[X86] Add SSE42 sat-add test coverage

Check SSE42 targets which have PCMPGTQ

diff --git a/llvm/test/CodeGen/X86/sat-add.ll b/llvm/test/CodeGen/X86/sat-add.ll
index 53df6d440830bd85233624154869b03eade3ba1e..72ccebedc7a848d7a768152a5c63f0e4dc86dd58 100644 (file)
--- a/llvm/test/CodeGen/X86/sat-add.ll
+++ b/llvm/test/CodeGen/X86/sat-add.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2   | FileCheck %s --check-prefixes=ANY,SSE2
-; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=ANY,SSE41
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=ANY,SSE4,SSE41
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.2 | FileCheck %s --check-prefixes=ANY,SSE4,SSE42
 
 ; There are at least 3 potential patterns corresponding to an unsigned saturated add: min, cmp with sum, cmp with not.
 ; Test each of those patterns with i8/i16/i32/i64.
@@ -401,11 +402,11 @@ define <8 x i16> @unsigned_sat_constant_v8i16_using_min(<8 x i16> %x) {
 ; SSE2-NEXT:    paddw {{.*}}(%rip), %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_constant_v8i16_using_min:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pminuw {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    paddw {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_constant_v8i16_using_min:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pminuw {{.*}}(%rip), %xmm0
+; SSE4-NEXT:    paddw {{.*}}(%rip), %xmm0
+; SSE4-NEXT:    retq
   %c = icmp ult <8 x i16> %x, <i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43>
   %s = select <8 x i1> %c, <8 x i16> %x, <8 x i16> <i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43>
   %r = add <8 x i16> %s, <i16 42, i16 42, i16 42, i16 42, i16 42, i16 42, i16 42, i16 42>
@@ -447,11 +448,11 @@ define <4 x i32> @unsigned_sat_constant_v4i32_using_min(<4 x i32> %x) {
 ; SSE2-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_constant_v4i32_using_min:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pminud {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    paddd {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_constant_v4i32_using_min:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pminud {{.*}}(%rip), %xmm0
+; SSE4-NEXT:    paddd {{.*}}(%rip), %xmm0
+; SSE4-NEXT:    retq
   %c = icmp ult <4 x i32> %x, <i32 -43, i32 -43, i32 -43, i32 -43>
   %s = select <4 x i1> %c, <4 x i32> %x, <4 x i32> <i32 -43, i32 -43, i32 -43, i32 -43>
   %r = add <4 x i32> %s, <i32 42, i32 42, i32 42, i32 42>
@@ -470,18 +471,18 @@ define <4 x i32> @unsigned_sat_constant_v4i32_using_cmp_sum(<4 x i32> %x) {
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_constant_v4i32_using_cmp_sum:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [42,42,42,42]
-; SSE41-NEXT:    paddd %xmm0, %xmm2
-; SSE41-NEXT:    movdqa %xmm0, %xmm1
-; SSE41-NEXT:    pminud %xmm2, %xmm1
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm1
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
-; SSE41-NEXT:    pxor %xmm0, %xmm1
-; SSE41-NEXT:    por %xmm2, %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_constant_v4i32_using_cmp_sum:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    movdqa {{.*#+}} xmm2 = [42,42,42,42]
+; SSE4-NEXT:    paddd %xmm0, %xmm2
+; SSE4-NEXT:    movdqa %xmm0, %xmm1
+; SSE4-NEXT:    pminud %xmm2, %xmm1
+; SSE4-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE4-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE4-NEXT:    pxor %xmm0, %xmm1
+; SSE4-NEXT:    por %xmm2, %xmm1
+; SSE4-NEXT:    movdqa %xmm1, %xmm0
+; SSE4-NEXT:    retq
   %a = add <4 x i32> %x, <i32 42, i32 42, i32 42, i32 42>
   %c = icmp ugt <4 x i32> %x, %a
   %r = select <4 x i1> %c, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, <4 x i32> %a
@@ -498,15 +499,15 @@ define <4 x i32> @unsigned_sat_constant_v4i32_using_cmp_notval(<4 x i32> %x) {
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_constant_v4i32_using_cmp_notval:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [42,42,42,42]
-; SSE41-NEXT:    paddd %xmm0, %xmm1
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [4294967254,4294967254,4294967254,4294967254]
-; SSE41-NEXT:    pmaxud %xmm0, %xmm2
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm0
-; SSE41-NEXT:    por %xmm1, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_constant_v4i32_using_cmp_notval:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    movdqa {{.*#+}} xmm1 = [42,42,42,42]
+; SSE4-NEXT:    paddd %xmm0, %xmm1
+; SSE4-NEXT:    movdqa {{.*#+}} xmm2 = [4294967254,4294967254,4294967254,4294967254]
+; SSE4-NEXT:    pmaxud %xmm0, %xmm2
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE4-NEXT:    por %xmm1, %xmm0
+; SSE4-NEXT:    retq
   %a = add <4 x i32> %x, <i32 42, i32 42, i32 42, i32 42>
   %c = icmp ugt <4 x i32> %x, <i32 -43, i32 -43, i32 -43, i32 -43>
   %r = select <4 x i1> %c, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, <4 x i32> %a
@@ -523,15 +524,15 @@ define <4 x i32> @unsigned_sat_constant_v4i32_using_cmp_notval_nonsplat(<4 x i32
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_constant_v4i32_using_cmp_notval_nonsplat:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [43,44,45,46]
-; SSE41-NEXT:    paddd %xmm0, %xmm1
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [4294967253,4294967252,4294967251,4294967250]
-; SSE41-NEXT:    pmaxud %xmm0, %xmm2
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm0
-; SSE41-NEXT:    por %xmm1, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_constant_v4i32_using_cmp_notval_nonsplat:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    movdqa {{.*#+}} xmm1 = [43,44,45,46]
+; SSE4-NEXT:    paddd %xmm0, %xmm1
+; SSE4-NEXT:    movdqa {{.*#+}} xmm2 = [4294967253,4294967252,4294967251,4294967250]
+; SSE4-NEXT:    pmaxud %xmm0, %xmm2
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE4-NEXT:    por %xmm1, %xmm0
+; SSE4-NEXT:    retq
   %a = add <4 x i32> %x, <i32 43, i32 44, i32 45, i32 46>
   %c = icmp ugt <4 x i32> %x, <i32 -44, i32 -45, i32 -46, i32 -47>
   %r = select <4 x i1> %c, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, <4 x i32> %a
@@ -575,6 +576,19 @@ define <2 x i64> @unsigned_sat_constant_v2i64_using_min(<2 x i64> %x) {
 ; SSE41-NEXT:    paddq {{.*}}(%rip), %xmm2
 ; SSE41-NEXT:    movdqa %xmm2, %xmm0
 ; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_constant_v2i64_using_min:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    movdqa %xmm0, %xmm1
+; SSE42-NEXT:    movapd {{.*#+}} xmm2 = [18446744073709551573,18446744073709551573]
+; SSE42-NEXT:    movdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
+; SSE42-NEXT:    pxor %xmm0, %xmm3
+; SSE42-NEXT:    movdqa {{.*#+}} xmm0 = [9223372036854775765,9223372036854775765]
+; SSE42-NEXT:    pcmpgtq %xmm3, %xmm0
+; SSE42-NEXT:    blendvpd %xmm0, %xmm1, %xmm2
+; SSE42-NEXT:    paddq {{.*}}(%rip), %xmm2
+; SSE42-NEXT:    movdqa %xmm2, %xmm0
+; SSE42-NEXT:    retq
   %c = icmp ult <2 x i64> %x, <i64 -43, i64 -43>
   %s = select <2 x i1> %c, <2 x i64> %x, <2 x i64> <i64 -43, i64 -43>
   %r = add <2 x i64> %s, <i64 42, i64 42>
@@ -582,23 +596,52 @@ define <2 x i64> @unsigned_sat_constant_v2i64_using_min(<2 x i64> %x) {
 }
 
 define <2 x i64> @unsigned_sat_constant_v2i64_using_cmp_sum(<2 x i64> %x) {
-; ANY-LABEL: unsigned_sat_constant_v2i64_using_cmp_sum:
-; ANY:       # %bb.0:
-; ANY-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
-; ANY-NEXT:    paddq %xmm0, %xmm1
-; ANY-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; ANY-NEXT:    pxor %xmm2, %xmm0
-; ANY-NEXT:    pxor %xmm1, %xmm2
-; ANY-NEXT:    movdqa %xmm0, %xmm3
-; ANY-NEXT:    pcmpgtd %xmm2, %xmm3
-; ANY-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; ANY-NEXT:    pcmpeqd %xmm0, %xmm2
-; ANY-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; ANY-NEXT:    pand %xmm4, %xmm2
-; ANY-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; ANY-NEXT:    por %xmm1, %xmm0
-; ANY-NEXT:    por %xmm2, %xmm0
-; ANY-NEXT:    retq
+; SSE2-LABEL: unsigned_sat_constant_v2i64_using_cmp_sum:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE2-NEXT:    paddq %xmm0, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm2
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE41-LABEL: unsigned_sat_constant_v2i64_using_cmp_sum:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE41-NEXT:    paddq %xmm0, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE41-NEXT:    pxor %xmm2, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm2
+; SSE41-NEXT:    movdqa %xmm0, %xmm3
+; SSE41-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE41-NEXT:    pand %xmm4, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    por %xmm2, %xmm0
+; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_constant_v2i64_using_cmp_sum:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE42-NEXT:    paddq %xmm0, %xmm1
+; SSE42-NEXT:    movdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
+; SSE42-NEXT:    pxor %xmm2, %xmm0
+; SSE42-NEXT:    pxor %xmm1, %xmm2
+; SSE42-NEXT:    pcmpgtq %xmm2, %xmm0
+; SSE42-NEXT:    por %xmm1, %xmm0
+; SSE42-NEXT:    retq
   %a = add <2 x i64> %x, <i64 42, i64 42>
   %c = icmp ugt <2 x i64> %x, %a
   %r = select <2 x i1> %c, <2 x i64> <i64 -1, i64 -1>, <2 x i64> %a
@@ -606,22 +649,48 @@ define <2 x i64> @unsigned_sat_constant_v2i64_using_cmp_sum(<2 x i64> %x) {
 }
 
 define <2 x i64> @unsigned_sat_constant_v2i64_using_cmp_notval(<2 x i64> %x) {
-; ANY-LABEL: unsigned_sat_constant_v2i64_using_cmp_notval:
-; ANY:       # %bb.0:
-; ANY-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
-; ANY-NEXT:    paddq %xmm0, %xmm1
-; ANY-NEXT:    pxor {{.*}}(%rip), %xmm0
-; ANY-NEXT:    movdqa {{.*#+}} xmm2 = [9223372034707292117,9223372034707292117]
-; ANY-NEXT:    movdqa %xmm0, %xmm3
-; ANY-NEXT:    pcmpgtd %xmm2, %xmm3
-; ANY-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; ANY-NEXT:    pcmpeqd %xmm2, %xmm0
-; ANY-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; ANY-NEXT:    pand %xmm4, %xmm2
-; ANY-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; ANY-NEXT:    por %xmm1, %xmm0
-; ANY-NEXT:    por %xmm2, %xmm0
-; ANY-NEXT:    retq
+; SSE2-LABEL: unsigned_sat_constant_v2i64_using_cmp_notval:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE2-NEXT:    paddq %xmm0, %xmm1
+; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [9223372034707292117,9223372034707292117]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE41-LABEL: unsigned_sat_constant_v2i64_using_cmp_notval:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE41-NEXT:    paddq %xmm0, %xmm1
+; SSE41-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [9223372034707292117,9223372034707292117]
