[Mlir-commits] [mlir] [mlir][TosaToLinalg] Lower tosa reduce operations to linalg reduce op (PR #68480)
Amir Bishara
llvmlistbot at llvm.org
Tue Oct 10 07:05:07 PDT 2023
https://github.com/amirBish updated https://github.com/llvm/llvm-project/pull/68480
>From 737cd8c06ae7899e30efff0f98260c4f17293bd2 Mon Sep 17 00:00:00 2001
From: Amir Bishara <amir.bishara at mobileye.com>
Date: Fri, 6 Oct 2023 18:43:01 +0300
Subject: [PATCH] [mlir][TosaToLinalg] Lower tosa reduce operations to linalg
reduce op
-lowering tosa reduce operations to linalg.reduce
op which indicates explicit reduce operation.
-the linalg reduce op doesn't have iterator
types or indexing maps instead it has a reduction
dimension which would provide the necessary
information.
---
.../Conversion/TosaToLinalg/TosaToLinalg.cpp | 19 +---
.../TosaToLinalg/tosa-to-linalg.mlir | 99 ++++++++-----------
2 files changed, 45 insertions(+), 73 deletions(-)
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
index 0fb3cddb0120810..3bf7bf12b5e96ff 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -980,8 +980,6 @@ static LogicalResult reduceMatchAndRewriteHelper(Operation *op, uint64_t axis,
}
}
- Type reduceTy = RankedTensorType::get(reduceShape, resultTy.getElementType());
-
// First fill the output buffer with the init value.
auto emptyTensor =
rewriter
@@ -1000,22 +998,9 @@ static LogicalResult reduceMatchAndRewriteHelper(Operation *op, uint64_t axis,
ValueRange{emptyTensor})
.result();
- SmallVector<AffineExpr, 2> srcExprs;
- SmallVector<AffineExpr, 2> dstExprs;
- SmallVector<utils::IteratorType, 4> iteratorTypes;
- for (unsigned int i = 0, rank = inputTy.getRank(); i != rank; ++i) {
- srcExprs.push_back(mlir::getAffineDimExpr(i, rewriter.getContext()));
-
- iteratorTypes.push_back(axis == i ? utils::IteratorType::reduction
- : utils::IteratorType::parallel);
- if (axis != i)
- dstExprs.push_back(mlir::getAffineDimExpr(i, rewriter.getContext()));
- }
-
bool didEncounterError = false;
- auto maps = AffineMap::inferFromExprList({srcExprs, dstExprs});
- auto linalgOp = rewriter.create<linalg::GenericOp>(
- loc, reduceTy, input, filledTensor, maps, iteratorTypes,
+ auto linalgOp = rewriter.create<linalg::ReduceOp>(
+ loc, input, filledTensor, axis,
[&](OpBuilder &nestedBuilder, Location nestedLoc, ValueRange blockArgs) {
auto result = createLinalgBodyCalculationForReduceOp(
op, blockArgs, elementTy, rewriter);
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
index 8e0307085f1ce26..aa53b366f6da684 100644
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
@@ -896,9 +896,6 @@ func.func @test_transpose_dyn_multiple(%arg0: tensor<?x?xf32>) -> () {
// -----
-// CHECK-DAG: #[[$MAP0:.*]] = affine_map<(d0, d1) -> (d0, d1)>
-// CHECK-DAG: #[[$MAP1:.*]] = affine_map<(d0, d1) -> (d1)>
-// CHECK-DAG: #[[$MAP2:.*]] = affine_map<(d0, d1) -> (d0)>
// CHECK-LABEL: @reduce_float
// CHECK-SAME: [[ARG0:%.+]]: tensor<5x4xf32>
@@ -906,38 +903,40 @@ func.func @reduce_float(%arg0: tensor<5x4xf32>) -> () {
// CHECK: [[INIT:%.+]] = tensor.empty() : tensor<4xf32>
