[Mlir-commits] [mlir] [MLIR][NVVM][Refactor] Refactor intrinsic lowering for NVVM Ops (PR #157079)

[Srinivasa Ravi]

================
@@ -19,14 +19,14 @@ llvm.func @tma_store_reduce_1d(%src : !llvm.ptr<3>, %tma_desc : !llvm.ptr, %d0 :
   nvvm.cp.async.bulk.tensor.reduce %tma_desc, %src, box[%d0] l2_cache_hint = %ch {redKind = #nvvm.tma_redux_kind<or>}  : !llvm.ptr, !llvm.ptr<3>
   nvvm.cp.async.bulk.tensor.reduce %tma_desc, %src, box[%d0] l2_cache_hint = %ch {redKind = #nvvm.tma_redux_kind<xor>} : !llvm.ptr, !llvm.ptr<3>
 
-  // CHECK: call void @llvm.nvvm.cp.async.bulk.tensor.reduce.add.tile.1d(ptr addrspace(3) %[[SRC]], ptr %[[DST]], i32 %[[D0]], i64 undef, i1 false)
----------------
Wolfram70 wrote:

Since `undef` is deprecated, while moving the code out of tablegen into C++, the CI complained that it should be replaced. So, I made commit https://github.com/llvm/llvm-project/pull/157079/commits/84b8e768329235b7c4d3d0188b7ef643b0be69db replacing it with `poison`.

https://github.com/llvm/llvm-project/pull/157079