Index: lib/CodeGen/SelectionDAG/SelectionDAG.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -3060,6 +3060,25 @@
       if (Tmp == SrcBits)
         return VTBits;
     }
+
+    // We often bitcast for bitwise ops before bitcasting back to the original
+    // value type, so we can peek through the bitcasts and use the minimum
+    // number of sign bits.
+    if ((N0.getOpcode() == ISD::AND || N0.getOpcode() == ISD::OR ||
+         N0.getOpcode() == ISD::XOR) &&
+        N0.getOperand(0).getOpcode() == ISD::BITCAST &&
+        N0.getOperand(1).getOpcode() == ISD::BITCAST) {
+      SDValue N00 = N0.getOperand(0).getOperand(0);
+      SDValue N01 = N0.getOperand(1).getOperand(0);
+      if (N00.getValueType() == VT && N01.getValueType() == VT) {
+        Tmp = ComputeNumSignBits(N00, Depth + 1);
+        if (Tmp == 1)
+          return Tmp;
+        Tmp2 = ComputeNumSignBits(N01, Depth + 1);
+        FirstAnswer = std::min(Tmp, Tmp2);
+      }
+    }
+
     break;
   }
 
Index: test/CodeGen/X86/bitcast-and-setcc-128.ll
===================================================================
--- test/CodeGen/X86/bitcast-and-setcc-128.ll
+++ test/CodeGen/X86/bitcast-and-setcc-128.ll
@@ -6,33 +6,22 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX512
 
 define i8 @v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c, <8 x i16> %d) {
-; SSE2-LABEL: v8i16:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSE2-NEXT:    pand %xmm0, %xmm2
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm2
-; SSE2-NEXT:    packuswb %xmm2, %xmm2
-; SSE2-NEXT:    pmovmskb %xmm2, %eax
-; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSE2-NEXT:    ret{{[l|q]}}
-;
-; SSSE3-LABEL: v8i16:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSSE3-NEXT:    pand %xmm0, %xmm2
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm2 = xmm2[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
-; SSSE3-NEXT:    pmovmskb %xmm2, %eax
-; SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSSE3-NEXT:    ret{{[l|q]}}
+; SSE2-SSSE3-LABEL: v8i16:
+; SSE2-SSSE3:       # BB#0:
+; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm2
+; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm2
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-SSSE3-NEXT:    ret{{[l|q]}}
 ;
 ; AVX12-LABEL: v8i16:
 ; AVX12:       # BB#0:
 ; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtw %xmm3, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX12-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX12-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    ret{{[l|q]}}
@@ -754,42 +743,23 @@
 }
 
 define i8 @v8i8(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d) {
-; SSE2-LABEL: v8i8:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    psllw $8, %xmm3
-; SSE2-NEXT:    psraw $8, %xmm3
-; SSE2-NEXT:    psllw $8, %xmm2
-; SSE2-NEXT:    psraw $8, %xmm2
-; SSE2-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSE2-NEXT:    psllw $8, %xmm1
-; SSE2-NEXT:    psraw $8, %xmm1
-; SSE2-NEXT:    psllw $8, %xmm0
-; SSE2-NEXT:    psraw $8, %xmm0
-; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-NEXT:    pand %xmm2, %xmm0
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE2-NEXT:    packuswb %xmm0, %xmm0
-; SSE2-NEXT:    pmovmskb %xmm0, %eax
-; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSE2-NEXT:    ret{{[l|q]}}
-;
-; SSSE3-LABEL: v8i8:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    psllw $8, %xmm3
-; SSSE3-NEXT:    psraw $8, %xmm3
-; SSSE3-NEXT:    psllw $8, %xmm2
-; SSSE3-NEXT:    psraw $8, %xmm2
-; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSSE3-NEXT:    psllw $8, %xmm1
-; SSSE3-NEXT:    psraw $8, %xmm1
-; SSSE3-NEXT:    psllw $8, %xmm0
-; SSSE3-NEXT:    psraw $8, %xmm0
-; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSSE3-NEXT:    pand %xmm2, %xmm0
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
-; SSSE3-NEXT:    pmovmskb %xmm0, %eax
-; SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSSE3-NEXT:    ret{{[l|q]}}
+; SSE2-SSSE3-LABEL: v8i8:
+; SSE2-SSSE3:       # BB#0:
+; SSE2-SSSE3-NEXT:    psllw $8, %xmm3
+; SSE2-SSSE3-NEXT:    psraw $8, %xmm3
+; SSE2-SSSE3-NEXT:    psllw $8, %xmm2
+; SSE2-SSSE3-NEXT:    psraw $8, %xmm2
+; SSE2-SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSE2-SSSE3-NEXT:    psllw $8, %xmm1
+; SSE2-SSSE3-NEXT:    psraw $8, %xmm1
+; SSE2-SSSE3-NEXT:    psllw $8, %xmm0
+; SSE2-SSSE3-NEXT:    psraw $8, %xmm0
+; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm0
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-SSSE3-NEXT:    ret{{[l|q]}}
 ;
 ; AVX12-LABEL: v8i8:
 ; AVX12:       # BB#0:
@@ -804,7 +774,7 @@
 ; AVX12-NEXT:    vpsraw $8, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX12-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX12-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    ret{{[l|q]}}
Index: test/CodeGen/X86/bitcast-and-setcc-256.ll
===================================================================
--- test/CodeGen/X86/bitcast-and-setcc-256.ll
+++ test/CodeGen/X86/bitcast-and-setcc-256.ll
@@ -210,34 +210,19 @@
 }
 