// CHECK: [[CST0:%.+]] = arith.constant 0.0
// CHECK: [[FILL:%.+]] = linalg.fill ins([[CST0]]{{.*}}outs([[INIT]]
- // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["reduction", "parallel"]} ins([[ARG0]] : tensor<5x4xf32>) outs([[FILL]] : tensor<4xf32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: [[REDUCE:%.+]] = linalg.reduce ins([[ARG0]] : tensor<5x4xf32>) outs([[FILL]] : tensor<4xf32>) dimensions = [0]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: [[RES:%.+]] = arith.addf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield [[RES]] : f32
- // CHECK: tensor.expand_shape [[GENERIC]] {{\[}}[0, 1]] : tensor<4xf32> into tensor<1x4xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape [[REDUCE]] {{\[}}[0, 1]] : tensor<4xf32> into tensor<1x4xf32>
%0 = tosa.reduce_sum %arg0 {axis = 0 : i32} : (tensor<5x4xf32>) -> tensor<1x4xf32>
// CHECK: [[INIT:%.+]] = tensor.empty() : tensor<5xf32>
// CHECK: [[CST0:%.+]] = arith.constant 0.0
// CHECK: [[FILL:%.+]] = linalg.fill ins([[CST0]]{{.*}}outs([[INIT]]
- // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP2]]], iterator_types = ["parallel", "reduction"]} ins([[ARG0]] : tensor<5x4xf32>) outs([[FILL]] : tensor<5xf32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: [[REDUCE:%.+]] = linalg.reduce ins([[ARG0]] : tensor<5x4xf32>) outs([[FILL]] : tensor<5xf32>) dimensions = [1]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: [[RES:%.+]] = arith.addf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield [[RES]] : f32
- // CHECK: tensor.expand_shape [[GENERIC]] {{\[}}[0, 1]] : tensor<5xf32> into tensor<5x1xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape [[REDUCE]] {{\[}}[0, 1]] : tensor<5xf32> into tensor<5x1xf32>
%1 = tosa.reduce_sum %arg0 {axis = 1 : i32} : (tensor<5x4xf32>) -> tensor<5x1xf32>
// CHECK: arith.constant 1.0
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.mulf
%2 = tosa.reduce_prod %arg0 {axis = 0 : i32} : (tensor<5x4xf32>) -> tensor<1x4xf32>
// CHECK: arith.constant 3.40282347E+38 : f32
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.minimumf
%3 = tosa.reduce_min %arg0 {axis = 0 : i32} : (tensor<5x4xf32>) -> tensor<1x4xf32>
// CHECK: arith.constant -3.40282347E+38 : f32
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.maximumf
%4 = tosa.reduce_max %arg0 {axis = 0 : i32} : (tensor<5x4xf32>) -> tensor<1x4xf32>
return
@@ -945,9 +944,6 @@ func.func @reduce_float(%arg0: tensor<5x4xf32>) -> () {
// -----
-// CHECK-DAG: #[[$MAP0:.*]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
-// CHECK-DAG: #[[$MAP1:.*]] = affine_map<(d0, d1, d2) -> (d0, d2)>
-
// CHECK-LABEL: @reduce_float_dyn
// CHECK-SAME: %[[ARG0:[0-9a-zA-Z_]*]]: tensor<?x5x4xf32>
func.func @reduce_float_dyn(%arg0: tensor<?x5x4xf32>) -> () {
@@ -956,40 +952,36 @@ func.func @reduce_float_dyn(%arg0: tensor<?x5x4xf32>) -> () {
// CHECK: %[[INIT:.+]] = tensor.empty(%[[DYN]]) : tensor<?x4xf32>
// CHECK: %[[CST0:.+]] = arith.constant 0.0
// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[CST0]]{{.*}}outs(%[[INIT]]
