From bdad9e197bcf7be73d4e121e6d297c5345295398 Mon Sep 17 00:00:00 2001
From: Mathias Vorreiter Pedersen <mathiasvp@github.com>
Date: Thu, 24 Aug 2023 15:57:39 +0100
Subject: [PATCH] C++: Fix more joins.

Before:
```
[2023-08-24 15:47:20] Evaluated non-recursive predicate _IRBlock#896e97af::IRBlock::dominates#1#dispred#ff_Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInpu__#antijoin_rhs@94b1847k in 9831ms (size: 70660).
Evaluated relational algebra for predicate _IRBlock#896e97af::IRBlock::dominates#1#dispred#ff_Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInpu__#antijoin_rhs@94b1847k with tuple counts:
    1121232  ~0%    {4} r1 = JOIN _DataFlowUtil#47741e1f::Cached::simpleLocalFlowStep#2#ff_10#join_rhs_DataFlowUtil#47741e1f::TSsaPhiN__#shared WITH Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInput#::DefinitionExt::definesAt#4#dispred#fffff ON FIRST 1 OUTPUT Rhs.2, Lhs.0, Lhs.1, Lhs.2
  265759166  ~2%    {4} r2 = JOIN r1 WITH IRBlock#896e97af::IRBlock::dominates#1#dispred#ff ON FIRST 1 OUTPUT Lhs.3, Rhs.1, Lhs.1, Lhs.2
      70684  ~5%    {3} r3 = JOIN r2 WITH project#DataFlowUtil#47741e1f::Node::hasIndexInBlock#fff ON FIRST 2 OUTPUT Lhs.2, Lhs.3, Lhs.0
                    return r3

[2023-08-24 15:47:29] Evaluated non-recursive predicate DataFlowUtil#47741e1f::SsaPhiNode::getAnInput#1#dispred#fff@b6f296tl in 8943ms (size: 1121232).
Evaluated relational algebra for predicate DataFlowUtil#47741e1f::SsaPhiNode::getAnInput#1#dispred#fff@b6f296tl with tuple counts:
    1050572   ~2%    {3} r1 = _DataFlowUtil#47741e1f::Cached::simpleLocalFlowStep#2#ff_10#join_rhs_DataFlowUtil#47741e1f::TSsaPhiN__#shared AND NOT _IRBlock#896e97af::IRBlock::dominates#1#dispred#ff_Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInpu__#antijoin_rhs(Lhs.0, Lhs.1, Lhs.2)
    1050572   ~3%    {3} r2 = SCAN r1 OUTPUT In.1, false, In.2

    1121232   ~0%    {3} r3 = JOIN _DataFlowUtil#47741e1f::Cached::simpleLocalFlowStep#2#ff_10#join_rhs_DataFlowUtil#47741e1f::TSsaPhiN__#shared WITH Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInput#::DefinitionExt::definesAt#4#dispred#fffff ON FIRST 1 OUTPUT Rhs.2, Lhs.1, Lhs.2
  265759166   ~1%    {3} r4 = JOIN r3 WITH IRBlock#896e97af::IRBlock::dominates#1#dispred#ff ON FIRST 1 OUTPUT Lhs.2, Rhs.1, Lhs.1
      70684   ~0%    {2} r5 = JOIN r4 WITH project#DataFlowUtil#47741e1f::Node::hasIndexInBlock#fff ON FIRST 2 OUTPUT Lhs.2, Lhs.0
      70684   ~0%    {3} r6 = SCAN r5 OUTPUT In.0, true, In.1

    1121256   ~2%    {3} r7 = r2 UNION r6
                      return r7
```

After:
```
Evaluated non-recursive predicate DataFlowUtil#47741e1f::SsaPhiNode::getAnInput#1#dispred#fff@59ab2a2e in 456ms (size: 1117096).
Evaluated relational algebra for predicate DataFlowUtil#47741e1f::SsaPhiNode::getAnInput#1#dispred#fff@59ab2a2e with tuple counts:
    384518   ~0%    {2} r1 = JOIN DataFlowUtil#47741e1f::TSsaPhiNode#ff WITH Ssa#da392372::Make#SsaInternals#7b362d2f::SsaInput#::DefinitionExt::definesAt#4#dispred#fffff ON FIRST 1 OUTPUT Lhs.1, Rhs.2
  1121232   ~0%    {3} r2 = JOIN r1 WITH DataFlowUtil#47741e1f::Cached::simpleLocalFlowStep#2#ff_10#join_rhs ON FIRST 1 OUTPUT Rhs.1, Lhs.0, Lhs.1

  1117447   ~0%    {4} r3 = JOIN r2 WITH project#DataFlowUtil#47741e1f::Node::hasIndexInBlock#fff ON FIRST 1 OUTPUT Lhs.2, Rhs.1, Lhs.1, Lhs.0
    70684   ~0%    {2} r4 = JOIN r3 WITH IRBlock#896e97af::IRBlock::dominates#1#dispred#ff ON FIRST 2 OUTPUT Lhs.2, Lhs.3
    70684   ~0%    {3} r5 = SCAN r4 OUTPUT In.0, true, In.1

  1117447   ~0%    {4} r6 = JOIN r2 WITH project#DataFlowUtil#47741e1f::Node::hasIndexInBlock#fff ON FIRST 1 OUTPUT Lhs.2, Rhs.1, Lhs.1, Lhs.0
                    {4} r7 = r6 AND NOT IRBlock#896e97af::IRBlock::dominates#1#dispred#ff(Lhs.0, Lhs.1)
  1046763   ~0%    {2} r8 = SCAN r7 OUTPUT In.2, In.3
  1046763   ~3%    {3} r9 = SCAN r8 OUTPUT In.0, false, In.1

  1117447   ~2%    {3} r10 = r5 UNION r9
                    return r10
```
---
 .../semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll b/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll
index 42198d12372..9355493303a 100644
--- a/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll
+++ b/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll
@@ -553,9 +553,11 @@ class SsaPhiNode extends Node, TSsaPhiNode {
   cached
   final Node getAnInput(boolean fromBackEdge) {
     localFlowStep(result, this) and
-    if phi.getBasicBlock().dominates(result.getBasicBlock())
-    then fromBackEdge = true
-    else fromBackEdge = false
+    exists(IRBlock bPhi, IRBlock bResult |
+      bPhi = phi.getBasicBlock() and bResult = result.getBasicBlock()
+    |
+      if bPhi.dominates(bResult) then fromBackEdge = true else fromBackEdge = false
+    )
   }
 
   /** Gets a node that is used as input to this phi node. */