141 lines · plain
1// RUN: mlir-opt %s -pass-pipeline='builtin.module(func.func(scf-parallel-loop-tiling{parallel-loop-tile-sizes=1,4}))' -split-input-file | FileCheck %s2 3func.func @parallel_loop(%arg0 : index, %arg1 : index, %arg2 : index,4 %arg3 : index, %arg4 : index, %arg5 : index,5 %A: memref<?x?xf32>, %B: memref<?x?xf32>,6 %C: memref<?x?xf32>, %result: memref<?x?xf32>) {7 scf.parallel (%i0, %i1) = (%arg0, %arg1) to (%arg2, %arg3) step (%arg4, %arg5) {8 %B_elem = memref.load %B[%i0, %i1] : memref<?x?xf32>9 %C_elem = memref.load %C[%i0, %i1] : memref<?x?xf32>10 %sum_elem = arith.addf %B_elem, %C_elem : f3211 memref.store %sum_elem, %result[%i0, %i1] : memref<?x?xf32>12 }13 return14}15 16// CHECK: #[[$MAP:.*]] = affine_map<(d0, d1, d2) -> (d0, d1 - d2)>17// CHECK-LABEL: func @parallel_loop(18// CHECK-SAME: [[ARG1:%.*]]: index, [[ARG2:%.*]]: index, [[ARG3:%.*]]: index, [[ARG4:%.*]]: index, [[ARG5:%.*]]: index, [[ARG6:%.*]]: index, [[ARG7:%.*]]: memref<?x?xf32>, [[ARG8:%.*]]: memref<?x?xf32>, [[ARG9:%.*]]: memref<?x?xf32>, [[ARG10:%.*]]: memref<?x?xf32>) {19// CHECK: [[C0:%.*]] = arith.constant 0 : index20// CHECK: [[C1:%.*]] = arith.constant 1 : index21// CHECK: [[C4:%.*]] = arith.constant 4 : index22// CHECK: [[V1:%.*]] = arith.muli [[ARG5]], [[C1]] : index23// CHECK: [[V2:%.*]] = arith.muli [[ARG6]], [[C4]] : index24// CHECK: scf.parallel ([[V3:%.*]], [[V4:%.*]]) = ([[ARG1]], [[ARG2]]) to ([[ARG3]], [[ARG4]]) step ([[V1]], [[V2]]) {25// CHECK: [[V5:%.*]] = affine.min #[[$MAP]]([[V1]], [[ARG3]], [[V3]])26// CHECK: [[V6:%.*]] = affine.min #[[$MAP]]([[V2]], [[ARG4]], [[V4]])27// CHECK: scf.parallel ([[V7:%.*]], [[V8:%.*]]) = ([[C0]], [[C0]]) to ([[V5]], [[V6]]) step ([[ARG5]], [[ARG6]]) {28// CHECK: [[V9:%.*]] = arith.addi [[V7]], [[V3]] : index29// CHECK: [[V10:%.*]] = arith.addi [[V8]], [[V4]] : index30// CHECK: [[V11:%.*]] = memref.load [[ARG8]]{{\[}}[[V9]], [[V10]]] : memref<?x?xf32>31// CHECK: [[V12:%.*]] = memref.load [[ARG9]]{{\[}}[[V9]], [[V10]]] : memref<?x?xf32>32// CHECK: [[V13:%.*]] = arith.addf [[V11]], [[V12]] : f3233// CHECK: memref.store [[V13]], [[ARG10]]{{\[}}[[V9]], [[V10]]] : memref<?x?xf32>34// CHECK: }35// CHECK: }36// CHECK: return37 38// -----39 40func.func @static_loop_with_step() {41 %c0 = arith.constant 0 : index42 %c3 = arith.constant 3 : index43 %c22 = arith.constant 22 : index44 %c24 = arith.constant 24 : index45 scf.parallel (%i0, %i1) = (%c0, %c0) to (%c22, %c24) step (%c3, %c3) {46 }47 return48}49 50// CHECK-LABEL: func @static_loop_with_step() {51// CHECK: [[C0:%.*]] = arith.constant 0 : index52// CHECK: [[C3:%.*]] = arith.constant 3 : index53// CHECK: [[C22:%.*]] = arith.constant 22 : index54// CHECK: [[C24:%.*]] = arith.constant 24 : index55// CHECK: [[C0_1:%.*]] = arith.constant 0 : index56// CHECK: [[C1:%.*]] = arith.constant 1 : index57// CHECK: [[C4:%.*]] = arith.constant 4 : index58// CHECK: [[V1:%.*]] = arith.muli [[C3]], [[C1]] : index59// CHECK: [[V2:%.*]] = arith.muli [[C3]], [[C4]] : index60// CHECK: scf.parallel ([[V3:%.