From b059941e416b7c44eb027791055b9f0accf45f1b Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@gmail.com>
Date: Wed, 4 Jan 2017 07:31:59 +0000
Subject: [PATCH] [AVX-512] Add more test cases for shuffles that should be
 handled with subvector insert instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@290945 91177308-0d34-0410-b5e6-96231b3b80d8
---
 test/CodeGen/X86/vector-shuffle-512-v16.ll |  54 ++++++
 test/CodeGen/X86/vector-shuffle-512-v8.ll  | 212 +++++++++++++++++++++
 2 files changed, 266 insertions(+)

diff --git a/test/CodeGen/X86/vector-shuffle-512-v16.ll b/test/CodeGen/X86/vector-shuffle-512-v16.ll
index 9075df41c6f..d11c789dfaf 100644
--- a/test/CodeGen/X86/vector-shuffle-512-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v16.ll
@@ -495,3 +495,57 @@ define <16 x i32> @test_vshufi32x4_512_mask(<16 x i32> %x, <16 x i32> %x1, <16 x
   %res = select <16 x i1> %mask, <16 x i32> %x2, <16 x i32> %y
   ret <16 x i32> %res
 }
+
+define <16 x float> @mask_shuffle_v16f32_00_01_02_03_04_05_06_07_16_17_18_19_20_21_22_23(<16 x float> %a, <16 x float> %b, <16 x float> %passthru, i16 %mask) {
+; ALL-LABEL: mask_shuffle_v16f32_00_01_02_03_04_05_06_07_16_17_18_19_20_21_22_23:
+; ALL:       # BB#0:
+; ALL-NEXT:    kmovw %edi, %k1
+; ALL-NEXT:    vinsertf32x8 $1, %ymm1, %zmm0, %zmm2 {%k1}
+; ALL-NEXT:    vmovaps %zmm2, %zmm0
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+  %mask.cast = bitcast i16 %mask to <16 x i1>
+  %res = select <16 x i1> %mask.cast, <16 x float> %shuffle, <16 x float> %passthru
+  ret <16 x float> %res
+}
+
+define <16 x float> @mask_shuffle_v16f32_00_01_02_03_16_17_18_19_08_09_10_11_12_13_14_15(<16 x float> %a, <16 x float> %b, <16 x float> %passthru, i16 %mask) {
+; ALL-LABEL: mask_shuffle_v16f32_00_01_02_03_16_17_18_19_08_09_10_11_12_13_14_15:
+; ALL:       # BB#0:
+; ALL-NEXT:    vmovapd {{.*#+}} zmm3 = [0,1,8,9,4,5,6,7]
+; ALL-NEXT:    vpermi2pd %zmm1, %zmm0, %zmm3
+; ALL-NEXT:    kmovw %edi, %k1
+; ALL-NEXT:    vblendmps %zmm3, %zmm2, %zmm0 {%k1}
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 16, i32 17, i32 18, i32 19, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %mask.cast = bitcast i16 %mask to <16 x i1>
+  %res = select <16 x i1> %mask.cast, <16 x float> %shuffle, <16 x float> %passthru
+  ret <16 x float> %res
+}
+
+define <16 x i32> @mask_shuffle_v16i32_00_01_02_03_04_05_06_07_16_17_18_19_20_21_22_23(<16 x i32> %a, <16 x i32> %b, <16 x i32> %passthru, i16 %mask) {
+; ALL-LABEL: mask_shuffle_v16i32_00_01_02_03_04_05_06_07_16_17_18_19_20_21_22_23:
+; ALL:       # BB#0:
+; ALL-NEXT:    kmovw %edi, %k1
+; ALL-NEXT:    vinserti32x8 $1, %ymm1, %zmm0, %zmm2 {%k1}
+; ALL-NEXT:    vmovdqa64 %zmm2, %zmm0
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+  %mask.cast = bitcast i16 %mask to <16 x i1>
+  %res = select <16 x i1> %mask.cast, <16 x i32> %shuffle, <16 x i32> %passthru
+  ret <16 x i32> %res
+}
+
+define <16 x i32> @mask_shuffle_v16i32_00_01_02_03_16_17_18_19_08_09_10_11_12_13_14_15(<16 x i32> %a, <16 x i32> %b, <16 x i32> %passthru, i16 %mask) {
+; ALL-LABEL: mask_shuffle_v16i32_00_01_02_03_16_17_18_19_08_09_10_11_12_13_14_15:
+; ALL:       # BB#0:
+; ALL-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,8,9,4,5,6,7]
+; ALL-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
+; ALL-NEXT:    kmovw %edi, %k1
+; ALL-NEXT:    vpblendmd %zmm3, %zmm2, %zmm0 {%k1}
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 16, i32 17, i32 18, i32 19, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %mask.cast = bitcast i16 %mask to <16 x i1>
+  %res = select <16 x i1> %mask.cast, <16 x i32> %shuffle, <16 x i32> %passthru
+  ret <16 x i32> %res
+}
diff --git a/test/CodeGen/X86/vector-shuffle-512-v8.ll b/test/CodeGen/X86/vector-shuffle-512-v8.ll
index 625681dc294..485c197d514 100644
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -2375,3 +2375,215 @@ define <8 x i64> @maskz_shuffle_v8i64_12345670(<8 x i64> %a, i8 %mask) {
   %res = select <8 x i1> %mask.cast, <8 x i64> %shuffle, <8 x i64> zeroinitializer
   ret <8 x i64> %res
 }
+
+define <8 x double> @shuffle_v8f64_012389AB(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_012389AB:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_012389AB:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+  ret <8 x double> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_89AB0123(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_89AB0123:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_89AB0123:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 0, i32 1, i32 2, i32 3>
+  ret <8 x double> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_01230123(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_01230123:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_01230123:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+  ret <8 x double> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_012389AB(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_012389AB:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_012389AB:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_89AB0123(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_89AB0123:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_89AB0123:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 0, i32 1, i32 2, i32 3>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_01230123(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_01230123:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_01230123:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_89234567(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_89234567:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovapd {{.*#+}} zmm2 = [8,9,2,3,4,5,6,7]
+; AVX512F-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_89234567:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovapd {{.*#+}} zmm2 = [8,0,9,0,2,0,3,0,4,0,5,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x double> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_01894567(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_01894567:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovapd {{.*#+}} zmm2 = [0,1,8,9,4,5,6,7]
+; AVX512F-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_01894567:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovapd {{.*#+}} zmm2 = [0,0,1,0,8,0,9,0,4,0,5,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x double> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_01238967(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_01238967:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovapd {{.*#+}} zmm2 = [0,1,2,3,8,9,6,7]
+; AVX512F-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_01238967:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovapd {{.*#+}} zmm2 = [0,0,1,0,2,0,3,0,8,0,9,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 6, i32 7>
+  ret <8 x double> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_01234589(<8 x double> %a, <8 x double> %b) {
+; AVX512F-LABEL: shuffle_v8f64_01234589:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovapd {{.*#+}} zmm2 = [0,1,2,3,4,5,8,9]
+; AVX512F-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8f64_01234589:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovapd {{.*#+}} zmm2 = [0,0,1,0,2,0,3,0,4,0,5,0,8,0,9,0]
+; AVX512F-32-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 9>
+  ret <8 x double> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_89234567(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_89234567:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [8,9,2,3,4,5,6,7]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_89234567:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [8,0,9,0,2,0,3,0,4,0,5,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_01894567(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_01894567:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,8,9,4,5,6,7]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_01894567:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,0,1,0,8,0,9,0,4,0,5,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 8, i32 9, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_01238967(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_01238967:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,8,9,6,7]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_01238967:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,0,1,0,2,0,3,0,8,0,9,0,6,0,7,0]
+; AVX512F-32-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 6, i32 7>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_01234589(<8 x i64> %a, <8 x i64> %b) {
+; AVX512F-LABEL: shuffle_v8i64_01234589:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,8,9]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_01234589:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,0,1,0,2,0,3,0,4,0,5,0,8,0,9,0]
+; AVX512F-32-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 9>
+  ret <8 x i64> %shuffle
+}
-- 
2.50.1