AMDGPU: Improve load/store of illegal types.
There was a combine before to handle the simple copy case.
Split this into handling loads and stores separately.
We might want to change how this handles some of the vector
extloads, since this can result in large code size increases.
llvm-svn: 274394
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-f32.ll b/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
index 17c09d1..77b5e3c 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
@@ -27,9 +27,10 @@
ret void
}
-; FIXME: should only do one b64 load
+; FIXME: should this do a read2_b64?
; FUNC-LABEL: {{^}}local_load_v3f32:
-; GCN: ds_read2_b64
+; GCN-DAG: ds_read_b32 v{{[0-9]+}}, v{{[0-9]+}} offset:8
+; GCN-DAG: ds_read_b64 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+$}}
; GCN: s_waitcnt
; GCN-DAG: ds_write_b64
; GCN-DAG: ds_write_b32 v{{[0-9]+}}, v{{[0-9]+}} offset:8{{$}}