InferAddressSpaces: Support memory intrinsics

llvm-svn: 293587

3 files changed, 249 insertions, 1 deletions

diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/intrinsics.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/intrinsics.ll
new file mode 100644
index 00000000000..36ec424267c
--- /dev/null
+++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/intrinsics.ll
@@ -0,0 +1,92 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -infer-address-spaces %s | FileCheck %s
+
+; CHECK-LABEL: @objectsize_group_to_flat_i32(
+; CHECK: %val = call i32 @llvm.objectsize.i32.p3i8(i8 addrspace(3)* %group.ptr, i1 true)
+define i32 @objectsize_group_to_flat_i32(i8 addrspace(3)* %group.ptr) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+  %val = call i32 @llvm.objectsize.i32.p4i8(i8 addrspace(4)* %cast, i1 true)
+  ret i32 %val
+}
+
+; CHECK-LABEL: @objectsize_global_to_flat_i64(
+; CHECK: %val = call i64 @llvm.objectsize.i64.p3i8(i8 addrspace(3)* %global.ptr, i1 true)
+define i64 @objectsize_global_to_flat_i64(i8 addrspace(3)* %global.ptr) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %global.ptr to i8 addrspace(4)*
+  %val = call i64 @llvm.objectsize.i64.p4i8(i8 addrspace(4)* %cast, i1 true)
+  ret i64 %val
+}
+
+; CHECK-LABEL: @atomicinc_global_to_flat_i32(
+; CHECK: call i32 @llvm.amdgcn.atomic.inc.i32.p1i32(i32 addrspace(1)* %global.ptr, i32 %y)
+define i32 @atomicinc_global_to_flat_i32(i32 addrspace(1)* %global.ptr, i32 %y) #0 {
+  %cast = addrspacecast i32 addrspace(1)* %global.ptr to i32 addrspace(4)*
+  %ret = call i32 @llvm.amdgcn.atomic.inc.i32.p4i32(i32 addrspace(4)* %cast, i32 %y)
+  ret i32 %ret
+}
+
+; CHECK-LABEL: @atomicinc_group_to_flat_i32(
+; CHECK: %ret = call i32 @llvm.amdgcn.atomic.inc.i32.p3i32(i32 addrspace(3)* %group.ptr, i32 %y)
+define i32 @atomicinc_group_to_flat_i32(i32 addrspace(3)* %group.ptr, i32 %y) #0 {
+  %cast = addrspacecast i32 addrspace(3)* %group.ptr to i32 addrspace(4)*
+  %ret = call i32 @llvm.amdgcn.atomic.inc.i32.p4i32(i32 addrspace(4)* %cast, i32 %y)
+  ret i32 %ret
+}
+
+; CHECK-LABEL: @atomicinc_global_to_flat_i64(
+; CHECK: call i64 @llvm.amdgcn.atomic.inc.i64.p1i64(i64 addrspace(1)* %global.ptr, i64 %y)
+define i64 @atomicinc_global_to_flat_i64(i64 addrspace(1)* %global.ptr, i64 %y) #0 {
+  %cast = addrspacecast i64 addrspace(1)* %global.ptr to i64 addrspace(4)*
+  %ret = call i64 @llvm.amdgcn.atomic.inc.i64.p4i64(i64 addrspace(4)* %cast, i64 %y)
+  ret i64 %ret
+}
+
+; CHECK-LABEL: @atomicinc_group_to_flat_i64(
+; CHECK: call i64 @llvm.amdgcn.atomic.inc.i64.p3i64(i64 addrspace(3)* %group.ptr, i64 %y)
+define i64 @atomicinc_group_to_flat_i64(i64 addrspace(3)* %group.ptr, i64 %y) #0 {
+  %cast = addrspacecast i64 addrspace(3)* %group.ptr to i64 addrspace(4)*
+  %ret = call i64 @llvm.amdgcn.atomic.inc.i64.p4i64(i64 addrspace(4)* %cast, i64 %y)
+  ret i64 %ret
+}
+
+; CHECK-LABEL: @atomicdec_global_to_flat_i32(
+; CHECK: call i32 @llvm.amdgcn.atomic.dec.i32.p1i32(i32 addrspace(1)* %global.ptr, i32 %val)
+define i32 @atomicdec_global_to_flat_i32(i32 addrspace(1)* %global.ptr, i32 %val) #0 {
+  %cast = addrspacecast i32 addrspace(1)* %global.ptr to i32 addrspace(4)*
+  %ret = call i32 @llvm.amdgcn.atomic.dec.i32.p4i32(i32 addrspace(4)* %cast, i32 %val)
+  ret i32 %ret
+}
+
+; CHECK-LABEL: @atomicdec_group_to_flat_i32(
+; CHECK: %ret = call i32 @llvm.amdgcn.atomic.dec.i32.p3i32(i32 addrspace(3)* %group.ptr, i32 %val)
