Index: lib/Transforms/InstCombine/InstCombineCasts.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -706,8 +706,9 @@
   if (SimplifyDemandedInstructionBits(CI))
     return &CI;
 
-  // Canonicalize trunc x to i1 -> (icmp ne (and x, 1), 0), likewise for vector.
-  if (DestTy->getScalarSizeInBits() == 1) {
+  // Canonicalize trunc x to i1 -> icmp ne (and x, 1), 0 (scalar only).
+  // TODO: Why is using an icmp preferable to the minimal form?
+  if (DestTy->getPrimitiveSizeInBits() == 1) {
     Constant *One = ConstantInt::get(SrcTy, 1);
     Src = Builder.CreateAnd(Src, One);
     Value *Zero = Constant::getNullValue(Src->getType());
Index: lib/Transforms/InstCombine/InstCombineCompares.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -1708,6 +1708,12 @@
   if (!Cmp.isEquality())
     return nullptr;
 
+  // For vectors: icmp ne (and X, 1), 0 --> trunc X to N x i1
+  // TODO: We canonicalize to the longer form for scalars. Why?
+  if (Cmp.getPredicate() == CmpInst::ICMP_NE && Cmp.getType()->isVectorTy() &&
+      match(Cmp.getOperand(1), m_Zero()) && match(And->getOperand(1), m_One()))
+    return CastInst::CreateTruncOrBitCast(And->getOperand(0), Cmp.getType());
+
   // X & -C == -C -> X >  u ~C
   // X & -C != -C -> X <= u ~C
   //   iff C is a power of 2
Index: test/Transforms/InstCombine/apint-shl-trunc.ll
===================================================================
--- test/Transforms/InstCombine/apint-shl-trunc.ll
+++ test/Transforms/InstCombine/apint-shl-trunc.ll
@@ -27,9 +27,8 @@
 
 define <2 x i1> @test0vec(<2 x i39> %X, <2 x i39> %A) {
 ; CHECK-LABEL: @test0vec(
-; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i39> <i39 1, i39 1>, [[A:%.*]]
-; CHECK-NEXT:    [[TMP2:%.*]] = and <2 x i39> [[TMP1]], [[X:%.*]]
-; CHECK-NEXT:    [[D:%.*]] = icmp ne <2 x i39> [[TMP2]], zeroinitializer
+; CHECK-NEXT:    [[B:%.*]] = lshr <2 x i39> [[X:%.*]], [[A:%.*]]
+; CHECK-NEXT:    [[D:%.*]] = trunc <2 x i39> [[B]] to <2 x i1>
 ; CHECK-NEXT:    ret <2 x i1> [[D]]
 ;
   %B = lshr <2 x i39> %X, %A
Index: test/Transforms/InstCombine/vector-casts.ll
===================================================================
--- test/Transforms/InstCombine/vector-casts.ll
+++ test/Transforms/InstCombine/vector-casts.ll
@@ -5,8 +5,7 @@
 
 define <2 x i1> @trunc(<2 x i64> %a) {
 ; CHECK-LABEL: @trunc(
-; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i64> [[A:%.*]], <i64 1, i64 1>
-; CHECK-NEXT:    [[T:%.*]] = icmp ne <2 x i64> [[TMP1]], zeroinitializer
+; CHECK-NEXT:    [[T:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i1>
 ; CHECK-NEXT:    ret <2 x i1> [[T]]
 ;
   %t = trunc <2 x i64> %a to <2 x i1>
@@ -15,8 +14,7 @@
 
 define <2 x i1> @and_cmp_is_trunc(<2 x i64> %a) {
 ; CHECK-LABEL: @and_cmp_is_trunc(
-; CHECK-NEXT:    [[T:%.*]] = and <2 x i64> [[A:%.*]], <i64 1, i64 1>
-; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i64> [[T]], zeroinitializer
+; CHECK-NEXT:    [[R:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %t = and <2 x i64> %a, <i64 1, i64 1>