diff options
Diffstat (limited to 'mlir/test/Dialect/Shard/sharding-propagation.mlir')
| -rw-r--r-- | mlir/test/Dialect/Shard/sharding-propagation.mlir | 301 |
1 files changed, 301 insertions, 0 deletions
diff --git a/mlir/test/Dialect/Shard/sharding-propagation.mlir b/mlir/test/Dialect/Shard/sharding-propagation.mlir new file mode 100644 index 0000000..34aaf05 --- /dev/null +++ b/mlir/test/Dialect/Shard/sharding-propagation.mlir @@ -0,0 +1,301 @@ +// RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation,cse))" %s | FileCheck %s + +shard.grid @grid_2(shape = 2) +shard.grid @grid_1d(shape = ?) +shard.grid @grid_2d(shape = 2x4) +shard.grid @grid_3d(shape = ?x?x?) + +// CHECK-LABEL: func.func @element_wise_empty_sharding_info +func.func @element_wise_empty_sharding_info(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: tosa.sigmoid + %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: return + return %0 : tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @element_wise_on_def +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @element_wise_on_def(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V1:.*]] = tosa.sigmoid %[[V0]] + %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]] : tensor<8x16xf32> + %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %1 = shard.shard %0 to %s1 : tensor<8x16xf32> + // CHECK-NEXT: return %[[V2]] + return %1 : tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @element_wise_on_use +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @element_wise_on_use(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + %s0 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %0 = shard.shard %arg0 to %s0 annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V1:.*]] = tosa.sigmoid %[[V0]] + %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]] : tensor<8x16xf32> + // CHECK-NEXT: return %[[V2]] + return %1 : tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @element_wise_on_graph_output +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @element_wise_on_graph_output(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V1:.*]] = tosa.sigmoid %[[V0]] + %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]] : tensor<8x16xf32> + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %1 = shard.shard %0 to %s1 annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: return %[[V3]] + return %1 : tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @element_wise_on_graph_input +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @element_wise_on_graph_input(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] : tensor<8x16xf32> + // CHECK-NEXT: %[[V1:.*]] = shard.shard %[[V0]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + %s0 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %0 = shard.shard %arg0 to %s0 : tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = tosa.sigmoid %[[V1]] + %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] : tensor<8x16xf32> + // CHECK-NEXT: return %[[V3]] + return %1 : tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @arrow_structure +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @arrow_structure(%arg0: tensor<8x16xf32>, %arg1: tensor<1xf32>, %arg2: tensor<1xf32>) -> (tensor<8x16xf32>, tensor<8x16xf32>) { + // CHECK-NEXT: %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V1:.*]] = shard.shard %[[ARG]] to %[[S1]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = tosa.tanh %[[V1]] + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S1]] : tensor<8x16xf32> + %0 = tosa.tanh %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V4:.*]] = shard.shard %[[V3]] to %[[S1]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V5:.*]] = tosa.abs %[[V4]] + // CHECK-NEXT: %[[V6:.*]] = shard.shard %[[V5]] to %[[S1]] : tensor<8x16xf32> + %1 = tosa.abs %0: (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[S3:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK-NEXT: %[[ZP1:.*]] = shard.shard %arg1 to %[[S3]] annotate_for_users : tensor<1xf32> + // CHECK-NEXT: %[[ZP2:.*]] = shard.shard %arg2 to %[[S3]] annotate_for_users : tensor<1xf32> + // CHECK-NEXT: %[[V7:.*]] = tosa.negate %[[V4]], %[[ZP1]], %[[ZP2]] + // CHECK-NEXT: %[[V8:.*]] = shard.shard %[[V7]] to %[[S1]] : tensor<8x16xf32> + %2 = tosa.negate %0, %arg1, %arg2 : (tensor<8x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<8x16xf32> + %s3 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %3 = shard.shard %2 to %s3 : tensor<8x16xf32> + // CHECK-NEXT: return %[[V6]], %[[V8]] + return %1, %3 : tensor<8x16xf32>, tensor<8x16xf32> +} + +// CHECK-LABEL: func.func @matmul_on_def_shard_batch_and_m +// CHECK-SAME: %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32> +func.func @matmul_on_def_shard_batch_and_m(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG0]] to %[[S0]] annotate_for_users : tensor<2x16x8xf32> + // CHECK-NEXT: %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0]] : !shard.sharding + // CHECK-NEXT: %[[V1:.