1 files changed, 28 insertions, 39 deletions
diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-512.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-512.ll
index 4ba208af470..9e6ac08ca0b 100644
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-512.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-512.ll
@@ -241,49 +241,38 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) nounwind {
 define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) nounwind {
 ; AVX512DQ-LABEL: constant_shift_v64i8:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vpsrlw $4, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX512DQ-NEXT:    vpand %ymm3, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [8192,24640,41088,57536,49376,32928,16480,32,8192,24640,41088,57536,49376,32928,16480,32]
-; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm2, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpsrlw $2, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm5 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
-; AVX512DQ-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpaddb %ymm4, %ymm4, %ymm6
-; AVX512DQ-NEXT:    vpblendvb %ymm6, %ymm2, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpsrlw $1, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm7 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512DQ-NEXT:    vpand %ymm7, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpaddb %ymm6, %ymm6, %ymm8
-; AVX512DQ-NEXT:    vpblendvb %ymm8, %ymm2, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpsrlw $4, %ymm1, %ymm2
-; AVX512DQ-NEXT:    vpand %ymm3, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm2, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpsrlw $2, %ymm1, %ymm2
-; AVX512DQ-NEXT:    vpand %ymm5, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendvb %ymm6, %ymm2, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpsrlw $1, %ymm1, %ymm2
-; AVX512DQ-NEXT:    vpand %ymm7, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendvb %ymm8, %ymm2, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11],ymm0[12],ymm2[12],ymm0[13],ymm2[13],ymm0[14],ymm2[14],ymm0[15],ymm2[15],ymm0[24],ymm2[24],ymm0[25],ymm2[25],ymm0[26],ymm2[26],ymm0[27],ymm2[27],ymm0[28],ymm2[28],ymm0[29],ymm2[29],ymm0[30],ymm2[30],ymm0[31],ymm2[31]
+; AVX512DQ-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [2,4,8,16,32,64,128,256,2,4,8,16,32,64,128,256]
+; AVX512DQ-NEXT:    # ymm4 = mem[0,1,0,1]
+; AVX512DQ-NEXT:    vpmullw %ymm4, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[4],ymm2[4],ymm0[5],ymm2[5],ymm0[6],ymm2[6],ymm0[7],ymm2[7],ymm0[16],ymm2[16],ymm0[17],ymm2[17],ymm0[18],ymm2[18],ymm0[19],ymm2[19],ymm0[20],ymm2[20],ymm0[21],ymm2[21],ymm0[22],ymm2[22],ymm0[23],ymm2[23]
+; AVX512DQ-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [256,128,64,32,16,8,4,2,256,128,64,32,16,8,4,2]
+; AVX512DQ-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX512DQ-NEXT:    vpmullw %ymm5, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpackuswb %ymm3, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
+; AVX512DQ-NEXT:    vpmullw %ymm4, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
+; AVX512DQ-NEXT:    vpmullw %ymm5, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vpackuswb %ymm3, %ymm1, %ymm1
 ; AVX512DQ-NEXT:    retq
 ;
 ; AVX512BW-LABEL: constant_shift_v64i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [8192,24640,41088,57536,49376,32928,16480,32,8192,24640,41088,57536,49376,32928,16480,32,8192,24640,41088,57536,49376,32928,16480,32,8192,24640,41088,57536,49376,32928,16480,32]
-; AVX512BW-NEXT:    vpmovb2m %zmm1, %k1
-; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm2
-; AVX512BW-NEXT:    vpandq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512BW-NEXT:    vmovdqu8 %zmm2, %zmm0 {%k1}
-; AVX512BW-NEXT:    vpsrlw $2, %zmm0, %zmm2
-; AVX512BW-NEXT:    vpandq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512BW-NEXT:    vpaddb %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpmovb2m %zmm1, %k1
-; AVX512BW-NEXT:    vmovdqu8 %zmm2, %zmm0 {%k1}
-; AVX512BW-NEXT:    vpsrlw $1, %zmm0, %zmm2
-; AVX512BW-NEXT:    vpandq {{.*}}(%rip), %zmm2, %zmm2
-; AVX512BW-NEXT:    vpaddb %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpmovb2m %zmm1, %k1
-; AVX512BW-NEXT:    vmovdqu8 %zmm2, %zmm0 {%k1}
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm1 = zmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
+; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpsllvw {{.*}}(%rip), %zmm1, %zmm1
+; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm0 = zmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllvw {{.*}}(%rip), %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift