diff options
Diffstat (limited to 'llvm/test/CodeGen/Hexagon/combiner-lts.ll')
-rw-r--r-- | llvm/test/CodeGen/Hexagon/combiner-lts.ll | 66 |
1 files changed, 66 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/Hexagon/combiner-lts.ll b/llvm/test/CodeGen/Hexagon/combiner-lts.ll new file mode 100644 index 00000000000..46c96d5dc1a --- /dev/null +++ b/llvm/test/CodeGen/Hexagon/combiner-lts.ll @@ -0,0 +1,66 @@ +; RUN: llc -march=hexagon -O2 < %s | FileCheck %s + +; In DAG combiner, eliminate a store in cases where the store is fed by a +; load from the same location. This is already done in cases where the store's +; chain reaches the "output chain" of the load, this tests for cases where +; the load's "input chain" is reached via an intervening node (eg. TokenFactor) +; that ensures ordering. + +target triple = "hexagon" + +%s.0 = type { [3 x i32] } + +; Function Attrs: nounwind +define void @f0(i32 %a0, i32 %a1, %s.0* nocapture %a2, %s.0* nocapture %a3) #0 { +b0: +; Pick one store that happens as a result. This isn't the best, but a regular +; expression for a register name matches some unrelated load. +; CHECK: %bb. +; CHECK: = memw(r3+#8) +; CHECK-NOT: memw(r3+#8) = +; CHECK: %bb. + %v0 = bitcast %s.0* %a2 to i8* + %v1 = bitcast %s.0* %a3 to i8* + call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %v0, i8* align 4 %v1, i32 12, i1 false) + %v2 = bitcast %s.0* %a2 to i96* + %v3 = zext i32 %a0 to i96 + %v4 = load i96, i96* %v2, align 4 + %v5 = shl nuw nsw i96 %v3, 48 + %v6 = and i96 %v5, 281474976710656 + %v7 = and i96 %v4, -281474976710657 + %v8 = or i96 %v7, %v6 + store i96 %v8, i96* %v2, align 4 + %v9 = icmp eq i32 %a1, 2147483647 + br i1 %v9, label %b1, label %b2 + +b1: ; preds = %b0 + %v10 = and i96 %v8, -12582913 + br label %b3 + +b2: ; preds = %b0 + %v11 = bitcast %s.0* %a3 to i96* + %v12 = load i96, i96* %v11, align 4 + %v13 = trunc i96 %v12 to i32 + %v14 = add i32 %v13, %a1 + %v15 = zext i32 %v14 to i96 + %v16 = and i96 %v15, 4194303 + %v17 = and i96 %v8, -4194304 + %v18 = or i96 %v16, %v17 + store i96 %v18, i96* %v2, align 4 + %v19 = load i96, i96* %v11, align 4 + %v20 = and i96 %v19, 12582912 + %v21 = and i96 %v18, -12582913 + %v22 = or i96 %v21, %v20 + br label %b3 + +b3: ; preds = %b2, %b1 + %v23 = phi i96 [ %v22, %b2 ], [ %v10, %b1 ] + store i96 %v23, i96* %v2, align 4 + ret void +} + +; Function Attrs: argmemonly nounwind +declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture writeonly, i8* nocapture readonly, i32, i1) #1 + +attributes #0 = { nounwind } +attributes #1 = { argmemonly nounwind } |