*]] = shard.shard %[[ARG1]] to %[[S1]] annotate_for_users : tensor<2x8x32xf32> + // CHECK-NEXT: %[[S2:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK-NEXT: %[[ZP:.*]] = shard.shard %[[ARG2]] to %[[S2]] annotate_for_users : tensor<1xf32> + // CHECK-NEXT: %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]] + %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x16x32xf32> + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] : tensor<2x16x32xf32> + %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding + %1 = shard.shard %0 to %s1 : tensor<2x16x32xf32> + // CHECK-NEXT: return %[[V3]] + return %1 : tensor<2x16x32xf32> +} + +// CHECK-LABEL: func.func @matmul_on_def_shard_m_and_n +// CHECK-SAME: [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32> +func.func @matmul_on_def_shard_m_and_n(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> { + // CHECK: [[vsharding:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding + // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32> + // CHECK: [[vsharding_0:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [], [1]] : !shard.sharding + // CHECK: [[vsharded_1:%.*]] = shard.shard [[varg1]] to [[vsharding_0]] annotate_for_users : tensor<2x8x32xf32> + // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK: [[vsharded_3:%.*]] = shard.shard [[varg2]] to [[vsharding_2]] annotate_for_users : tensor<1xf32> + // CHECK: [[v0:%.*]] = tosa.matmul + %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x16x32xf32> + // CHECK: [[vsharding_4:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0], [1]] : !shard.sharding + // CHECK: [[vsharded_5:%.*]] = shard.shard [[v0]] to [[vsharding_4]] : tensor<2x16x32xf32> + %s1 = shard.sharding @grid_2d split_axes = [[], [0], [1]] : !shard.sharding + %1 = shard.shard %0 to %s1 : tensor<2x16x32xf32> + // CHECK-NEXT: return [[vsharded_5]] + return %1 : tensor<2x16x32xf32> +} + +// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_k +// CHECK-SAME: [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32> +func.func @matmul_on_use_shard_m_and_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> { + // CHECK: [[vsharding:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0], [1]] : !shard.sharding + %s0 = shard.sharding @grid_2d split_axes = [[], [0], [1]] : !shard.sharding + // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x16x8xf32> + %arg0_s = shard.shard %arg0 to %s0 : tensor<2x16x8xf32> + // CHECK: [[vsharded_0:%.*]] = shard.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32> + // CHECK: [[vsharding_1:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1]] : !shard.sharding + // CHECK: [[vsharded_2:%.*]] = shard.shard [[varg1]] to [[vsharding_1]] annotate_for_users : tensor<2x8x32xf32> + // CHECK: [[vsharding_3:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK: [[vsharded_4:%.*]] = shard.shard [[varg2]] to [[vsharding_3]] annotate_for_users : tensor<1xf32> + // CHECK: [[v0:%.*]] = tosa.matmul + // CHECK: [[vsharding_5:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding + // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding_5]] : tensor<2x16x32xf32> + %0 = tosa.matmul %arg0_s, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x16x32xf32> + // CHECK: return [[vsharded_6]] + return %0 : tensor<2x16x32xf32> +} + +// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_duplicted_k +// CHECK-SAME: %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32> +func.func @matmul_on_use_shard_m_and_duplicted_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1], [0]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG0]] to %[[S0]] annotate_for_users : tensor<2x16x8xf32> + %s0 = shard.sharding @grid_2d split_axes = [[], [1], [0]] : !shard.sharding + %0 = shard.shard %arg0 to %s0 annotate_for_users : tensor<2x16x8xf32> + // CHECK-NEXT: %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding + // CHECK-NEXT: %[[V1:.*]] = shard.shard %[[ARG1]] to %[[S1]] annotate_for_users : tensor<2x8x32xf32> + %s1 = shard.sharding @grid_2d split_axes = [[], [0]] : !shard.sharding + %1 = shard.shard %arg1 to %s1 annotate_for_users : tensor<2x8x32xf32> + // CHECK-NEXT: %[[S2:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK-NEXT: %[[ZP:.*]] = shard.shard %[[ARG2]] to %[[S2]] annotate_for_users : tensor<1xf32> + // CHECK-NEXT: %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]] + %2 = tosa.matmul %0, %1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x16x32xf32> + // CHECK-NEXT: %[[S3:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1]] : !shard.sharding + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S3]] : tensor<2x16x32xf32> + // CHECK-NEXT: return %[[V3]] + return %2 : tensor<2x16x32xf32> +} + +// CHECK-LABEL: func.func @resolve_conflicting_annotations +func.func @resolve_conflicting_annotations( + // CHECK-SAME: %[[IN1:.*]]: tensor<2x3xf32>, + %arg0: tensor<2x3xf32>, + // CHECK-SAME: %[[IN2:.*]]: tensor<3x2xf32>, + %arg1: tensor<3x2xf32>, + // CHECK-SAME: %[[OUT_DPS:.*]]: tensor<2x2xf32> + %out_dps: tensor<2x2xf32> +// CHECK-SAME: ) -> tensor<2x2xf32> { +) -> tensor<2x2xf32> { + // CHECK: %[[SIN1_SHARDED1:.*]] = shard.sharding @grid_2 split_axes = {{\[\[}}0]] : !shard.sharding + // CHECK-NEXT: %[[IN1_SHARDED1:.*]] = shard.shard %[[IN1]] to %[[SIN1_SHARDED1]] : tensor<2x3xf32> + // CHECK: %[[SIN2_SHARDED:.*]] = shard.sharding @grid_2 split_axes = {{\[\[}}]] : !shard.sharding + // CHECK-NEXT: %[[IN1_SHARDED2:.*]] = shard.shard %[[IN1_SHARDED1]] to %[[SIN2_SHARDED]] annotate_for_users : tensor<2x3xf32> + // CHECK-NEXT: %[[IN2_SHARDED:.*]] = shard.shard %[[IN2]] to %[[SIN2_SHARDED]] annotate_for_users : tensor<3x2xf32> + // CHECK-NEXT: %[[OUT_DPS_SHARDED:.*]] = shard.shard %[[OUT_DPS]] to %[[SIN2_SHARDED]] annotate_for_users : tensor<2x2xf32> + %sarg0_sharded = shard.sharding @grid_2 split_axes = [[0]] : !shard.sharding + %arg0_sharded = shard.shard %arg0 to %sarg0_sharded : tensor<2x3xf32> + // CHECK: %[[MATMUL:.*]] = linalg.matmul ins(%[[IN1_SHARDED2]], %[[IN2_SHARDED]] : tensor<2x3xf32>, tensor<3x2xf32>) + // CHECK-SAME: outs(%[[OUT_DPS_SHARDED]] : tensor<2x2xf32>) -> tensor<2x2xf32> + %res = linalg.matmul ins(%arg0_sharded, %arg1 : tensor<2x3xf32>, tensor<3x2xf32>) + outs(%out_dps : tensor<2x2xf32>) -> tensor<2x2xf32> + // CHECK-NEXT: %[[RES:.*]] = shard.shard %[[MATMUL]] to %[[SIN2_SHARDED]] : tensor<2x2xf32> + %sres_sharded = shard.sharding @grid_2 split_axes = [[]] : !shard.sharding + %res_sharded = shard.shard %res to %sres_sharded : tensor<2x2xf32> + // CHECK: return %[[RES]] : tensor<2x2xf32> + return %res_sharded : tensor<2x2xf32> +} + +// https://arxiv.org/abs/2211.05102 Figure 2(a) +// The sharding propagation results in unnecessary reshards, +// an optimization pass should be able to remove them. +// CHECK-LABEL: func.func @mlp_1d_weight_stationary +// CHECK-SAME: [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32> +func.func @mlp_1d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> { + %s0 = shard.sharding @grid_1d split_axes = [[], [], [0, 1, 2]] : !shard.sharding + %sharded0 = shard.shard %arg0 to %s0 : tensor<2x4x8xf32> + %sharded1 = shard.shard %arg1 to %s0 : tensor<2x8x32xf32> + // CHECK: [[vsharding:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}], [], [0, 1, 2]] : !shard.sharding + // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32> + // CHECK: [[vsharded_0:%.*]] = shard.shard [[varg1]] to [[vsharding]] : tensor<2x8x32xf32> + // CHECK: [[vsharded_1:%.*]] = shard.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32> + // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}], [0, 1, 2]] : !shard.sharding + // CHECK: [[vsharded_3:%.*]] = shard.shard [[vsharded_0]] to [[vsharding_2]] annotate_for_users : tensor<2x8x32xf32> + // CHECK: [[vsharding_4:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK: [[vsharded_5:%.*]] = shard.shard [[varg3]] to [[vsharding_4]] annotate_for_users : tensor<1xf32> + // CHECK: [[v0:%.*]] = tosa.matmul + %1 = tosa.matmul %sharded0, %sharded1, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x4x32xf32> + // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding_4]] : tensor<2x4x32xf32> + // CHECK: [[vsharded_7:%.*]] = shard.shard [[vsharded_6]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32> + // CHECK: [[v1:%.*]] = tosa.sigmoid [[vsharded_7]] : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32> + // CHECK: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding_4]] : tensor<2x4x32xf32> + %2 = tosa.sigmoid %1 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32> + %sharding = shard.sharding @grid_1d split_axes = [[], [0, 1, 2]] : !shard.sharding + // CHECK: [[vsharded_9:%.*]] = shard.shard [[varg2]] to [[vsharding_2]] : tensor<2x32x8xf32> + %sharded2 = shard.shard %arg2 to %sharding : tensor<2x32x8xf32> + // CHECK: [[vsharded_10:%.*]] = shard.shard [[vsharded_8]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32> + // CHECK: [[vsharded_11:%.