- // CHECK: %[[GENERIC:.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["parallel", "reduction", "parallel"]} ins(%[[ARG0]] : tensor<?x5x4xf32>) outs(%[[FILL]] : tensor<?x4xf32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: %[[REDUCE:.+]] = linalg.reduce ins(%[[ARG0]] : tensor<?x5x4xf32>) outs(%[[FILL]] : tensor<?x4xf32>) dimensions = [1]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: %[[RES:.+]] = arith.addf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield %[[RES]] : f32
- // CHECK: tensor.expand_shape %[[GENERIC]] {{\[}}[0], [1, 2]] : tensor<?x4xf32> into tensor<?x1x4xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape %[[REDUCE]] {{\[}}[0], [1, 2]] : tensor<?x4xf32> into tensor<?x1x4xf32>
%0 = tosa.reduce_sum %arg0 {axis = 1 : i32} : (tensor<?x5x4xf32>) -> tensor<?x1x4xf32>
return
}
// -----
-// CHECK-DAG: #[[$MAP0:.*]] = affine_map<(d0) -> (d0)>
-// CHECK-DAG: #[[$MAP1:.*]] = affine_map<(d0) -> ()>
-
// CHECK-LABEL: @reduce_float_dyn_rank_1
// CHECK-SAME: %[[ARG0:[0-9a-zA-Z_]*]]: tensor<?xf32>
func.func @reduce_float_dyn_rank_1(%arg0: tensor<?xf32>) -> () {
// CHECK-DAG: %[[INIT:.+]] = tensor.empty() : tensor<f32>
// CHECK-DAG: %[[CST0:.+]] = arith.constant 0.0
// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[CST0]]{{.*}}outs(%[[INIT]]
- // CHECK: %[[GENERIC:.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["reduction"]} ins(%[[ARG0]] : tensor<?xf32>) outs(%[[FILL]] : tensor<f32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: %[[REDUCE:.+]] = linalg.reduce ins(%[[ARG0]] : tensor<?xf32>) outs(%[[FILL]] : tensor<f32>) dimensions = [0]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: %[[RES:.+]] = arith.addf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield %[[RES]] : f32
- // CHECK: tensor.expand_shape %[[GENERIC]] {{\[}}] : tensor<f32> into tensor<1xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape %[[REDUCE]] {{\[}}] : tensor<f32> into tensor<1xf32>
%0 = tosa.reduce_sum %arg0 {axis = 0 : i32} : (tensor<?xf32>) -> tensor<1xf32>
return
}
// -----
-// CHECK-DAG: #[[$MAP0:.*]] = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
-// CHECK-DAG: #[[$MAP1:.*]] = affine_map<(d0, d1, d2) -> (d0, d1)>
-
// CHECK-LABEL: @reduce_float_dyn_nonzero_batch
// CHECK-SAME: (%[[ARG0:[0-9a-zA-Z_]*]]:
func.func @reduce_float_dyn_nonzero_batch(%arg0: tensor<5x?x4xf32>) -> () {
@@ -998,20 +990,18 @@ func.func @reduce_float_dyn_nonzero_batch(%arg0: tensor<5x?x4xf32>) -> () {
// CHECK: %[[INIT:.+]] = tensor.empty(%[[DYN]]) : tensor<5x?xf32>
// CHECK: %[[CST1:.+]] = arith.constant 1.0
// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[CST1]]{{.*}}outs(%[[INIT]]
- // CHECK: %[[GENERIC:.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["parallel", "parallel", "reduction"]} ins(%[[ARG0]] : tensor<5x?x4xf32>) outs(%[[FILL]] : tensor<5x?xf32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: %[[REDUCE:.+]] = linalg.reduce ins(%[[ARG0]] : tensor<5x?x4xf32>) outs(%[[FILL]] : tensor<5x?xf32>) dimensions = [2]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: %[[RES:.+]] = arith.mulf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield %[[RES]] : f32
