[Mlir-commits] [mlir] [mlir][mpi] Lowering MPI_Allreduce (PR #133133)

[Jeff Hammond]

================
@@ -94,6 +116,20 @@ class MPICHImplTraits : public MPIImplTraits {
   static constexpr int MPI_UINT16_T = 0x4c00023c;
   static constexpr int MPI_UINT32_T = 0x4c00043d;
   static constexpr int MPI_UINT64_T = 0x4c00083e;
+  static constexpr int MPI_MAX = 0x58000001;
+  static constexpr int MPI_MIN = 0x58000002;
+  static constexpr int MPI_SUM = 0x58000003;
+  static constexpr int MPI_PROD = 0x58000004;
+  static constexpr int MPI_LAND = 0x58000005;
+  static constexpr int MPI_BAND = 0x58000006;
+  static constexpr int MPI_LOR = 0x58000007;
+  static constexpr int MPI_BOR = 0x58000008;
+  static constexpr int MPI_LXOR = 0x58000009;
+  static constexpr int MPI_BXOR = 0x5800000a;
+  static constexpr int MPI_MINLOC = 0x5800000b;
+  static constexpr int MPI_MAXLOC = 0x5800000c;
+  static constexpr int MPI_REPLACE = 0x5800000d;
+  static constexpr int MPI_NO_OP = 0x5800000e;
----------------
jeffhammond wrote:

@mofeing As I said already, MPICH supports the MPI-5 ABI already (you must enable it with configure, as it is not the default).  I will sync Mukautuva to the final MPI-5 ABI in a few days.
https://github.com/llvm/llvm-project/pull/133280#issuecomment-2761189904

https://github.com/llvm/llvm-project/pull/133133