diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp --- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp @@ -3364,6 +3364,23 @@ llvm_unreachable("All possible folds are handled."); } + // The mask value may be a vector constant that has undefined elements. But it + // may not be safe to propagate those undefs into the new compare, so replace + // those elements by copying an existing, defined, and safe scalar constant. + Type *OpTy = M->getType(); + auto *VecC = dyn_cast(M); + if (OpTy->isVectorTy() && VecC && VecC->containsUndefElement()) { + Constant *SafeReplacementConstant = nullptr; + for (unsigned i = 0, e = OpTy->getVectorNumElements(); i != e; ++i) { + if (!isa(VecC->getAggregateElement(i))) { + SafeReplacementConstant = VecC->getAggregateElement(i); + break; + } + } + assert(SafeReplacementConstant && "Failed to find undef replacement"); + M = Constant::replaceUndefsWith(VecC, SafeReplacementConstant); + } + return Builder.CreateICmp(DstPred, X, M); } diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll --- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll +++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll @@ -82,7 +82,7 @@ define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_splat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x, @@ -92,7 +92,7 @@ define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_nonsplat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x, diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll --- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll +++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll @@ -76,7 +76,7 @@ define <3 x i1> @p3_vec_splat_undef() { ; CHECK-LABEL: @p3_vec_splat_undef( ; CHECK-NEXT: [[X:%.*]] = call <3 x i8> @gen3x8() -; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %x = call <3 x i8> @gen3x8() @@ -88,7 +88,7 @@ define <3 x i1> @p3_vec_nonsplat_undef() { ; CHECK-LABEL: @p3_vec_nonsplat_undef( ; CHECK-NEXT: [[X:%.*]] = call <3 x i8> @gen3x8() -; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %x = call <3 x i8> @gen3x8() diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll --- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll +++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll @@ -61,7 +61,7 @@ define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_splat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x, @@ -71,7 +71,7 @@ define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_nonsplat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x, diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll --- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll +++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll @@ -99,7 +99,7 @@ define <3 x i1> @p3_vec_splat_undef() { ; CHECK-LABEL: @p3_vec_splat_undef( ; CHECK-NEXT: [[X:%.*]] = call <3 x i8> @gen3x8() -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %x = call <3 x i8> @gen3x8() @@ -111,7 +111,7 @@ define <3 x i1> @p3_vec_nonsplat_undef() { ; CHECK-LABEL: @p3_vec_nonsplat_undef( ; CHECK-NEXT: [[X:%.*]] = call <3 x i8> @gen3x8() -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %x = call <3 x i8> @gen3x8() diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll --- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll +++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll @@ -83,7 +83,7 @@ define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_splat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x, @@ -93,7 +93,7 @@ define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) { ; CHECK-LABEL: @p3_vec_nonsplat_undef( -; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], +; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <3 x i8> [[X:%.*]], ; CHECK-NEXT: ret <3 x i1> [[TMP1]] ; %tmp0 = and <3 x i8> %x,