[llvm-bugs] [Bug 48640] New: Missed optimization for ((X ashr C1) & C2) == 0

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=48640

            Bug ID: 48640
           Summary: Missed optimization for ((X ashr C1) & C2) == 0
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: david.bolvansky at gmail.com
                CC: llvm-bugs at lists.llvm.org

void foo();

void bad1(int e)
{
   if (((e >> 31) & 64) == 0) foo();
}

void bad2(int e)
{
   if (((e >> 31) & 64) != 0) foo();
}


void ok1(unsigned e)
{
   if (((e >> 31) & 64) == 0) foo();
}

void ok2(unsigned e)
{
   if (((e >> 31) & 64) != 0) foo();
}


LLVM:
bad1(int): # @bad1(int)
  shr edi, 25
  test dil, 64
  jne .LBB0_1
  jmp foo() # TAILCALL
.LBB0_1:
  ret
bad2(int): # @bad2(int)
  shr edi, 25
  test dil, 64
  jne .LBB1_2
  ret
.LBB1_2:
  jmp foo() # TAILCALL
ok1(unsigned int): # @ok1(unsigned int)
  jmp foo() # TAILCALL
ok2(unsigned int): # @ok2(unsigned int)
  ret


GCC:
bad1(int):
        test    edi, edi
        jns     .L4
        ret
.L4:
        jmp     foo()
bad2(int):
        test    edi, edi
        js      .L7
        ret
.L7:
        jmp     foo()
ok1(unsigned int):
        jmp     foo()
ok2(unsigned int):
        ret


https://godbolt.org/z/x85bf6


So..
((X ashr C1) & C2) != 0 to X < 0

..and
((X ashr C1) & C2) == 0 to X >= 0?


define i1 @src(i32 %0) {
%1:
  %2 = ashr i32 %0, 31
  %3 = and i32 %2, 64
  %4 = icmp eq i32 %3, 0
  ret i1 %4
}
=>
define i1 @tgt(i32 %0) {
%1:
  %r = icmp sge i32 %0, 0
  ret i1 %r
}
Transformation seems to be correct!


define i1 @src(i32 %0) {
%1:
  %2 = ashr i32 %0, 31
  %3 = and i32 %2, 64
  %4 = icmp ne i32 %3, 0
  ret i1 %4
}
=>
define i1 @tgt(i32 %0) {
%1:
  %r = icmp slt i32 %0, 0
  ret i1 %r
}
Transformation seems to be correct!

-- 
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/20201231/d8b313a7/attachment-0001.html>