+define i32 @atomicdec_group_to_flat_i32(i32 addrspace(3)* %group.ptr, i32 %val) #0 {
+  %cast = addrspacecast i32 addrspace(3)* %group.ptr to i32 addrspace(4)*
+  %ret = call i32 @llvm.amdgcn.atomic.dec.i32.p4i32(i32 addrspace(4)* %cast, i32 %val)
+  ret i32 %ret
+}
+
+; CHECK-LABEL: @atomicdec_global_to_flat_i64(
+; CHECK: call i64 @llvm.amdgcn.atomic.dec.i64.p1i64(i64 addrspace(1)* %global.ptr, i64 %y)
+define i64 @atomicdec_global_to_flat_i64(i64 addrspace(1)* %global.ptr, i64 %y) #0 {
+  %cast = addrspacecast i64 addrspace(1)* %global.ptr to i64 addrspace(4)*
+  %ret = call i64 @llvm.amdgcn.atomic.dec.i64.p4i64(i64 addrspace(4)* %cast, i64 %y)
+  ret i64 %ret
+}
+
+; CHECK-LABEL: @atomicdec_group_to_flat_i64(
+; CHECK: call i64 @llvm.amdgcn.atomic.dec.i64.p3i64(i64 addrspace(3)* %group.ptr, i64 %y)
+define i64 @atomicdec_group_to_flat_i64(i64 addrspace(3)* %group.ptr, i64 %y) #0 {
+  %cast = addrspacecast i64 addrspace(3)* %group.ptr to i64 addrspace(4)*
+  %ret = call i64 @llvm.amdgcn.atomic.dec.i64.p4i64(i64 addrspace(4)* %cast, i64 %y)
+  ret i64 %ret
+}
+
+declare i32 @llvm.objectsize.i32.p4i8(i8 addrspace(4)*, i1) #1
+declare i64 @llvm.objectsize.i64.p4i8(i8 addrspace(4)*, i1) #1
+declare i32 @llvm.amdgcn.atomic.inc.i32.p4i32(i32 addrspace(4)* nocapture, i32) #2
+declare i64 @llvm.amdgcn.atomic.inc.i64.p4i64(i64 addrspace(4)* nocapture, i64) #2
+declare i32 @llvm.amdgcn.atomic.dec.i32.p4i32(i32 addrspace(4)* nocapture, i32) #2
+declare i64 @llvm.amdgcn.atomic.dec.i64.p4i64(i64 addrspace(4)* nocapture, i64) #2
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind argmemonly }
diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll
new file mode 100644
index 00000000000..afd1493fc0e
--- /dev/null
+++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll
@@ -0,0 +1,134 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -infer-address-spaces %s | FileCheck %s
+
+; CHECK-LABEL: @memset_group_to_flat(
+; CHECK: call void @llvm.memset.p3i8.i64(i8 addrspace(3)* %group.ptr, i8 4, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memset_group_to_flat(i8 addrspace(3)* %group.ptr, i32 %y) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memset_global_to_flat(
+; CHECK: call void @llvm.memset.p1i8.i64(i8 addrspace(1)* %global.ptr, i8 4, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memset_global_to_flat(i8 addrspace(1)* %global.ptr, i32 %y) #0 {
+  %cast = addrspacecast i8 addrspace(1)* %global.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memset_group_to_flat_no_md(
+; CHECK: call void @llvm.memset.p3i8.i64(i8 addrspace(3)* %group.ptr, i8 4, i64 %size, i32 4, i1 false){{$}}
+define void @memset_group_to_flat_no_md(i8 addrspace(3)* %group.ptr, i64 %size) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 %size, i32 4, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: @memset_global_to_flat_no_md(
+; CHECK: call void @llvm.memset.p1i8.i64(i8 addrspace(1)* %global.ptr, i8 4, i64 %size, i32 4, i1 false){{$}}
+define void @memset_global_to_flat_no_md(i8 addrspace(1)* %global.ptr, i64 %size) #0 {
+  %cast = addrspacecast i8 addrspace(1)* %global.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 %size, i32 4, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_src_with_group(
+; CHCK: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_flat_to_flat_replace_src_with_group(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_dest_with_group(
