| commit | 12e46a006d731c8a79a99049e770a60a126fd8a0 | |
| author | Mark Williams <mwilliams@fb.com> | |
| Fri, 17 Mar 2017 21:57:40 +0000 (17 14:57 -0700) | ||
| committer | Hhvm Bot <hhvm-bot@users.noreply.github.com> | |
| Mon, 20 Mar 2017 23:32:56 +0000 (20 16:32 -0700) | ||
| tree | f270f94117ebf9dad70196bc4a59636319f85932 | treesnapshot (tar.gz zip) |
| parent | 5bf38b7c2e33d0bf98ef11080e375ec3411d9dc9 | commitdiff |
Deleted instructions don't do anything
Summary:
Even after D4716282 I was seeing SetLs that appeared to be dead, but
weren't deleted. I noticed that CGetL called AddLocGen even when it
was going to be removed - so stop doing that. Auditing the rest of the
AddLocGens turned up one more case.
There are still a few places where we are probably going to delete the
instruction, but have to mark its effects anyway, just in case.
I have more dce diffs that will hopefully give us a way to deal with
that.
Reviewed By: ricklavoie
Differential Revision: D4728909
fbshipit-source-id: c33e812fe4d59090e5bcbcd750b433132154ff28
Summary:
Even after D4716282 I was seeing SetLs that appeared to be dead, but
weren't deleted. I noticed that CGetL called AddLocGen even when it
was going to be removed - so stop doing that. Auditing the rest of the
AddLocGens turned up one more case.
There are still a few places where we are probably going to delete the
instruction, but have to mark its effects anyway, just in case.
I have more dce diffs that will hopefully give us a way to deal with
that.
Reviewed By: ricklavoie
Differential Revision: D4728909
fbshipit-source-id: c33e812fe4d59090e5bcbcd750b433132154ff28
| hphp/hhbbc/dce.cpp | diffblobblamehistory |