-
Notifications
You must be signed in to change notification settings - Fork 1
/
Algorithm_2_DLS.m
223 lines (207 loc) · 8.75 KB
/
Algorithm_2_DLS.m
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
function nSchedule = Algorithm_2_DLS(SDFgraph, verbose)
if nargin < 2
verbose = 1;
end
%number of buffer
matrix_buffers = gen_init_buffers(SDFgraph);
%occurence of each actor
actor_occ = cal_occurrence_of_actors(SDFgraph);
for nProcessors = 1:1000
numIter = floor(30*sqrt(nProcessors));
[nSchedule, ~] = Algorithm_2_DLS_multi(SDFgraph, nProcessors, numIter, matrix_buffers, actor_occ, verbose);
if(~strcmp(nSchedule.type, 'unset'))
break;
end
end
end
function [nSchedule, maxBuff]= Algorithm_2_DLS_multi(SDFgraph, nProcessors, numIter, matrix_buffers, actor_occ, verbose)
nSchedule.type = 'unset';
maxBuff = 10^6;
nValid = 0;
for iter = 1:numIter
Schedule = Algorithm_2_DLS_impl(SDFgraph, nProcessors, matrix_buffers, actor_occ);
[Schedule, constraint_OK, nProc, nBuff] = Schedule_evaluate(SDFgraph, Schedule, 0);
if (constraint_OK==1 && nProc==nProcessors)
nValid = nValid + 1;
if( nBuff < maxBuff)
disp_verbose(verbose,['update ' num2str(nBuff)]);
maxBuff = nBuff;
nSchedule = Schedule;
end
end
end
disp_verbose(verbose,['num. of processors: ' num2str(nProcessors) '. Valid rate: ' num2str(double(nValid)/double(numIter))]);
end
function nSchedule = Algorithm_2_DLS_impl(SDFgraph, nProcessors, matrix_buffers, actor_occ)
nSchedule.type = 'PaSTA';
nSchedule.xmlns = 'http://peace.snu.ac.kr/CICXMLSchema';
nSchedule.taskGroup.name = 'task';
nSchedule.taskGroup.buffer = 0;
scheduleGroups = [];
%begin of the algorithm
max_buffers = matrix_buffers;
p_start = -1* ones(nProcessors, 1); %store next starting time
pool = zeros(nProcessors, 1); %store next available time
events = table;
nevent.type = "start";
nevent.time = 0;
nevent.actor= 0;
events = [events; struct2table(nevent)];
while(sum(actor_occ(:))>0 || size(events,1)>0)
events = sortrows(events,'time');
event = events(1,:);
if(event.type == "start")
runable_actor = runable_actors(SDFgraph, matrix_buffers, actor_occ);
avail_procs = procs_priority(p_start, pool, shuffle(available_procs(pool, event.time)));
for idx=1:length(avail_procs)
n_actors = length(runable_actor);
if(1 <= n_actors)
%add to scheduleGroups, change pool available time, change
%matrix_buffers, add new event to event pool, reduce actor
%occurence
rproc = avail_procs(idx);
runable_actor = actors_priority(shuffle(runable_actor), SDFgraph, matrix_buffers, max_buffers);
ractor_idxs = shuffle([runable_actor(1) runable_actor(randi(n_actors)) runable_actor(randi(n_actors))]);
ractor_idx = ractor_idxs(1);
runable_actor = runable_actor(find(runable_actor~=ractor_idx));
mActor = SDFgraph.actors(ractor_idx);
scheduleGroups = add_task_to_schedule(scheduleGroups, rproc, mActor, event.time);
pool(rproc) = event.time + mActor.execTime;
if(p_start(rproc) == -1)
p_start(rproc) = event.time;
end
actor_occ(ractor_idx) = actor_occ(ractor_idx) - 1;
for jdx = 1:size(SDFgraph.channels,1)
if(~isempty(SDFgraph.channels{jdx, ractor_idx}))
matrix_buffers(jdx, ractor_idx) = matrix_buffers(jdx, ractor_idx) - SDFgraph.channels{jdx, ractor_idx}.rate_out;
end
end
nevent.type = "finish";
nevent.time = pool(rproc);
nevent.actor= ractor_idx;
events = [events; struct2table(nevent)];
end
end
else
%matrix_buffers, add new event to event pool
ractor_idx = event.actor;
for rdx = 1:size(SDFgraph.channels,2)
if(~isempty(SDFgraph.channels{ractor_idx, rdx}))
matrix_buffers(ractor_idx, rdx) = matrix_buffers(ractor_idx, rdx) + SDFgraph.channels{ractor_idx, rdx}.rate_in;