+; CHECK: call void @llvm.memcpy.p3i8.p4i8.i64(i8 addrspace(3)* %dest.group.ptr, i8 addrspace(4)* %src.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_flat_to_flat_replace_dest_with_group(i8 addrspace(3)* %dest.group.ptr, i8 addrspace(4)* %src.ptr, i64 %size) #0 {
+  %cast.dest = addrspacecast i8 addrspace(3)* %dest.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %cast.dest, i8 addrspace(4)* %src.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_dest_src_with_group(
+; CHECK: call void @llvm.memcpy.p3i8.p3i8.i64(i8 addrspace(3)* %src.group.ptr, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_flat_to_flat_replace_dest_src_with_group(i8 addrspace(3)* %dest.group.ptr, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  %cast.dest = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %cast.dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_dest_group_src_global(
+; CHECK: call void @llvm.memcpy.p3i8.p1i8.i64(i8 addrspace(3)* %dest.group.ptr, i8 addrspace(1)* %src.global.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_flat_to_flat_replace_dest_group_src_global(i8 addrspace(3)* %dest.group.ptr, i8 addrspace(1)* %src.global.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(1)* %src.global.ptr to i8 addrspace(4)*
+  %cast.dest = addrspacecast i8 addrspace(3)* %dest.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %cast.dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_group_to_flat_replace_dest_global(
+; CHECK: call void @llvm.memcpy.p1i8.p3i8.i32(i8 addrspace(1)* %dest.global.ptr, i8 addrspace(3)* %src.group.ptr, i32 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_group_to_flat_replace_dest_global(i8 addrspace(1)* %dest.global.ptr, i8 addrspace(3)* %src.group.ptr, i32 %size) #0 {
+  %cast.dest = addrspacecast i8 addrspace(1)* %dest.global.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p3i8.i32(i8 addrspace(4)* %cast.dest, i8 addrspace(3)* %src.group.ptr, i32 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_src_with_group_tbaa_struct(
+; CHECK: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false), !tbaa.struct !7
+define void @memcpy_flat_to_flat_replace_src_with_group_tbaa_struct(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false), !tbaa.struct !7
+  ret void
+}
+
+; CHECK-LABEL: @memcpy_flat_to_flat_replace_src_with_group_no_md(
+; CHECK: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false){{$}}
+define void @memcpy_flat_to_flat_replace_src_with_group_no_md(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: @multiple_memcpy_flat_to_flat_replace_src_with_group_no_md(
+; CHECK: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* %dest0, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false){{$}}
+; CHECK: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* %dest1, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false){{$}}
+define void @multiple_memcpy_flat_to_flat_replace_src_with_group_no_md(i8 addrspace(4)* %dest0, i8 addrspace(4)* %dest1, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %dest0, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false)
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %dest1, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false)
+  ret void
+}
+
+; Check for iterator problems if the pointer has 2 uses in the same call
