# This reproduces a bug with shrink wrapping when trying to move
# push-pops where restores were not validated to be POPs (and could be
# a regular load, which violated our assumptions). Check that we
# identify those cases.
# REQUIRES: system-linux
# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown \
# RUN: %s -o %t.o
# RUN: link_fdata %s %t.o %t.fdata
# Delete our BB symbols so BOLT doesn't mark them as entry points
# RUN: llvm-strip --strip-unneeded %t.o
# RUN: %clang %cflags -no-pie %t.o -o %t.exe -Wl,-q
# RUN: llvm-bolt %t.exe --relocs=1 --frame-opt=all --print-finalized \
# RUN: --print-only=main --data %t.fdata -o %t.out | FileCheck %s
# RUN: %t.out
# CHECK: BOLT-INFO: Shrink wrapping moved 1 spills inserting load/stores and 0 spills inserting push/pops
.text
.globl main
.type main, %function
.p2align 4
main:
# FDATA: 0 [unknown] 0 1 main 0 0 510
pushq %rbp
movq %rsp, %rbp
pushq %rbx # We save rbx here, but there is an
# opportunity to move it to .LBB2
subq $0x18, %rsp
cmpl $0x2, %edi
.J1:
jb .BBend
# FDATA: 1 main #.J1# 1 main #.BB2# 0 10
# FDATA: 1 main #.J1# 1 main #.BBend# 0 500
.BB2:
movq $2, %rbx # Use rbx in a cold block
xorq %rax, %rax
movb mystring, %al
addq %rbx, %rax
movb %al, mystring
leaq mystring, %rdi
# Avoid making the actual call here to allow push-pop mode to operate
# without fear of an unknown function that may require aligned stack
# callq puts
.BBend:
mov -0x08(%rbp), %rbx # Original restore is here. Instead of a pop
# we use a load to reproduce gcc behavior while
# using leave in epilogue. Ordinarily it would
# addq $0x18, %rsp followed by pop %rbx
xorq %rax, %rax
leaveq
retq
.size main, .-main
.data
mystring: .asciz "0 is rbx mod 10 contents in decimal\n"
Codebase Browser
llvm