[DAGCombiner] Added BSWAP(BSWAP(x)) -> x combine pattern.

llvm-svn: 239682

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 1e33f1e..33f5f52 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -4773,6 +4773,9 @@ SDValue DAGCombiner::visitBSWAP(SDNode *N) {
   // fold (bswap c1) -> c2
   if (isConstantIntBuildVectorOrConstantInt(N0))
     return DAG.getNode(ISD::BSWAP, SDLoc(N), VT, N0);
+  // fold (bswap (bswap x)) -> x
+  if (N0.getOpcode() == ISD::BSWAP)
+    return N0->getOperand(0);
   return SDValue();
 }
 

diff --git a/llvm/test/CodeGen/X86/bswap-vector.ll b/llvm/test/CodeGen/X86/bswap-vector.ll
index 16da162..5376601 100644 (file)
--- a/llvm/test/CodeGen/X86/bswap-vector.ll
+++ b/llvm/test/CodeGen/X86/bswap-vector.ll
@@ -287,6 +287,80 @@ entry:
 }
 
 ;
+; Double BSWAP -> Identity
+;
+
+define <8 x i16> @identity_v8i16(<8 x i16> %v) {
+; CHECK-ALL-LABEL: identity_v8i16:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL:         retq
+entry:
+  %bs1 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %v)
+  %bs2 = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %bs1)
+  ret <8 x i16> %bs2
+}
+
+define <4 x i32> @identity_v4i32(<4 x i32> %v) {
+; CHECK-ALL-LABEL: identity_v4i32:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %v)
+  %bs2 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %bs1)
+  ret <4 x i32> %bs2
+}
+
+define <2 x i64> @identity_v2i64(<2 x i64> %v) {
+; CHECK-ALL-LABEL: identity_v2i64:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %v)
+  %bs2 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %bs1)
+  ret <2 x i64> %bs2
+}
+
+define <16 x i16> @identity_v16i16(<16 x i16> %v) {
+; CHECK-ALL-LABEL: identity_v16i16:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %v)
+  %bs2 = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %bs1)
+  ret <16 x i16> %bs2
+}
+
+define <8 x i32> @identity_v8i32(<8 x i32> %v) {
+; CHECK-ALL-LABEL: identity_v8i32:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %v)
+  %bs2 = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %bs1)
+  ret <8 x i32> %bs2
+}
+
+define <4 x i64> @identity_v4i64(<4 x i64> %v) {
+; CHECK-ALL-LABEL: identity_v4i64:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %v)
+  %bs2 = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %bs1)
+  ret <4 x i64> %bs2
+}
+
+define <4 x i16> @identity_v4i16(<4 x i16> %v) {
+; CHECK-ALL-LABEL: identity_v4i16:
+; CHECK-ALL:       # BB#0: # %entry
+; CHECK-ALL-NEXT:    retq
+entry:
+  %bs1 = call <4 x i16> @llvm.bswap.v4i16(<4 x i16> %v)
+  %bs2 = call <4 x i16> @llvm.bswap.v4i16(<4 x i16> %bs1)
+  ret <4 x i16> %bs2
+}
+
+;
 ; Constant Folding
 ;