AMDGPU: Handle i64->v2i32 loads/stores in PreprocessISelDAG
This fixes a select error when the i64 source was also
bitcasted to v2i32 in the original source.
Instead of awkwardly trying to select the modified source value and
the store, replace before isel begins.
Uses a worklist to avoid possible problems from mutating the DAG,
although it seems to work OK without it.
llvm-svn: 248589
diff --git a/llvm/test/CodeGen/AMDGPU/extract-vector-elt-i64.ll b/llvm/test/CodeGen/AMDGPU/extract-vector-elt-i64.ll
new file mode 100644
index 0000000..24301cc
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/extract-vector-elt-i64.ll
@@ -0,0 +1,19 @@
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+
+; How the replacement of i64 stores with v2i32 stores resulted in
+; breaking other users of the bitcast if they already existed
+
+; GCN-LABEL: {{^}}extract_vector_elt_select_error:
+; GCN: buffer_store_dword
+; GCN: buffer_store_dword
+; GCN: buffer_store_dwordx2
+define void @extract_vector_elt_select_error(i32 addrspace(1)* %out, i64 addrspace(1)* %in, i64 %val) nounwind {
+ %vec = bitcast i64 %val to <2 x i32>
+ %elt0 = extractelement <2 x i32> %vec, i32 0
+ %elt1 = extractelement <2 x i32> %vec, i32 1
+
+ store volatile i32 %elt0, i32 addrspace(1)* %out
+ store volatile i32 %elt1, i32 addrspace(1)* %out
+ store volatile i64 %val, i64 addrspace(1)* %in
+ ret void
+}