[NVPTX] deal with all aggregate return types.

[tra]

Summary

Fixes a crash in llvm_unreachable when a function has array return type.

[tra]

renamed vector-return.ll -> aggregate-return.ll and added tests for array and structure return types.

[jlebar]

test/CodeGen/NVPTX/aggregate-return.ll

9 ↗

(On Diff #64539)

Do we need this line? Same below.

[tra]

Removed unneeded entry label and fixed names of patterns in the last test.

lib/Target/NVPTX/NVPTXAsmPrinter.cpp

if (Ty->isFloatingPointTy() || Ty->isIntegerTy()) {
         assert(Ty->isFloatingPointTy() && "Floating point type expected here");
         size = Ty->getPrimitiveSizeInBits();
       }
 
       O << ".param .b" << size << " func_retval0";
     } else if (isa<PointerType>(Ty)) {
       O << ".param .b" << TLI->getPointerTy(DL).getSizeInBits()
         << " func_retval0";
-    } else if ((Ty->getTypeID() == Type::StructTyID) || isa<VectorType>(Ty)) {
+    } else if (Ty->isAggregateType() || Ty->isVectorTy()) {
       unsigned totalsz = DL.getTypeAllocSize(Ty);
-       unsigned retAlignment = 0;
-       if (!llvm::getAlign(*F, 0, retAlignment))
-         retAlignment = DL.getABITypeAlignment(Ty);
-       O << ".param .align " << retAlignment << " .b8 func_retval0[" << totalsz
-         << "]";
     } else
       llvm_unreachable("Unknown return type");
   } else {
     SmallVector<EVT, 16> vtparts;
     ComputeValueVTs(*TLI, DL, Ty, vtparts);
     unsigned idx = 0;
     for (unsigned i = 0, e = vtparts.size(); i != e; ++i) {
       unsigned elems = 1;

test/CodeGen/NVPTX/arg-lowering.ll

 ; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
 
 ; CHECK: .visible .func  (.param .align 16 .b8 func_retval0[16]) foo0(
 ; CHECK:          .param .align 4 .b8 foo0_param_0[8]
 define <4 x float> @foo0({float, float} %arg0) {
   ret <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>
 }
 
 ; CHECK: .visible .func  (.param .align 8 .b8 func_retval0[8]) foo1(
 ; CHECK:          .param .align 8 .b8 foo1_param_0[16]
 define <2 x float> @foo1({float, float, i64} %arg0) {
   ret <2 x float> <float 1.0, float 1.0>
 }
+
+; CHECK: .visible .func  (.param .align 4 .b8 func_retval0[16]) foo2(
+; CHECK:          .param .align 4 .b8 foo2_param_0[8]
+define [4 x float] @foo2({float, float} %arg0) {
+  ret [4 x float] [float 1.0, float 1.0, float 1.0, float 1.0]
+}