[X86] Add test cases for suboptimal insertion of a vXi1 vector into a larger vector...

Currently we generate kshifts to clear both the upper and lower
elements, but we only need one kshift.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@369223 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/avx512-skx-insert-subvec.ll b/test/CodeGen/X86/avx512-skx-insert-subvec.ll
index 5338eb3c3a16d980efdeb3b245bdcc0c6019ca0e..bd0057e0d829e1fae8f6d23eb79246ced32fc9bc 100644 (file)
--- a/test/CodeGen/X86/avx512-skx-insert-subvec.ll
+++ b/test/CodeGen/X86/avx512-skx-insert-subvec.ll
@@ -163,3 +163,42 @@ define <8 x i1> @test11(<4 x i1> %a, <4 x i1>%b) {
   %res = shufflevector <4 x i1> %a, <4 x i1> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 2, i32 3>
   ret <8 x i1> %res
 }
+
+define <16 x i1> @test12(<2 x i1> %a) {
+; CHECK-LABEL: test12:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpsllq $63, %xmm0, %xmm0
+; CHECK-NEXT:    vpmovq2m %xmm0, %k0
+; CHECK-NEXT:    kshiftlw $14, %k0, %k0
+; CHECK-NEXT:    kshiftrw $4, %k0, %k0
+; CHECK-NEXT:    vpmovm2b %k0, %xmm0
+; CHECK-NEXT:    retq
+  %res = shufflevector <2 x i1> %a, <2 x i1> zeroinitializer, <16 x i32> <i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <16 x i1> %res
+}
+
+define <32 x i1> @test13(<2 x i1> %a) {
+; CHECK-LABEL: test13:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpsllq $63, %xmm0, %xmm0
+; CHECK-NEXT:    vpmovq2m %xmm0, %k0
+; CHECK-NEXT:    kshiftld $30, %k0, %k0
+; CHECK-NEXT:    kshiftrd $20, %k0, %k0
+; CHECK-NEXT:    vpmovm2b %k0, %ymm0
+; CHECK-NEXT:    retq
+  %res = shufflevector <2 x i1> %a, <2 x i1> zeroinitializer, <32 x i32> <i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <32 x i1> %res
+}
+
+define <64 x i1> @test14(<2 x i1> %a) {
+; CHECK-LABEL: test14:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpsllq $63, %xmm0, %xmm0
+; CHECK-NEXT:    vpmovq2m %xmm0, %k0
+; CHECK-NEXT:    kshiftlq $62, %k0, %k0
+; CHECK-NEXT:    kshiftrq $52, %k0, %k0
+; CHECK-NEXT:    vpmovm2b %k0, %zmm0
+; CHECK-NEXT:    retq
+  %res = shufflevector <2 x i1> %a, <2 x i1> zeroinitializer, <64 x i32> <i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+  ret <64 x i1> %res
+}