[x86] fix uses check in broadcast transform (PR38949)

https://bugs.llvm.org/show_bug.cgi?id=38949

It's not clear to me that we even need a one-use check in this fold.
Ie, 2 independent loads might be better than a load+dependent shuffle.

Note that the existing re-use tests are not affected. We actually do form a
broadcast node in those tests now because there's no extra use of the
insert_subvector node in those cases. But something later in isel pattern
matching decides that it is not worth using a broadcast for the full load in
those tests:

Legalized selection DAG: %bb.0 'test_broadcast_2f64_4f64_reuse:'
  t7: v2f64,ch = load<(load 16 from %ir.p0)> t0, t2, undef:i64
      t4: i64,ch = CopyFromReg t0, Register:i64 %1
    t10: ch = store<(store 16 into %ir.p1)> t7:1, t7, t4, undef:i64
      t18: v4f64 = insert_subvector undef:v4f64, t7, Constant:i64<0>
    t20: v4f64 = insert_subvector t18, t7, Constant:i64<2>

Becomes:
  t7: v2f64,ch = load<(load 16 from %ir.p0)> t0, t2, undef:i64
      t4: i64,ch = CopyFromReg t0, Register:i64 %1
    t10: ch = store<(store 16 into %ir.p1)> t7:1, t7, t4, undef:i64
    t21: v4f64 = X86ISD::SUBV_BROADCAST t7

ISEL: Starting selection on root node: t21: v4f64 = X86ISD::SUBV_BROADCAST t7
...
  Created node: t27: v4f64 = INSERT_SUBREG IMPLICIT_DEF:v4f64, t7, TargetConstant:i32<7>
  Morphed node: t21: v4f64 = VINSERTF128rr t27, t7, TargetConstant:i8<1>

llvm-svn: 342347

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 14880e4..c6a79a2 100644 (file)
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -40114,11 +40114,10 @@ static SDValue combineInsertSubvector(SDNode *N, SelectionDAG &DAG,
             return Ld;
         }
       }
-      // If lower/upper loads are the same and the only users of the load, then
-      // lower to a VBROADCASTF128/VBROADCASTI128/etc.
+      // If lower/upper loads are the same and there's no other use of the lower
+      // load, then splat the loaded value with a broadcast.
       if (auto *Ld = dyn_cast<LoadSDNode>(peekThroughOneUseBitcasts(SubVec2)))
-        if (SubVec2 == SubVec && ISD::isNormalLoad(Ld) &&
-            SDNode::areOnlyUsersOf({N, Vec.getNode()}, SubVec2.getNode()))
+        if (SubVec2 == SubVec && ISD::isNormalLoad(Ld) && Vec.hasOneUse())
           return DAG.getNode(X86ISD::SUBV_BROADCAST, dl, OpVT, SubVec);
 
       // If this is subv_broadcast insert into both halves, use a larger

diff --git a/llvm/test/CodeGen/X86/avx-vbroadcastf128.ll b/llvm/test/CodeGen/X86/avx-vbroadcastf128.ll
index e35919d..9bd4179 100644 (file)
--- a/llvm/test/CodeGen/X86/avx-vbroadcastf128.ll
+++ b/llvm/test/CodeGen/X86/avx-vbroadcastf128.ll
@@ -106,16 +106,14 @@ define void @subv_reuse_is_ok(<4 x float>* %a, <8 x float>* %b) {
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    vmovups (%ecx), %xmm0
-; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
 ; X32-NEXT:    vmovups %ymm0, (%eax)
 ; X32-NEXT:    vzeroupper
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: subv_reuse_is_ok:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovups (%rdi), %xmm0
-; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
 ; X64-NEXT:    vmovups %ymm0, (%rsi)
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq