[clang] [llvm] [AMDGPU] Add support for `v_tanh_bf16` on gfx1250 (PR #147425)

[Shilei Tian via llvm-commits]

================
@@ -13658,6 +13658,7 @@ bool SITargetLowering::isCanonicalized(Register Reg, const MachineFunction &MF,
     case Intrinsic::amdgcn_frexp_mant:
     case Intrinsic::amdgcn_fdot2:
     case Intrinsic::amdgcn_trig_preop:
+    case Intrinsic::amdgcn_tanh:
----------------
shiltian wrote:

We can't do it at this moment because the support for `v_cvt_pk_bf16_f32` on gfx1250 has not been upstreamed. Without it `llvm.canonicalize.bf16` doesn't work.

I'll add the test to downstream though.

https://github.com/llvm/llvm-project/pull/147425