[llvm] [X86][AVX] Add missing X86ISD::VBROADCAST(v2i64 -> v4i64) isel pattern for AVX1 targets (PR #102853)

[Mészáros Gergely via llvm-commits]

https://github.com/Maetveis created https://github.com/llvm/llvm-project/pull/102853

An internal bug revealed that this form can be formed from existing combines, and then fail to select. Use the same pattern as v2f64 -> v4f64 (the instructions are moving bits around the actual type does not matter here).

>From 20fdbeec021e82c5b71a29637d376276407c755e Mon Sep 17 00:00:00 2001
From: Gergely Meszaros <gergely.meszaros at intel.com>
Date: Fri, 9 Aug 2024 08:05:11 -0700
Subject: [PATCH] [X86][AVX] Add missing X86ISD::VBROADCAST(v2i64 -> v4i64)
 isel pattern for AVX1 targets

An internal bug revealed that this form can be formed from existing
combines, and then fail to select. Add the same pattern as v2f64 -> v4f64
(the instructions are moving bits around the actual type does not matter here).
---
 llvm/lib/Target/X86/X86InstrSSE.td            |  5 ++++
 .../test/CodeGen/X86/combine-concatvectors.ll | 23 +++++++++++++++++++
 2 files changed, 28 insertions(+)

diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 5f9211edfa161b..515e7d80a39637 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -7797,6 +7797,11 @@ let Predicates = [HasAVX1Only] in {
             (VPSHUFDri (VMOV64toPQIrr GR64:$src), 0x44)>;
   def : Pat<(v2i64 (X86VBroadcastld64 addr:$src)),
             (VMOVDDUPrm addr:$src)>;
+
+  def : Pat<(v4i64 (X86VBroadcast v2i64:$src)),
+            (VINSERTF128rr (INSERT_SUBREG (v4i64 (IMPLICIT_DEF)),
+              (v2i64 (VMOVDDUPrr VR128:$src)), sub_xmm),
+              (v2i64 (VMOVDDUPrr VR128:$src)), 1)>;
 }
 
 //===----------------------------------------------------------------------===//
diff --git a/llvm/test/CodeGen/X86/combine-concatvectors.ll b/llvm/test/CodeGen/X86/combine-concatvectors.ll
index 17d22607cfea88..5b06a8091dce7e 100644
--- a/llvm/test/CodeGen/X86/combine-concatvectors.ll
+++ b/llvm/test/CodeGen/X86/combine-concatvectors.ll
@@ -118,3 +118,26 @@ define <4 x float> @concat_of_broadcast_v4f32_v8f32(ptr %a0, ptr %a1, ptr %a2) {
   %shuffle1 = shufflevector <8 x float> %ld2, <8 x float> %shuffle, <4 x i32> <i32 6, i32 15, i32 12, i32 3>
   ret <4 x float> %shuffle1
 }
+
+define <4 x i64> @broadcast_of_shuffle_v2i64_v4i64(<16 x i8> %vecinit.i) {
+; AVX1-LABEL: broadcast_of_shuffle_v2i64_v4i64:
+; AVX1:       # %bb.0: # %entry
+; AVX1-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: broadcast_of_shuffle_v2i64_v4i64:
+; AVX2:       # %bb.0: # %entry
+; AVX2-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastq %xmm0, %ymm0
+; AVX2-NEXT:    retq
+entry:
+  %vecinit15.i = shufflevector <16 x i8> %vecinit.i, <16 x i8> poison, <16 x i32> zeroinitializer
+  %0 = bitcast <16 x i8> %vecinit15.i to <2 x i64>
+  %1 = extractelement <2 x i64> %0, i64 0
+  %2 = and i64 %1, -72057594037927936 ; 0xFF00 0000 0000 0000
+  %3 = insertelement <4 x i64> poison, i64 %2, i64 0
+  %4 = shufflevector <4 x i64> %3, <4 x i64> poison, <4 x i32> zeroinitializer
+  ret <4 x i64> %4
+}