*]] = shard.shard [[vsharded_9]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32> + // CHECK: [[v2:%.*]] = tosa.matmul + %3 = tosa.matmul %2, %sharded2, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x4x8xf32> + // CHECK: [[vsharded_12:%.*]] = shard.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32> + %s4 = shard.sharding @grid_1d split_axes = [[], [], [0, 1, 2]] : !shard.sharding + %4 = shard.shard %3 to %s4 : tensor<2x4x8xf32> + // CHECK: return [[vsharded_12]] + return %4 : tensor<2x4x8xf32> +} + +// https://arxiv.org/abs/2211.05102 Figure 2(b) +// The sharding propagation results in unnecessary reshards, +// an optimization pass should be able to remove them. +// CHECK-LABEL: func.func @mlp_2d_weight_stationary +// CHECK-SAME: [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32> +func.func @mlp_2d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> { + // CHECK: [[vsharding:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [], [0, 1, 2]] : !shard.sharding + %s0 = shard.sharding @grid_3d split_axes = [[], [], [0, 1, 2]] : !shard.sharding + // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32> + %arg0_s = shard.shard %arg0 to %s0 : tensor<2x4x8xf32> + // CHECK: [[vsharding_0:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [0], [1, 2]] : !shard.sharding + %s1 = shard.sharding @grid_3d split_axes = [[], [0], [1, 2]] : !shard.sharding + // CHECK: [[vsharded_1:%.*]] = shard.shard [[varg1]] to [[vsharding_0]] : tensor<2x8x32xf32> + %arg1_s = shard.shard %arg1 to %s1 : tensor<2x8x32xf32> + // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK: [[vsharded_3:%.*]] = shard.shard [[vsharded]] to [[vsharding_2]] annotate_for_users : tensor<2x4x8xf32> + // CHECK: [[vsharded_4:%.*]] = shard.shard [[vsharded_1]] to [[vsharding]] annotate_for_users : tensor<2x8x32xf32> + // CHECK: [[vsharded_5:%.*]] = shard.shard [[varg3]] to [[vsharding_2]] annotate_for_users : tensor<1xf32> + // CHECK: [[v0:%.*]] = tosa.matmul + %1 = tosa.matmul %arg0_s, %arg1_s, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x4x32xf32> + // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding]] : tensor<2x4x32xf32> + %2 = shard.shard %1 to %s0 : tensor<2x4x32xf32> + // CHECK: [[vsharded_7:%.*]] = shard.shard [[vsharded_6]] to [[vsharding]] annotate_for_users : tensor<2x4x32xf32> + // CHECK: [[v1:%.*]] = tosa.sigmoid + // CHECK: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding]] : tensor<2x4x32xf32> + %3 = tosa.sigmoid %2 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32> + // CHECK: [[vsharding_9:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [1, 2], [0]] : !shard.sharding + %s2 = shard.sharding @grid_3d split_axes = [[], [1, 2], [0]] : !shard.sharding + // CHECK: [[vsharded_10:%.*]] = shard.shard [[varg2]] to [[vsharding_9]] : tensor<2x32x8xf32> + %arg2_s = shard.shard %arg2 to %s2 : tensor<2x32x8xf32> + // CHECK: [[vsharded_11:%.*]] = shard.shard [[vsharded_8]] to [[vsharding_2]] annotate_for_users : tensor<2x4x32xf32> + // CHECK: [[vsharded_12:%.*]] = shard.shard [[vsharded_10]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32> + // CHECK: [[v2:%.*]] = tosa.matmul + %4 = tosa.matmul %3, %arg2_s, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<2x4x8xf32> + // CHECK: [[vsharded_13:%.*]] = shard.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32> + %5 = shard.shard %4 to %s0 : tensor<2x4x8xf32> + // CHECK: [[vsharded_14:%.*]] = shard.shard [[vsharded_13]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32> + %6 = shard.shard %5 to %s0 annotate_for_users : tensor<2x4x8xf32> + // CHECK: return [[vsharded_14]] + return %6 : tensor<2x4x8xf32> +} + +// CHECK-LABEL: func.func @elementwise_duplicated_chain +// CHECK-SAME: %[[ARG:.*]]: tensor<8x16xf32> +func.func @elementwise_duplicated_chain(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> { + // CHECK-NEXT: %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding + // CHECK-NEXT: %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V1:.*]] = tosa.sigmoid %[[V0]] + %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]] : tensor<8x16xf32> + // CHECK-NEXT: %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] annotate_for_users : tensor<8x16xf32> + // CHECK-NEXT: %[[V4:.*]] = tosa.sigmoid %[[V3]] + %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + // CHECK-NEXT: %[[V5:.*]] = shard.shard %[[V4]] to %[[S0]] : tensor<8x16xf32> + %s0 = shard.sharding @grid_2d split_axes = [[]] : !shard.sharding + %2 = shard.shard %1 to %s0 : tensor<8x16xf32> + // CHECK-NEXT: return %[[V5]] + return %2 : tensor<8x16xf32> +} |