[Mlir-commits] [llvm] [mlir] [mlir][spirv] Support `memref` in `convert-to-spirv` pass (PR #102534)

[Jakub Kuderski]

================
@@ -0,0 +1,36 @@
+// RUN: mlir-opt -convert-to-spirv="run-signature-conversion=false run-vector-unrolling=false" -cse -split-input-file %s | FileCheck %s
+
+module attributes {
+  spirv.target_env = #spirv.target_env<
+    #spirv.vce<v1.0, [Shader], [SPV_KHR_storage_buffer_storage_class]>, #spirv.resource_limits<>>
+} {
+
+// CHECK-LABEL: @load_store_float_rank_zero
+//  CHECK-SAME: %[[ARG0:.*]]: !spirv.ptr<!spirv.struct<(!spirv.array<1 x f32, stride=4> [0])>, StorageBuffer>, %[[ARG1:.*]]: !spirv.ptr<!spirv.struct<(!spirv.array<1 x f32, stride=4> [0])>, StorageBuffer>
+//       CHECK: %[[CST0:.*]] = spirv.Constant 0 : i32
+//       CHECK: %[[AC0:.*]] = spirv.AccessChain %[[ARG0]][%[[CST0]], %[[CST0]]] : !spirv.ptr<!spirv.struct<(!spirv.array<1 x f32, stride=4> [0])>, StorageBuffer>, i32, i32
+//       CHECK: %[[LOAD:.*]] = spirv.Load "StorageBuffer" %[[AC0]] : f32
+//       CHECK: %[[AC1:.*]] = spirv.AccessChain %[[ARG1]][%[[CST0]], %[[CST0]]] : !spirv.ptr<!spirv.struct<(!spirv.array<1 x f32, stride=4> [0])>, StorageBuffer>, i32, i32
+//       CHECK: spirv.Store "StorageBuffer" %[[AC1]], %[[LOAD]] : f32
+//       CHECK: spirv.Return
+func.func @load_store_float_rank_zero(%arg0: memref<f32>, %arg1: memref<f32>) {
+  %0 = memref.load %arg0[] : memref<f32>
+  memref.store %0, %arg1[] : memref<f32>
+  return
+}
+
+// CHECK-LABEL: @load_store_int_rank_one
+//  CHECK-SAME: %[[ARG0:.*]]: !spirv.ptr<!spirv.struct<(!spirv.array<4 x i32, stride=4> [0])>, StorageBuffer>, %[[ARG1:.*]]: !spirv.ptr<!spirv.struct<(!spirv.array<4 x i32, stride=4> [0])>, StorageBuffer>, %[[ARG2:.*]]: i32
+//       CHECK: %[[CST0:.*]] = spirv.Constant 0 : i32
+//       CHECK: %[[AC0:.*]] = spirv.AccessChain %[[ARG0]][%[[CST0]], %[[ARG2]]] : !spirv.ptr<!spirv.struct<(!spirv.array<4 x i32, stride=4> [0])>, StorageBuffer>, i32, i32
+//       CHECK: %[[LOAD:.*]] = spirv.Load "StorageBuffer" %[[AC0]] : i32
+//       CHECK: %[[AC1:.*]] = spirv.AccessChain %[[ARG1]][%[[CST0]], %[[ARG2]]] : !spirv.ptr<!spirv.struct<(!spirv.array<4 x i32, stride=4> [0])>, StorageBuffer>, i32, i32
+//       CHECK: spirv.Store "StorageBuffer" %[[AC1]], %[[LOAD]] : i32
+//       CHECK: spirv.Return
+func.func @load_store_int_rank_one(%arg0: memref<4xi32>, %arg1: memref<4xi32>, %arg2 : index) {
+  %0 = memref.load %arg0[%arg2] : memref<4xi32>
+  memref.store %0, %arg1[%arg2] : memref<4xi32>
+  return
+}
+
----------------
kuhar wrote:

Could you add a test case that attempts to address into a larger memref?
What happens when the loaded type is a vector?

https://github.com/llvm/llvm-project/pull/102534