[PATCH] D32056: AMDGPU: Move trap lowering to DAG

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Apr 24 10:01:35 PDT 2017


arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:2404
+                                     Op.getDebugLoc(),
+                                     DS_Warning);
+    LLVMContext &Ctx = MF.getFunction()->getContext();
----------------
kzhuravl wrote:
> t-tye wrote:
> > arsenm wrote:
> > > kzhuravl wrote:
> > > > Can we make it an error?
> > > We decided this should be a warning before, this isn't changing that
> > If there is no trap handler (namely it is not amdhsa) then the code generated is simply ignoring the llvm.debugtrap. So in this case is it better to give a warning or an error? If the llvm.debugtrap was present for a reason then ignoring it without giving an error may not be very helpful.
> was the question answered?
Changing the behavior is beyond the scope of this verifier error fixing patch


https://reviews.llvm.org/D32056





More information about the llvm-commits mailing list