[CodeGen] Canonicalise adds/subs of i1 vectors using XOR

[david-arm]

Summary

When calling SelectionDAG::getNode() to create an ADD or SUB
of two vectors with i1 element types we can canonicalise this
to use XOR instead, where 1+1 is treated as wrapping around
to 0 and 0-1 wraps to 1.

I've added the following tests for SVE targets:

CodeGen/AArch64/sve-pred-arith.ll

and modified some X86 tests to reflect the much simpler codegen
required.

[craig.topper]

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

5318

Isn't this also possible for sub?

[david-arm]

• Also canonicalised SUB of two predicate vectors to using XOR.

[david-arm]

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

5318

Yes it is - this is what the X86 target already seems to do. Thanks for the suggestion!

[paulwalker-arm]

Not sure of the value of the "ILLEGAL" tests given they're actually testing we can type legalise ISD::XOR, which I'm presuming is already well tested. That said it's nothing I cannot live with.

[nikic]

Why is this in getNode() rather than DAGCombine? https://github.com/llvm/llvm-project/commit/6f5a805bbbed5d0cdaaf67846dffa7f044afb407 for example does something very similar in DAGCombine. What's the guideline for correct placement here?

[paulwalker-arm]

In D97276#2584555, @nikic wrote:

Why is this in getNode() rather than DAGCombine? https://github.com/llvm/llvm-project/commit/6f5a805bbbed5d0cdaaf67846dffa7f044afb407 for example does something very similar in DAGCombine. What's the guideline for correct placement here?

We've followed similar logic as used for i1 based min/max and reductions. My thinking is that the earlier you can canonicalise expressions the better, and you cannot get earlier than never creating them in the first place. Ultimately the goal here is that for some operations there is never a need to consider i1 based vectors. This is partially true today because i1 vectors are rarely a legal type. However i1 vectors are legal for SVE and thus we're hitting corner cases that have not been hit before (for aarch64 at least).

[RKSimon]

In D97276#2584555, @nikic wrote:

Why is this in getNode() rather than DAGCombine? https://github.com/llvm/llvm-project/commit/6f5a805bbbed5d0cdaaf67846dffa7f044afb407 for example does something very similar in DAGCombine. What's the guideline for correct placement here?

Something that purely depends on the VT should be handled in getNode() - I can't recall the reasons I didn't put it there at the the time

[nikic]

Thanks for the clarification!

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
   case ISD::SUB:
     assert(VT.isInteger() && "This operator does not apply to FP types!");
     assert(N1.getValueType() == N2.getValueType() &&
            N1.getValueType() == VT && "Binary operator types must match!");
     // (X ^|+- 0) -> X.  This commonly occurs when legalizing i64 values, so
     // it's worth handling here.
     if (N2C && N2C->isNullValue())
       return N1;
+    if ((Opcode == ISD::ADD || Opcode == ISD::SUB) && VT.isVector() &&
+        VT.getVectorElementType() == MVT::i1)
+      return getNode(ISD::XOR, DL, VT, N1, N2);
     break;
   case ISD::MUL:
     assert(VT.isInteger() && "This operator does not apply to FP types!");
     assert(N1.getValueType() == N2.getValueType() &&
            N1.getValueType() == VT && "Binary operator types must match!");
     if (N2C && (N1.getOpcode() == ISD::VSCALE) && Flags.hasNoSignedWrap()) {
       const APInt &MulImm = N1->getConstantOperandAPInt(0);
       const APInt &N2CImm = N2C->getAPIntValue();

