[CostModel][X86] Enable non-uniform vector division by constants costs.

Non-uniform division/remainder handling was added back at D49248/D50765 - so share the 'mul+sub' costs that already exist for uniform cases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@345164 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 598 insertions, 102 deletions

diff --git a/test/Analysis/CostModel/X86/div.ll b/test/Analysis/CostModel/X86/div.ll
index 63fb25dd2f6..7a53db1b26b 100644
--- a/test/Analysis/CostModel/X86/div.ll
+++ b/test/Analysis/CostModel/X86/div.ll
@@ -136,24 +136,157 @@ define i32 @udiv() {
 }
 
 define i32 @sdiv_const() {
-; CHECK-LABEL: 'sdiv_const'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'sdiv_const'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'sdiv_const'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'sdiv_const'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'sdiv_const'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'sdiv_const'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512F-LABEL: 'sdiv_const'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512BW-LABEL: 'sdiv_const'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SLM-LABEL: 'sdiv_const'
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'sdiv_const'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, 7
@@ -161,13 +294,13 @@ define i32 @sdiv_const() {
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = sdiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = sdiv i32 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = sdiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = sdiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = sdiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = sdiv i16 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = sdiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = sdiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = sdiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = sdiv i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = sdiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = sdiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
@@ -198,24 +331,119 @@ define i32 @sdiv_const() {
 }
 
 define i32 @udiv_const() {
-; CHECK-LABEL: 'udiv_const'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE-LABEL: 'udiv_const'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'udiv_const'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'udiv_const'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512F-LABEL: 'udiv_const'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512BW-LABEL: 'udiv_const'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SLM-LABEL: 'udiv_const'
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'udiv_const'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, 7
@@ -223,13 +451,13 @@ define i32 @udiv_const() {
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = udiv <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = udiv i32 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4i32 = udiv <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = udiv <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16i32 = udiv <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = udiv i16 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8i16 = udiv <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16i16 = udiv <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32i16 = udiv <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = udiv i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = udiv <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = udiv <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
diff --git a/test/Analysis/CostModel/X86/rem.ll b/test/Analysis/CostModel/X86/rem.ll
index fd7e83d74ff..e28b4d9583f 100644
--- a/test/Analysis/CostModel/X86/rem.ll
+++ b/test/Analysis/CostModel/X86/rem.ll
@@ -136,24 +136,176 @@ define i32 @urem() {
 }
 
 define i32 @srem_const() {
-; CHECK-LABEL: 'srem_const'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'srem_const'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSSE3-LABEL: 'srem_const'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE42-LABEL: 'srem_const'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'srem_const'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'srem_const'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512F-LABEL: 'srem_const'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512BW-LABEL: 'srem_const'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SLM-LABEL: 'srem_const'
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; SLM-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; GLM-LABEL: 'srem_const'
+; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = srem <2 x i64> undef, <i64 6, i64 7>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
+; GLM-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = srem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'srem_const'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = srem i64 undef, 7
@@ -161,13 +313,13 @@ define i32 @srem_const() {
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = srem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = srem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = srem i32 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = srem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8i32 = srem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V16i32 = srem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = srem i16 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = srem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = srem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = srem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = srem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = srem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = srem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
@@ -198,24 +350,100 @@ define i32 @srem_const() {
 }
 
 define i32 @urem_const() {
-; CHECK-LABEL: 'urem_const'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE-LABEL: 'urem_const'
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; SSE-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX1-LABEL: 'urem_const'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX2-LABEL: 'urem_const'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512F-LABEL: 'urem_const'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512BW-LABEL: 'urem_const'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = urem <2 x i64> undef, <i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1280 for instruction: %V64i8 = urem <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'urem_const'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = urem i64 undef, 7
@@ -223,13 +451,13 @@ define i32 @urem_const() {
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = urem <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i64 = urem <8 x i64> undef, <i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = urem i32 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4i32 = urem <4 x i32> undef, <i32 4, i32 5, i32 6, i32 7>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8i32 = urem <8 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V16i32 = urem <16 x i32> undef, <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = urem i16 undef, 7
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = urem <8 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16i16 = urem <16 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32i16 = urem <32 x i16> undef, <i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15, i16 16, i16 17, i16 18, i16 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = urem i8 undef, 7
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %V16i8 = urem <16 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 640 for instruction: %V32i8 = urem <32 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
diff --git a/test/Analysis/CostModel/X86/vdiv-cost.ll b/test/Analysis/CostModel/X86/vdiv-cost.ll
index 7bb935cbcec..d5b404227aa 100644
--- a/test/Analysis/CostModel/X86/vdiv-cost.ll
+++ b/test/Analysis/CostModel/X86/vdiv-cost.ll
@@ -162,18 +162,58 @@ define <8 x i32> @test9(<8 x i32> %a) {
 }
 
 define <8 x i32> @test10(<8 x i32> %a) {
-; CHECK-LABEL: 'test10'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+; SSE2-LABEL: 'test10'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+;
+; SSSE3-LABEL: 'test10'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+;
+; SSE42-LABEL: 'test10'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+;
+; AVX1-LABEL: 'test10'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+;
+; AVX2-LABEL: 'test10'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
+;
+; AVX512-LABEL: 'test10'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %div
 ;
   %div = sdiv <8 x i32> %a, <i32 8, i32 7, i32 7, i32 7,i32 7, i32 7, i32 7, i32 7>
   ret <8 x i32> %div
 }
 
 define <16 x i32> @test11(<16 x i32> %a) {
-; CHECK-LABEL: 'test11'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+; SSE2-LABEL: 'test11'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+;
+; SSSE3-LABEL: 'test11'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+;
+; SSE42-LABEL: 'test11'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+;
+; AVX1-LABEL: 'test11'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+;
+; AVX2-LABEL: 'test11'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
+;
+; AVX512-LABEL: 'test11'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %div
 ;
   %div = sdiv <16 x i32> %a, <i32 8, i32 7, i32 7, i32 7,i32 7, i32 7, i32 7, i32 7, i32 8, i32 7, i32 7, i32 7,i32 7, i32 7, i32 7, i32 7>
   ret <16 x i32> %div