[AVX-512] Add test cases to show missed opportunities for using VALIGND/Q to handle shuffles.

llvm-svn: 286425
2026-01-26 12:26:52 +08:00 · 2016-11-10 03:39:19 +00:00
parent 85f9e8b316
commit 924c5ec472
4 changed files with 145 additions and 0 deletions
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -1445,3 +1445,46 @@ define <4 x i64> @insert_dup_mem_v4i64(i64* %ptr) {
  %tmp2 = shufflevector <2 x i64> %tmp1, <2 x i64> undef, <4 x i32> zeroinitializer
  ret <4 x i64> %tmp2
 }
+
+define <4 x i64> @shuffle_v4i64_1234(<4 x i64> %a, <4 x i64> %b) {
+; AVX1-LABEL: shuffle_v4i64_1234:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shuffle_v4i64_1234:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v4i64_1234:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,0]
+; AVX512VL-NEXT:    retq
+  %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
+  ret <4 x i64> %shuffle
+}
+
+define <4 x i64> @shuffle_v4i64_1230(<4 x i64> %a) {
+; AVX1-LABEL: shuffle_v4i64_1230:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shuffle_v4i64_1230:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,0]
+; AVX2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v4i64_1230:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,0]
+; AVX512VL-NEXT:    retq
+  %shuffle = shufflevector <4 x i64> %a, <4 x i64> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
+  ret <4 x i64> %shuffle
+}
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -2543,3 +2543,51 @@ define <8 x i32> @insert_dup_mem_v8i32(i32* %ptr) {
  %tmp2 = shufflevector <4 x i32> %tmp1, <4 x i32> undef, <8 x i32> zeroinitializer
  ret <8 x i32> %tmp2
 }
+
+define <8 x i32> @shuffle_v8i32_12345678(<8 x i32> %a, <8 x i32> %b) {
+; AVX1-LABEL: shuffle_v8i32_12345678:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shuffle_v8i32_12345678:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7]
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,2,3,4,5,6,7,0]
+; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i32_12345678:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vmovdqa32 {{.*#+}} ymm2 = [1,2,3,4,5,6,7,8]
+; AVX512VL-NEXT:    vpermt2d %ymm1, %ymm2, %ymm0
+; AVX512VL-NEXT:    retq
+  %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+  ret <8 x i32> %shuffle
+}
+
+define <8 x i32> @shuffle_v8i32_12345670(<8 x i32> %a) {
+; AVX1-LABEL: shuffle_v8i32_12345670:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shuffle_v8i32_12345670:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,2,3,4,5,6,7,0]
+; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i32_12345670:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vmovdqa32 {{.*#+}} ymm1 = [1,2,3,4,5,6,7,0]
+; AVX512VL-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX512VL-NEXT:    retq
+  %shuffle = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0>
+  ret <8 x i32> %shuffle
+}
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
@@ -338,3 +338,23 @@ define <16 x i32> @shuffle_v16i32_16_zz_17_zz_18_zz_19_zz_20_zz_21_zz_22_zz_23_z
  %shuffle = shufflevector <16 x i32> zeroinitializer, <16 x i32> %a, <16 x i32> <i32 16, i32 0, i32 17, i32 0, i32 18, i32 0, i32 19, i32 0, i32 20, i32 0, i32 21, i32 0, i32 22, i32 0, i32 23, i32 0>
  ret <16 x i32> %shuffle
 }
+
+define <16 x i32> @shuffle_v16i32_01_02_03_04_05_06_07_08_09_10_11_12_13_14_15_16(<16 x i32> %a, <16 x i32> %b) {
+; ALL-LABEL: shuffle_v16i32_01_02_03_04_05_06_07_08_09_10_11_12_13_14_15_16:
+; ALL:       # BB#0:
+; ALL-NEXT:    vmovdqa32 {{.*#+}} zmm2 = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
+; ALL-NEXT:    vpermt2d %zmm1, %zmm2, %zmm0
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32><i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
+  ret <16 x i32> %shuffle
+}
+
+define <16 x i32> @shuffle_v16i32_01_02_03_04_05_06_07_08_09_10_11_12_13_14_15_00(<16 x i32> %a) {
+; ALL-LABEL: shuffle_v16i32_01_02_03_04_05_06_07_08_09_10_11_12_13_14_15_00:
+; ALL:       # BB#0:
+; ALL-NEXT:    vmovdqa32 {{.*#+}} zmm1 = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,0]
+; ALL-NEXT:    vpermd %zmm0, %zmm1, %zmm0
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <16 x i32> %a, <16 x i32> undef, <16 x i32><i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 0>
+  ret <16 x i32> %shuffle
+}
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -2275,3 +2275,37 @@ define <8 x double> @shuffle_v8f64_0zzzzzzz(<8 x double> %a) {
  %shuffle = shufflevector <8 x double> %a, <8 x double> zeroinitializer, <8 x i32> <i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
  ret <8 x double> %shuffle
 }
+
+define <8 x i64> @shuffle_v8i64_12345678(<8 x i64> %a, <8 x i64> %b) {
+;
+; AVX512F-LABEL: shuffle_v8i64_12345678:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [1,2,3,4,5,6,7,8]
+; AVX512F-NEXT:    vpermt2q %zmm1, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_12345678:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [1,0,2,0,3,0,4,0,5,0,6,0,7,0,8,0]
+; 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 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_12345670(<8 x i64> %a) {
+;
+; AVX512F-LABEL: shuffle_v8i64_12345670:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [1,2,3,4,5,6,7,0]
+; AVX512F-NEXT:    vpermq %zmm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512F-32-LABEL: shuffle_v8i64_12345670:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [1,0,2,0,3,0,4,0,5,0,6,0,7,0,0,0]
+; AVX512F-32-NEXT:    vpermq %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> undef, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0>
+  ret <8 x i64> %shuffle
+}