[Mlir-commits] [mlir] [MLIR][Conversion] Add convert-xevm-to-llvm pass. (PR #147375)

Adam Siemieniuk llvmlistbot at llvm.org
Thu Jul 10 13:47:27 PDT 2025


================
@@ -0,0 +1,244 @@
+// RUN: mlir-opt --convert-xevm-to-llvm --split-input-file %s | FileCheck %s
+
+// Same below, but using the `ConvertToLLVMPatternInterface` entry point
+// and the generic `convert-to-llvm` pass.
+// RUN: mlir-opt --convert-to-llvm --split-input-file %s | FileCheck %s
+
+// CHECK-LABEL:      llvm.func spir_funccc @_Z41intel_sub_group_2d_block_read_16b_8r16x1cPU3AS1viiiDv2_iPt(
+// CHECK-SAME:   !llvm.ptr<1> {llvm.nonnull, llvm.readonly}, i32, i32, i32, vector<2xi32>,
+// CHECK-SAME:   !llvm.ptr {llvm.nonnull, llvm.writeonly}) attributes {no_unwind, will_return}
+// CHECK:      llvm.func @blockload2d(%[[ARG0:.*]]: !llvm.ptr<1>,
+// CHECK-SAME:   %[[ARG1:.*]]: i32, %[[ARG2:.*]]: i32, %[[ARG3:.*]]: i32, %[[ARG4:.*]]: i32, %[[ARG5:.*]]: i32)
+llvm.func @blockload2d(%a: !llvm.ptr<1>, %base_width_a: i32, %base_height_a: i32, %base_pitch_a: i32, %x: i32, %y: i32) -> vector<8xi16> {
----------------
adam-smnk wrote:

nit: it'd be also nice to add a bit more context to these test case names to bring focus on what's the difference between them
sth like `blockload2d_v_blocks_2`, `blockload2d_transpose` etc.

https://github.com/llvm/llvm-project/pull/147375


More information about the Mlir-commits mailing list