[X86] Add XOR(X, MIN_SIGNED_VALUE) -> ADD(X, MIN_SIGNED_VALUE) isel patterns (PR52267)

[RKSimon]

Summary

Improve chances of folding to LEA patterns

[craig.topper]

llvm/lib/Target/X86/X86InstrCompiler.td

1521

I'm not sure the "If safe" is needed here. I think this was copied from or_is_add where "safe" meant checking known bits. Here you explicitly wrote min_signed_value so there's nothing additional to check.

1525

Drop curly braces?

You could even write

ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N->getOperand(1));
return CN && CN->isMinSignedValue()

[RKSimon]

address feedback

[craig.topper]

llvm/lib/Target/X86/X86InstrCompiler.td

1529

It occurs to me you could avoid the AddedComplexity by writing

def : Pat<(xor GR8:$src1, 128),
          (ADD8ri GR8:$src1, 128)>;
def : Pat<(xor GR16:$src1, 32768),
          (ADD16ri GR16:$src1, 32768)>;
def : Pat<(xor GR32:$src1, 2147483648),
          (ADD32ri GR32:$src1, 2147483648)>;

[craig.topper]

llvm/lib/Target/X86/X86InstrCompiler.td

1529

Those might need to be negative constants.

[RKSimon]

remove PatFrags

[craig.topper]

LGTM

llvm/lib/Target/X86/X86InstrCompiler.td

def ADD64ri32_DB : I<0, Pseudo,
                      (outs GR64:$dst), (ins GR64:$src1, i64i32imm:$src2),
                      "", // orq/addq REG, imm
                      [(set GR64:$dst, (or_is_add GR64:$src1,
                                                  i64immSExt32:$src2))]>;
 }
 } // AddedComplexity, SchedRW
 
 //===----------------------------------------------------------------------===//
+// Pattern match XOR as ADD
+//===----------------------------------------------------------------------===//
+
+// Prefer to pattern match XOR with min_signed_value as ADD at isel time.
+// ADD can be 3-addressified into an LEA instruction to avoid copies.
+let AddedComplexity = 5 in {
+def : Pat<(xor GR8:$src1, -128),
+          (ADD8ri GR8:$src1, -128)>;
+def : Pat<(xor GR16:$src1, -32768),
+          (ADD16ri GR16:$src1, -32768)>;
+def : Pat<(xor GR32:$src1, -2147483648),
+          (ADD32ri GR32:$src1, -2147483648)>;
+}
+
+//===----------------------------------------------------------------------===//
 // Pattern match SUB as XOR
 //===----------------------------------------------------------------------===//
 
 // An immediate in the LHS of a subtract can't be encoded in the instruction.
 // If there is no possibility of a borrow we can use an XOR instead of a SUB
 // to enable the immediate to be folded.
 // TODO: Move this to a DAG combine?
 

llvm/test/CodeGen/X86/avx512-cmp.ll

entry:
   %or = zext i1 %0 to i32
   ret i32 %or
 }
 
 define i32 @test8(i32 %a1, i32 %a2, i32 %a3) {
 ; ALL-LABEL: test8:
 ; ALL:       ## %bb.0:
 ; ALL-NEXT:    notl %edi
-; ALL-NEXT:    xorl $-2147483648, %esi ## imm = 0x80000000
+; ALL-NEXT:    addl $-2147483648, %esi ## imm = 0x80000000
 ; ALL-NEXT:    testl %edx, %edx
 ; ALL-NEXT:    movl $1, %eax
 ; ALL-NEXT:    cmovel %eax, %edx
 ; ALL-NEXT:    orl %edi, %esi
 ; ALL-NEXT:    cmovnel %edx, %eax
 ; ALL-NEXT:    retq
   %tmp1 = icmp eq i32 %a1, -1
   %tmp2 = icmp eq i32 %a2, -2147483648

llvm/test/CodeGen/X86/combine-add-ssat.ll

 
 ; fold (sadd_sat c, x) -> (sadd_sat x, c)
 define i32 @combine_constant_i32(i32 %a0) {
 ; CHECK-LABEL: combine_constant_i32:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    # kill: def $edi killed $edi def $rdi
 ; CHECK-NEXT:    leal 1(%rdi), %eax
 ; CHECK-NEXT:    sarl $31, %eax
-; CHECK-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; CHECK-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; CHECK-NEXT:    incl %edi
 ; CHECK-NEXT:    cmovnol %edi, %eax
 ; CHECK-NEXT:    retq
   %res = call i32 @llvm.sadd.sat.i32(i32 1, i32 %a0)
   ret i32 %res
 }
 
 define <8 x i16> @combine_constant_v8i16(<8 x i16> %a0) {

 ; CHECK-LABEL: combine_no_overflow_i32:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    # kill: def $esi killed $esi def $rsi
 ; CHECK-NEXT:    # kill: def $edi killed $edi def $rdi
 ; CHECK-NEXT:    sarl $16, %edi
 ; CHECK-NEXT:    shrl $16, %esi
 ; CHECK-NEXT:    leal (%rdi,%rsi), %eax
 ; CHECK-NEXT:    sarl $31, %eax
-; CHECK-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; CHECK-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; CHECK-NEXT:    addl %edi, %esi
 ; CHECK-NEXT:    cmovnol %esi, %eax
 ; CHECK-NEXT:    retq
   %1 = ashr i32 %a0, 16
   %2 = lshr i32 %a1, 16
   %3 = call i32 @llvm.sadd.sat.i32(i32 %1, i32 %2)
   ret i32 %3
 }

