llvm/test/Transforms/SLPVectorizer/X86/store-chains-shared-values.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 < %s | FileCheck %s

; The shorter chain of stores comes first and shares the stored values with the
; longer one. The longer chain has to be vectorized first, the shorter one then
; stores a part of its vector.

define void @test(ptr %p, ptr %q, ptr %r) {
; CHECK-LABEL: define void @test(
; CHECK-SAME: ptr [[P:%.*]], ptr [[Q:%.*]], ptr [[R:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i64>, ptr [[R]], align 8
; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i64> [[TMP0]], <i64 1, i64 2, i64 3, i64 4>
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i64> [[TMP1]], <4 x i64> poison, <2 x i32> <i32 0, i32 1>
; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[P]], align 8
; CHECK-NEXT:    store <4 x i64> [[TMP1]], ptr [[Q]], align 8
; CHECK-NEXT:    ret void
;
entry:
  %r1 = getelementptr inbounds i64, ptr %r, i64 1
  %r2 = getelementptr inbounds i64, ptr %r, i64 2
  %r3 = getelementptr inbounds i64, ptr %r, i64 3
  %t0 = load i64, ptr %r, align 8
  %t1 = load i64, ptr %r1, align 8
  %t2 = load i64, ptr %r2, align 8
  %t3 = load i64, ptr %r3, align 8
  %a0 = add i64 %t0, 1
  %a1 = add i64 %t1, 2
  %a2 = add i64 %t2, 3
  %a3 = add i64 %t3, 4
  %p1 = getelementptr inbounds i64, ptr %p, i64 1
  store i64 %a0, ptr %p, align 8
  store i64 %a1, ptr %p1, align 8
  %q1 = getelementptr inbounds i64, ptr %q, i64 1
  %q2 = getelementptr inbounds i64, ptr %q, i64 2
  %q3 = getelementptr inbounds i64, ptr %q, i64 3
  store i64 %a0, ptr %q, align 8
  store i64 %a1, ptr %q1, align 8
  store i64 %a2, ptr %q2, align 8
  store i64 %a3, ptr %q3, align 8
  ret void
}