[X86] Fold (trunc (i32 (zextload i16))) into vbroadcast.

[oota-llvm.git] / test / CodeGen / X86 / vector-shuffle-256-v16.ll

diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 4d2bcd9bc8dc3f2dcc0de3d15655f86a2cd3bd99..10e8e3a680c30cb8153e87234e98a3efe9ff0ca9 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -3342,9 +3342,7 @@ define <16 x i16> @insert_dup_elt1_mem_v16i16_i32(i32* %ptr) #0 {
 ;
 ; AVX2-LABEL: insert_dup_elt1_mem_v16i16_i32:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    movzwl 2(%rdi), %eax
-; AVX2-NEXT:    vmovd %eax, %xmm0
-; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX2-NEXT:    vpbroadcastw 2(%rdi), %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
@@ -3363,9 +3361,7 @@ define <16 x i16> @insert_dup_elt3_mem_v16i16_i32(i32* %ptr) #0 {
 ;
 ; AVX2-LABEL: insert_dup_elt3_mem_v16i16_i32:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    movzwl 2(%rdi), %eax
-; AVX2-NEXT:    vmovd %eax, %xmm0
-; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX2-NEXT:    vpbroadcastw 2(%rdi), %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 1