llvm/test/CodeGen/AArch64/sve-pred-arith.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s 2>%t | FileCheck %s
+; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t
+
+; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it.
+; WARN-NOT: warning
+
+; LEGAL ADDS
+
+define <vscale x 16 x i1> @add_nxv16i1(<vscale x 16 x i1> %a, <vscale x 16 x i1> %b) {
+; CHECK-LABEL: add_nxv16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.b
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = add <vscale x 16 x i1> %a, %b
+  ret <vscale x 16 x i1> %res;
+}
+
+define <vscale x 8 x i1> @add_nxv8i1(<vscale x 8 x i1> %a, <vscale x 8 x i1> %b) {
+; CHECK-LABEL: add_nxv8i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.h
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = add <vscale x 8 x i1> %a, %b
+  ret <vscale x 8 x i1> %res;
+}
+
+define <vscale x 4 x i1> @add_nxv4i1(<vscale x 4 x i1> %a, <vscale x 4 x i1> %b) {
+; CHECK-LABEL: add_nxv4i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.s
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = add <vscale x 4 x i1> %a, %b
+  ret <vscale x 4 x i1> %res;
+}
+
+define <vscale x 2 x i1> @add_nxv2i1(<vscale x 2 x i1> %a, <vscale x 2 x i1> %b) {
+; CHECK-LABEL: add_nxv2i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.d
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = add <vscale x 2 x i1> %a, %b
+  ret <vscale x 2 x i1> %res;
+}
+
+
+; ILLEGAL ADDS
+
+define aarch64_sve_vector_pcs <vscale x 64 x i1> @add_nxv64i1(<vscale x 64 x i1> %a, <vscale x 64 x i1> %b) {
+; CHECK-LABEL: add_nxv64i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #3, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    ldr p4, [x3]
+; CHECK-NEXT:    ldr p5, [x0]
+; CHECK-NEXT:    ldr p6, [x1]
+; CHECK-NEXT:    ldr p7, [x2]
+; CHECK-NEXT:    ptrue p8.b
+; CHECK-NEXT:    eor p0.b, p8/z, p0.b, p5.b
+; CHECK-NEXT:    eor p1.b, p8/z, p1.b, p6.b
+; CHECK-NEXT:    eor p2.b, p8/z, p2.b, p7.b
+; CHECK-NEXT:    eor p3.b, p8/z, p3.b, p4.b
+; CHECK-NEXT:    ldr p8, [sp, #3, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %res = add <vscale x 64 x i1> %a, %b
+  ret <vscale x 64 x i1> %res;
+}
+
+
+; LEGAL SUBS
+
+define <vscale x 16 x i1> @sub_xv16i1(<vscale x 16 x i1> %a, <vscale x 16 x i1> %b) {
+; CHECK-LABEL: sub_xv16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.b
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = sub <vscale x 16 x i1> %a, %b
+  ret <vscale x 16 x i1> %res;
+}
+
+define <vscale x 8 x i1> @sub_xv8i1(<vscale x 8 x i1> %a, <vscale x 8 x i1> %b) {
+; CHECK-LABEL: sub_xv8i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.h
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = sub <vscale x 8 x i1> %a, %b
+  ret <vscale x 8 x i1> %res;
+}
+
+define <vscale x 4 x i1> @sub_xv4i1(<vscale x 4 x i1> %a, <vscale x 4 x i1> %b) {
+; CHECK-LABEL: sub_xv4i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.s
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = sub <vscale x 4 x i1> %a, %b
+  ret <vscale x 4 x i1> %res;
+}
+
+define <vscale x 2 x i1> @sub_xv2i1(<vscale x 2 x i1> %a, <vscale x 2 x i1> %b) {
+; CHECK-LABEL: sub_xv2i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p2.d
+; CHECK-NEXT:    eor p0.b, p2/z, p0.b, p1.b
+; CHECK-NEXT:    ret
+  %res = sub <vscale x 2 x i1> %a, %b
+  ret <vscale x 2 x i1> %res;
+}
+
+
+; ILLEGAL SUBGS
+
+
+define aarch64_sve_vector_pcs <vscale x 64 x i1> @sub_nxv64i1(<vscale x 64 x i1> %a, <vscale x 64 x i1> %b) {
+; CHECK-LABEL: sub_nxv64i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #3, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    ldr p4, [x3]
+; CHECK-NEXT:    ldr p5, [x0]
+; CHECK-NEXT:    ldr p6, [x1]
+; CHECK-NEXT:    ldr p7, [x2]
+; CHECK-NEXT:    ptrue p8.b
+; CHECK-NEXT:    eor p0.b, p8/z, p0.b, p5.b
+; CHECK-NEXT:    eor p1.b, p8/z, p1.b, p6.b
+; CHECK-NEXT:    eor p2.b, p8/z, p2.b, p7.b
+; CHECK-NEXT:    eor p3.b, p8/z, p3.b, p4.b
+; CHECK-NEXT:    ldr p8, [sp, #3, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %res = sub <vscale x 64 x i1> %a, %b
+  ret <vscale x 64 x i1> %res;
+}

llvm/test/CodeGen/X86/avx512-mask-op.ll

; X86-NEXT:    retl
    %v1 = icmp eq <16 x i32> %a, zeroinitializer
    %mask1 = bitcast <16 x i1> %v1 to i16
    %val = zext i16 %mask1 to i32
    %val1 = add i32 %val, %val
    ret i32 %val1
 }
 
