Index: lib/Target/ARM/ARMISelLowering.cpp
===================================================================
--- lib/Target/ARM/ARMISelLowering.cpp
+++ lib/Target/ARM/ARMISelLowering.cpp
@@ -4049,6 +4049,33 @@
     std::swap(LHS, RHS);
   }
 
+  // Thumb1 has very limited immediate modes, so turning an "and" into a
+  // shift can save multiple instructions.
+  //
+  // If we have (x & C1), and C1 is an appropriate mask, we can transform it
+  // into "((x << n) >> n)".  But that isn't necessarily profitable on its
+  // own. If it's the operand to an unsigned comparison with an immediate,
+  // we can eliminate one of the shifts: we transform
+  // "((x << n) >> n) == C2" to "(x << n) == (C2 << n)".
+  //
+  // We avoid transforming cases where the mask is small because some cases
+  // aren't obviously profitable:
+  //
+  // 1. C2 fits into the immediate field of a cmp.
+  // 2. C1 is 255 or 65535, so we can use "uxtb" or "uxth".
+  if (Subtarget->isThumb1Only() && LHS->getOpcode() == ISD::AND &&
+      LHS->hasOneUse() && isa<ConstantSDNode>(LHS.getOperand(1)) &&
+      LHS.getValueType() == MVT::i32 && isa<ConstantSDNode>(RHS) &&
+      !isSignedIntSetCC(CC)) {
+    unsigned Mask = cast<ConstantSDNode>(LHS.getOperand(1))->getZExtValue();
+    auto *RHSC = cast<ConstantSDNode>(RHS.getNode());
+    if (isMask_32(Mask) && (RHSC->getZExtValue() & ~Mask) == 0 && Mask > 65535) {
+      unsigned ShiftBits = countLeadingZeros(Mask);
+      SDValue ShiftAmt = DAG.getConstant(ShiftBits, dl, MVT::i32);
+      LHS = DAG.getNode(ISD::SHL, dl, MVT::i32, LHS.getOperand(0), ShiftAmt);
+      RHS = DAG.getNode(ISD::SHL, dl, MVT::i32, RHS, ShiftAmt);
+    }
+  }
   ARMCC::CondCodes CondCode = IntCCToARMCC(CC);
 
   // If the RHS is a constant zero then the V (overflow) flag will never be
Index: test/CodeGen/Thumb/cmp-and-fold.ll
===================================================================
--- /dev/null
+++ test/CodeGen/Thumb/cmp-and-fold.ll
@@ -0,0 +1,157 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=thumbv6-eabi %s -o - | FileCheck %s
+
+define void @test1(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    lsls r0, r0, #2
+; CHECK-NEXT:    cmp r0, #68
+; CHECK-NEXT:    beq .LBB0_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB0_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = and i32 %x, 1073741823
+  %cmp = icmp eq i32 %a, 17
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+define void @test2(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test2:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    movs r2, #1
+; CHECK-NEXT:    lsls r2, r2, #31
+; CHECK-NEXT:    lsls r0, r0, #7
+; CHECK-NEXT:    cmp r0, r2
+; CHECK-NEXT:    bhi .LBB1_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB1_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = shl i32 %x, 7
+  %cmp = icmp ugt i32 %a, 2147483648
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+define void @test3(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test3:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    movs r2, #1
+; CHECK-NEXT:    lsls r2, r2, #31
+; CHECK-NEXT:    lsls r0, r0, #2
+; CHECK-NEXT:    cmp r0, r2
+; CHECK-NEXT:    bhi .LBB2_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB2_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = and i32 %x, 1073741823
+  %cmp = icmp ugt i32 %a, 536870912
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+define void @test4(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test4:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    uxtb r0, r0
+; CHECK-NEXT:    cmp r0, #17
+; CHECK-NEXT:    beq .LBB3_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB3_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = and i32 %x, 255
+  %cmp = icmp eq i32 %a, 17
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+define void @test5(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test5:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    uxth r0, r0
+; CHECK-NEXT:    cmp r0, #17
+; CHECK-NEXT:    beq .LBB4_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB4_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = and i32 %x, 65535
+  %cmp = icmp eq i32 %a, 17
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}
+
+define void @test6(i32 %x, void ()* %f)  {
+; CHECK-LABEL: test6:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    movs r2, #32
+; CHECK-NEXT:    ands r2, r0
+; CHECK-NEXT:    cmp r2, #17
+; CHECK-NEXT:    beq .LBB5_2
+; CHECK-NEXT:  @ %bb.1: @ %if.then
+; CHECK-NEXT:    blx r1
+; CHECK-NEXT:  .LBB5_2: @ %if.end
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a = and i32 %x, 32
+  %cmp = icmp eq i32 %a, 17
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:
+  tail call void %f()
+  br label %if.end
+
+if.end:
+  ret void
+}