- // CHECK: tensor.expand_shape %[[GENERIC]] {{\[}}[0], [1, 2]] : tensor<5x?xf32> into tensor<5x?x1xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape %[[REDUCE]] {{\[}}[0], [1, 2]] : tensor<5x?xf32> into tensor<5x?x1xf32>
%0 = tosa.reduce_prod %arg0 {axis = 2 : i32} : (tensor<5x?x4xf32>) -> tensor<5x?x1xf32>
return
}
// -----
-// CHECK-DAG: #[[$MAP0:.*]] = affine_map<(d0, d1) -> (d0, d1)>
-// CHECK-DAG: #[[$MAP1:.*]] = affine_map<(d0, d1) -> (d0)>
-
// CHECK-LABEL: @reduce_float_dyn_multiple
// CHECK-SAME: (%[[ARG0:[0-9a-zA-Z_]*]]:
func.func @reduce_float_dyn_multiple(%arg0: tensor<?x?xf32>) -> () {
@@ -1020,60 +1010,59 @@ func.func @reduce_float_dyn_multiple(%arg0: tensor<?x?xf32>) -> () {
// CHECK: %[[INIT:.+]] = tensor.empty(%[[DYN]])
// CHECK: %[[CMIN:.+]] = arith.constant -3.40282347E+38
// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[CMIN]]{{.*}}outs(%[[INIT]]
- // CHECK: %[[GENERIC:.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["parallel", "reduction"]} ins(%[[ARG0]] : tensor<?x?xf32>) outs(%[[FILL]] : tensor<?xf32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32)
+ // CHECK: %[[REDUCE:.+]] = linalg.reduce ins(%[[ARG0]] : tensor<?x?xf32>) outs(%[[FILL]] : tensor<?xf32>) dimensions = [1]
+ // CHECK: (%[[ARG1:.*]]: f32, %[[ARG2:.*]]: f32) {
// CHECK: %[[MAX:.+]] = arith.maximumf %[[ARG1]], %[[ARG2]] : f32
// CHECK: linalg.yield %[[MAX]] : f32
- // CHECK: tensor.expand_shape %[[GENERIC]] {{\[}}[0, 1]] : tensor<?xf32> into tensor<?x1xf32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape %[[REDUCE]] {{\[}}[0, 1]] : tensor<?xf32> into tensor<?x1xf32>
%0 = tosa.reduce_max %arg0 {axis = 1 : i32} : (tensor<?x?xf32>) -> tensor<?x1xf32>
return
}
// -----
-// CHECK: #[[$MAP0:.*]] = affine_map<(d0, d1) -> (d0, d1)>
-// CHECK: #[[$MAP1:.*]] = affine_map<(d0, d1) -> (d1)>
-// CHECK: #[[$MAP2:.*]] = affine_map<(d0, d1) -> (d0)>
-
// CHECK-LABEL: @reduce_int
// CHECK-SAME: [[ARG0:%.+]]: tensor<5x4xi32>
func.func @reduce_int(%arg0: tensor<5x4xi32>) -> () {
// CHECK: [[INIT:%.+]] = tensor.empty()
// CHECK: [[CST0:%.+]] = arith.constant 0
// CHECK: [[FILL:%.+]] = linalg.fill ins([[CST0]]{{.*}}outs([[INIT]]
- // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["reduction", "parallel"]} ins([[ARG0]] : tensor<5x4xi32>) outs([[FILL]] : tensor<4xi32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: i32, %[[ARG2:.*]]: i32)
+ // CHECK: [[REDUCE:%.+]] = linalg.reduce ins([[ARG0]] : tensor<5x4xi32>) outs([[FILL]] : tensor<4xi32>) dimensions = [0]
+ // CHECK: (%[[ARG1:.*]]: i32, %[[ARG2:.*]]: i32) {
// CHECK: [[RES:%.+]] = arith.addi %[[ARG1]], %[[ARG2]] : i32
// CHECK: linalg.yield [[RES]] : i32
- // CHECK: tensor.expand_shape [[GENERIC]] {{\[}}[0, 1]] : tensor<4xi32> into tensor<1x4xi32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape [[REDUCE]] {{\[}}[0, 1]] : tensor<4xi32> into tensor<1x4xi32>
%0 = tosa.reduce_sum %arg0 {axis = 0 : i32} : (tensor<5x4xi32>) -> tensor<1x4xi32>
