[llvm-bugs] [Bug 35487] New: Poor code generation for rotates

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=35487

            Bug ID: 35487
           Summary: Poor code generation for rotates
           Product: libraries
           Version: 5.0
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: fabiang at radgametools.com
                CC: llvm-bugs at lists.llvm.org

This: https://godbolt.org/g/b2t1a9

unsigned long f(unsigned long x, int amt)
{
    x += (x << amt) | (x >> (64 - amt));
    return x & 0xffffffffu;
}

unsigned long g(unsigned long x, int amt)
{
    x += (x << amt) | (x >> (64 - amt));
    return x;
}

produces:

f(unsigned long, int): # @f(unsigned long, int)
  mov rax, rdi
  mov ecx, esi
  shl rax, cl
  mov ecx, 64
  sub ecx, esi
  mov rdx, rdi
  shr rdx, cl
  or edx, eax
  add edi, edx
  mov rax, rdi
  ret
g(unsigned long, int): # @g(unsigned long, int)
  mov rax, rdi
  mov ecx, esi
  rol rax, cl
  lea rax, [rax + rdi]
  ret

Found while investigating a significant perf difference between LLVM and gcc on
a hash function. (Yes, "amt" is variable in the code in question.)

-- 
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/20171201/3d38130f/attachment.html>