[DAGCombine] Improve Lifetime node chains.

[niravd]

Summary

Improve both start and end lifetime nodes chain dependencies.

[courbet]

Nice !

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

15575 ↗

(On Diff #192184)

Please refactor with the code in visitLIFETIME_END

15585 ↗

(On Diff #192184)

I have a hard time convincing myself that we can omit this check. Can you add a comment to the code to explain why ?

15594 ↗

(On Diff #192184)

please clang-format the patch.

[niravd]

Extract copied code into helper and format

[niravd]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

15585 ↗

(On Diff #192184)

It's not strictly; we could accidentally skip a store because the TokenFactors aren't properly minimized. There needs to be an extra check that the found store is the last semantically meaningful before the LIFETIME_END. I'll factor this out separately from the chain improvement logic.

[courbet]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

1600 ↗

(On Diff #192343)

Maybe avoid clang-formatting this for consistency with the rest of the function ? (sigh...)

15585 ↗

(On Diff #192184)

we could accidentally skip a store because the TokenFactors aren't properly minimized.

What happens then ? :)

Can you add this comment to the code so that future readers are aware of the choices made here ?

[niravd]

Now that lifetime node hashing and TokenFactor cleanup has happened, factor out Dead Store improvements as they are no longer needed to prevent regressions.

[courbet]

Thanks, just one nit.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

15670 ↗

(On Diff #192682)

This can be simplified to:

return ImproveLifetimeNodeChain(N);

[hans]

This was reverted together with r357309 in r357563 as it caused
https://bugs.llvm.org/show_bug.cgi?id=41352

llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

private:
     SDValue visitBRCOND(SDNode *N);
     SDValue visitBR_CC(SDNode *N);
     SDValue visitLOAD(SDNode *N);
 
     SDValue replaceStoreChain(StoreSDNode *ST, SDValue BetterChain);
     SDValue replaceStoreOfFPConstant(StoreSDNode *ST);
 
     SDValue visitSTORE(SDNode *N);
+
+    SDValue ImproveLifetimeNodeChain(SDNode *N);
+
+    SDValue visitLIFETIME_START(SDNode *N);
     SDValue visitLIFETIME_END(SDNode *N);
     SDValue visitINSERT_VECTOR_ELT(SDNode *N);
     SDValue visitEXTRACT_VECTOR_ELT(SDNode *N);
     SDValue visitBUILD_VECTOR(SDNode *N);
     SDValue visitCONCAT_VECTORS(SDNode *N);
     SDValue visitEXTRACT_SUBVECTOR(SDNode *N);
     SDValue visitVECTOR_SHUFFLE(SDNode *N);
     SDValue visitSCALAR_TO_VECTOR(SDNode *N);
SDValue DAGCombiner::visit(SDNode *N) {
   case ISD::EXTRACT_SUBVECTOR:  return visitEXTRACT_SUBVECTOR(N);
   case ISD::VECTOR_SHUFFLE:     return visitVECTOR_SHUFFLE(N);
   case ISD::SCALAR_TO_VECTOR:   return visitSCALAR_TO_VECTOR(N);
   case ISD::INSERT_SUBVECTOR:   return visitINSERT_SUBVECTOR(N);
   case ISD::MGATHER:            return visitMGATHER(N);
   case ISD::MLOAD:              return visitMLOAD(N);
   case ISD::MSCATTER:           return visitMSCATTER(N);
   case ISD::MSTORE:             return visitMSTORE(N);
+  case ISD::LIFETIME_START:     return visitLIFETIME_START(N);
   case ISD::LIFETIME_END:       return visitLIFETIME_END(N);
   case ISD::FP_TO_FP16:         return visitFP_TO_FP16(N);
   case ISD::FP16_TO_FP:         return visitFP16_TO_FP(N);
   case ISD::VECREDUCE_FADD:
   case ISD::VECREDUCE_FMUL:
   case ISD::VECREDUCE_ADD:
   case ISD::VECREDUCE_MUL:
   case ISD::VECREDUCE_AND:
SDValue DAGCombiner::visitSTORE(SDNode *N) {
   }
 
   if (SDValue NewSt = splitMergedValStore(ST))
     return NewSt;
 
   return ReduceLoadOpStoreWidth(N);
 }
 
+SDValue DAGCombiner::ImproveLifetimeNodeChain(SDNode *N) {
+  auto Chain = N->getOperand(0);
+  auto NewChain = FindBetterChain(N, Chain);
+  if (NewChain != Chain) {
+    SDNode *N2 = DAG.UpdateNodeOperands(N, NewChain, N->getOperand(1));
+    // Make sure users of new N still depend on Chain
+    auto TF = DAG.getNode(ISD::TokenFactor, SDLoc(N2), MVT::Other, Chain,
+                          SDValue(N2, 0));
+    DAG.ReplaceAllUsesOfValueWith(SDValue(N2, 0), TF);
+    AddToWorklist(DAG.UpdateNodeOperands(TF.getNode(), Chain, SDValue(N2, 0)));
+    AddToWorklist(N2);
+    return SDValue(N, 0);
+  }
+  return SDValue();
+}
+
+SDValue DAGCombiner::visitLIFETIME_START(SDNode *N) {
+  if (SDValue V = ImproveLifetimeNodeChain(N))
+    return V;
+  return SDValue();
+}
+
 SDValue DAGCombiner::visitLIFETIME_END(SDNode *N) {
+  if (SDValue V = ImproveLifetimeNodeChain(N))
+    return V;
+
   const auto *LifetimeEnd = cast<LifetimeSDNode>(N);
   if (!LifetimeEnd->hasOffset())
     return SDValue();
 
   const BaseIndexOffset LifetimeEndBase(N->getOperand(1), SDValue(),
                                         LifetimeEnd->getOffset(), false);
 
   // We walk up the chains to find stores.

llvm/trunk/test/CodeGen/X86/swap.ll

 ; NOAA-NEXT:    vmovups (%rdi), %xmm0
 ; NOAA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; NOAA-NEXT:    vmovups %xmm0, (%rsi)
 ; NOAA-NEXT:    retq
 ;
 ; AA-LABEL: twoallocs:
 ; AA:       # %bb.0: # %entry
 ; AA-NEXT:    vmovups (%rdi), %xmm0
-; AA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; AA-NEXT:    vmovups %xmm0, (%rsi)
 ; AA-NEXT:    retq
 entry:
   %alloc1 = alloca [16 x i8], align 1
   %alloc2 = alloca [16 x i8], align 1
   %part1 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc1, i64 0, i64 0
   %part2 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc2, i64 0, i64 0
   %part2_alias = getelementptr inbounds [16 x i8], [16 x i8]* %alloc2, i64 0, i64 0

 ; NOAA-NEXT:    vmovups (%rsi), %xmm0
 ; NOAA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; NOAA-NEXT:    vmovups %xmm0, (%rdi)
 ; NOAA-NEXT:    retq
 ;
 ; AA-LABEL: onealloc_readback_1:
 ; AA:       # %bb.0: # %entry
 ; AA-NEXT:    vmovups (%rsi), %xmm0
-; AA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; AA-NEXT:    vmovups %xmm0, (%rdi)
 ; AA-NEXT:    retq
 entry:
   %alloc = alloca [16 x i8], i8 2, align 1
   %part2 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc, i64 0, i64 0
   %part1 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc, i64 1, i64 0
   call void @llvm.lifetime.start.p0i8(i64 16, i8* nonnull %part1)
   call void @llvm.lifetime.start.p0i8(i64 16, i8* nonnull %part2)

 ; NOAA-NEXT:    vmovups (%rsi), %xmm0
 ; NOAA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; NOAA-NEXT:    vmovups %xmm0, (%rdi)
 ; NOAA-NEXT:    retq
 ;
 ; AA-LABEL: onealloc_readback_2:
 ; AA:       # %bb.0: # %entry
 ; AA-NEXT:    vmovups (%rsi), %xmm0
-; AA-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; AA-NEXT:    vmovups %xmm0, (%rdi)
 ; AA-NEXT:    retq
 entry:
   %alloc = alloca [16 x i8], i8 2, align 1
   %part1 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc, i64 0, i64 0
   %part2 = getelementptr inbounds [16 x i8], [16 x i8]* %alloc, i64 1, i64 0
   call void @llvm.lifetime.start.p0i8(i64 16, i8* nonnull %part1)
   call void @llvm.lifetime.start.p0i8(i64 16, i8* nonnull %part2)

llvm/trunk/test/DebugInfo/COFF/lexicalblock.ll

 ; CHECK: }
 ; CHECK: BlockSym {
 ; CHECK:   Kind: S_BLOCK32 {{.*}}
 ; CHECK:   BlockName: 
 ; CHECK: }
 ; CHECK: ScopeEndSym {
 ; CHECK:   Kind: S_END {{.*}}
 ; CHECK: }
-; CHECK: BlockSym {
-; CHECK:   Kind: S_BLOCK32 {{.*}}
-; CHECK:   BlockName: 
-; CHECK: }
-; CHECK: ScopeEndSym {
-; CHECK: }
-; CHECK: BlockSym {
-; CHECK:   Kind: S_BLOCK32 {{.*}}
-; CHECK:   BlockName: 
-; CHECK: }
-; CHECK: ScopeEndSym {
-; CHECK:   Kind: S_END {{.*}}
-; CHECK: }
 ; CHECK: ProcEnd {
 ; CHECK: }
 ;
 ; ModuleID = 'lexicalblock.cxx'
 source_filename = "lexicalblock.cxx"
 target datalayout = "e-m:x-p:32:32-i64:64-f80:32-n8:16:32-a:0:32-S32"
 target triple = "i686-pc-windows-msvc"