[llvm-bugs] [Bug 40878] New: Failure to reduce indvarsimplify modulo loop output to a simple AND mask

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=40878

            Bug ID: 40878
           Summary: Failure to reduce indvarsimplify modulo loop output to
                    a simple AND mask
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Loop Optimizer
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: greened at obbligato.org, lebedev.ri at gmail.com,
                    llvm-bugs at lists.llvm.org, spatel+llvm at rotateright.com
            Blocks: 38280

Pulled out of [Bug #38280 Comment #1]:

> On a perhaps related note, the very simple example
> (https://godbolt.org/g/MToVLm):
> unsigned func(unsigned count)
> {
>     while (count >= 32)
>         count -= 32;
>     return count;
> }
> 
> Currently produces something like (thanks to indvarsimplify):
> 
> define i32 @test(i32 %size) {
> entry:
>   %0 = xor i32 %size, -1
>   %1 = icmp ugt i32 %0, -32
>   %umax = select i1 %1, i32 %0, i32 -32
>   %2 = add i32 %umax, %size
>   %3 = add i32 %2, 32
>   %4 = and i32 %3, -32
>   %5 = sub i32 %size, %4
>   ret i32 %5
> }
> 
> Which is really just:
> https://rise4fun.com/Alive/lKL
> 
> define i32 @test(i32 %size) {
>   %0 = and i32 %size, 31
>   ret i32 %0
> }


Referenced Bugs:

https://bugs.llvm.org/show_bug.cgi?id=38280
[Bug 38280] Pointless loop unroll / vectorization
-- 
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/20190227/b8d525b6/attachment-0001.html>