[llvm-bugs] [Bug 33763] New: Potential code size reduction optimization: reusing function tails

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=33763

            Bug ID: 33763
           Summary: Potential code size reduction optimization: reusing
                    function tails
           Product: clang
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: -New Bugs
          Assignee: unassignedclangbugs at nondot.org
          Reporter: jpakkane at gmail.com
                CC: llvm-bugs at lists.llvm.org

Suppose you have code that looks like this:

int func1();
int func2();
int func3();
int func4();

int func5() {
  int i = 0;
  i+=func2();
  i+=func3();
  i+=func4();
  return i+func1();
}

int func6() {
  int i=1;
  i+=func3();
  i+=func3();
  i+=func4();
  return i+func1();
}

It gets compiled to the following assembly when using -Os:

func5():
        pushq   %rbx
        call    func2()
        movl    %eax, %ebx
        call    func3()
        addl    %eax, %ebx
        call    func4()
        addl    %eax, %ebx
        call    func1()
        addl    %ebx, %eax
        popq    %rbx
        ret
func6():
        pushq   %rbx
        call    func3()
        leal    1(%rax), %ebx
        call    func3()
        addl    %eax, %ebx
        call    func4()
        addl    %eax, %ebx
        call    func1()
        addl    %ebx, %eax
        popq    %rbx
        ret


However the ends of the two functions are identical. This could be compiled
into the following which is functionally the same but takes less space:

func5():
        pushq   %rbx
        call    func2()
        movl    %eax, %ebx
common_tail:
        call    func3()
        addl    %eax, %ebx
        call    func4()
        addl    %eax, %ebx
        call    func1()
        addl    %ebx, %eax
        popq    %rbx
        ret
func6():
        pushq   %rbx
        call    func3()
        leal    1(%rax), %ebx
        jmp common_tail

Testing with compiler explorer says that neither GCC, Clang, MSVC nor ICC do
this optimization but some embedded compilers do.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20170712/7efe23fd/attachment.html>