*]], [[V4:%.*]]) = ([[C0]], [[C0]]) to ([[C22]], [[C24]]) step ([[V1]], [[V2]]) {61// CHECK: scf.parallel ([[V5:%.*]], [[V6:%.*]]) = ([[C0_1]], [[C0_1]]) to ([[V1]], [[V2]]) step ([[C3]], [[C3]]) {62// CHECK: = arith.addi [[V5]], [[V3]] : index63// CHECK: = arith.addi [[V6]], [[V4]] : index64// CHECK: }65// CHECK: }66// CHECK: return67 68// -----69 70func.func @tile_nested_innermost() {71 %c2 = arith.constant 2 : index72 %c0 = arith.constant 0 : index73 %c1 = arith.constant 1 : index74 scf.parallel (%i, %j) = (%c0, %c0) to (%c2, %c2) step (%c1, %c1) {75 scf.parallel (%k, %l) = (%c0, %c0) to (%c2, %c2) step (%c1, %c1) {76 }77 }78 scf.parallel (%i, %j) = (%c0, %c0) to (%c2, %c2) step (%c1, %c1) {79 }80 return81}82 83// CHECK-LABEL: func @tile_nested_innermost() {84// CHECK: [[C2:%.*]] = arith.constant 2 : index85// CHECK: [[C0:%.*]] = arith.constant 0 : index86// CHECK: [[C1:%.*]] = arith.constant 1 : index87// CHECK: scf.parallel ([[V1:%.*]], [[V2:%.*]]) = ([[C0]], [[C0]]) to ([[C2]], [[C2]]) step ([[C1]], [[C1]]) {88// CHECK: [[C0_1:%.*]] = arith.constant 0 : index89// CHECK: [[C1_1:%.*]] = arith.constant 1 : index90// CHECK: [[C4:%.*]] = arith.constant 4 : index91// CHECK: [[V3:%.*]] = arith.muli [[C1]], [[C1_1]] : index92// CHECK: [[V4:%.*]] = arith.muli [[C1]], [[C4]] : index93// CHECK: scf.parallel ([[V5:%.*]], [[V6:%.*]]) = ([[C0]], [[C0]]) to ([[C2]], [[C2]]) step ([[V3]], [[V4]]) {94// CHECK: [[V7:%.*]] = affine.min #{{.*}}([[V4]], [[C2]], [[V6]])95// CHECK: scf.parallel ([[V8:%.*]], [[V9:%.*]]) = ([[C0_1]], [[C0_1]]) to ([[V3]], [[V7]]) step ([[C1]], [[C1]]) {96// CHECK: = arith.addi [[V8]], [[V5]] : index97// CHECK: = arith.addi [[V9]], [[V6]] : index98// CHECK: }99// CHECK: }100// CHECK: }101// CHECK: [[C0_2:%.*]] = arith.constant 0 : index102// CHECK: [[C1_2:%.*]] = arith.constant 1 : index103// CHECK: [[C4_1:%.*]] = arith.constant 4 : index104// CHECK: [[V10:%.*]] = arith.muli [[C1]], [[C1_2]] : index105// CHECK: [[V11:%.*]] = arith.muli [[C1]], [[C4_1]] : index106// CHECK: scf.parallel ([[V12:%.*]], [[V13:%.*]]) = ([[C0]], [[C0]]) to ([[C2]], [[C2]]) step ([[V10]], [[V11]]) {107// CHECK: [[V14:%.*]] = affine.min #{{.*}}([[V11]], [[C2]], [[V13]])108// CHECK: scf.parallel ([[V15:%.*]], [[V16:%.*]]) = ([[C0_2]], [[C0_2]]) to ([[V10]], [[V14]]) step ([[C1]], [[C1]]) {109// CHECK: = arith.addi [[V15]], [[V12]] : index110// CHECK: = arith.addi [[V16]], [[V13]] : index111// CHECK: }112// CHECK: }113// CHECK: return114// CHECK: }115 116// -----117 118func.func @tile_nested_in_non_ploop() {119 %c0 = arith.constant 0 : index120 %c1 = arith.constant 1 : index121 %c2 = arith.constant 2 : index122 scf.for %i = %c0 to %c2 step %c1 {123 scf.for %j = %c0 to %c2 step %c1 {124 scf.parallel (%k, %l) = (%c0, %c0) to (%c2, %c2) step (%c1, %c1) {125 }126 }127 }128 return129}130 131// CHECK-LABEL: func @tile_nested_in_non_ploop132// CHECK: scf.for133// CHECK: scf.for134// CHECK: scf.parallel135// CHECK: scf.parallel136// CHECK: }137// CHECK: }138// CHECK: }139// CHECK: }140// CHECK: }141