llvm/test/CodeGen/X86/horizontal-reduce-smin.ll

 ; X86-SSE42-LABEL: test_reduce_v16i8:
 ; X86-SSE42:       ## %bb.0:
 ; X86-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE42-NEXT:    psrlw $8, %xmm1
 ; X86-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X86-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X86-SSE42-NEXT:    movd %xmm0, %eax
-; X86-SSE42-NEXT:    xorb $-128, %al
+; X86-SSE42-NEXT:    addb $-128, %al
 ; X86-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-SSE42-NEXT:    retl
 ;
 ; X86-AVX-LABEL: test_reduce_v16i8:
 ; X86-AVX:       ## %bb.0:
 ; X86-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX-NEXT:    vmovd %xmm0, %eax
-; X86-AVX-NEXT:    xorb $-128, %al
+; X86-AVX-NEXT:    addb $-128, %al
 ; X86-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX-NEXT:    retl
 ;
 ; X64-SSE2-LABEL: test_reduce_v16i8:
 ; X64-SSE2:       ## %bb.0:
 ; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; X64-SSE2-NEXT:    pcmpgtb %xmm0, %xmm2

 ; X64-SSE42-LABEL: test_reduce_v16i8:
 ; X64-SSE42:       ## %bb.0:
 ; X64-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X64-SSE42-NEXT:    psrlw $8, %xmm1
 ; X64-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X64-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X64-SSE42-NEXT:    movd %xmm0, %eax
-; X64-SSE42-NEXT:    xorb $-128, %al
+; X64-SSE42-NEXT:    addb $-128, %al
 ; X64-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-SSE42-NEXT:    retq
 ;
 ; X64-AVX-LABEL: test_reduce_v16i8:
 ; X64-AVX:       ## %bb.0:
 ; X64-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX-NEXT:    vmovd %xmm0, %eax
-; X64-AVX-NEXT:    xorb $-128, %al
+; X64-AVX-NEXT:    addb $-128, %al
 ; X64-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX-NEXT:    retq
   %1  = shufflevector <16 x i8> %a0, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %2  = icmp slt <16 x i8> %a0, %1
   %3  = select <16 x i1> %2, <16 x i8> %a0, <16 x i8> %1
   %4  = shufflevector <16 x i8> %3, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %5  = icmp slt <16 x i8> %3, %4
   %6  = select <16 x i1> %5, <16 x i8> %3, <16 x i8> %4

 ; X86-SSE42:       ## %bb.0:
 ; X86-SSE42-NEXT:    pminsb %xmm1, %xmm0
 ; X86-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE42-NEXT:    psrlw $8, %xmm1
 ; X86-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X86-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X86-SSE42-NEXT:    movd %xmm0, %eax
-; X86-SSE42-NEXT:    xorb $-128, %al
+; X86-SSE42-NEXT:    addb $-128, %al
 ; X86-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-SSE42-NEXT:    retl
 ;
 ; X86-AVX1-LABEL: test_reduce_v32i8:
 ; X86-AVX1:       ## %bb.0:
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
-; X86-AVX1-NEXT:    xorb $-128, %al
+; X86-AVX1-NEXT:    addb $-128, %al
 ; X86-AVX1-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX1-NEXT:    vzeroupper
 ; X86-AVX1-NEXT:    retl
 ;
 ; X86-AVX2-LABEL: test_reduce_v32i8:
 ; X86-AVX2:       ## %bb.0:
 ; X86-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
-; X86-AVX2-NEXT:    xorb $-128, %al
+; X86-AVX2-NEXT:    addb $-128, %al
 ; X86-AVX2-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX2-NEXT:    vzeroupper
 ; X86-AVX2-NEXT:    retl
 ;
 ; X64-SSE2-LABEL: test_reduce_v32i8:
 ; X64-SSE2:       ## %bb.0:
 ; X64-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; X64-SSE2-NEXT:    pcmpgtb %xmm0, %xmm2

 ; X64-SSE42:       ## %bb.0:
 ; X64-SSE42-NEXT:    pminsb %xmm1, %xmm0
 ; X64-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X64-SSE42-NEXT:    psrlw $8, %xmm1
 ; X64-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X64-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X64-SSE42-NEXT:    movd %xmm0, %eax
-; X64-SSE42-NEXT:    xorb $-128, %al
+; X64-SSE42-NEXT:    addb $-128, %al
 ; X64-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-SSE42-NEXT:    retq
 ;
 ; X64-AVX1-LABEL: test_reduce_v32i8:
 ; X64-AVX1:       ## %bb.0:
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vmovd %xmm0, %eax
-; X64-AVX1-NEXT:    xorb $-128, %al
+; X64-AVX1-NEXT:    addb $-128, %al
 ; X64-AVX1-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX1-NEXT:    vzeroupper
 ; X64-AVX1-NEXT:    retq
 ;
 ; X64-AVX2-LABEL: test_reduce_v32i8:
 ; X64-AVX2:       ## %bb.0:
 ; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vmovd %xmm0, %eax
-; X64-AVX2-NEXT:    xorb $-128, %al
+; X64-AVX2-NEXT:    addb $-128, %al
 ; X64-AVX2-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
 ;
 ; X64-AVX512-LABEL: test_reduce_v32i8:
 ; X64-AVX512:       ## %bb.0:
 ; X64-AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX512-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vmovd %xmm0, %eax
-; X64-AVX512-NEXT:    xorb $-128, %al
+; X64-AVX512-NEXT:    addb $-128, %al
 ; X64-AVX512-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX512-NEXT:    vzeroupper
 ; X64-AVX512-NEXT:    retq
   %1  = shufflevector <32 x i8> %a0, <32 x i8> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %2  = icmp slt <32 x i8> %a0, %1
   %3  = select <32 x i1> %2, <32 x i8> %a0, <32 x i8> %1
   %4  = shufflevector <32 x i8> %3, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %5  = icmp slt <32 x i8> %3, %4

 ; X86-SSE42-NEXT:    pminsb %xmm2, %xmm1
 ; X86-SSE42-NEXT:    pminsb %xmm0, %xmm1
 ; X86-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
 ; X86-SSE42-NEXT:    movdqa %xmm1, %xmm0
 ; X86-SSE42-NEXT:    psrlw $8, %xmm0
 ; X86-SSE42-NEXT:    pminub %xmm1, %xmm0
 ; X86-SSE42-NEXT:    phminposuw %xmm0, %xmm0
 ; X86-SSE42-NEXT:    movd %xmm0, %eax
-; X86-SSE42-NEXT:    xorb $-128, %al
+; X86-SSE42-NEXT:    addb $-128, %al
 ; X86-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-SSE42-NEXT:    retl
 ;
 ; X86-AVX1-LABEL: test_reduce_v64i8:
 ; X86-AVX1:       ## %bb.0:
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; X86-AVX1-NEXT:    vpminsb %xmm2, %xmm3, %xmm2
 ; X86-AVX1-NEXT:    vpminsb %xmm2, %xmm1, %xmm1
 ; X86-AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
-; X86-AVX1-NEXT:    xorb $-128, %al
+; X86-AVX1-NEXT:    addb $-128, %al
 ; X86-AVX1-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX1-NEXT:    vzeroupper
 ; X86-AVX1-NEXT:    retl
 ;
 ; X86-AVX2-LABEL: test_reduce_v64i8:
 ; X86-AVX2:       ## %bb.0:
 ; X86-AVX2-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; X86-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
-; X86-AVX2-NEXT:    xorb $-128, %al
+; X86-AVX2-NEXT:    addb $-128, %al
 ; X86-AVX2-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX2-NEXT:    vzeroupper
 ; X86-AVX2-NEXT:    retl
 ;
 ; X64-SSE2-LABEL: test_reduce_v64i8:
 ; X64-SSE2:       ## %bb.0:
 ; X64-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; X64-SSE2-NEXT:    pcmpgtb %xmm0, %xmm4

 ; X64-SSE42-NEXT:    pminsb %xmm2, %xmm1
 ; X64-SSE42-NEXT:    pminsb %xmm0, %xmm1
 ; X64-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
 ; X64-SSE42-NEXT:    movdqa %xmm1, %xmm0
 ; X64-SSE42-NEXT:    psrlw $8, %xmm0
 ; X64-SSE42-NEXT:    pminub %xmm1, %xmm0
 ; X64-SSE42-NEXT:    phminposuw %xmm0, %xmm0
 ; X64-SSE42-NEXT:    movd %xmm0, %eax
-; X64-SSE42-NEXT:    xorb $-128, %al
+; X64-SSE42-NEXT:    addb $-128, %al
 ; X64-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-SSE42-NEXT:    retq
 ;
 ; X64-AVX1-LABEL: test_reduce_v64i8:
 ; X64-AVX1:       ## %bb.0:
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; X64-AVX1-NEXT:    vpminsb %xmm2, %xmm3, %xmm2
 ; X64-AVX1-NEXT:    vpminsb %xmm2, %xmm1, %xmm1
 ; X64-AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vmovd %xmm0, %eax
-; X64-AVX1-NEXT:    xorb $-128, %al
+; X64-AVX1-NEXT:    addb $-128, %al
 ; X64-AVX1-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX1-NEXT:    vzeroupper
 ; X64-AVX1-NEXT:    retq
 ;
 ; X64-AVX2-LABEL: test_reduce_v64i8:
 ; X64-AVX2:       ## %bb.0:
 ; X64-AVX2-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX2-NEXT:    vmovd %xmm0, %eax
-; X64-AVX2-NEXT:    xorb $-128, %al
+; X64-AVX2-NEXT:    addb $-128, %al
 ; X64-AVX2-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
 ;
 ; X64-AVX512-LABEL: test_reduce_v64i8:
 ; X64-AVX512:       ## %bb.0:
 ; X64-AVX512-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; X64-AVX512-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; X64-AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX512-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX512-NEXT:    vmovd %xmm0, %eax
-; X64-AVX512-NEXT:    xorb $-128, %al
+; X64-AVX512-NEXT:    addb $-128, %al
 ; X64-AVX512-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX512-NEXT:    vzeroupper
 ; X64-AVX512-NEXT:    retq
   %1  = shufflevector <64 x i8> %a0, <64 x i8> undef, <64 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %2  = icmp slt <64 x i8> %a0, %1
   %3  = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %1
   %4  = shufflevector <64 x i8> %3, <64 x i8> undef, <64 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %5  = icmp slt <64 x i8> %3, %4