 define i8 @v8i32(<8 x i32> %a, <8 x i32> %b, <8 x i32> %c, <8 x i32> %d) {
-; SSE2-LABEL: v8i32:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm1
-; SSE2-NEXT:    pcmpgtd %xmm2, %xmm0
-; SSE2-NEXT:    packssdw %xmm1, %xmm0
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm6, %xmm4
-; SSE2-NEXT:    packssdw %xmm5, %xmm4
-; SSE2-NEXT:    pand %xmm0, %xmm4
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm4
-; SSE2-NEXT:    packuswb %xmm4, %xmm4
-; SSE2-NEXT:    pmovmskb %xmm4, %eax
-; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSE2-NEXT:    ret{{[l|q]}}
-;
-; SSSE3-LABEL: v8i32:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm1
-; SSSE3-NEXT:    pcmpgtd %xmm2, %xmm0
-; SSSE3-NEXT:    packssdw %xmm1, %xmm0
-; SSSE3-NEXT:    pcmpgtd %xmm7, %xmm5
-; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm4
-; SSSE3-NEXT:    packssdw %xmm5, %xmm4
-; SSSE3-NEXT:    pand %xmm0, %xmm4
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm4 = xmm4[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
-; SSSE3-NEXT:    pmovmskb %xmm4, %eax
-; SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSSE3-NEXT:    ret{{[l|q]}}
+; SSE2-SSSE3-LABEL: v8i32:
+; SSE2-SSSE3:       # BB#0:
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm1
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm7, %xmm5
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm6, %xmm4
+; SSE2-SSSE3-NEXT:    packssdw %xmm5, %xmm4
+; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm4
+; SSE2-SSSE3-NEXT:    packsswb %xmm0, %xmm4
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm4, %eax
+; SSE2-SSSE3-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-SSSE3-NEXT:    ret{{[l|q]}}
 ;
 ; AVX1-LABEL: v8i32:
 ; AVX1:       # BB#0:
Index: test/CodeGen/X86/widen_arith-2.ll
===================================================================
--- test/CodeGen/X86/widen_arith-2.ll
+++ test/CodeGen/X86/widen_arith-2.ll
@@ -24,7 +24,7 @@
 ; CHECK-NEXT:    pmovzxbw {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; CHECK-NEXT:    psubw %xmm0, %xmm2
 ; CHECK-NEXT:    pand %xmm1, %xmm2
-; CHECK-NEXT:    packsswb %xmm0, %xmm2
+; CHECK-NEXT:    packuswb %xmm0, %xmm2
 ; CHECK-NEXT:    movq %xmm2, (%edx,%eax,8)
 ; CHECK-NEXT:    incl (%esp)
 ; CHECK-NEXT:  .LBB0_1: # %forcond