Skip to content

[GlobalISel][IRTranslator] Port switch binary tree search optimization. #77279

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jan 8, 2024

Conversation

aemerson
Copy link
Contributor

@aemerson aemerson commented Jan 8, 2024

This re-uses some code extracted earlier from SelectionDAG into SwitchLoweringUtils

Much of the code is a straight port from SDAG's splitWorkItem(), with minor changes needed for GISel.

This re-uses some code extracted earlier from SelectionDAG into SwitchLoweringUtils

Much of the code is a straight port from SDAG's splitWorkItem(), with minor
changes needed for GISel.
@llvmbot
Copy link
Member

llvmbot commented Jan 8, 2024

@llvm/pr-subscribers-backend-aarch64

@llvm/pr-subscribers-llvm-globalisel

Author: Amara Emerson (aemerson)

Changes

This re-uses some code extracted earlier from SelectionDAG into SwitchLoweringUtils

Much of the code is a straight port from SDAG's splitWorkItem(), with minor changes needed for GISel.


Full diff: https://github.com/llvm/llvm-project/pull/77279.diff

3 Files Affected:

  • (modified) llvm/include/llvm/CodeGen/GlobalISel/IRTranslator.h (+4)
  • (modified) llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp (+77-2)
  • (modified) llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-switch-split.ll (+18-16)
diff --git a/llvm/include/llvm/CodeGen/GlobalISel/IRTranslator.h b/llvm/include/llvm/CodeGen/GlobalISel/IRTranslator.h
index bffc03ed0187e6..1b094d9d9fe771 100644
--- a/llvm/include/llvm/CodeGen/GlobalISel/IRTranslator.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/IRTranslator.h
@@ -366,6 +366,10 @@ class IRTranslator : public MachineFunctionPass {
                        BranchProbability BranchProbToNext, Register Reg,
                        SwitchCG::BitTestCase &B, MachineBasicBlock *SwitchBB);
 
+  void splitWorkItem(SwitchCG::SwitchWorkList &WorkList,
+                     const SwitchCG::SwitchWorkListItem &W, Value *Cond,
+                     MachineBasicBlock *SwitchMBB, MachineIRBuilder &MIB);
+
   bool lowerJumpTableWorkItem(
       SwitchCG::SwitchWorkListItem W, MachineBasicBlock *SwitchMBB,
       MachineBasicBlock *CurMBB, MachineBasicBlock *DefaultMBB,
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index 9c11113902a24e..6708f2baa5ed5f 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -751,16 +751,91 @@ bool IRTranslator::translateSwitch(const User &U, MachineIRBuilder &MIB) {
   auto DefaultProb = getEdgeProbability(SwitchMBB, DefaultMBB);
   WorkList.push_back({SwitchMBB, First, Last, nullptr, nullptr, DefaultProb});
 
-  // FIXME: At the moment we don't do any splitting optimizations here like
-  // SelectionDAG does, so this worklist only has one entry.
   while (!WorkList.empty()) {
     SwitchWorkListItem W = WorkList.pop_back_val();
+
+    unsigned NumClusters = W.LastCluster - W.FirstCluster + 1;
+    // For optimized builds, lower large range as a balanced binary tree.
+    if (NumClusters > 3 &&
+        MF->getTarget().getOptLevel() != CodeGenOptLevel::None &&
+        !DefaultMBB->getParent()->getFunction().hasMinSize()) {
+      splitWorkItem(WorkList, W, SI.getCondition(), SwitchMBB, MIB);
+      continue;
+    }
+
     if (!lowerSwitchWorkItem(W, SI.getCondition(), SwitchMBB, DefaultMBB, MIB))
       return false;
   }
   return true;
 }
 