+; SSE41-NEXT:    movdqa %xmm0, %xmm3
+; SSE41-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm4, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    por %xmm2, %xmm0
+; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_constant_v2i64_using_cmp_notval:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    movdqa {{.*#+}} xmm1 = [42,42]
+; SSE42-NEXT:    paddq %xmm0, %xmm1
+; SSE42-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE42-NEXT:    pcmpgtq {{.*}}(%rip), %xmm0
+; SSE42-NEXT:    por %xmm1, %xmm0
+; SSE42-NEXT:    retq
   %a = add <2 x i64> %x, <i64 42, i64 42>
   %c = icmp ugt <2 x i64> %x, <i64 -43, i64 -43>
   %r = select <2 x i1> %c, <2 x i64> <i64 -1, i64 -1>, <2 x i64> %a
@@ -684,13 +753,13 @@ define <8 x i16> @unsigned_sat_variable_v8i16_using_min(<8 x i16> %x, <8 x i16>
 ; SSE2-NEXT:    paddw %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_variable_v8i16_using_min:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT:    pxor %xmm1, %xmm2
-; SSE41-NEXT:    pminuw %xmm2, %xmm0
-; SSE41-NEXT:    paddw %xmm1, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_variable_v8i16_using_min:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE4-NEXT:    pxor %xmm1, %xmm2
+; SSE4-NEXT:    pminuw %xmm2, %xmm0
+; SSE4-NEXT:    paddw %xmm1, %xmm0
+; SSE4-NEXT:    retq
   %noty = xor <8 x i16> %y, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
   %c = icmp ult <8 x i16> %x, %noty
   %s = select <8 x i1> %c, <8 x i16> %x, <8 x i16> %noty
@@ -720,17 +789,17 @@ define <8 x i16> @unsigned_sat_variable_v8i16_using_cmp_notval(<8 x i16> %x, <8
 ; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_variable_v8i16_using_cmp_notval:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT:    movdqa %xmm0, %xmm3
-; SSE41-NEXT:    paddw %xmm1, %xmm3
-; SSE41-NEXT:    pxor %xmm2, %xmm1
-; SSE41-NEXT:    pminuw %xmm0, %xmm1
-; SSE41-NEXT:    pcmpeqw %xmm1, %xmm0
-; SSE41-NEXT:    pxor %xmm2, %xmm0
-; SSE41-NEXT:    por %xmm3, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_variable_v8i16_using_cmp_notval:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE4-NEXT:    movdqa %xmm0, %xmm3
+; SSE4-NEXT:    paddw %xmm1, %xmm3
+; SSE4-NEXT:    pxor %xmm2, %xmm1
+; SSE4-NEXT:    pminuw %xmm0, %xmm1
+; SSE4-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE4-NEXT:    pxor %xmm2, %xmm0
+; SSE4-NEXT:    por %xmm3, %xmm0
+; SSE4-NEXT:    retq
   %noty = xor <8 x i16> %y, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
   %a = add <8 x i16> %x, %y
   %c = icmp ugt <8 x i16> %x, %noty
@@ -755,13 +824,13 @@ define <4 x i32> @unsigned_sat_variable_v4i32_using_min(<4 x i32> %x, <4 x i32>
 ; SSE2-NEXT:    paddd %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_variable_v4i32_using_min:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT:    pxor %xmm1, %xmm2
-; SSE41-NEXT:    pminud %xmm2, %xmm0
-; SSE41-NEXT:    paddd %xmm1, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_variable_v4i32_using_min:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE4-NEXT:    pxor %xmm1, %xmm2
+; SSE4-NEXT:    pminud %xmm2, %xmm0
+; SSE4-NEXT:    paddd %xmm1, %xmm0
+; SSE4-NEXT:    retq
   %noty = xor <4 x i32> %y, <i32 -1, i32 -1, i32 -1, i32 -1>
   %c = icmp ult <4 x i32> %x, %noty
   %s = select <4 x i1> %c, <4 x i32> %x, <4 x i32> %noty
@@ -780,17 +849,17 @@ define <4 x i32> @unsigned_sat_variable_v4i32_using_cmp_sum(<4 x i32> %x, <4 x i
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_variable_v4i32_using_cmp_sum:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    paddd %xmm0, %xmm1
-; SSE41-NEXT:    movdqa %xmm0, %xmm2
-; SSE41-NEXT:    pminud %xmm1, %xmm2
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm2
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
-; SSE41-NEXT:    pxor %xmm0, %xmm2
-; SSE41-NEXT:    por %xmm1, %xmm2
-; SSE41-NEXT:    movdqa %xmm2, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_variable_v4i32_using_cmp_sum:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    paddd %xmm0, %xmm1
+; SSE4-NEXT:    movdqa %xmm0, %xmm2
+; SSE4-NEXT:    pminud %xmm1, %xmm2
+; SSE4-NEXT:    pcmpeqd %xmm0, %xmm2
+; SSE4-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE4-NEXT:    pxor %xmm0, %xmm2
+; SSE4-NEXT:    por %xmm1, %xmm2
+; SSE4-NEXT:    movdqa %xmm2, %xmm0
+; SSE4-NEXT:    retq
   %a = add <4 x i32> %x, %y
   %c = icmp ugt <4 x i32> %x, %a
   %r = select <4 x i1> %c, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, <4 x i32> %a
@@ -808,17 +877,17 @@ define <4 x i32> @unsigned_sat_variable_v4i32_using_cmp_notval(<4 x i32> %x, <4
 ; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSE41-LABEL: unsigned_sat_variable_v4i32_using_cmp_notval:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT:    movdqa %xmm0, %xmm3
-; SSE41-NEXT:    paddd %xmm1, %xmm3
-; SSE41-NEXT:    pxor %xmm2, %xmm1
-; SSE41-NEXT:    pminud %xmm0, %xmm1
-; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
-; SSE41-NEXT:    pxor %xmm2, %xmm0
-; SSE41-NEXT:    por %xmm3, %xmm0
-; SSE41-NEXT:    retq
+; SSE4-LABEL: unsigned_sat_variable_v4i32_using_cmp_notval:
+; SSE4:       # %bb.0:
+; SSE4-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE4-NEXT:    movdqa %xmm0, %xmm3
+; SSE4-NEXT:    paddd %xmm1, %xmm3
+; SSE4-NEXT:    pxor %xmm2, %xmm1
+; SSE4-NEXT:    pminud %xmm0, %xmm1
+; SSE4-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE4-NEXT:    pxor %xmm2, %xmm0
+; SSE4-NEXT:    por %xmm3, %xmm0
+; SSE4-NEXT:    retq
   %noty = xor <4 x i32> %y, <i32 -1, i32 -1, i32 -1, i32 -1>
   %a = add <4 x i32> %x, %y
   %c = icmp ugt <4 x i32> %x, %noty
@@ -869,6 +938,21 @@ define <2 x i64> @unsigned_sat_variable_v2i64_using_min(<2 x i64> %x, <2 x i64>
 ; SSE41-NEXT:    paddq %xmm1, %xmm3
 ; SSE41-NEXT:    movdqa %xmm3, %xmm0
 ; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_variable_v2i64_using_min:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    movdqa %xmm0, %xmm2
+; SSE42-NEXT:    pcmpeqd %xmm3, %xmm3
+; SSE42-NEXT:    pxor %xmm1, %xmm3
+; SSE42-NEXT:    movdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808]
+; SSE42-NEXT:    pxor %xmm0, %xmm4
+; SSE42-NEXT:    movdqa {{.*#+}} xmm0 = [9223372036854775807,9223372036854775807]
+; SSE42-NEXT:    pxor %xmm1, %xmm0
+; SSE42-NEXT:    pcmpgtq %xmm4, %xmm0
+; SSE42-NEXT:    blendvpd %xmm0, %xmm2, %xmm3
+; SSE42-NEXT:    paddq %xmm1, %xmm3
+; SSE42-NEXT:    movdqa %xmm3, %xmm0
+; SSE42-NEXT:    retq
   %noty = xor <2 x i64> %y, <i64 -1, i64 -1>
   %c = icmp ult <2 x i64> %x, %noty
   %s = select <2 x i1> %c, <2 x i64> %x, <2 x i64> %noty
@@ -877,22 +961,49 @@ define <2 x i64> @unsigned_sat_variable_v2i64_using_min(<2 x i64> %x, <2 x i64>
 }
 
 define <2 x i64> @unsigned_sat_variable_v2i64_using_cmp_sum(<2 x i64> %x, <2 x i64> %y) {
-; ANY-LABEL: unsigned_sat_variable_v2i64_using_cmp_sum:
-; ANY:       # %bb.0:
-; ANY-NEXT:    paddq %xmm0, %xmm1
-; ANY-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
-; ANY-NEXT:    pxor %xmm2, %xmm0
-; ANY-NEXT:    pxor %xmm1, %xmm2
-; ANY-NEXT:    movdqa %xmm0, %xmm3
-; ANY-NEXT:    pcmpgtd %xmm2, %xmm3
-; ANY-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; ANY-NEXT:    pcmpeqd %xmm0, %xmm2
-; ANY-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; ANY-NEXT:    pand %xmm4, %xmm2
-; ANY-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; ANY-NEXT:    por %xmm1, %xmm0
-; ANY-NEXT:    por %xmm2, %xmm0
-; ANY-NEXT:    retq
+; SSE2-LABEL: unsigned_sat_variable_v2i64_using_cmp_sum:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    paddq %xmm0, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm2
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE41-LABEL: unsigned_sat_variable_v2i64_using_cmp_sum:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    paddq %xmm0, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
+; SSE41-NEXT:    pxor %xmm2, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm2
+; SSE41-NEXT:    movdqa %xmm0, %xmm3
+; SSE41-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE41-NEXT:    pand %xmm4, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    por %xmm2, %xmm0
+; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_variable_v2i64_using_cmp_sum:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    paddq %xmm0, %xmm1
+; SSE42-NEXT:    movdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
+; SSE42-NEXT:    pxor %xmm2, %xmm0
+; SSE42-NEXT:    pxor %xmm1, %xmm2
+; SSE42-NEXT:    pcmpgtq %xmm2, %xmm0
+; SSE42-NEXT:    por %xmm1, %xmm0
+; SSE42-NEXT:    retq
   %a = add <2 x i64> %x, %y
   %c = icmp ugt <2 x i64> %x, %a
   %r = select <2 x i1> %c, <2 x i64> <i64 -1, i64 -1>, <2 x i64> %a
@@ -900,22 +1011,49 @@ define <2 x i64> @unsigned_sat_variable_v2i64_using_cmp_sum(<2 x i64> %x, <2 x i
 }
 
 define <2 x i64> @unsigned_sat_variable_v2i64_using_cmp_notval(<2 x i64> %x, <2 x i64> %y) {
-; ANY-LABEL: unsigned_sat_variable_v2i64_using_cmp_notval:
-; ANY:       # %bb.0:
-; ANY-NEXT:    movdqa %xmm0, %xmm2
-; ANY-NEXT:    paddq %xmm1, %xmm2
-; ANY-NEXT:    pxor {{.*}}(%rip), %xmm1
-; ANY-NEXT:    pxor {{.*}}(%rip), %xmm0
-; ANY-NEXT:    movdqa %xmm0, %xmm3
-; ANY-NEXT:    pcmpgtd %xmm1, %xmm3
-; ANY-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
-; ANY-NEXT:    pcmpeqd %xmm1, %xmm0
-; ANY-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; ANY-NEXT:    pand %xmm4, %xmm1
-; ANY-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; ANY-NEXT:    por %xmm2, %xmm0
-; ANY-NEXT:    por %xmm1, %xmm0
-; ANY-NEXT:    retq
+; SSE2-LABEL: unsigned_sat_variable_v2i64_using_cmp_notval:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    paddq %xmm1, %xmm2
+; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm4, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE41-LABEL: unsigned_sat_variable_v2i64_using_cmp_notval:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    paddq %xmm1, %xmm2
+; SSE41-NEXT:    pxor {{.*}}(%rip), %xmm1
+; SSE41-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa %xmm0, %xmm3
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm4, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE41-NEXT:    por %xmm2, %xmm0
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; SSE42-LABEL: unsigned_sat_variable_v2i64_using_cmp_notval:
+; SSE42:       # %bb.0:
+; SSE42-NEXT:    movdqa %xmm0, %xmm2
+; SSE42-NEXT:    paddq %xmm1, %xmm2
+; SSE42-NEXT:    pxor {{.*}}(%rip), %xmm1
+; SSE42-NEXT:    pxor {{.*}}(%rip), %xmm0
+; SSE42-NEXT:    pcmpgtq %xmm1, %xmm0
+; SSE42-NEXT:    por %xmm2, %xmm0
+; SSE42-NEXT:    retq
   %noty = xor <2 x i64> %y, <i64 -1, i64 -1>
   %a = add <2 x i64> %x, %y
   %c = icmp ugt <2 x i64> %x, %noty