replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[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
This commit is contained in:
Sanjay Patel 2018-09-16 15:41:56 +00:00
parent 3e095174b0
commit bfee5a9b42
2 changed files with 5 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -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

6
llvm/test/CodeGen/X86/avx-vbroadcastf128.ll
View File

@ -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