// CHECK: [[INIT:%.+]] = tensor.empty()
// CHECK: [[CST0:%.+]] = arith.constant 0
// CHECK: [[FILL:%.+]] = linalg.fill ins([[CST0]]{{.*}}outs([[INIT]]
- // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP2]]], iterator_types = ["parallel", "reduction"]} ins([[ARG0]] : tensor<5x4xi32>) outs([[FILL]] : tensor<5xi32>)
- // CHECK: ^bb0(%[[ARG1:.*]]: i32, %[[ARG2:.*]]: i32)
+ // CHECK: [[REDUCE:%.+]] = linalg.reduce ins([[ARG0]] : tensor<5x4xi32>) outs([[FILL]] : tensor<5xi32>) dimensions = [1]
+ // CHECK: (%[[ARG1:.*]]: i32, %[[ARG2:.*]]: i32) {
// CHECK: [[RES:%.+]] = arith.addi %[[ARG1]], %[[ARG2]] : i32
// CHECK: linalg.yield [[RES]] : i32
- // CHECK: tensor.expand_shape [[GENERIC]] {{\[}}[0, 1]] : tensor<5xi32> into tensor<5x1xi32>
+ // CHECK: }
+ // CHECK: tensor.expand_shape [[REDUCE]] {{\[}}[0, 1]] : tensor<5xi32> into tensor<5x1xi32>
%1 = tosa.reduce_sum %arg0 {axis = 1 : i32} : (tensor<5x4xi32>) -> tensor<5x1xi32>
// CHECK: arith.constant 1
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.muli
%2 = tosa.reduce_prod %arg0 {axis = 0 : i32} : (tensor<5x4xi32>) -> tensor<1x4xi32>
// CHECK: arith.constant 2147483647 : i32
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.cmpi slt
// CHECK: select
%3 = tosa.reduce_min %arg0 {axis = 0 : i32} : (tensor<5x4xi32>) -> tensor<1x4xi32>
// CHECK: arith.constant -2147483648 : i32
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: arith.cmpi sgt
// CHECK: select
%4 = tosa.reduce_max %arg0 {axis = 0 : i32} : (tensor<5x4xi32>) -> tensor<1x4xi32>
@@ -1082,25 +1071,23 @@ func.func @reduce_int(%arg0: tensor<5x4xi32>) -> () {
// -----
-// CHECK: #[[$MAP0:.*]] = affine_map<(d0, d1) -> (d0, d1)>
-// CHECK: #[[$MAP1:.*]] = affine_map<(d0, d1) -> (d1)>
-
// CHECK-LABEL: @reduce_bool
// CHECK-SAME: [[ARG0:%.+]]: tensor<5x4xi1>
func.func @reduce_bool(%arg0: tensor<5x4xi1>) -> () {
// CHECK: [[INIT:%.+]] = tensor.empty()
// CHECK: [[CST0:%.+]] = arith.constant true
// CHECK: [[FILL:%.+]] = linalg.fill ins([[CST0]]{{.*}}outs([[INIT]]
- // CHECK: [[GENERIC:%.+]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["reduction", "parallel"]} ins([[ARG0]] : tensor<5x4xi1>) outs([[FILL]] : tensor<4xi1>)
- // CHECK: ^bb0(%[[ARG1:[0-9a-zA-Z_]+]]: i1, %[[ARG2:[0-9a-zA-Z_]+]]: i1)
+ // CHECK: [[REDUCE:%.+]] = linalg.reduce ins([[ARG0]] : tensor<5x4xi1>) outs([[FILL]] : tensor<4xi1>) dimensions = [0]
+ // CHECK: (%[[ARG1:[0-9a-zA-Z_]+]]: i1, %[[ARG2:[0-9a-zA-Z_]+]]: i1) {
// CHECK: [[RES:%.+]] = arith.andi %[[ARG1]], %[[ARG2]] : i1
// CHECK: linalg.yield [[RES]] : i1
- // CHECK: tensor.expand_shape [[GENERIC]] {{\[}}[0, 1]] : tensor<4xi1> into tensor<1x4xi1>
+ // CHECK: }
+ // CHECK: tensor.expand_shape [[REDUCE]] {{\[}}[0, 1]] : tensor<4xi1> into tensor<1x4xi1>
%0 = tosa.reduce_all %arg0 {axis = 0 : i32} : (tensor<5x4xi1>) -> tensor<1x4xi1>
// CHECK: arith.constant false
// CHECK: linalg.fill
- // CHECK: linalg.generic
+ // CHECK: linalg.reduce
// CHECK: or
%1 = tosa.reduce_any %arg0 {axis = 0 : i32} : (tensor<5x4xi1>) -> tensor<1x4xi1>
More information about the Mlir-commits
mailing list