[X86][AVX512DQVL] Improved testing of vcvtqq2ps/vcvtuqq2ps implicit zeroing of upper...

Ensure that masked instruction doesn't assume implicit zeroing.

llvm-svn: 288209

diff --git a/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
index c6ab1c3..0d6cc19 100644 (file)
--- a/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
@@ -182,14 +182,17 @@ define <4 x float>@test_int_x86_avx512_mask_cvt_qq2ps_128_zext(<2 x i64> %x0, <4
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovb %edi, %k1 ## encoding: [0xc5,0xf9,0x92,0xcf]
 ; CHECK-NEXT:    vcvtqq2ps %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0xfc,0x09,0x5b,0xc8]
+; CHECK-NEXT:    vmovq %xmm1, %xmm1 ## encoding: [0x62,0xf1,0xfe,0x08,0x7e,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],zero
 ; CHECK-NEXT:    vcvtqq2ps %xmm0, %xmm0 ## encoding: [0x62,0xf1,0xfc,0x08,0x5b,0xc0]
 ; CHECK-NEXT:    vaddps %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x74,0x08,0x58,0xc0]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 %x2)
-  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1)
-  %res2 = shufflevector <4 x float> %res1, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
-  %res3 = fadd <4 x float> %res, %res2
-  ret <4 x float> %res3
+  %res1 = shufflevector <4 x float> %res, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %res2 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1)
+  %res3 = shufflevector <4 x float> %res2, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %res4 = fadd <4 x float> %res1, %res3
+  ret <4 x float> %res4
 }
 
 declare <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64>, <4 x float>, i8)
@@ -357,14 +360,17 @@ define <4 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_128_zext(<2 x i64> %x0, <
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovb %edi, %k1 ## encoding: [0xc5,0xf9,0x92,0xcf]
 ; CHECK-NEXT:    vcvtuqq2ps %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0xff,0x09,0x7a,0xc8]
+; CHECK-NEXT:    vmovq %xmm1, %xmm1 ## encoding: [0x62,0xf1,0xfe,0x08,0x7e,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],zero
 ; CHECK-NEXT:    vcvtuqq2ps %xmm0, %xmm0 ## encoding: [0x62,0xf1,0xff,0x08,0x7a,0xc0]
 ; CHECK-NEXT:    vaddps %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x74,0x08,0x58,0xc0]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 %x2)
-  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1)
-  %res2 = shufflevector <4 x float> %res1, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
-  %res3 = fadd <4 x float> %res, %res2
-  ret <4 x float> %res3
+  %res1 = shufflevector <4 x float> %res, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %res2 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1)
+  %res3 = shufflevector <4 x float> %res2, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %res4 = fadd <4 x float> %res1, %res3
+  ret <4 x float> %res4
 }
 
 declare <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64>, <4 x float>, i8)