Lower statepoints with multi-def targets.

[pgavlin]

Summary

Statepoint lowering currently expects that the target method of a statepoint only defines a single value. This precludes using statepoints with ABIs that return values in multiple registers (e.g. the SysV AMD64 ABI). This change adds support for lowering statepoints with mutli-def targets.

[reames]

LGTM - I can't really assess the correctness of the code for the struct return, but this doesn't appear to be breaking anything outside that use case.

[pgavlin]

Committed in r253339.

lib/CodeGen/SelectionDAG/StatepointLowering.cpp

lowerCallFromStatepoint(ImmutableStatepoint ISP, const BasicBlock *EHPadBB,
   // We are expecting DAG to have the following form:
   //
   // ch = eh_label (only in case of invoke statepoint)
   //   ch, glue = callseq_start ch
   //   ch, glue = X86::Call ch, glue
   //   ch, glue = callseq_end ch, glue
   //   get_return_value ch, glue
   //
-  // get_return_value can either be a CopyFromReg to grab the return value from
-  // %RAX, or it can be a LOAD to load a value returned by reference via a stack
-  // slot.
+  // get_return_value can either be a sequence of CopyFromReg instructions
+  // to grab the return value from the return register(s), or it can be a LOAD
+  // to load a value returned by reference via a stack slot.
 
-  if (HasDef && (CallEnd->getOpcode() == ISD::CopyFromReg ||
-                 CallEnd->getOpcode() == ISD::LOAD))
+  if (HasDef) {
+    if (CallEnd->getOpcode() == ISD::LOAD)
       CallEnd = CallEnd->getOperand(0).getNode();
+    else
+      while (CallEnd->getOpcode() == ISD::CopyFromReg)
+        CallEnd = CallEnd->getOperand(0).getNode();
+  }
 
   assert(CallEnd->getOpcode() == ISD::CALLSEQ_END && "expected!");
 
   // Export the result value if needed
   const Instruction *GCResult = ISP.getGCResult();
   if (HasDef && GCResult) {
     if (GCResult->getParent() != CS.getParent()) {
       // Result value will be used in a different basic block so we need to

test/CodeGen/X86/statepoint-call-lowering.ll

 ; RUN: llc < %s | FileCheck %s
 ; This file contains a collection of basic tests to ensure we didn't
 ; screw up normal call lowering when there are no deopt or gc arguments.
 
 target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-pc-linux-gnu"
 
+%struct = type { i64, i64 }
+
 declare zeroext i1 @return_i1()
 declare zeroext i32 @return_i32()
 declare i32* @return_i32ptr()
 declare float @return_float()
+declare %struct @return_struct()
 declare void @varargf(i32, ...)
 
 define i1 @test_i1_return() gc "statepoint-example" {
 ; CHECK-LABEL: test_i1_return
 ; This is just checking that a i1 gets lowered normally when there's no extra
 ; state arguments to the statepoint
 ; CHECK: pushq %rax
 ; CHECK: callq return_i1

 ; CHECK: popq %rax
 ; CHECK: retq
 entry:
   %safepoint_token = tail call i32 (i64, i32, float ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_f32f(i64 0, i32 0, float ()* @return_float, i32 0, i32 0, i32 0, i32 0)
   %call1 = call float @llvm.experimental.gc.result.f32(i32 %safepoint_token)
   ret float %call1
 }
 
+define %struct @test_struct_return() gc "statepoint-example" {
+; CHECK-LABEL: test_struct_return
+; CHECK: pushq %rax
+; CHECK: callq return_struct
+; CHECK: popq %rcx
+; CHECK: retq
+entry:
+  %safepoint_token = tail call i32 (i64, i32, %struct ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_structf(i64 0, i32 0, %struct ()* @return_struct, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call %struct @llvm.experimental.gc.result.struct(i32 %safepoint_token)
+  ret %struct %call1
+}
+
 define i1 @test_relocate(i32 addrspace(1)* %a) gc "statepoint-example" {
 ; CHECK-LABEL: test_relocate
 ; Check that an ununsed relocate has no code-generation impact
 ; CHECK: pushq %rax
 ; CHECK: callq return_i1
-; CHECK-NEXT: .Ltmp9:
+; CHECK-NEXT: .Ltmp11:
 ; CHECK-NEXT: popq %rdx
 ; CHECK-NEXT: retq
 entry:
   %safepoint_token = tail call i32 (i64, i32, i1 ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_i1f(i64 0, i32 0, i1 ()* @return_i1, i32 0, i32 0, i32 0, i32 0, i32 addrspace(1)* %a)
   %call1 = call i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(i32 %safepoint_token,  i32 7, i32 7)
   %call2 = call zeroext i1 @llvm.experimental.gc.result.i1(i32 %safepoint_token)
   ret i1 %call2
 }

 declare i32 @llvm.experimental.gc.result.i32(i32)
 
 declare i32 @llvm.experimental.gc.statepoint.p0f_p0i32f(i64, i32, i32* ()*, i32, i32, ...)
 declare i32* @llvm.experimental.gc.result.p0i32(i32)
 
 declare i32 @llvm.experimental.gc.statepoint.p0f_f32f(i64, i32, float ()*, i32, i32, ...)
 declare float @llvm.experimental.gc.result.f32(i32)
 
+declare i32 @llvm.experimental.gc.statepoint.p0f_structf(i64, i32, %struct ()*, i32, i32, ...)
+declare %struct @llvm.experimental.gc.result.struct(i32)
+
 declare i32 @llvm.experimental.gc.statepoint.p0f_isVoidi32varargf(i64, i32, void (i32, ...)*, i32, i32, ...)
 
 declare i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(i32, i32, i32)