 ; X86-SSE42-LABEL: test_reduce_v32i8_v16i8:
 ; X86-SSE42:       ## %bb.0:
 ; X86-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE42-NEXT:    psrlw $8, %xmm1
 ; X86-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X86-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X86-SSE42-NEXT:    movd %xmm0, %eax
-; X86-SSE42-NEXT:    xorb $-128, %al
+; X86-SSE42-NEXT:    addb $-128, %al
 ; X86-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-SSE42-NEXT:    retl
 ;
 ; X86-AVX-LABEL: test_reduce_v32i8_v16i8:
 ; X86-AVX:       ## %bb.0:
 ; X86-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX-NEXT:    vmovd %xmm0, %eax
-; X86-AVX-NEXT:    xorb $-128, %al
+; X86-AVX-NEXT:    addb $-128, %al
 ; X86-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX-NEXT:    vzeroupper
 ; X86-AVX-NEXT:    retl
 ;
 ; X64-SSE2-LABEL: test_reduce_v32i8_v16i8:
 ; X64-SSE2:       ## %bb.0:
 ; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-SSE2-NEXT:    movdqa %xmm1, %xmm2

 ; X64-SSE42-LABEL: test_reduce_v32i8_v16i8:
 ; X64-SSE42:       ## %bb.0:
 ; X64-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X64-SSE42-NEXT:    psrlw $8, %xmm1
 ; X64-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X64-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X64-SSE42-NEXT:    movd %xmm0, %eax
-; X64-SSE42-NEXT:    xorb $-128, %al
+; X64-SSE42-NEXT:    addb $-128, %al
 ; X64-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-SSE42-NEXT:    retq
 ;
 ; X64-AVX-LABEL: test_reduce_v32i8_v16i8:
 ; X64-AVX:       ## %bb.0:
 ; X64-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX-NEXT:    vmovd %xmm0, %eax
-; X64-AVX-NEXT:    xorb $-128, %al
+; X64-AVX-NEXT:    addb $-128, %al
 ; X64-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX-NEXT:    vzeroupper
 ; X64-AVX-NEXT:    retq
   %1  = shufflevector <32 x i8> %a0, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %2  = icmp slt <32 x i8> %a0, %1
   %3  = select <32 x i1> %2, <32 x i8> %a0, <32 x i8> %1
   %4  = shufflevector <32 x i8> %3, <32 x i8> undef, <32 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %5  = icmp slt <32 x i8> %3, %4

 ; X86-SSE42-LABEL: test_reduce_v64i8_v16i8:
 ; X86-SSE42:       ## %bb.0:
 ; X86-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE42-NEXT:    psrlw $8, %xmm1
 ; X86-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X86-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X86-SSE42-NEXT:    movd %xmm0, %eax
-; X86-SSE42-NEXT:    xorb $-128, %al
+; X86-SSE42-NEXT:    addb $-128, %al
 ; X86-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-SSE42-NEXT:    retl
 ;
 ; X86-AVX-LABEL: test_reduce_v64i8_v16i8:
 ; X86-AVX:       ## %bb.0:
 ; X86-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X86-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X86-AVX-NEXT:    vmovd %xmm0, %eax
-; X86-AVX-NEXT:    xorb $-128, %al
+; X86-AVX-NEXT:    addb $-128, %al
 ; X86-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X86-AVX-NEXT:    vzeroupper
 ; X86-AVX-NEXT:    retl
 ;
 ; X64-SSE2-LABEL: test_reduce_v64i8_v16i8:
 ; X64-SSE2:       ## %bb.0:
 ; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-SSE2-NEXT:    movdqa %xmm1, %xmm2

 ; X64-SSE42-LABEL: test_reduce_v64i8_v16i8:
 ; X64-SSE42:       ## %bb.0:
 ; X64-SSE42-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-SSE42-NEXT:    movdqa %xmm0, %xmm1
 ; X64-SSE42-NEXT:    psrlw $8, %xmm1
 ; X64-SSE42-NEXT:    pminub %xmm0, %xmm1
 ; X64-SSE42-NEXT:    phminposuw %xmm1, %xmm0
 ; X64-SSE42-NEXT:    movd %xmm0, %eax
-; X64-SSE42-NEXT:    xorb $-128, %al
+; X64-SSE42-NEXT:    addb $-128, %al
 ; X64-SSE42-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-SSE42-NEXT:    retq
 ;
 ; X64-AVX-LABEL: test_reduce_v64i8_v16i8:
 ; X64-AVX:       ## %bb.0:
 ; X64-AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; X64-AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; X64-AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; X64-AVX-NEXT:    vmovd %xmm0, %eax
-; X64-AVX-NEXT:    xorb $-128, %al
+; X64-AVX-NEXT:    addb $-128, %al
 ; X64-AVX-NEXT:    ## kill: def $al killed $al killed $eax
 ; X64-AVX-NEXT:    vzeroupper
 ; X64-AVX-NEXT:    retq
   %1  = shufflevector <64 x i8> %a0, <64 x i8> undef, <64 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %2  = icmp slt <64 x i8> %a0, %1
   %3  = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %1
   %4  = shufflevector <64 x i8> %3, <64 x i8> undef, <64 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %5  = icmp slt <64 x i8> %3, %4

llvm/test/CodeGen/X86/sadd_sat.ll

 
 define i32 @func(i32 %x, i32 %y) nounwind {
 ; X86-LABEL: func:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    leal (%eax,%ecx), %edx
 ; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    addl %ecx, %eax
 ; X86-NEXT:    cmovol %edx, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $esi killed $esi def $rsi
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    leal (%rdi,%rsi), %eax
 ; X64-NEXT:    sarl $31, %eax
-; X64-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X64-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; X64-NEXT:    addl %esi, %edi
 ; X64-NEXT:    cmovnol %edi, %eax
 ; X64-NEXT:    retq
   %tmp = call i32 @llvm.sadd.sat.i32(i32 %x, i32 %y);
   ret i32 %tmp;
 }
 
 define i64 @func2(i64 %x, i64 %y) nounwind {
 ; X86-LABEL: func2:
 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    seto %bl
 ; X86-NEXT:    movl %ecx, %edx
 ; X86-NEXT:    sarl $31, %edx
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovnel %edx, %eax
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovel %ecx, %edx
 ; X86-NEXT:    popl %ebx
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    leaq (%rdi,%rsi), %rcx

 define signext i8 @func8(i8 signext %x, i8 signext %y) nounwind {
 ; X86-LABEL: func8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
 ; X86-NEXT:    movl %eax, %edx
 ; X86-NEXT:    addb %cl, %dl
 ; X86-NEXT:    sarb $7, %dl
-; X86-NEXT:    xorb $-128, %dl
+; X86-NEXT:    addb $-128, %dl
 ; X86-NEXT:    addb %cl, %al
 ; X86-NEXT:    movzbl %al, %ecx
 ; X86-NEXT:    movzbl %dl, %eax
 ; X86-NEXT:    cmovnol %ecx, %eax
 ; X86-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func8:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $esi killed $esi def $rsi
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    leal (%rdi,%rsi), %eax
 ; X64-NEXT:    sarb $7, %al
-; X64-NEXT:    xorb $-128, %al
+; X64-NEXT:    addb $-128, %al
 ; X64-NEXT:    addb %sil, %dil
 ; X64-NEXT:    movzbl %dil, %ecx
 ; X64-NEXT:    movzbl %al, %eax
 ; X64-NEXT:    cmovnol %ecx, %eax
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %tmp = call i8 @llvm.sadd.sat.i8(i8 %x, i8 %y)
   ret i8 %tmp

 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    leal (%ecx,%eax), %esi
 ; X86-NEXT:    sarl $31, %esi
-; X86-NEXT:    xorl $-2147483648, %esi # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %esi # imm = 0x80000000
 ; X86-NEXT:    addl %eax, %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    cmovol %esi, %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    leal (%edx,%eax), %edi
 ; X86-NEXT:    sarl $31, %edi
-; X86-NEXT:    xorl $-2147483648, %edi # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edi # imm = 0x80000000
 ; X86-NEXT:    addl %eax, %edx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X86-NEXT:    cmovol %edi, %edx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    leal (%esi,%eax), %ebx
 ; X86-NEXT:    sarl $31, %ebx
-; X86-NEXT:    xorl $-2147483648, %ebx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %ebx # imm = 0x80000000
 ; X86-NEXT:    addl %eax, %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
 ; X86-NEXT:    cmovol %ebx, %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    leal (%edi,%eax), %ebx
 ; X86-NEXT:    sarl $31, %ebx
-; X86-NEXT:    xorl $-2147483648, %ebx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %ebx # imm = 0x80000000
 ; X86-NEXT:    addl %eax, %edi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    cmovol %ebx, %edi
 ; X86-NEXT:    movl %ecx, 12(%eax)
 ; X86-NEXT:    movl %edx, 8(%eax)
 ; X86-NEXT:    movl %esi, 4(%eax)
 ; X86-NEXT:    movl %edi, (%eax)
 ; X86-NEXT:    popl %esi

llvm/test/CodeGen/X86/sadd_sat_plus.ll

 define i32 @func32(i32 %x, i32 %y, i32 %z) nounwind {
 ; X86-LABEL: func32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    imull {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    leal (%eax,%ecx), %edx
 ; X86-NEXT:    sarl $31, %edx
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    addl %ecx, %eax
 ; X86-NEXT:    cmovol %edx, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func32:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $esi killed $esi def $rsi
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    imull %edx, %esi
 ; X64-NEXT:    leal (%rdi,%rsi), %eax
 ; X64-NEXT:    sarl $31, %eax
-; X64-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X64-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; X64-NEXT:    addl %edi, %esi
 ; X64-NEXT:    cmovnol %esi, %eax
 ; X64-NEXT:    retq
   %a = mul i32 %y, %z
   %tmp = call i32 @llvm.sadd.sat.i32(i32 %x, i32 %a)
   ret i32 %tmp
 }
 
 define i64 @func64(i64 %x, i64 %y, i64 %z) nounwind {
 ; X86-LABEL: func64:
 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    seto %bl
 ; X86-NEXT:    movl %ecx, %edx
 ; X86-NEXT:    sarl $31, %edx
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovnel %edx, %eax
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovel %ecx, %edx
 ; X86-NEXT:    popl %ebx
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    leaq (%rdi,%rdx), %rcx

 ; X86-LABEL: func8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    mulb {{[0-9]+}}(%esp)
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
 ; X86-NEXT:    movl %eax, %edx
 ; X86-NEXT:    addb %cl, %dl
 ; X86-NEXT:    sarb $7, %dl
-; X86-NEXT:    xorb $-128, %dl
+; X86-NEXT:    addb $-128, %dl
 ; X86-NEXT:    addb %cl, %al
 ; X86-NEXT:    movzbl %al, %ecx
 ; X86-NEXT:    movzbl %dl, %eax
 ; X86-NEXT:    cmovnol %ecx, %eax
 ; X86-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func8:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movl %esi, %eax
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    mulb %dl
 ; X64-NEXT:    # kill: def $al killed $al def $rax
 ; X64-NEXT:    leal (%rdi,%rax), %ecx
 ; X64-NEXT:    sarb $7, %cl
-; X64-NEXT:    xorb $-128, %cl
+; X64-NEXT:    addb $-128, %cl
 ; X64-NEXT:    addb %al, %dil
 ; X64-NEXT:    movzbl %dil, %edx
 ; X64-NEXT:    movzbl %cl, %eax
 ; X64-NEXT:    cmovnol %edx, %eax
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %a = mul i8 %y, %z
   %tmp = call i8 @llvm.sadd.sat.i8(i8 %x, i8 %a)

llvm/test/CodeGen/X86/sadd_sat_vec.ll

 
 define void @v1i8(<1 x i8>* %px, <1 x i8>* %py, <1 x i8>* %pz) nounwind {
 ; SSE-LABEL: v1i8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movb (%rdi), %al
 ; SSE-NEXT:    movb (%rsi), %cl
 ; SSE-NEXT:    leal (%rax,%rcx), %esi
 ; SSE-NEXT:    sarb $7, %sil
-; SSE-NEXT:    xorb $-128, %sil
+; SSE-NEXT:    addb $-128, %sil
 ; SSE-NEXT:    addb %cl, %al
 ; SSE-NEXT:    movzbl %al, %eax
 ; SSE-NEXT:    movzbl %sil, %ecx
 ; SSE-NEXT:    cmovnol %eax, %ecx
 ; SSE-NEXT:    movb %cl, (%rdx)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: v1i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    movb (%rdi), %al
 ; AVX-NEXT:    movb (%rsi), %cl
 ; AVX-NEXT:    leal (%rax,%rcx), %esi
 ; AVX-NEXT:    sarb $7, %sil
-; AVX-NEXT:    xorb $-128, %sil
+; AVX-NEXT:    addb $-128, %sil
 ; AVX-NEXT:    addb %cl, %al
 ; AVX-NEXT:    movzbl %al, %eax
 ; AVX-NEXT:    movzbl %sil, %ecx
 ; AVX-NEXT:    cmovnol %eax, %ecx
 ; AVX-NEXT:    movb %cl, (%rdx)
 ; AVX-NEXT:    retq
   %x = load <1 x i8>, <1 x i8>* %px
   %y = load <1 x i8>, <1 x i8>* %py

llvm/test/CodeGen/X86/ssub_sat.ll

 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    sbbl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    seto %bl
 ; X86-NEXT:    movl %ecx, %edx
 ; X86-NEXT:    sarl $31, %edx
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovnel %edx, %eax
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovel %ecx, %edx
 ; X86-NEXT:    popl %ebx
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %ecx, %ecx

llvm/test/CodeGen/X86/ssub_sat_plus.ll

 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    sbbl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    seto %bl
 ; X86-NEXT:    movl %ecx, %edx
 ; X86-NEXT:    sarl $31, %edx
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovnel %edx, %eax
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    testb %bl, %bl
 ; X86-NEXT:    cmovel %ecx, %edx
 ; X86-NEXT:    popl %ebx
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %ecx, %ecx

llvm/test/CodeGen/X86/vector-reduce-smin.ll

 ; SSE4-LABEL: test_v16i8:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; SSE4-NEXT:    movdqa %xmm0, %xmm1
 ; SSE4-NEXT:    psrlw $8, %xmm1
 ; SSE4-NEXT:    pminub %xmm0, %xmm1
 ; SSE4-NEXT:    phminposuw %xmm1, %xmm0
 ; SSE4-NEXT:    movd %xmm0, %eax
-; SSE4-NEXT:    xorb $-128, %al
+; SSE4-NEXT:    addb $-128, %al
 ; SSE4-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE4-NEXT:    retq
 ;
 ; AVX-LABEL: test_v16i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX-NEXT:    vmovd %xmm0, %eax
-; AVX-NEXT:    xorb $-128, %al
+; AVX-NEXT:    addb $-128, %al
 ; AVX-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v16i8:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    xorb $-128, %al
+; AVX512-NEXT:    addb $-128, %al
 ; AVX512-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX512-NEXT:    retq
   %1 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> %a0)
   ret i8 %1
 }
 
 define i8 @test_v32i8(<32 x i8> %a0) {
 ; SSE2-LABEL: test_v32i8:

 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    pminsb %xmm1, %xmm0
 ; SSE4-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; SSE4-NEXT:    movdqa %xmm0, %xmm1
 ; SSE4-NEXT:    psrlw $8, %xmm1
 ; SSE4-NEXT:    pminub %xmm0, %xmm1
 ; SSE4-NEXT:    phminposuw %xmm1, %xmm0
 ; SSE4-NEXT:    movd %xmm0, %eax
-; SSE4-NEXT:    xorb $-128, %al
+; SSE4-NEXT:    addb $-128, %al
 ; SSE4-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE4-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v32i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    xorb $-128, %al
+; AVX1-NEXT:    addb $-128, %al
 ; AVX1-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v32i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    xorb $-128, %al
+; AVX2-NEXT:    addb $-128, %al
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v32i8:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    xorb $-128, %al
+; AVX512-NEXT:    addb $-128, %al
 ; AVX512-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %1 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> %a0)
   ret i8 %1
 }
 
 define i8 @test_v64i8(<64 x i8> %a0) {

 ; SSE4-NEXT:    pminsb %xmm2, %xmm1
 ; SSE4-NEXT:    pminsb %xmm0, %xmm1
 ; SSE4-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    psrlw $8, %xmm0
 ; SSE4-NEXT:    pminub %xmm1, %xmm0
 ; SSE4-NEXT:    phminposuw %xmm0, %xmm0
 ; SSE4-NEXT:    movd %xmm0, %eax
-; SSE4-NEXT:    xorb $-128, %al
+; SSE4-NEXT:    addb $-128, %al
 ; SSE4-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE4-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v64i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; AVX1-NEXT:    vpminsb %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpminsb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    xorb $-128, %al
+; AVX1-NEXT:    addb $-128, %al
 ; AVX1-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v64i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    xorb $-128, %al
+; AVX2-NEXT:    addb $-128, %al
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v64i8:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; AVX512-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    xorb $-128, %al
+; AVX512-NEXT:    addb $-128, %al
 ; AVX512-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %1 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> %a0)
   ret i8 %1
 }
 
 define i8 @test_v128i8(<128 x i8> %a0) {

 ; SSE4-NEXT:    pminsb %xmm3, %xmm2
 ; SSE4-NEXT:    pminsb %xmm0, %xmm2
 ; SSE4-NEXT:    pxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
 ; SSE4-NEXT:    movdqa %xmm2, %xmm0
 ; SSE4-NEXT:    psrlw $8, %xmm0
 ; SSE4-NEXT:    pminub %xmm2, %xmm0
 ; SSE4-NEXT:    phminposuw %xmm0, %xmm0
 ; SSE4-NEXT:    movd %xmm0, %eax
-; SSE4-NEXT:    xorb $-128, %al
+; SSE4-NEXT:    addb $-128, %al
 ; SSE4-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE4-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v128i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm5
 ; AVX1-NEXT:    vpminsb %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm5
 ; AVX1-NEXT:    vpminsb %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
 ; AVX1-NEXT:    vpminsb %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vpminsb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpminsb %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpminsb %xmm4, %xmm1, %xmm1
 ; AVX1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    xorb $-128, %al
+; AVX1-NEXT:    addb $-128, %al
 ; AVX1-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v128i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpminsb %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    vpminsb %ymm1, %ymm2, %ymm1
 ; AVX2-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX2-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    xorb $-128, %al
+; AVX2-NEXT:    addb $-128, %al
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v128i8:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vpminsb %zmm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; AVX512-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512-NEXT:    vpminsb %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpxor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm1
 ; AVX512-NEXT:    vpminub %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vphminposuw %xmm0, %xmm0
 ; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    xorb $-128, %al
+; AVX512-NEXT:    addb $-128, %al
 ; AVX512-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %1 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> %a0)
   ret i8 %1
 }
 
 declare i64 @llvm.vector.reduce.smin.v2i64(<2 x i64>)

llvm/test/CodeGen/X86/xor-lea.ll

 ;
 ; XOR(X,MIN_SIGNED_VALUE)
 ;
 
 define i8 @xor_sminval_i8(i8 %x) {
 ; X86-LABEL: xor_sminval_i8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    xorb $-128, %al
+; X86-NEXT:    addb $-128, %al
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_sminval_i8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    xorb $-128, %al
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -128(%rdi), %eax
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %r = xor i8 %x, 128
   ret i8 %r
 }
 
