test/Transforms/PhaseOrdering/inline-store-to-load.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S -passes="default<O3>" < %s | FileCheck %s

; Test that the CGSCC inliner forwards stores to load arguments after
; inlining. This addresses a phase ordering issue: the constructor is
; inlined first (producing stores), and then the destructor's load
; arguments should see the stored constants. Without this forwarding,
; the recursive erase function appears too expensive to inline.
;
; This models the std::set<int>{} pattern where the constructor stores
; null to the root pointer and the destructor checks it before recursive
; tree deletion (_Rb_tree::_M_erase).

target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"

declare void @free(ptr)
declare void @use(i64)

; Models _Rb_tree constructor: stores null root + zero node count.
define internal void @ctor(ptr %this) {
  store ptr null, ptr %this, align 8
  %count = getelementptr inbounds i8, ptr %this, i64 8
  store i64 0, ptr %count, align 8
  ret void
}

; Models ~_Rb_tree: loads root and calls the recursive eraser.
define internal void @dtor(ptr %this) {
  %root = load ptr, ptr %this, align 8
  call void @erase(ptr %root)
  ret void
}

; Models _Rb_tree::_M_erase — recursive node deletion.
; With null: icmp + branch to done → trivially cheap.
; With unknown ptr: recursive calls + external calls → too expensive.
define internal void @erase(ptr %node) {
  %is_null = icmp eq ptr %node, null
  br i1 %is_null, label %done, label %recurse

recurse:
  %left_val = load ptr, ptr %node, align 8
  call void @erase(ptr %left_val)
  %right_ptr = getelementptr inbounds i8, ptr %node, i64 8
  %right_val = load ptr, ptr %right_ptr, align 8
  call void @erase(ptr %right_val)
  %d1 = getelementptr inbounds i8, ptr %node, i64 16
  %v1 = load i64, ptr %d1, align 8
  call void @use(i64 %v1)
  %d2 = getelementptr inbounds i8, ptr %node, i64 24
  %v2 = load i64, ptr %d2, align 8
  call void @use(i64 %v2)
  %d3 = getelementptr inbounds i8, ptr %node, i64 32
  %v3 = load i64, ptr %d3, align 8
  call void @use(i64 %v3)
  %d4 = getelementptr inbounds i8, ptr %node, i64 40
  %v4 = load i64, ptr %d4, align 8
  call void @use(i64 %v4)
  %d5 = getelementptr inbounds i8, ptr %node, i64 48
  %v5 = load i64, ptr %d5, align 8
  call void @use(i64 %v5)
  call void @free(ptr %node)
  br label %done

done:
  ret void
}

define void @test_empty_tree() {
; CHECK-LABEL: define void @test_empty_tree(
; CHECK-SAME: ) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; CHECK-NEXT:    ret void
;
  %tree = alloca ptr, align 8
  call void @ctor(ptr %tree)
  call void @dtor(ptr %tree)
  ret void
}