 define i16 @test_v16i1_add(i16 %x, i16 %y) {
-; KNL-LABEL: test_v16i1_add:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    ## kill: def $ax killed $ax killed $eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v16i1_add:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kmovd %edi, %k0
-; SKX-NEXT:    kmovd %esi, %k1
-; SKX-NEXT:    kxorw %k1, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    ## kill: def $ax killed $ax killed $eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v16i1_add:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    ## kill: def $ax killed $ax killed $eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v16i1_add:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kmovw %edi, %k0
-; AVX512DQ-NEXT:    kmovw %esi, %k1
-; AVX512DQ-NEXT:    kxorw %k1, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    ## kill: def $ax killed $ax killed $eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v16i1_add:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $ax killed $ax killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v16i1_add:
 ; X86:       ## %bb.0:
-; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorw %k1, %k0, %k0
-; X86-NEXT:    kmovd %k0, %eax
-; X86-NEXT:    ## kill: def $ax killed $ax killed $eax
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorw {{[0-9]+}}(%esp), %ax
 ; X86-NEXT:    retl
   %m0 = bitcast i16 %x to <16 x i1>
   %m1 = bitcast i16 %y to <16 x i1>
   %m2 = add <16 x i1> %m0,  %m1
   %ret = bitcast <16 x i1> %m2 to i16
   ret i16 %ret
 }
 
 define i16 @test_v16i1_sub(i16 %x, i16 %y) {
-; KNL-LABEL: test_v16i1_sub:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    ## kill: def $ax killed $ax killed $eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v16i1_sub:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kmovd %edi, %k0
-; SKX-NEXT:    kmovd %esi, %k1
-; SKX-NEXT:    kxorw %k1, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    ## kill: def $ax killed $ax killed $eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v16i1_sub:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    ## kill: def $ax killed $ax killed $eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v16i1_sub:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kmovw %edi, %k0
-; AVX512DQ-NEXT:    kmovw %esi, %k1
-; AVX512DQ-NEXT:    kxorw %k1, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    ## kill: def $ax killed $ax killed $eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v16i1_sub:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $ax killed $ax killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v16i1_sub:
 ; X86:       ## %bb.0:
-; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorw %k1, %k0, %k0
-; X86-NEXT:    kmovd %k0, %eax
-; X86-NEXT:    ## kill: def $ax killed $ax killed $eax
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorw {{[0-9]+}}(%esp), %ax
 ; X86-NEXT:    retl
   %m0 = bitcast i16 %x to <16 x i1>
   %m1 = bitcast i16 %y to <16 x i1>
   %m2 = sub <16 x i1> %m0,  %m1
   %ret = bitcast <16 x i1> %m2 to i16
   ret i16 %ret
 }
 
; X86-NEXT:    retl
   %m0 = bitcast i16 %x to <16 x i1>
   %m1 = bitcast i16 %y to <16 x i1>
   %m2 = mul <16 x i1> %m0,  %m1
   %ret = bitcast <16 x i1> %m2 to i16
   ret i16 %ret
 }
 
 define i8 @test_v8i1_add(i8 %x, i8 %y) {
-; KNL-LABEL: test_v8i1_add:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    ## kill: def $al killed $al killed $eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v8i1_add:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kmovd %edi, %k0
-; SKX-NEXT:    kmovd %esi, %k1
-; SKX-NEXT:    kxorb %k1, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    ## kill: def $al killed $al killed $eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v8i1_add:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v8i1_add:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kmovw %edi, %k0
-; AVX512DQ-NEXT:    kmovw %esi, %k1
-; AVX512DQ-NEXT:    kxorb %k1, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v8i1_add:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v8i1_add:
 ; X86:       ## %bb.0:
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorb %k1, %k0, %k0
-; X86-NEXT:    kmovd %k0, %eax
-; X86-NEXT:    ## kill: def $al killed $al killed $eax
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    xorb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    retl
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = bitcast i8 %y to <8 x i1>
   %m2 = add <8 x i1> %m0,  %m1
   %ret = bitcast <8 x i1> %m2 to i8
   ret i8 %ret
 }
 
 define i8 @test_v8i1_sub(i8 %x, i8 %y) {
-; KNL-LABEL: test_v8i1_sub:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    ## kill: def $al killed $al killed $eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v8i1_sub:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kmovd %edi, %k0
-; SKX-NEXT:    kmovd %esi, %k1
-; SKX-NEXT:    kxorb %k1, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    ## kill: def $al killed $al killed $eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v8i1_sub:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v8i1_sub:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kmovw %edi, %k0
-; AVX512DQ-NEXT:    kmovw %esi, %k1
-; AVX512DQ-NEXT:    kxorb %k1, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v8i1_sub:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v8i1_sub:
 ; X86:       ## %bb.0:
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorb %k1, %k0, %k0
-; X86-NEXT:    kmovd %k0, %eax
-; X86-NEXT:    ## kill: def $al killed $al killed $eax
+; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-NEXT:    xorb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    retl
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = bitcast i8 %y to <8 x i1>
   %m2 = sub <8 x i1> %m0,  %m1
   %ret = bitcast <8 x i1> %m2 to i8
   ret i8 %ret
 }
 

 ; X86-NEXT:    vpmovm2b %k0, %zmm0
 ; X86-NEXT:    retl
   %a_i = trunc i32 %a to i1
   %maskv = insertelement <64 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, i1 %a_i, i32 0
   ret <64 x i1> %maskv
 }
 
 define i1 @test_v1i1_add(i1 %x, i1 %y) {
-; KNL-LABEL: test_v1i1_add:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kshiftlw $15, %k0, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; KNL-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v1i1_add:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    andl $1, %edi
-; SKX-NEXT:    movb %dil, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    andl $1, %esi
-; SKX-NEXT:    movb %sil, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k0
-; SKX-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k1
-; SKX-NEXT:    kxorw %k1, %k0, %k0
-; SKX-NEXT:    kshiftlb $7, %k0, %k0
-; SKX-NEXT:    kshiftrb $7, %k0, %k0
-; SKX-NEXT:    kmovb %k0, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v1i1_add:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kshiftlw $15, %k0, %k0
-; AVX512BW-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX512BW-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v1i1_add:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    andl $1, %edi
-; AVX512DQ-NEXT:    movb %dil, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    andl $1, %esi
-; AVX512DQ-NEXT:    movb %sil, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k0
-; AVX512DQ-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k1
-; AVX512DQ-NEXT:    kxorw %k1, %k0, %k0
-; AVX512DQ-NEXT:    kshiftlb $7, %k0, %k0
-; AVX512DQ-NEXT:    kshiftrb $7, %k0, %k0
-; AVX512DQ-NEXT:    kmovb %k0, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v1i1_add:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v1i1_add:
 ; X86:       ## %bb.0:
-; X86-NEXT:    pushl %eax
-; X86-NEXT:    .cfi_def_cfa_offset 8
-; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    andb $1, %al
-; X86-NEXT:    movb %al, {{[0-9]+}}(%esp)
-; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    andb $1, %al
-; X86-NEXT:    movb %al, {{[0-9]+}}(%esp)
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorw %k1, %k0, %k0
-; X86-NEXT:    kshiftlb $7, %k0, %k0
-; X86-NEXT:    kshiftrb $7, %k0, %k0
-; X86-NEXT:    kmovb %k0, {{[0-9]+}}(%esp)
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    popl %ecx
+; X86-NEXT:    xorb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    retl
   %m0 = bitcast i1 %x to <1 x i1>
   %m1 = bitcast i1 %y to <1 x i1>
   %m2 = add <1 x i1> %m0,  %m1
   %ret = bitcast <1 x i1> %m2 to i1
   ret i1 %ret
 }
 