 ; negative test
 define i8 @xor_notsminval_i8(i8 %x) {

 ; X86-LABEL: xor_sminval_i32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
 ; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_sminval_i32:
 ; X64:       # %bb.0:
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -2147483648(%rdi), %eax
 ; X64-NEXT:    retq
   %r = xor i32 %x, 2147483648
   ret i32 %r
 }
 
 ; negative test
 define i32 @xor_notsminval_i32(i32 %x) {
 ; X86-LABEL: xor_notsminval_i32:

 ; XOR(ADD/SUB(X,C),MIN_SIGNED_VALUE)
 ;
 
 define i8 @xor_add_sminval_i8(i8 %x, i8 %y) {
 ; X86-LABEL: xor_add_sminval_i8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    addb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    xorb $-128, %al
+; X86-NEXT:    addb $-128, %al
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_add_sminval_i8:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $esi killed $esi def $rsi
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    leal (%rdi,%rsi), %eax
-; X64-NEXT:    xorb $-128, %al
+; X64-NEXT:    addb $-128, %al
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %s = add i8 %x, %y
   %r = xor i8 %s, 128
   ret i8 %r
 }
 
 define i16 @xor_sub_sminval_i16(i16 %x) {
; X64-NEXT:    retq
   ret i16 %r
 }
 
 define i32 @xor_add_sminval_i32(i32 %x) {
 ; X86-LABEL: xor_add_sminval_i32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl $512, %eax # imm = 0x200
 ; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_add_sminval_i32:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    leal 512(%rdi), %eax
-; X64-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X64-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; X64-NEXT:    retq
   %s = add i32 %x, 512
   %r = xor i32 %s, 2147483648
   ret i32 %r
 }
 
 define i64 @xor_add_sminval_i64(i64 %x, i64 %y) {
 ; X86-LABEL: xor_add_sminval_i64:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    adcl {{[0-9]+}}(%esp), %edx
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_add_sminval_i64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    leaq (%rdi,%rsi), %rcx
 ; X64-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
 ; X64-NEXT:    xorq %rcx, %rax
 ; X64-NEXT:    retq
   %s = add i64 %x, %y
   %r = xor i64 %s, -9223372036854775808
   ret i64 %r
 }
 
 ;
 ; ADD/SUB(XOR(X,MIN_SIGNED_VALUE),C)
 ;
 
 define i8 @sub_xor_sminval_i8(i8 %x, i8 %y) {
 ; X86-LABEL: sub_xor_sminval_i8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    xorb $-128, %al
+; X86-NEXT:    addb $-128, %al
 ; X86-NEXT:    subb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: sub_xor_sminval_i8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    xorb $-128, %al
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -128(%rdi), %eax
 ; X64-NEXT:    subb %sil, %al
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %r = xor i8 %x, 128
   %s = sub i8 %r, %y
   ret i8 %s
 }
 

 ; XOR(SHL(X,C),MIN_SIGNED_VALUE)
 ;
 
 define i8 @xor_shl_sminval_i8(i8 %x) {
 ; X86-LABEL: xor_shl_sminval_i8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    addb %al, %al
-; X86-NEXT:    xorb $-128, %al
+; X86-NEXT:    addb $-128, %al
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_shl_sminval_i8:
 ; X64:       # %bb.0:
 ; X64-NEXT:    # kill: def $edi killed $edi def $rdi
 ; X64-NEXT:    leal (%rdi,%rdi), %eax
-; X64-NEXT:    xorb $-128, %al
+; X64-NEXT:    addb $-128, %al
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %s = shl i8 %x, 1
   %r = xor i8 %s, 128
   ret i8 %r
 }
 
 define i16 @xor_shl_sminval_i16(i16 %x) {

 }
 
 ; negative test
 define i32 @xor_bigshl_sminval_i32(i32 %x) {
 ; X86-LABEL: xor_bigshl_sminval_i32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    shll $8, %eax
-; X86-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %eax # imm = 0x80000000
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_bigshl_sminval_i32:
 ; X64:       # %bb.0:
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    shll $8, %eax
-; X64-NEXT:    xorl $-2147483648, %eax # imm = 0x80000000
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    shll $8, %edi
+; X64-NEXT:    leal -2147483648(%rdi), %eax
 ; X64-NEXT:    retq
   %s = shl i32 %x, 8
   %r = xor i32 %s, 2147483648
   ret i32 %r
 }
 
 define i64 @xor_shl_sminval_i64(i64 %x) {
 ; X86-LABEL: xor_shl_sminval_i64:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    shldl $2, %eax, %edx
 ; X86-NEXT:    shll $2, %eax
-; X86-NEXT:    xorl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    addl $-2147483648, %edx # imm = 0x80000000
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: xor_shl_sminval_i64:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
 ; X64-NEXT:    leaq (%rax,%rdi,4), %rax
 ; X64-NEXT:    retq
   %s = shl i64 %x, 2
   %r = xor i64 %s, -9223372036854775808
   ret i64 %r
 }