[llvm] [InstCombine] Fold shifts + selects with -1 to scmp(X, 0) (PR #164129)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-transforms

Author: AZero13 (AZero13)

<details>
<summary>Changes</summary>

This is because the sign function with 0 tends to be folded to ashr and other things.

---
Full diff: https://github.com/llvm/llvm-project/pull/164129.diff


2 Files Affected:

- (modified) llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp (+25) 
- (modified) llvm/test/Transforms/InstCombine/scmp.ll (+39-3) 


``````````diff
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index a8eb9b9cf6a84..71aa94045f6ec 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -4680,5 +4680,30 @@ Instruction *InstCombinerImpl::visitSelectInst(SelectInst &SI) {
                                      Align(MaskedLoadAlignment->getZExtValue()),
                                      CondVal, FalseVal));
 
+  // Canonicalize sign function ashr pattern: select (icmp slt X, 1), ashr X,
+  // bitwidth-1, 1 -> scmp(X, 0)
+  // Also handles: select (icmp sgt X, 0), 1, ashr X, bitwidth-1 -> scmp(X, 0)
+  Value *X;
+  unsigned BitWidth = SI.getType()->getScalarSizeInBits();
+  CmpPredicate Pred;
+  Value *CmpLHS, *CmpRHS;
+
+  // Canonicalize sign function ashr patterns:
+  // select (icmp slt X, 1), ashr X, bitwidth-1, 1 -> scmp(X, 0)
+  // select (icmp sgt X, 0), 1, ashr X, bitwidth-1 -> scmp(X, 0)
+  if (match(&SI, m_Select(m_ICmp(Pred, m_Value(CmpLHS), m_Value(CmpRHS)),
+                          m_Value(TrueVal), m_Value(FalseVal))) &&
+      ((Pred == ICmpInst::ICMP_SLT && match(CmpLHS, m_Value(X)) && match(CmpRHS, m_One()) &&
+        match(TrueVal, m_AShr(m_Deferred(X), m_SpecificInt(BitWidth - 1))) &&
+        match(FalseVal, m_One())) ||
+       (Pred == ICmpInst::ICMP_SGT && match(CmpLHS, m_Value(X)) && match(CmpRHS, m_Zero()) &&
+        match(TrueVal, m_One()) &&
+        match(FalseVal, m_AShr(m_Deferred(X), m_SpecificInt(BitWidth - 1)))))) {
+
+    Function *Scmp = Intrinsic::getOrInsertDeclaration(
+        SI.getModule(), Intrinsic::scmp, {SI.getType(), SI.getType()});
+    return CallInst::Create(Scmp, {X, ConstantInt::get(SI.getType(), 0)});
+  }
+  
   return nullptr;
 }
diff --git a/llvm/test/Transforms/InstCombine/scmp.ll b/llvm/test/Transforms/InstCombine/scmp.ll
index c0be5b986b7fd..cada97aeadbad 100644
--- a/llvm/test/Transforms/InstCombine/scmp.ll
+++ b/llvm/test/Transforms/InstCombine/scmp.ll
@@ -519,9 +519,7 @@ define <3 x i2> @scmp_unary_shuffle_ops(<3 x i8> %x, <3 x i8> %y) {
 define i32 @scmp_sgt_slt(i32 %a) {
 ; CHECK-LABEL: define i32 @scmp_sgt_slt(
 ; CHECK-SAME: i32 [[A:%.*]]) {
-; CHECK-NEXT:    [[A_LOBIT:%.*]] = ashr i32 [[A]], 31
-; CHECK-NEXT:    [[CMP_INV:%.*]] = icmp slt i32 [[A]], 1
-; CHECK-NEXT:    [[RETVAL_0:%.*]] = select i1 [[CMP_INV]], i32 [[A_LOBIT]], i32 1
+; CHECK-NEXT:    [[RETVAL_0:%.*]] = call i32 @llvm.scmp.i32.i32(i32 [[A]], i32 0)
 ; CHECK-NEXT:    ret i32 [[RETVAL_0]]
 ;
   %cmp = icmp sgt i32 %a, 0
@@ -747,3 +745,41 @@ define i8 @scmp_from_select_eq_and_gt_neg3(i32 %x, i32 %y) {
   %r = select i1 %eq, i8 0, i8 %sel1
   ret i8 %r
 }
+
+define i32 @scmp_ashr(i32 %a) {
+; CHECK-LABEL: define i32 @scmp_ashr(
+; CHECK-SAME: i32 [[A:%.*]]) {
+; CHECK-NEXT:    [[RETVAL_0:%.*]] = call i32 @llvm.scmp.i32.i32(i32 [[A]], i32 0)
+; CHECK-NEXT:    ret i32 [[RETVAL_0]]
+;
+  %a.lobit = ashr i32 %a, 31
+  %cmp.inv = icmp slt i32 %a, 1
+  %retval.0 = select i1 %cmp.inv, i32 %a.lobit, i32 1
+  ret i32 %retval.0
+}
+
+; Test the new SGT pattern: select (icmp sgt X, 0), 1, ashr X, bitwidth-1 -> scmp(X, 0)
+define i8 @scmp_ashr_sgt_pattern(i8 %a) {
+; CHECK-LABEL: define i8 @scmp_ashr_sgt_pattern(
+; CHECK-SAME: i8 [[A:%.*]]) {
+; CHECK-NEXT:    [[R:%.*]] = call i8 @llvm.scmp.i8.i8(i8 [[A]], i8 0)
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %a.lobit = ashr i8 %a, 7
+  %cmp = icmp sgt i8 %a, 0
+  %retval = select i1 %cmp, i8 1, i8 %a.lobit
+  ret i8 %retval
+}
+
+; Test the SLT pattern: select (icmp slt X, 1), ashr X, bitwidth-1, 1 -> scmp(X, 0)
+define i8 @scmp_ashr_slt_pattern(i8 %a) {
+; CHECK-LABEL: define i8 @scmp_ashr_slt_pattern(
+; CHECK-SAME: i8 [[A:%.*]]) {
+; CHECK-NEXT:    [[R:%.*]] = call i8 @llvm.scmp.i8.i8(i8 [[A]], i8 0)
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %a.lobit = ashr i8 %a, 7
+  %cmp = icmp slt i8 %a, 1
+  %retval = select i1 %cmp, i8 %a.lobit, i8 1
+  ret i8 %retval
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/164129