Use "standard" load and stores in LowerVectorTransfers

Clipping creates non-affine memory accesses, use std_load and std_store instead of affine_load and affine_store.
In the future we may also want a fill with the neutral element rather than clip, this would make the accesses affine if we wanted more analyses and transformations to happen post lowering to pointwise copies.

PiperOrigin-RevId: 260110503

diff --git a/mlir/include/mlir/EDSC/Intrinsics.h b/mlir/include/mlir/EDSC/Intrinsics.h
index 274541a..872c3b2 100644 (file)
--- a/mlir/include/mlir/EDSC/Intrinsics.h
+++ b/mlir/include/mlir/EDSC/Intrinsics.h
@@ -198,6 +198,8 @@ using dim = ValueBuilder<DimOp>;
 using muli = ValueBuilder<MulIOp>;
 using ret = OperationBuilder<ReturnOp>;
 using select = ValueBuilder<SelectOp>;
+using std_load = ValueBuilder<LoadOp>;
+using std_store = OperationBuilder<StoreOp>;
 using subi = ValueBuilder<SubIOp>;
 using vector_type_cast = ValueBuilder<VectorTypeCastOp>;
 

diff --git a/mlir/lib/Transforms/LowerVectorTransfers.cpp b/mlir/lib/Transforms/LowerVectorTransfers.cpp
index 1ecde11..3585e2b 100644 (file)
--- a/mlir/lib/Transforms/LowerVectorTransfers.cpp
+++ b/mlir/lib/Transforms/LowerVectorTransfers.cpp
@@ -263,6 +263,8 @@ VectorTransferRewriter<VectorTransferReadOp>::matchAndRewrite(
   using namespace mlir::edsc;
   using namespace mlir::edsc::op;
   using namespace mlir::edsc::intrinsics;
+  using IndexedValue =
+      TemplatedIndexedValue<intrinsics::std_load, intrinsics::std_store>;
 
   VectorTransferReadOp transfer = cast<VectorTransferReadOp>(op);
 
@@ -289,7 +291,7 @@ VectorTransferRewriter<VectorTransferReadOp>::matchAndRewrite(
     // Computes clippedScalarAccessExprs in the loop nest scope (ivs exist).
     local(ivs) = remote(clip(transfer, view, ivs));
   });
-  ValueHandle vectorValue = affine_load(vec, {constant_index(0)});
+  ValueHandle vectorValue = std_load(vec, {constant_index(0)});
   (dealloc(tmp)); // vexing parse
 
   // 3. Propagate.
@@ -322,6 +324,8 @@ VectorTransferRewriter<VectorTransferWriteOp>::matchAndRewrite(
   using namespace mlir::edsc;
   using namespace mlir::edsc::op;
   using namespace mlir::edsc::intrinsics;
+  using IndexedValue =
+      TemplatedIndexedValue<intrinsics::std_load, intrinsics::std_store>;
 
   VectorTransferWriteOp transfer = cast<VectorTransferWriteOp>(op);
 
@@ -345,7 +349,7 @@ VectorTransferRewriter<VectorTransferWriteOp>::matchAndRewrite(
   ValueHandle tmp = alloc(tmpMemRefType(transfer));
   IndexedValue local(tmp);
   ValueHandle vec = vector_type_cast(tmp, vectorMemRefType(transfer));
-  affine_store(vectorValue, vec, {constant_index(0)});
+  std_store(vectorValue, vec, {constant_index(0)});
   LoopNestBuilder(pivs, lbs, ubs, steps)([&] {
     // Computes clippedScalarAccessExprs in the loop nest scope (ivs exist).
     remote(clip(transfer, view, ivs)) = local(ivs);

diff --git a/mlir/test/Transforms/Vectorize/lower_vector_transfers.mlir b/mlir/test/Transforms/Vectorize/lower_vector_transfers.mlir
index 7295d05..5d8acea 100644 (file)
--- a/mlir/test/Transforms/Vectorize/lower_vector_transfers.mlir
+++ b/mlir/test/Transforms/Vectorize/lower_vector_transfers.mlir
@@ -20,7 +20,7 @@ func @materialize_read_1d() {
       // CHECK: %[[FILTERED1:.*]] = select
       // CHECK: {{.*}} = select
       // CHECK: %[[FILTERED2:.*]] = select
-      // CHECK-NEXT: %{{.*}} = affine.load {{.*}}[%[[FILTERED1]], %[[FILTERED2]]] : memref<7x42xf32>
+      // CHECK-NEXT: %{{.*}} = load {{.*}}[%[[FILTERED1]], %[[FILTERED2]]] : memref<7x42xf32>
     }
   }
   return
@@ -94,12 +94,12 @@ func @materialize_read(%M: index, %N: index, %O: index, %P: index) {
   // CHECK-NEXT:                {{.*}} = cmpi "slt", {{.*}}, %[[C0]] : index
   // CHECK-NEXT:                %[[L3:.*]] = select
   //
-  // CHECK-NEXT:                {{.*}} = affine.load %{{.*}}[%[[L0]], %[[L1]], %[[L2]], %[[L3]]] : memref<?x?x?x?xf32>
+  // CHECK-NEXT:                {{.*}} = load %{{.*}}[%[[L0]], %[[L1]], %[[L2]], %[[L3]]] : memref<?x?x?x?xf32>
   // CHECK-NEXT:                store {{.*}}, %[[ALLOC]][%[[I6]], %[[I5]], %[[I4]]] : memref<5x4x3xf32>
   // CHECK-NEXT:              }
   // CHECK-NEXT:            }
   // CHECK-NEXT:          }
-  //      CHECK:          {{.*}} = affine.load %[[VECTOR_VIEW]][{{.*}}] : memref<1xvector<5x4x3xf32>>
+  //      CHECK:          {{.*}} = load %[[VECTOR_VIEW]][{{.*}}] : memref<1xvector<5x4x3xf32>>
   // CHECK-NEXT:          dealloc %[[ALLOC]] : memref<5x4x3xf32>
   // CHECK-NEXT:        }
   // CHECK-NEXT:      }
@@ -170,7 +170,7 @@ func @materialize_write(%M: index, %N: index, %O: index, %P: index) {
   // CHECK-NEXT:                {{.*}} = cmpi "slt", {{.*}}, %[[C0]] : index
   // CHECK-NEXT:                %[[S3:.*]] = select {{.*}}, %[[C0]], {{.*}} : index
   //
-  // CHECK-NEXT:                {{.*}} = affine.load {{.*}}[%[[I6]], %[[I5]], %[[I4]]] : memref<5x4x3xf32>
+  // CHECK-NEXT:                {{.*}} = load {{.*}}[%[[I6]], %[[I5]], %[[I4]]] : memref<5x4x3xf32>
   //      CHECK:                store {{.*}}, {{.*}}[%[[S0]], %[[S1]], %[[S2]], %[[S3]]] : memref<?x?x?x?xf32>
   // CHECK-NEXT:              }
   // CHECK-NEXT:            }