[llvm] [NVPTX] Add support for atomic add for f16 type (PR #84295)

[Adrian Kuegel via llvm-commits]

================
@@ -0,0 +1,17 @@
+; RUN: llc < %s -march=nvptx -mcpu=sm_70 | FileCheck %s
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_70 | FileCheck %s
----------------
akuegel wrote:

You are right, and it also shows the failure in presubmit testing. Sorry for missing that. And thanks for the tip with setting -mattr=+ptx63. I also added tests with -mattr=+ptx62 to show that it is not using atom.add.noftz.f16 in that case.

https://github.com/llvm/llvm-project/pull/84295