Coverage for  / home / jenkins / .local / lib / python3.10 / site-packages / hyper_parallel / core / shard / ops / parallel_sort.py: 94%

1# Copyright 2026 Huawei Technologies Co., Ltd 

2# 

3# Licensed under the Apache License, Version 2.0 (the "License"); 

4# you may not use this file except in compliance with the License. 

5# You may obtain a copy of the License at 

6# 

7# http://www.apache.org/licenses/LICENSE-2.0 

8# 

9# Unless required by applicable law or agreed to in writing, software 

10# distributed under the License is distributed on an "AS IS" BASIS, 

11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

12# See the License for the specific language governing permissions and 

13# limitations under the License. 

14# ============================================================================ 

15""" 

16Distributed implementation for Sort operator. 

17""" 

18 

19from typing import Tuple 

20 

21from .parallel_ops import DistributedOp 

22 

23 

24def _normalize_sort_args(x, dim=-1, descending=False, stable=False): 

25 return (x,), {'dim': dim, 'descending': descending, 'stable': stable} 

26 

27 

28class SortDistributedOp(DistributedOp): 

29 """Distributed implementation for Sort operator.""" 

30 

31 def preprocess(self, args: tuple, kwargs: dict) -> tuple: 

32 """ 

33 Preprocess arguments for Sort operator. 

34 

35 Args: 

36 args (tuple): Input arguments, first element is the input tensor. 

37 kwargs (dict): Keyword arguments (dim, descending, stable). 

38 

39 Returns: 

40 tuple: (local_args, local_kwargs, cache_values) 

41 """ 

42 args, kwargs = _normalize_sort_args(*args, **kwargs) 

43 input_tensor = args[0] 

44 dim = kwargs['dim'] 

45 descending = kwargs['descending'] 

46 stable = kwargs['stable'] 

47 

48 local_args = (input_tensor.to_local(),) 

49 local_kwargs = {'dim': dim, 'descending': descending, 'stable': stable} 

50 

51 cache_values = [input_tensor.layout, dim] 

52 return local_args, local_kwargs, cache_values 

53 

54 def infer_layout(self, cache_values: list) -> Tuple[tuple, None]: 

55 """ 

56 Infer output layouts for Sort operator. 

57 

58 Rules: 

59 1. Input must not have Partial status. 

60 2. dim must be an integer within the valid range [-ndim, ndim-1]. 

61 3. The sort dimension must not be sharded (including StridedShard multi-axis mappings). 

62 4. Output values and indices layouts are identical to the input layout. 

63 

64 Args: 

65 cache_values (list): [input_layout, dim] where dim is the sort dimension. 

66 

67 Returns: 

68 tuple: ((values_layout, indices_layout), None) 

69 

70 Raises: 

71 ValueError: If input has Partial status, dim is out of range, or the sort dimension 

72 is sharded. 

73 """ 

74 layout = cache_values[0] 

75 dim = cache_values[1] 

76 

77 self._check_partial_inputs([layout]) 

78 

79 if not isinstance(dim, int): 

80 raise ValueError( 

81 f"For {self.op_name}, dimension should be int, but got {type(dim)}" 

82 ) 

83 

84 # Get tensor map to check sharding status 

85 in_tensor_map = layout.tensor_map 

86 ndim = len(in_tensor_map) 

87 

88 if dim < -ndim or dim >= ndim: 

89 raise ValueError( 

90 f"For {self.op_name}, dimension out of range " 

91 f"(expected to be in range of [{-ndim}, {ndim - 1}], but got {dim})" 

92 ) 

93 

94 if dim < 0: 

95 dim += ndim 

96 

97 # Check if the sorting dimension is sharded. 

98 # In tensor_map, -1 means Replicate (not sharded); any other value implies sharding. 

99 mapping = in_tensor_map[dim] 

100 if isinstance(mapping, (list, tuple)): 

101 is_sharded = any(m != -1 for m in mapping) 

102 else: 

103 is_sharded = mapping != -1 

104 

105 if is_sharded: 

106 raise ValueError( 

107 f"For {self.op_name}, sorting along a sharded dimension " 

108 f"(dim {dim} mapped to {mapping}) is not supported. " 

109 f"Please redistribute the tensor to Replicate on this dimension before sorting." 

110 ) 

111 

112 return ((layout, layout), None)