT: local.tee       $push11=, $1=, $pop12
+; CHECK-NEXT: global.set      __stack_pointer@GLOBAL, $pop11
+; CHECK-NEXT: i32.const       $push0=, 24
+; CHECK-NEXT: i32.add         $push1=, $1, $pop0
+; CHECK-NEXT: i64.const       $push2=, -9223372036854775808
+; CHECK-NEXT: i64.store       0($pop1), $pop2
+; CHECK-NEXT: i32.const       $push3=, 16
+; CHECK-NEXT: i32.add         $push4=, $1, $pop3
+; CHECK-NEXT: i64.const       $push5=, 1
+; CHECK-NEXT: i64.store       0($pop4), $pop5
+; CHECK-NEXT: i32.const       $push6=, 7
+; CHECK-NEXT: i32.store       0($1), $pop6
+; CHECK-NEXT: call            callee@FUNCTION, $1
+define void @call_fp128_alignment(i8* %p) {
+entry:
+  call void (...) @callee(i8 7, fp128 0xL00000000000000018000000000000000)
+  ret void
+}
+
 declare void @llvm.va_start(i8*)
 declare void @llvm.va_end(i8*)
 declare void @llvm.va_copy(i8*, i8*)