AMDGPU/SI: Remove calling convention assertion from LowerFormalArguments()

[• tstellarAMD]

Summary

We currently ignore the calling convention, so there is no real reason to
assert on the calling convention of functions.

[arsenm]

LGTM. Ideally we would never get a C calling convention. I think the test should also include a spir_func

llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp

SDValue SITargetLowering::LowerFormalArguments(
     SmallVectorImpl<SDValue> &InVals) const {
   const SIRegisterInfo *TRI =
       static_cast<const SIRegisterInfo *>(Subtarget->getRegisterInfo());
 
   MachineFunction &MF = DAG.getMachineFunction();
   FunctionType *FType = MF.getFunction()->getFunctionType();
   SIMachineFunctionInfo *Info = MF.getInfo<SIMachineFunctionInfo>();
 
-  assert(CallConv == CallingConv::C);
+  // FIXME: We currently assume all calling conventions are kernels.
 
   SmallVector<ISD::InputArg, 16> Splits;
   BitVector Skipped(Ins.size());
 
   for (unsigned i = 0, e = Ins.size(), PSInputNum = 0; i != e; ++i) {
     const ISD::InputArg &Arg = Ins[i];
 
     // First check if it's a PS input addr

llvm/trunk/test/CodeGen/AMDGPU/calling-conventions.ll

+; RUN: llc < %s -march=amdgcn -mcpu=SI -verify-machineinstrs | FileCheck --check-prefix=SI %s
+
+; Make sure we don't crash or assert on spir_kernel calling convention.
+
+; SI-LABEL: {{^}}kernel:
+; SI: s_endpgm
+define spir_kernel void @kernel(i32 addrspace(1)* %out) {
+entry:
+  store i32 0, i32 addrspace(1)* %out
+  ret void
+}
+
+; FIXME: This is treated like a kernel
+; SI-LABEL: {{^}}func:
+; SI: s_endpgm
+define spir_func void @func(i32 addrspace(1)* %out) {
+entry:
+  store i32 0, i32 addrspace(1)* %out
+  ret void
+}