DAG: Legalize truncstores to illegal int types

Truncate to a legal int type, and produce a new
truncstore from a narrower type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@319185 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index 6974e7006ce2eac239787f87e749bcb7a4b25aa5..fdbd3e1105be500c7088f10cb17f2845f573705e 100644 (file)
--- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -624,13 +624,23 @@ void SelectionDAGLegalize::LegalizeStoreOps(SDNode *Node) {
       assert(!StVT.isVector() &&
              "Vector Stores are handled in LegalizeVectorOps");
 
+      SDValue Result;
+
       // TRUNCSTORE:i16 i32 -> STORE i16
-      assert(TLI.isTypeLegal(StVT) &&
-             "Do not know how to expand this store!");
-      Value = DAG.getNode(ISD::TRUNCATE, dl, StVT, Value);
-      SDValue Result =
-          DAG.getStore(Chain, dl, Value, Ptr, ST->getPointerInfo(),
-                       Alignment, MMOFlags, AAInfo);
+      if (TLI.isTypeLegal(StVT)) {
+        Value = DAG.getNode(ISD::TRUNCATE, dl, StVT, Value);
+        Result = DAG.getStore(Chain, dl, Value, Ptr, ST->getPointerInfo(),
+                              Alignment, MMOFlags, AAInfo);
+      } else {
+        // The in-memory type isn't legal. Truncate to the type it would promote
+        // to, and then do a truncstore.
+        Value = DAG.getNode(ISD::TRUNCATE, dl,
+                            TLI.getTypeToTransformTo(*DAG.getContext(), StVT),
+                            Value);
+        Result = DAG.getTruncStore(Chain, dl, Value, Ptr, ST->getPointerInfo(),
+                                   StVT, Alignment, MMOFlags, AAInfo);
+      }
+
       ReplaceNode(SDValue(Node, 0), Result);
       break;
     }

diff --git a/test/CodeGen/AMDGPU/store-weird-sizes.ll b/test/CodeGen/AMDGPU/store-weird-sizes.ll
new file mode 100644 (file)
index 0000000..8b60500
--- /dev/null
+++ b/test/CodeGen/AMDGPU/store-weird-sizes.ll
@@ -0,0 +1,56 @@
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CIVI %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CIVI %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9 %s
+
+; GCN-LABEL: {{^}}local_store_i56:
+; GCN-DAG: ds_write_b8 v0, v{{[0-9]+}} offset:6
+; GCN-DAG: ds_write_b16 v0, v{{[0-9]+}} offset:4
+; GCN-DAG: ds_write_b32 v0, v{{[0-9]+$}}
+define void @local_store_i56(i56 addrspace(3)* %ptr, i56 %arg) #0 {
+  store i56 %arg, i56 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+; GCN-LABEL: {{^}}local_store_i55:
+; GCN-DAG: ds_write_b8 v0, v{{[0-9]+}} offset:6
+; GCN-DAG: ds_write_b16 v0, v{{[0-9]+}} offset:4
+; GCN-DAG: ds_write_b32 v0, v{{[0-9]+$}}
+define amdgpu_kernel void @local_store_i55(i55 addrspace(3)* %ptr, i55 %arg) #0 {
+  store i55 %arg, i55 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+; GCN-LABEL: {{^}}local_store_i48:
+; GCN-DAG: ds_write_b16 v0, v{{[0-9]+}} offset:4
+; GCN-DAG: ds_write_b32 v0, v{{[0-9]+$}}
+define amdgpu_kernel void @local_store_i48(i48 addrspace(3)* %ptr, i48 %arg) #0 {
+  store i48 %arg, i48 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+; GCN-LABEL: {{^}}local_store_i65:
+; GCN-DAG: ds_write_b8 v{{[0-9]+}}, v0 offset:8
+; GCN-DAG: ds_write_b64
+define amdgpu_kernel void @local_store_i65(i65 addrspace(3)* %ptr, i65 %arg) #0 {
+  store i65 %arg, i65 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+; GCN-LABEL: {{^}}local_store_i13:
+; GCN: v_and_b32_e32 [[TRUNC:v[0-9]+]], 0x1fff, v1
+; GCN: ds_write_b16 v0, [[TRUNC]]
+define void @local_store_i13(i13 addrspace(3)* %ptr, i13 %arg) #0 {
+  store i13 %arg, i13 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+; GCN-LABEL: {{^}}local_store_i17:
+; GCN: ds_write_b16 v0
+; CIVI: ds_write_b8 v0, v{{[0-9]+}} offset:2
+; GFX9: ds_write_b8_d16_hi v0, v{{[0-9]+}} offset:2
+define void @local_store_i17(i17 addrspace(3)* %ptr, i17 %arg) #0 {
+  store i17 %arg, i17 addrspace(3)* %ptr, align 8
+  ret void
+}
+
+attributes #0 = { nounwind }