1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
|
/* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111-1301 USA */
#include "mariadb.h"
#include "filesort_utils.h"
#include "sql_const.h"
#include "sql_sort.h"
#include "table.h"
namespace {
/**
A local helper function. See comments for get_merge_buffers_cost().
*/
double get_merge_cost(ha_rows num_elements, ha_rows num_buffers, uint elem_size)
{
return
2.0 * ((double) num_elements * elem_size) / IO_SIZE
+ (double) num_elements * log((double) num_buffers) /
(TIME_FOR_COMPARE_ROWID * M_LN2);
}
}
/**
This is a simplified, and faster version of @see get_merge_many_buffs_cost().
We calculate the cost of merging buffers, by simulating the actions
of @see merge_many_buff. For explanations of formulas below,
see comments for get_merge_buffers_cost().
TODO: Use this function for Unique::get_use_cost().
*/
double get_merge_many_buffs_cost_fast(ha_rows num_rows,
ha_rows num_keys_per_buffer,
uint elem_size)
{
ha_rows num_buffers= num_rows / num_keys_per_buffer;
ha_rows last_n_elems= num_rows % num_keys_per_buffer;
double total_cost;
// Calculate CPU cost of sorting buffers.
total_cost=
( num_buffers * num_keys_per_buffer * log(1.0 + num_keys_per_buffer) +
last_n_elems * log(1.0 + last_n_elems) )
/ TIME_FOR_COMPARE_ROWID;
// Simulate behavior of merge_many_buff().
while (num_buffers >= MERGEBUFF2)
{
// Calculate # of calls to merge_buffers().
const ha_rows loop_limit= num_buffers - MERGEBUFF*3/2;
const ha_rows num_merge_calls= 1 + loop_limit/MERGEBUFF;
const ha_rows num_remaining_buffs=
num_buffers - num_merge_calls * MERGEBUFF;
// Cost of merge sort 'num_merge_calls'.
total_cost+=
num_merge_calls *
get_merge_cost(num_keys_per_buffer * MERGEBUFF, MERGEBUFF, elem_size);
// # of records in remaining buffers.
last_n_elems+= num_remaining_buffs * num_keys_per_buffer;
// Cost of merge sort of remaining buffers.
total_cost+=
get_merge_cost(last_n_elems, 1 + num_remaining_buffs, elem_size);
num_buffers= num_merge_calls;
num_keys_per_buffer*= MERGEBUFF;
}
// Simulate final merge_buff call.
last_n_elems+= num_keys_per_buffer * num_buffers;
total_cost+= get_merge_cost(last_n_elems, 1 + num_buffers, elem_size);
return total_cost;
}
/*
alloc_sort_buffer()
Allocate buffer for sorting keys.
Try to reuse old buffer if possible.
@return
0 Error
# Pointer to allocated buffer
*/
uchar **Filesort_buffer::alloc_sort_buffer(uint num_records,
uint record_length)
{
size_t buff_size;
uchar **sort_keys, **start_of_data;
DBUG_ENTER("alloc_sort_buffer");
DBUG_EXECUTE_IF("alloc_sort_buffer_fail",
DBUG_SET("+d,simulate_out_of_memory"););
buff_size= ((size_t)num_records) * (record_length + sizeof(uchar*));
set_if_bigger(buff_size, record_length * MERGEBUFF2);
if (!m_idx_array.is_null())
{
/*
Reuse old buffer if exists and is large enough
Note that we don't make the buffer smaller, as we want to be
prepared for next subquery iteration.
*/
sort_keys= m_idx_array.array();
if (buff_size > allocated_size)
{
/*
Better to free and alloc than realloc as we don't have to remember
the old values
*/
my_free(sort_keys);
if (!(sort_keys= (uchar**) my_malloc(buff_size,
MYF(MY_THREAD_SPECIFIC))))
{
reset();
DBUG_RETURN(0);
}
allocated_size= buff_size;
}
}
else
{
if (!(sort_keys= (uchar**) my_malloc(buff_size, MYF(MY_THREAD_SPECIFIC))))
DBUG_RETURN(0);
allocated_size= buff_size;
}
m_idx_array= Idx_array(sort_keys, num_records);
m_record_length= record_length;
start_of_data= m_idx_array.array() + m_idx_array.size();
m_start_of_data= reinterpret_cast<uchar*>(start_of_data);
DBUG_RETURN(m_idx_array.array());
}
void Filesort_buffer::free_sort_buffer()
{
my_free(m_idx_array.array());
m_idx_array.reset();
m_start_of_data= NULL;
}
void Filesort_buffer::sort_buffer(const Sort_param *param, uint count)
{
size_t size= param->sort_length;
if (count <= 1 || size == 0)
return;
uchar **keys= get_sort_keys();
uchar **buffer= NULL;
if (radixsort_is_appliccable(count, param->sort_length) &&
(buffer= (uchar**) my_malloc(count*sizeof(char*),
MYF(MY_THREAD_SPECIFIC))))
{
radixsort_for_str_ptr(keys, count, param->sort_length, buffer);
my_free(buffer);
return;
}
my_qsort2(keys, count, sizeof(uchar*), get_ptr_compare(size), &size);
}
|
