| ; Test that DAGCombiner gets helped by computeKnownBitsForTargetNode(). |
| ; |
| ; RUN: llc -mtriple=s390x-linux-gnu -mcpu=z13 < %s | FileCheck %s |
| |
| ; SystemZISD::REPLICATE |
| define i32 @f0() { |
| ; CHECK-LABEL: f0: |
| ; CHECK-LABEL: # %bb.0: |
| ; CHECK: vlgvf |
| ; CHECK-NOT: lhi %r2, 0 |
| ; CHECK-NOT: chi %r0, 0 |
| ; CHECK-NOT: lochilh %r2, 1 |
| ; CHECK: br %r14 |
| %cmp0 = icmp ne <4 x i32> undef, zeroinitializer |
| %zxt0 = zext <4 x i1> %cmp0 to <4 x i32> |
| %ext0 = extractelement <4 x i32> %zxt0, i32 3 |
| br label %exit |
| |
| exit: |
| ; The vector icmp+zext involves a REPLICATE of 1's. If KnownBits reflects |
| ; this, DAGCombiner can see that the i32 icmp and zext here are not needed. |
| %cmp1 = icmp ne i32 %ext0, 0 |
| %zxt1 = zext i1 %cmp1 to i32 |
| ret i32 %zxt1 |
| } |
| |
| ; SystemZISD::JOIN_DWORDS (and REPLICATE) |
| define void @f1() { |
| ; The DAG XOR has JOIN_DWORDS and REPLICATE operands. With KnownBits properly set |
| ; for both these nodes, ICMP is used instead of TM during lowering because |
| ; adjustForRedundantAnd() succeeds. |
| ; CHECK-LABEL: f1: |
| ; CHECK-LABEL: # %bb.0: |
| ; CHECK-NOT: tmll |
| ; CHECK-NOT: jne |
| ; CHECK: cijlh |
| %1 = load i16, i16* null, align 2 |
| %2 = icmp eq i16 %1, 0 |
| %3 = insertelement <2 x i1> undef, i1 %2, i32 0 |
| %4 = insertelement <2 x i1> %3, i1 true, i32 1 |
| %5 = xor <2 x i1> %4, <i1 true, i1 true> |
| %6 = extractelement <2 x i1> %5, i32 0 |
| %7 = or i1 %6, undef |
| br i1 %7, label %9, label %8 |
| |
| ; <label>:8: ; preds = %0 |
| unreachable |
| |
| ; <label>:9: ; preds = %0 |
| unreachable |
| } |