if (matrix_buffers(ractor_idx, rdx) > max_buffers(ractor_idx, rdx))
max_buffers(ractor_idx, rdx) = matrix_buffers(ractor_idx, rdx);
end
end
end
nevent.type = "start";
nevent.time = event.time;
nevent.actor= 0;
events = [events; struct2table(nevent)];
end
events = events(2:end,:);
end
nSchedule.taskGroup.scheduleGroups = scheduleGroups;
end
function result = procs_priority(p_start, pool, procs)
criteria = pool(procs) - p_start(procs);
[~,I] = sort(criteria);
result = procs(I);
end
function result = actors_priority(runable_actor, SDFgraph, matrix_buffers, max_buffers)
criteria = zeros(size(runable_actor));
for idx=1:length(runable_actor)
ractor_idx = runable_actor(idx);
for rdx = 1:size(SDFgraph.channels,2)
if(~isempty(SDFgraph.channels{ractor_idx, rdx}))
matrix_buffers(ractor_idx, rdx) = matrix_buffers(ractor_idx, rdx) + SDFgraph.channels{ractor_idx, rdx}.rate_in;
if (matrix_buffers(ractor_idx, rdx) > max_buffers(ractor_idx, rdx))
criteria(idx) = criteria(idx) + matrix_buffers(ractor_idx, rdx) - max_buffers(ractor_idx, rdx);
else
criteria(idx) = criteria(idx) - 1.0/double(max_buffers(ractor_idx, rdx)-matrix_buffers(ractor_idx, rdx));
end
end
end
end
[~,I] = sort(criteria);
result = runable_actor(I);
end
function matrix_buffers = gen_init_buffers(SDFgraph)
matrix_buffers = zeros(size(SDFgraph.channels));
for idx = 1:size(matrix_buffers,1)
for jdx = 1:size(matrix_buffers,2)
if(~isempty(SDFgraph.channels{idx,jdx}))
matrix_buffers(idx, jdx) = SDFgraph.channels{idx,jdx}.initialTokens;
end
end
end
end
function runable_actor = runable_actors(SDFgraph, matrix_buffers, actor_occ)
runable_actor = [];
for ractor_idx = 1:size(SDFgraph.channels,2)
buffer_col = matrix_buffers(:, ractor_idx);
valid = 1;
for jdx = 1:size(SDFgraph.channels,1)
if(~isempty(SDFgraph.channels{jdx, ractor_idx}))
buffer_col(jdx) = buffer_col(jdx) - SDFgraph.channels{jdx, ractor_idx}.rate_out;
if (buffer_col(jdx) < 0)
valid = 0;
end
end
end
if (valid && (actor_occ(ractor_idx) > 0))
runable_actor = [runable_actor; ractor_idx];
end
end
end
function actor_occ = cal_occurrence_of_actors(SDFgraph)
actor_occ = ones(length(SDFgraph.actors), 1);
unconverged = 1;
while (unconverged)
unconverged = 0;
for idx = 1:size(SDFgraph.channels,1)
for jdx = 1:size(SDFgraph.channels,2)
if(~isempty(SDFgraph.channels{idx,jdx}))
s1 = lcm(actor_occ(idx), actor_occ(jdx));
s2 = lcm(SDFgraph.channels{idx,jdx}.rate_in, SDFgraph.channels{idx,jdx}.rate_out);
ss = lcm(s1,s2);
if(actor_occ(idx) ~= ss/SDFgraph.channels{idx,jdx}.rate_in)
actor_occ(idx) = ss/SDFgraph.channels{idx,jdx}.rate_in;
unconverged = 1;
end
if(actor_occ(jdx) ~= ss/SDFgraph.channels{idx,jdx}.rate_out)
actor_occ(jdx) = ss/SDFgraph.channels{idx,jdx}.rate_out;
unconverged = 1;
end
end
end
end
end
end
function procs = available_procs(pool, time)
procs = find(time >= pool);
end
function scheduleGroups = add_task_to_schedule(scheduleGroups, proc_id, mActor, time)
proc_idx = 0;
for idx=1:length(scheduleGroups)
schedule = scheduleGroups(:,idx);
if (schedule.localId == proc_id -1)
proc_idx = idx;
break;
end
end
clear idx schedule
if(proc_idx == 0)
schedule.localId = proc_id -1;
schedule.name = 'sg0';
schedule.poolName = 'p1';
schedule.scheduleType = 'static';
schedule.tasks = [];
scheduleGroups = [scheduleGroups schedule];
proc_idx = length(scheduleGroups);
end
task.name = mActor.name;
task.repetition = 1;
task.startTime = time;
task.endTime = time + mActor.execTime;
scheduleGroups(proc_idx).tasks = [scheduleGroups(proc_idx).tasks task];
end