statepoint-call-lowering: relocate array and vector

[artagnon]

Summary

This is mainly to present the direction in which future work
is going, and gather early feedback.

[reames]

This looks like it overlaps with http://reviews.llvm.org/D6915?

[artagnon]

In D6938#107800, @reames wrote:

This looks like it overlaps with http://reviews.llvm.org/D6915?

Yes, it incidentally does: D6915 is meant to test that the name mangling happens correctly, and incidentally uses gc.statepoint, gc.relocate to achieve its ends. This is specifically about testing the assembly of lowering.

Can you give me a hint about how vector relocations are supposed to look in assembly? I aim to include changes to SelectionDAG/StatepointLowering with this patch: the first step is to figure out what is wrong with the lowering.

[artagnon]

This conflates two different things:

1. Adding vector support to gc.relocate.
2. Adding iAny array support for intrinsics.

Expect two different review requests.

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

 ; CHECK: popq %rax
 ; CHECK: retq
 entry:
   %safepoint_token = tail call i32 (float ()*, i32, i32, ...)* @llvm.experimental.gc.statepoint.p0f_f32f(float ()* @return_float, i32 0, i32 0, i32 0)
   %call1 = call float @llvm.experimental.gc.result.float.f32(i32 %safepoint_token)
   ret float %call1
 }
 
-define i1 @test_relocate(i32* %a) {
+define i1 @test_relocate_noop(i32* %a) {
 ; CHECK-LABEL: test_relocate
 ; Check that an ununsed relocate has no code-generation impact
 ; CHECK: pushq %rax
 ; CHECK: callq return_i1
 ; CHECK-NEXT: .Ltmp13:
 ; CHECK-NEXT: popq %rdx
 ; CHECK-NEXT: retq
 entry:
   %safepoint_token = tail call i32 (i1 ()*, i32, i32, ...)* @llvm.experimental.gc.statepoint.p0f_i1f(i1 ()* @return_i1, i32 0, i32 0, i32 0, i32* %a)
   %call1 = call i32* @llvm.experimental.gc.relocate.p0i32(i32 %safepoint_token, i32 4, i32 4)
   %call2 = call zeroext i1 @llvm.experimental.gc.result.int.i1(i32 %safepoint_token)
   ret i1 %call2
 }
 
+define i32 @test_relocate_integer(i32* %i) {
+; CHECK-LABEL: test_relocate_integer
+; Check that relocation of an integer is lowered correctly
+; CHECK: pushq %rax
+; CHECK: callq return_i1
+; CHECK-NEXT: .Ltmp16:
+; CHECK-NEXT: movq (%rsp), %rax
+; CHECK-NEXT: movl (%rax), %eax
+; CHECK-NEXT: popq %rdx
+; CHECK-NEXT: retq
+entry:
+  %tok = tail call i32 (i1 ()*, i32, i32, ...)* @llvm.experimental.gc.statepoint.p0f_i1f(i1 ()* @return_i1, i32 0, i32 0, i32 0, i32* %i)
+  %i-new = call i32* @llvm.experimental.gc.relocate.p0i32(i32 %tok, i32 4, i32 4)
+  %ret = load i32* %i-new
+  ret i32 %ret
+}
+
+define [3 x i32] @test_relocate_array([3 x i32]* %v) {
+; CHECK-LABEL: test_relocate_array
+; Check that relocation of an array of integers is lowered correctly
+; CHECK: pushq %rax
+; CHECK: callq return_i1
+; CHECK-NEXT: .Ltmp19:
+; CHECK-NEXT: movq (%rsp), %rcx
+; CHECK-NEXT: movl (%rcx), %eax
+; CHECK-NEXT: movl 4(%rcx), %edx
+; CHECK-NEXT: movl 8(%rcx), %ecx
+; CHECK-NEXT: popq %rsi
+; CHECK-NEXT: retq
+  %tok = call i32 (i1 ()*, i32, i32, ...)* @llvm.experimental.gc.statepoint.p0f_i1f(i1 ()* @return_i1, i32 0, i32 0, i32 0, [3 x i32]* %v)
+  %v-new = call [3 x i32]* @llvm.experimental.gc.relocate.p0a3i32(i32 %tok, i32 4, i32 4)
+  %ret = load [3 x i32]* %v-new
+  ret [3 x i32] %ret
+}
+
+define <3 x i32> @test_relocate_vector(<3 x i32>* %v) {
+; CHECK-LABEL: test_relocate_vector
+; Check that relocation of a vector of integers is lowered correctly
+; CHECK: pushq %rax
+; CHECK: callq return_i1
+; CHECK-NEXT: .Ltmp22:
+; CHECK-NEXT: movq (%rsp), %rax
+; TODO: How should this be lowered?
+; CHECK: popq %rax
+; CHECK-NEXT: retq
+  %tok = call i32 (i1 ()*, i32, i32, ...)* @llvm.experimental.gc.statepoint.p0f_i1f(i1 ()* @return_i1, i32 0, i32 0, i32 0, <3 x i32>* %v)
+  %v-new = call <3 x i32>* @llvm.experimental.gc.relocate.p0v3i32(i32 %tok, i32 4, i32 4)
+  %ret = load <3 x i32>* %v-new
+  ret <3 x i32> %ret
+}
+
 declare i32 @llvm.experimental.gc.statepoint.p0f_i1f(i1 ()*, i32, i32, ...)
 declare i1 @llvm.experimental.gc.result.int.i1(i32)
 
 declare i32 @llvm.experimental.gc.statepoint.p0f_i32f(i32 ()*, i32, i32, ...)
 declare i32 @llvm.experimental.gc.result.int.i32(i32)
 
 declare i32 @llvm.experimental.gc.statepoint.p0f_p0i32f(i32* ()*, i32, i32, ...)
 declare i32* @llvm.experimental.gc.result.ptr.p0i32(i32)
 
 declare i32 @llvm.experimental.gc.statepoint.p0f_f32f(float ()*, i32, i32, ...)
 declare float @llvm.experimental.gc.result.float.f32(i32)
 
 declare i32* @llvm.experimental.gc.relocate.p0i32(i32, i32, i32)
+declare [3 x i32]* @llvm.experimental.gc.relocate.p0a3i32(i32, i32, i32)
+declare <3 x i32>* @llvm.experimental.gc.relocate.p0v3i32(i32, i32, i32)