[ConstProp] Handle insertelement constants

Previously ConstantFoldExtractElementInstruction() would only work with
insertelement instructions, not contants. This properly handles
insertelement constants as well.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D85865

diff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp
index f02246c..2d3afca 100644 (file)
--- a/llvm/lib/IR/ConstantFold.cpp
+++ b/llvm/lib/IR/ConstantFold.cpp
@@ -853,6 +853,15 @@ Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val,
       }
       return CE->getWithOperands(Ops, ValVTy->getElementType(), false,
                                  Ops[0]->getType()->getPointerElementType());
+    } else if (CE->getOpcode() == Instruction::InsertElement) {
+      if (const auto *IEIdx = dyn_cast<ConstantInt>(CE->getOperand(2))) {
+        if (APSInt::isSameValue(APSInt(IEIdx->getValue()),
+                                APSInt(CIdx->getValue()))) {
+          return CE->getOperand(1);
+        } else {
+          return ConstantExpr::getExtractElement(CE->getOperand(0), CIdx);
+        }
+      }
     }
   }
 

diff --git a/llvm/test/Transforms/InstSimplify/vscale.ll b/llvm/test/Transforms/InstSimplify/vscale.ll
index a96b943..dd4ca47 100644 (file)
--- a/llvm/test/Transforms/InstSimplify/vscale.ll
+++ b/llvm/test/Transforms/InstSimplify/vscale.ll
@@ -95,6 +95,22 @@ define i32 @insert_extract_element_same_vec_idx_2() {
   ret i32 %r
 }
 
+define i32 @insert_extract_element_same_vec_idx_3() {
+; CHECK-LABEL: @insert_extract_element_same_vec_idx_3(
+; CHECK-NEXT:    ret i32 1
+;
+  %r = extractelement <vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i64 4), i64 4
+  ret i32 %r
+}
+
+define i32 @insert_extract_element_same_vec_idx_4() {
+; CHECK-LABEL: @insert_extract_element_same_vec_idx_4(
+; CHECK-NEXT:    ret i32 1
+;
+  %r = extractelement <vscale x 4 x i32> insertelement (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i32 4), i32 2, i64 3), i64 4
+  ret i32 %r
+}
+
 ; more complicated expressions
 
 define <vscale x 2 x i1> @cmp_le_smax_always_true(<vscale x 2 x i64> %x) {