[PATCH] D49874: AMDGPU: Add clamp bit to dot intrinsics

[Konstantin Zhuravlyov via Phabricator via llvm-commits]

kzhuravl added inline comments.


================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:7550
+      return DAG.getNode(AMDGPUISD::FDOT2, SL, MVT::f32, Vec1, Vec2, FMAAcc,
+                         DAG.getConstant(0, SL, MVT::i1));
+    }
----------------
arsenm wrote:
> should be getTargetConstant. Is there any reason we also can't match clamp here?
> should be getTargetConstant

Done.

> Is there any reason we also can't match clamp here?

This goes out of scope of the ticket. Can this be done in a separate change?


https://reviews.llvm.org/D49874