[PATCH] D39011: [SimplifyCFG] try harder to forward switch condition to phi (PR34471)

[Daniel Berlin via llvm-commits]

result of opt -gvn on pr34471.ll:
; ModuleID = 'foo.ll'
source_filename = "foo.ll"

define i32 @PR34471(i32 %x) {
entry:
  switch i32 %x, label %else3 [
    i32 17, label %return
    i32 19, label %if19
    i32 42, label %if42
  ]

if19:                                             ; preds = %entry
  br label %return

if42:                                             ; preds = %entry
  br label %return

else3:                                            ; preds = %entry
  br label %return

return:                                           ; preds = %else3, %if42,
%if19, %entry
  %r = phi i32 [ 19, %if19 ], [ 42, %if42 ], [ 0, %else3 ], [ 17, %entry ]
  ret i32 %r
}

-correlated-propagation gives the same answer, as do -newgvn and a few
other passes.



On Sun, Oct 22, 2017 at 10:06 AM, Daniel Berlin via Phabricator <
reviews at reviews.llvm.org> wrote:

> dberlin added a comment.
>
> Hey Sanjay,
> GVN/NewGVN/et al will undo this transform, and propagate constants. In
> fact, i expect pretty much any pass that can, will.
>
> This definitely belongs in the late simplification category at best.
>
> More than that, i'm concerned this is the wrong fix.
>
> Why is the right answer  to undo the transform (which will cause fights
> between passes)  in simplifycfg instead of improving the codegen to
> recognize the issue?
>
> It seems the same code you are using to detect and undo the transforrm in
> simplification could be used to codegen it better.
>
>
> Repository:
>   rL LLVM
>
> https://reviews.llvm.org/D39011
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20171022/125f6233/attachment.html>