[llvm] [NVPTX] Add patterns for fma.relu.{f16|f16x2|bf16|bf16x2} (PR #114977)

[Artem Belevich via llvm-commits]

================
@@ -3917,3 +3904,63 @@ def atomic_thread_fence_seq_cst_cta :
 def atomic_thread_fence_acq_rel_cta :
   NVPTXInst<(outs), (ins), "fence.acq_rel.cta;", []>,
   Requires<[hasPTX<60>, hasSM<70>]>;
+
+def fpimm_pos_neg_zero : FPImmLeaf<fAny, [{
----------------
Artem-B wrote:

`fpimm_any_zero` ?

https://github.com/llvm/llvm-project/pull/114977