[SVE] Add fatal error for unnamed SVE variadic arguments

[david-arm]

Summary

We don't currently support passing unnamed variadic SVE arguments
so I've added a fatal error if we hit such cases to prevent any
silent ABI issues in future.

[sdesmalen]

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

4802

Does this also report a fatal error if ArgVT is a named argument?

i.e. declare i32 @sve_printf(i8*, <vscale x 4 x i32> %v, ...) and then calling with:

call i32 (i8*, <vscale x 4 x i32>, ...) @sve_printf(i8* %f, <vscale x 4 x i32> %x)

[efriedma]

We should have corresponding handling for va_arg. (clang doesn't produce va_arg instructions, but other frontends do.)

[david-arm]

• Added fatal errors on the callee side when one of the variadic arguments is a scalable vector.

[efriedma]

LGTM

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
 
   if (IsVarArg) {
     // Handle fixed and variable vector arguments differently.
     // Variable vector arguments always go into memory.
     unsigned NumArgs = Outs.size();
 
     for (unsigned i = 0; i != NumArgs; ++i) {
       MVT ArgVT = Outs[i].VT;
+      if (!Outs[i].IsFixed && ArgVT.isScalableVector())
+        report_fatal_error("Passing SVE types to variadic functions is "
+                           "currently not supported");
+
       ISD::ArgFlagsTy ArgFlags = Outs[i].Flags;
       CCAssignFn *AssignFn = CCAssignFnForCall(CallConv,
                                                /*IsVarArg=*/ !Outs[i].IsFixed);
       bool Res = AssignFn(i, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, CCInfo);
       assert(!Res && "Call operand has unhandled type");
       (void)Res;
     }
   } else {

llvm/test/CodeGen/AArch64/sve-varargs-broken.ll

+; RUN: not --crash llc -mtriple aarch64-linux-gnu -mattr=+sve <%s 2>&1 | FileCheck %s
+
+declare i32 @sve_printf(i8*, <vscale x 4 x i32>, ...)
+
+@.str_1 = internal constant [6 x i8] c"boo!\0A\00"
+
+; CHECK: Passing SVE types to variadic functions is currently not supported
+define void @foo(<vscale x 4 x i32> %x) {
+  %f = getelementptr [6 x i8], [6 x i8]* @.str_1, i64 0, i64 0
+  call i32 (i8*, <vscale x 4 x i32>, ...) @sve_printf(i8* %f, <vscale x 4 x i32> %x, <vscale x 4 x i32> %x)
+  ret void
+}

llvm/test/CodeGen/AArch64/sve-varargs.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64--linux-gnu -mattr=+sve < %s 2>%t | FileCheck %s
+; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t
+
+; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it.
+; WARN-NOT: warning
+
+declare i32 @sve_printf(i8*, <vscale x 4 x i32>, ...)
+
+@.str_1 = internal constant [6 x i8] c"boo!\0A\00"
+
+define void @foo(<vscale x 4 x i32> %x) {
+; CHECK-LABEL: foo:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -16
+; CHECK-NEXT:    adrp x0, .str_1
+; CHECK-NEXT:    add x0, x0, :lo12:.str_1
+; CHECK-NEXT:    bl sve_printf
+; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %f = getelementptr [6 x i8], [6 x i8]* @.str_1, i64 0, i64 0
+  call i32 (i8*, <vscale x 4 x i32>, ...) @sve_printf(i8* %f, <vscale x 4 x i32> %x)
+  ret void
+}