[Mlir-commits] [mlir] [MLIR][XeVM] Add XeVM to LLVMIR translation. (PR #150696)

[Adam Siemieniuk]

================
@@ -0,0 +1,101 @@
+// RUN: mlir-translate --split-input-file -mlir-to-llvmir %s | FileCheck %s
+
+module {
+  llvm.func spir_funccc @_Z41intel_sub_group_2d_block_read_16b_8r16x1cPU3AS1viiiDv2_iPt(!llvm.ptr<1> {llvm.nonnull, llvm.readonly}, i32, i32, i32, vector<2xi32>, !llvm.ptr {llvm.nonnull, llvm.writeonly}) attributes {no_unwind, will_return}
+  llvm.func @blockload2d_cache_control(%arg0: !llvm.ptr<1>, %arg1: i32, %arg2: i32, %arg3: i32, %arg4: i32, %arg5: i32) -> vector<8xi16> {
----------------
adam-smnk wrote:

Does the metadata translation itself depend on the operation it's attached to?

If not, could you rename these cases to highlight what's exactly being tested? Maybe only a single or two cases would be sufficient?
Just thinking in terms of test noise reduction here.

https://github.com/llvm/llvm-project/pull/150696