 define i1 @test_v1i1_sub(i1 %x, i1 %y) {
-; KNL-LABEL: test_v1i1_sub:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kmovw %edi, %k0
-; KNL-NEXT:    kmovw %esi, %k1
-; KNL-NEXT:    kxorw %k1, %k0, %k0
-; KNL-NEXT:    kshiftlw $15, %k0, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; KNL-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test_v1i1_sub:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    andl $1, %edi
-; SKX-NEXT:    movb %dil, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    andl $1, %esi
-; SKX-NEXT:    movb %sil, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k0
-; SKX-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k1
-; SKX-NEXT:    kxorw %k1, %k0, %k0
-; SKX-NEXT:    kshiftlb $7, %k0, %k0
-; SKX-NEXT:    kshiftrb $7, %k0, %k0
-; SKX-NEXT:    kmovb %k0, -{{[0-9]+}}(%rsp)
-; SKX-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test_v1i1_sub:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kmovd %edi, %k0
-; AVX512BW-NEXT:    kmovd %esi, %k1
-; AVX512BW-NEXT:    kxorw %k1, %k0, %k0
-; AVX512BW-NEXT:    kshiftlw $15, %k0, %k0
-; AVX512BW-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX512BW-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test_v1i1_sub:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    andl $1, %edi
-; AVX512DQ-NEXT:    movb %dil, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    andl $1, %esi
-; AVX512DQ-NEXT:    movb %sil, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k0
-; AVX512DQ-NEXT:    kmovb -{{[0-9]+}}(%rsp), %k1
-; AVX512DQ-NEXT:    kxorw %k1, %k0, %k0
-; AVX512DQ-NEXT:    kshiftlb $7, %k0, %k0
-; AVX512DQ-NEXT:    kshiftrb $7, %k0, %k0
-; AVX512DQ-NEXT:    kmovb %k0, -{{[0-9]+}}(%rsp)
-; AVX512DQ-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test_v1i1_sub:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
+; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test_v1i1_sub:
 ; X86:       ## %bb.0:
-; X86-NEXT:    pushl %eax
-; X86-NEXT:    .cfi_def_cfa_offset 8
-; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    andb $1, %al
-; X86-NEXT:    movb %al, {{[0-9]+}}(%esp)
-; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    andb $1, %al
-; X86-NEXT:    movb %al, {{[0-9]+}}(%esp)
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k0
-; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1
-; X86-NEXT:    kxorw %k1, %k0, %k0
-; X86-NEXT:    kshiftlb $7, %k0, %k0
-; X86-NEXT:    kshiftrb $7, %k0, %k0
-; X86-NEXT:    kmovb %k0, {{[0-9]+}}(%esp)
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    popl %ecx
+; X86-NEXT:    xorb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    retl
   %m0 = bitcast i1 %x to <1 x i1>
   %m1 = bitcast i1 %y to <1 x i1>
   %m2 = sub <1 x i1> %m0,  %m1
   %ret = bitcast <1 x i1> %m2 to i1
   ret i1 %ret
 }
 

llvm/test/CodeGen/X86/avx512bw-mask-op.ll

; CHECK-NEXT:    retq
   %me = or <64 x i1> %mc, %md
   %ret = bitcast <64 x i1> %me to i64
   ret i64 %ret
 }
 
 define i32 @test_v32i1_add(i32 %x, i32 %y) {
 ; CHECK-LABEL: test_v32i1_add:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovd %edi, %k0
-; CHECK-NEXT:    kmovd %esi, %k1
-; CHECK-NEXT:    kxord %k1, %k0, %k0
-; CHECK-NEXT:    kmovd %k0, %eax
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i32 %x to <32 x i1>
   %m1 = bitcast i32 %y to <32 x i1>
   %m2 = add <32 x i1> %m0,  %m1
   %ret = bitcast <32 x i1> %m2 to i32
   ret i32 %ret
 }
 
 define i32 @test_v32i1_sub(i32 %x, i32 %y) {
 ; CHECK-LABEL: test_v32i1_sub:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovd %edi, %k0
-; CHECK-NEXT:    kmovd %esi, %k1
-; CHECK-NEXT:    kxord %k1, %k0, %k0
-; CHECK-NEXT:    kmovd %k0, %eax
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    xorl %esi, %eax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i32 %x to <32 x i1>
   %m1 = bitcast i32 %y to <32 x i1>
   %m2 = sub <32 x i1> %m0,  %m1
   %ret = bitcast <32 x i1> %m2 to i32
   ret i32 %ret
 }
 
; CHECK-NEXT:    retq
   %m2 = mul <32 x i1> %m0,  %m1
   %ret = bitcast <32 x i1> %m2 to i32
   ret i32 %ret
 }
 
 define i64 @test_v64i1_add(i64 %x, i64 %y) {
 ; CHECK-LABEL: test_v64i1_add:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovq %rdi, %k0
-; CHECK-NEXT:    kmovq %rsi, %k1
-; CHECK-NEXT:    kxorq %k1, %k0, %k0
-; CHECK-NEXT:    kmovq %k0, %rax
+; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    xorq %rsi, %rax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i64 %x to <64 x i1>
   %m1 = bitcast i64 %y to <64 x i1>
   %m2 = add <64 x i1> %m0,  %m1
   %ret = bitcast <64 x i1> %m2 to i64
   ret i64 %ret
 }
 
 define i64 @test_v64i1_sub(i64 %x, i64 %y) {
 ; CHECK-LABEL: test_v64i1_sub:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovq %rdi, %k0
-; CHECK-NEXT:    kmovq %rsi, %k1
-; CHECK-NEXT:    kxorq %k1, %k0, %k0
-; CHECK-NEXT:    kmovq %k0, %rax
+; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    xorq %rsi, %rax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i64 %x to <64 x i1>
   %m1 = bitcast i64 %y to <64 x i1>
   %m2 = sub <64 x i1> %m0,  %m1
   %ret = bitcast <64 x i1> %m2 to i64
   ret i64 %ret
 }