[PATCH] [IC] Turn non-null MD on pointer loads to range MD on integer loads.

Charles Davis cdavis5x at gmail.com
Tue Feb 24 12:29:10 PST 2015


================
Comment at: test/Transforms/InstCombine/loadstore-metadata.ll:98
@@ -99,2 +97,3 @@
 
+; CHECK: ![[MD]] = !{i64 1, i64 -1}
 !0 = !{ !1, !1, i64 0 }
----------------
chandlerc wrote:
> cdavis5x wrote:
> > chandlerc wrote:
> > > Does this work even if the above test case isn't the last one in the file? I wonder if we need to make the IR printing denormalize this somewhat for easier checking.
> > Probably not, given that it's a plain ol' `CHECK` line. If any more `CHECK` lines are added after this, they will fail, because `FileCheck` will have already scanned past them.
> > 
> > Frankly, I was kind of worried about that when I wrote this. (I should've brought it up then...) I considered using `CHECK-DAG`, but I think that only works between the previous and next `CHECK-LABEL` lines. We might need to do what you said and have the IR pretty printer emit MD tuples immediately after the function that uses them (for instance) rather than always at the end of the file.
> Are you planning to look into that?
> 
> Either way, can you leave a comment here to help the next person to debug this?
This doesn't seem to be pressing. Perhaps later.

I've left a FIXME in the code for now.

http://reviews.llvm.org/D7621

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the llvm-commits mailing list