+; CHECK-LABEL: @memcpy_group_flat_to_flat_self(
+; CHECK: call void @llvm.memcpy.p3i8.p3i8.i64(i8 addrspace(3)* %group.ptr, i8 addrspace(3)* %group.ptr, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memcpy_group_flat_to_flat_self(i8 addrspace(3)* %group.ptr) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+  call void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* %cast, i8 addrspace(4)* %cast, i64 32, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+; CHECK-LABEL: @memmove_flat_to_flat_replace_src_with_group(
+; CHECK: call void @llvm.memmove.p4i8.p3i8.i64(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+define void @memmove_flat_to_flat_replace_src_with_group(i8 addrspace(4)* %dest, i8 addrspace(3)* %src.group.ptr, i64 %size) #0 {
+  %cast.src = addrspacecast i8 addrspace(3)* %src.group.ptr to i8 addrspace(4)*
+  call void @llvm.memmove.p4i8.p4i8.i64(i8 addrspace(4)* %dest, i8 addrspace(4)* %cast.src, i64 %size, i32 4, i1 false), !tbaa !0, !alias.scope !3, !noalias !4
+  ret void
+}
+
+declare void @llvm.memset.p4i8.i64(i8 addrspace(4)* nocapture writeonly, i8, i64, i32, i1) #1
+declare void @llvm.memcpy.p4i8.p4i8.i64(i8 addrspace(4)* nocapture writeonly, i8 addrspace(4)* nocapture readonly, i64, i32, i1) #1
+declare void @llvm.memcpy.p4i8.p3i8.i32(i8 addrspace(4)* nocapture writeonly, i8 addrspace(3)* nocapture readonly, i32, i32, i1) #1
+declare void @llvm.memmove.p4i8.p4i8.i64(i8 addrspace(4)* nocapture writeonly, i8 addrspace(4)* nocapture readonly, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }
+
+!0 = !{!1, !1, i64 0}
+!1 = !{!"A", !2}
+!2 = !{!"tbaa root"}
+!3 = !{!"B", !2}
+!4 = !{!5}
+!5 = distinct !{!5, !6, !"some scope"}
+!6 = distinct !{!6, !"some domain"}
+!7 = !{i64 0, i64 8, null}
\ No newline at end of file
diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/volatile.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/volatile.ll
index f32d65b66ae..d9b80e99bf0 100644
--- a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/volatile.ll
+++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/volatile.ll
@@ -115,4 +115,26 @@ define { i32, i1 } @volatile_cmpxchg_group_to_flat(i32 addrspace(3)* %group.ptr,
   ret { i32, i1 } %ret
 }
 
-attributes #0 = { nounwind }
\ No newline at end of file
+; FIXME: Shouldn't be losing names
+; CHECK-LABEL: @volatile_memset_group_to_flat(
+; CHECK: addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+; CHECK: call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %1, i8 4, i64 32, i32 4, i1 true)
+define void @volatile_memset_group_to_flat(i8 addrspace(3)* %group.ptr, i32 %y) #0 {
+  %cast = addrspacecast i8 addrspace(3)* %group.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 32, i32 4, i1 true)
+  ret void
+}
+
+; CHECK-LABEL: @volatile_memset_global_to_flat(
+; CHECK: addrspacecast i8 addrspace(1)* %global.ptr to i8 addrspace(4)*
+; CHECK: call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %1, i8 4, i64 32, i32 4, i1 true)
+define void @volatile_memset_global_to_flat(i8 addrspace(1)* %global.ptr, i32 %y) #0 {
+  %cast = addrspacecast i8 addrspace(1)* %global.ptr to i8 addrspace(4)*
+  call void @llvm.memset.p4i8.i64(i8 addrspace(4)* %cast, i8 4, i64 32, i32 4, i1 true)
+  ret void
+}
+
+declare void @llvm.memset.p4i8.i64(i8 addrspace(4)* nocapture writeonly, i8, i64, i32, i1) #1
+
+attributes #0 = { nounwind }
+attributes #1 = { argmemonly nounwind }