[X86][AVX512BF16] Add a few missing insert/extract patterns

These are really the same as the f16 (and i16) instructions, but we need
them for any type that can occur.

2 files changed, 33 insertions, 0 deletions

diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index b610560..a76561f0 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -494,6 +494,8 @@ defm : vinsert_for_size_lowering<"VINSERTI32x4Z256", v16i8x_info, v32i8x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
 defm : vinsert_for_size_lowering<"VINSERTF32x4Z256", v8f16x_info, v16f16x_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
+defm : vinsert_for_size_lowering<"VINSERTF32x4Z256", v8bf16x_info, v16bf16x_info,
+              vinsert128_insert, INSERT_get_vinsert128_imm, [HasVLX]>;
 // Codegen pattern with the alternative types insert VEC128 into VEC512
 defm : vinsert_for_size_lowering<"VINSERTI32x4Z", v8i16x_info, v32i16_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
@@ -501,6 +503,8 @@ defm : vinsert_for_size_lowering<"VINSERTI32x4Z", v16i8x_info, v64i8_info,
                vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
 defm : vinsert_for_size_lowering<"VINSERTF32x4Z", v8f16x_info, v32f16_info,
               vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
+defm : vinsert_for_size_lowering<"VINSERTF32x4Z", v8bf16x_info, v32bf16_info,
+              vinsert128_insert, INSERT_get_vinsert128_imm, [HasAVX512]>;
 // Codegen pattern with the alternative types insert VEC256 into VEC512
 defm : vinsert_for_size_lowering<"VINSERTI64x4Z", v16i16x_info, v32i16_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
@@ -508,6 +512,8 @@ defm : vinsert_for_size_lowering<"VINSERTI64x4Z", v32i8x_info, v64i8_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
 defm : vinsert_for_size_lowering<"VINSERTF64x4Z", v16f16x_info, v32f16_info,
               vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
+defm : vinsert_for_size_lowering<"VINSERTF64x4Z", v16bf16x_info, v32bf16_info,
+              vinsert256_insert, INSERT_get_vinsert256_imm, [HasAVX512]>;
 
 
 multiclass vinsert_for_mask_cast<string InstrStr, X86VectorVTInfo From,
@@ -795,6 +801,8 @@ defm : vextract_for_size_lowering<"VEXTRACTI32x4Z256", v32i8x_info, v16i8x_info,
           vextract128_extract, EXTRACT_get_vextract128_imm, [HasVLX]>;
 defm : vextract_for_size_lowering<"VEXTRACTF32x4Z256", v16f16x_info, v8f16x_info,
           vextract128_extract, EXTRACT_get_vextract128_imm, [HasVLX]>;
+defm : vextract_for_size_lowering<"VEXTRACTF32x4Z256", v16bf16x_info, v8bf16x_info,
+          vextract128_extract, EXTRACT_get_vextract128_imm, [HasVLX]>;
 
 // Codegen pattern with the alternative types extract VEC128 from VEC512
 defm : vextract_for_size_lowering<"VEXTRACTI32x4Z", v32i16_info, v8i16x_info,
@@ -803,6 +811,8 @@ defm : vextract_for_size_lowering<"VEXTRACTI32x4Z", v64i8_info, v16i8x_info,
                  vextract128_extract, EXTRACT_get_vextract128_imm, [HasAVX512]>;
 defm : vextract_for_size_lowering<"VEXTRACTF32x4Z", v32f16_info, v8f16x_info,
                  vextract128_extract, EXTRACT_get_vextract128_imm, [HasAVX512]>;
+defm : vextract_for_size_lowering<"VEXTRACTF32x4Z", v32bf16_info, v8bf16x_info,
+                 vextract128_extract, EXTRACT_get_vextract128_imm, [HasAVX512]>;
 // Codegen pattern with the alternative types extract VEC256 from VEC512
 defm : vextract_for_size_lowering<"VEXTRACTI64x4Z", v32i16_info, v16i16x_info,
                  vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
@@ -810,6 +820,8 @@ defm : vextract_for_size_lowering<"VEXTRACTI64x4Z", v64i8_info, v32i8x_info,
                  vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
 defm : vextract_for_size_lowering<"VEXTRACTF64x4Z", v32f16_info, v16f16x_info,
                  vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
+defm : vextract_for_size_lowering<"VEXTRACTF64x4Z", v32bf16_info, v16bf16x_info,
+                 vextract256_extract, EXTRACT_get_vextract256_imm, [HasAVX512]>;
 
 
 // A 128-bit extract from bits [255:128] of a 512-bit vector should use a
diff --git a/llvm/test/CodeGen/X86/bfloat.ll b/llvm/test/CodeGen/X86/bfloat.ll
index 0042d47..6e3020d 100644
--- a/llvm/test/CodeGen/X86/bfloat.ll
+++ b/llvm/test/CodeGen/X86/bfloat.ll
@@ -2805,3 +2805,24 @@ define <16 x bfloat> @concat_zero_v8bf16(<8 x bfloat> %x, <8 x bfloat> %y) {
   %a = shufflevector <8 x bfloat> %x, <8 x bfloat> zeroinitializer, <16 x i32> <i32 0, 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>
   ret <16 x bfloat> %a
 }
+
+define <16 x bfloat> @concat_dup_v8bf16(<8 x bfloat> %x, <8 x bfloat> %y) {
+; X86-LABEL: concat_dup_v8bf16:
+; X86:       # %bb.0:
+; X86-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X86-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; X86-NEXT:    retl
+;
+; SSE2-LABEL: concat_dup_v8bf16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0,0]
+; SSE2-NEXT:    retq
+;
+; AVX-LABEL: concat_dup_v8bf16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT:    retq
+  %a = shufflevector <8 x bfloat> %x, <8 x bfloat> %y, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <16 x bfloat> %a
+}