+void IRTranslator::splitWorkItem(SwitchCG::SwitchWorkList &WorkList,
+                                 const SwitchCG::SwitchWorkListItem &W,
+                                 Value *Cond, MachineBasicBlock *SwitchMBB,
+                                 MachineIRBuilder &MIB) {
+  using namespace SwitchCG;
+  assert(W.FirstCluster->Low->getValue().slt(W.LastCluster->Low->getValue()) &&
+         "Clusters not sorted?");
+  assert(W.LastCluster - W.FirstCluster + 1 >= 2 && "Too small to split!");
+
+  auto [LastLeft, FirstRight, LeftProb, RightProb] =
+      SL->computeSplitWorkItemInfo(W);
+
+  // Use the first element on the right as pivot since we will make less-than
+  // comparisons against it.
+  CaseClusterIt PivotCluster = FirstRight;
+  assert(PivotCluster > W.FirstCluster);
+  assert(PivotCluster <= W.LastCluster);
+
+  CaseClusterIt FirstLeft = W.FirstCluster;
+  CaseClusterIt LastRight = W.LastCluster;
+
+  const ConstantInt *Pivot = PivotCluster->Low;
+
+  // New blocks will be inserted immediately after the current one.
+  MachineFunction::iterator BBI(W.MBB);
+  ++BBI;
+
+  // We will branch to the LHS if Value < Pivot. If LHS is a single cluster,
+  // we can branch to its destination directly if it's squeezed exactly in
+  // between the known lower bound and Pivot - 1.
+  MachineBasicBlock *LeftMBB;
+  if (FirstLeft == LastLeft && FirstLeft->Kind == CC_Range &&
+      FirstLeft->Low == W.GE &&
+      (FirstLeft->High->getValue() + 1LL) == Pivot->getValue()) {
+    LeftMBB = FirstLeft->MBB;
+  } else {
+    LeftMBB = FuncInfo.MF->CreateMachineBasicBlock(W.MBB->getBasicBlock());
+    FuncInfo.MF->insert(BBI, LeftMBB);
+    WorkList.push_back(
+        {LeftMBB, FirstLeft, LastLeft, W.GE, Pivot, W.DefaultProb / 2});
+  }
+
+  // Similarly, we will branch to the RHS if Value >= Pivot. If RHS is a
+  // single cluster, RHS.Low == Pivot, and we can branch to its destination
+  // directly if RHS.High equals the current upper bound.
+  MachineBasicBlock *RightMBB;
+  if (FirstRight == LastRight && FirstRight->Kind == CC_Range && W.LT &&
+      (FirstRight->High->getValue() + 1ULL) == W.LT->getValue()) {
+    RightMBB = FirstRight->MBB;
+  } else {
+    RightMBB = FuncInfo.MF->CreateMachineBasicBlock(W.MBB->getBasicBlock());
+    FuncInfo.MF->insert(BBI, RightMBB);
+    WorkList.push_back(
+        {RightMBB, FirstRight, LastRight, Pivot, W.LT, W.DefaultProb / 2});
+  }
+
+  // Create the CaseBlock record that will be used to lower the branch.
+  CaseBlock CB(ICmpInst::Predicate::ICMP_SLT, false, Cond, Pivot, nullptr,
+               LeftMBB, RightMBB, W.MBB, MIB.getDebugLoc(), LeftProb,
+               RightProb);
+
+  if (W.MBB == SwitchMBB)
+    emitSwitchCase(CB, SwitchMBB, MIB);
+  else
+    SL->SwitchCases.push_back(CB);
+}
+
 void IRTranslator::emitJumpTable(SwitchCG::JumpTable &JT,
                                  MachineBasicBlock *MBB) {
   // Emit the code for the jump table
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-switch-split.ll b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-switch-split.ll
index 54c8eb913d5d40..55cf48ed2245fa 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-switch-split.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-switch-split.ll
@@ -17,31 +17,33 @@ define i32 @scanfile(i32 %call148) {
 ; CHECK-NEXT:    .cfi_offset w30, -8
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    cmp w0, #1
 ; CHECK-NEXT:    mov w0, wzr
-; CHECK-NEXT:    cbz w8, LBB0_7
+; CHECK-NEXT:    b.ge LBB0_3
 ; CHECK-NEXT:  ; %bb.1: ; %entry
-; CHECK-NEXT:    cmp w8, #1
-; CHECK-NEXT:    b.eq LBB0_7
-; CHECK-NEXT:  ; %bb.2: ; %entry
+; CHECK-NEXT:    cbnz w8, LBB0_7
+; CHECK-NEXT:  LBB0_2: ; %common.ret1
+; CHECK-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-NEXT:    ret
+; CHECK-NEXT:  LBB0_3: ; %entry
+; CHECK-NEXT:    b.eq LBB0_2
+; CHECK-NEXT:  ; %bb.4: ; %entry
 ; CHECK-NEXT:    cmp w8, #2
-; CHECK-NEXT:    b.eq LBB0_4
-; CHECK-NEXT:  ; %bb.3: ; %entry
+; CHECK-NEXT:    b.eq LBB0_6
+; CHECK-NEXT:  ; %bb.5: ; %entry
 ; CHECK-NEXT:    cmp w8, #3
-; CHECK-NEXT:    b.ne LBB0_5
-; CHECK-NEXT:  LBB0_4: ; %sw.bb300
+; CHECK-NEXT:    b.ne LBB0_2
+; CHECK-NEXT:  LBB0_6: ; %sw.bb300
 ; CHECK-NEXT:    bl _logg
 ; CHECK-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  LBB0_5: ; %entry
+; CHECK-NEXT:  LBB0_7: ; %entry
 ; CHECK-NEXT:    cmn w8, #2
-; CHECK-NEXT:    b.eq LBB0_8
-; CHECK-NEXT:  ; %bb.6: ; %entry
+; CHECK-NEXT:    b.eq LBB0_9
+; CHECK-NEXT:  ; %bb.8: ; %entry
 ; CHECK-NEXT:    cmn w8, #1
-; CHECK-NEXT:    b.eq LBB0_8
-; CHECK-NEXT:  LBB0_7: ; %common.ret1
-; CHECK-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; CHECK-NEXT:    ret
-; CHECK-NEXT:  LBB0_8: ; %sw.bb150
+; CHECK-NEXT:    b.ne LBB0_2
+; CHECK-NEXT:  LBB0_9: ; %sw.bb150
 ; CHECK-NEXT:    bl _logg
 ; CHECK-NEXT:    brk #0x1
 entry:

@aemerson aemerson merged commit ca20c99 into llvm:main Jan 8, 2024
justinfargnoli pushed a commit to justinfargnoli/llvm-project that referenced this pull request Jan 28, 2024
…n. (llvm#77279)

This re-uses some code extracted earlier from SelectionDAG into
SwitchLoweringUtils

Much of the code is a straight port from SDAG's splitWorkItem(), with
minor changes needed for GISel.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants