[CostModel][X86] Add truncate cost tests to cover all legal destination types

We were only testing costs for legal source vector element counts

llvm-svn: 350323

1 files changed, 137 insertions, 9 deletions

diff --git a/llvm/test/Analysis/CostModel/X86/trunc.ll b/llvm/test/Analysis/CostModel/X86/trunc.ll
index e2b2cfa732d..9668da56413 100644
--- a/llvm/test/Analysis/CostModel/X86/trunc.ll
+++ b/llvm/test/Analysis/CostModel/X86/trunc.ll
@@ -16,35 +16,41 @@ define i32 @trunc_vXi32() {
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i32>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i32>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i32>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i32>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'trunc_vXi32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i32>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'trunc_vXi32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i32>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'trunc_vXi32'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i32>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'trunc_vXi32'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i32>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2i64 = trunc <2 x i64> undef to <2 x i32>
   %V4i64 = trunc <4 x i64> undef to <4 x i32>
   %V8i64 = trunc <8 x i64> undef to <8 x i32>
+  %V16i64 = trunc <16 x i64> undef to <16 x i32>
   ret i32 undef
 }
 
@@ -53,79 +59,117 @@ define i32 @trunc_vXi16() {
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'trunc_vXi16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'trunc_vXi16'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'trunc_vXi16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'trunc_vXi16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
-; AVX512-LABEL: 'trunc_vXi16'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; AVX512F-LABEL: 'trunc_vXi16'
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; AVX512BW-LABEL: 'trunc_vXi16'
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'trunc_vXi16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i16>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2i64 = trunc <2 x i64> undef to <2 x i16>
   %V4i64 = trunc <4 x i64> undef to <4 x i16>
   %V8i64 = trunc <8 x i64> undef to <8 x i16>
+  %V16i64 = trunc <16 x i64> undef to <16 x i16>
+  %V32i64 = trunc <32 x i64> undef to <32 x i16>
+
   %V2i32 = trunc <2 x i32> undef to <2 x i16>
   %V4i32 = trunc <4 x i32> undef to <4 x i16>
   %V8i32 = trunc <8 x i32> undef to <8 x i16>
   %V16i32 = trunc <16 x i32> undef to <16 x i16>
+  %V32i32 = trunc <32 x i32> undef to <32 x i16>
   ret i32 undef
 }
 
@@ -134,136 +178,190 @@ define i32 @trunc_vXi8() {
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'trunc_vXi8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'trunc_vXi8'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'trunc_vXi8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'trunc_vXi8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'trunc_vXi8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'trunc_vXi8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'trunc_vXi8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i8>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i8>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2i64 = trunc <2 x i64> undef to <2 x i8>
   %V4i64 = trunc <4 x i64> undef to <4 x i8>
   %V8i64 = trunc <8 x i64> undef to <8 x i8>
+  %V16i64 = trunc <16 x i64> undef to <16 x i8>
+  %V32i64 = trunc <32 x i64> undef to <32 x i8>
+  %V64i64 = trunc <64 x i64> undef to <64 x i8>
 
   %V2i32 = trunc <2 x i32> undef to <2 x i8>
   %V4i32 = trunc <4 x i32> undef to <4 x i8>
   %V8i32 = trunc <8 x i32> undef to <8 x i8>
   %V16i32 = trunc <16 x i32> undef to <16 x i8>
+  %V32i32 = trunc <32 x i32> undef to <32 x i8>
+  %V64i32 = trunc <64 x i32> undef to <64 x i8>
 
   %V2i16 = trunc <2 x i16> undef to <2 x i8>
   %V4i16 = trunc <4 x i16> undef to <4 x i8>
   %V8i16 = trunc <8 x i16> undef to <8 x i8>
   %V16i16 = trunc <16 x i16> undef to <16 x i8>
   %V32i16 = trunc <32 x i16> undef to <32 x i8>
+  %V64i16 = trunc <64 x i16> undef to <64 x i8>
 
   ret i32 undef
 }
@@ -273,15 +371,21 @@ define i32 @trunc_vXi1() {
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i1>
+; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i8 = trunc <32 x i8> undef to <32 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i8 = trunc <64 x i8> undef to <64 x i1>
 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
@@ -290,15 +394,21 @@ define i32 @trunc_vXi1() {
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i1>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i8 = trunc <32 x i8> undef to <32 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i8 = trunc <64 x i8> undef to <64 x i1>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
@@ -307,15 +417,21 @@ define i32 @trunc_vXi1() {
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i1>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i8 = trunc <32 x i8> undef to <32 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i8 = trunc <64 x i8> undef to <64 x i1>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
@@ -324,15 +440,21 @@ define i32 @trunc_vXi1() {
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i64 = trunc <2 x i64> undef to <2 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i64 = trunc <4 x i64> undef to <4 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i64 = trunc <8 x i64> undef to <8 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i64 = trunc <16 x i64> undef to <16 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i64 = trunc <32 x i64> undef to <32 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i64 = trunc <64 x i64> undef to <64 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i32 = trunc <2 x i32> undef to <2 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i32 = trunc <4 x i32> undef to <4 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i32 = trunc <8 x i32> undef to <8 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i32 = trunc <16 x i32> undef to <16 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i32 = trunc <32 x i32> undef to <32 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i32 = trunc <64 x i32> undef to <64 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V2i16 = trunc <2 x i16> undef to <2 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V4i16 = trunc <4 x i16> undef to <4 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V8i16 = trunc <8 x i16> undef to <8 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V16i16 = trunc <16 x i16> undef to <16 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = trunc <32 x i16> undef to <32 x i1>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64i16 = trunc <64 x i16> undef to <64 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V32i8 = trunc <32 x i8> undef to <32 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V64i8 = trunc <64 x i8> undef to <64 x i1>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
@@ -340,17 +462,23 @@ define i32 @trunc_vXi1() {
   %V2i64 = trunc <2 x i64> undef to <2 x i1>
   %V4i64 = trunc <4 x i64> undef to <4 x i1>
   %V8i64 = trunc <8 x i64> undef to <8 x i1>
+  %V16i64 = trunc <16 x i64> undef to <16 x i1>
+  %V32i64 = trunc <32 x i64> undef to <32 x i1>
+  %V64i64 = trunc <64 x i64> undef to <64 x i1>
 
   %V2i32 = trunc <2 x i32> undef to <2 x i1>
   %V4i32 = trunc <4 x i32> undef to <4 x i1>
   %V8i32 = trunc <8 x i32> undef to <8 x i1>
   %V16i32 = trunc <16 x i32> undef to <16 x i1>
+  %V32i32 = trunc <32 x i32> undef to <32 x i1>
+  %V64i32 = trunc <64 x i32> undef to <64 x i1>
 
   %V2i16 = trunc <2 x i16> undef to <2 x i1>
   %V4i16 = trunc <4 x i16> undef to <4 x i1>
   %V8i16 = trunc <8 x i16> undef to <8 x i1>
   %V16i16 = trunc <16 x i16> undef to <16 x i1>
   %V32i16 = trunc <32 x i16> undef to <32 x i1>
+  %V64i16 = trunc <64 x i16> undef to <64 x i1>
 
   %V32i8 = trunc <32 x i8> undef to <32 x i1>
   %V64i8 = trunc <64 x i8> undef to <64 x i1>