tapasco.1 8.47 KB
Newer Older
Jens Korinth's avatar
Jens Korinth committed
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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
.TH tapasco 1 "May 11, 2017" "version 2017.1"  "USER COMMANDS"
.SH NAME
tapasco \- The Task Parallel System Composer
.SH SYNOPSIS
.B tapasco
[global options]* [jobs]*
.SH DESCRIPTION
The
.B Task Parallel System Composer (TaPaSCo, or TPC)
is an FPGA toolchain to compose a task parallel on\-chip microarchitecture for
multiple instances of application\-specific processing elements (PEs). TaPaSCo
features an automatic design space exploration approach: It can vary both target
operating frequency, as well as number of PE instances to aim for the ideal
trade\-off between area utilization and frequency (the performance sweet spot
for FPGA architectures).

." *****************************************************************************
.SH GLOBAL OPTIONS
Global options affect all jobs in this run of tapasco.
.TP
\-\-archDir [PATH]
.RS
Base directory for architecture descriptions
.RE
.TP
\-\-compositionDir [PATH]
.RS
Output base directory for Compose jobs
.RE
.TP
\-\-coreDir [PATH}
.RS
Output base directory for HLS jobs, synthesized cores
.RE
.TP
\-\-kernelDir [PATH]
.RS
Base directory for kernel descriptions (HLS)
.RE
.TP
\-\-platformDir [PATH]
.RS
Base directory for platform descriptions
.RE
.TP
\-\-logFile [FILE]
.RS
Path to output log file
.RE
.TP
\-\-configFile [FILE]
.RS
Path to Json file with Configuration
.RE
.TP
\-\-jobsFile [FILE]
.RS
Path to Json file with Jobs array
.RE
.TP
\-\-slurm
.RS
Activate SLURM cluster execution (requires SLURM/sbatch)
.RE

." *****************************************************************************
.SH COMPOSITION SYNTAX
Compositions are a sequence of pairs of processing elements and their instance
count. They can be specified as separate Json files, or inline:
.PP
\'[\' [NAME] \'x\' [COUNT] (\',\' [NAME] \'x\' [COUNT])* \']\'
.TP
where NAME is a valid PE identifier and COUNT a positive number > 0.
.SS Example
[counter x 24, arrayinit x 42]

." *****************************************************************************
.SH JOBS
A
.B job
is a set of parameters for one of the activities TaPaSCo can perform:
bulkimport, compose, corestats, dse, hls, import.
See below for descriptions of the jobs and their corresponding activities.
.PP
Any number of consecutive jobs can be specified and will be executed
sequentially. There is a Json format for all jobs, and a
.B job file
containing a Json array of jobs can be specified instead.

." -----------------------------------------------------------------------------
.SS bulkimport \-\-csv [FILE]
Parses a file in
.I  CSV (comma-separated values)
format containing the following header:
.PP
.SM "Zip, ID, Description, Architecture, Platform, Avg Runtime (clock cycles)"
.PP
where Zip is the path to zip file, ID is the TaPaSCO function ID (used to call
the PE in the APIs), Description is an optional text, Architecture and Platform
are the names of the import Target.
.PP
.TP
bulkimport \-\-csv [FILE]
.RS
[FILE] should be in comma-separated values (CSV) format and must contain the
header line.
.RE

." -----------------------------------------------------------------------------
.SS compose [composition] [frequency] [implementation]? [options]*
Composes a microarchitecture for the given Composition. Will produce a .bit
bitstream file in the \-\-compositionDir.
.PP
.B Mandatory Arguments
.TP
\-\-composition {[FILE] | [DEF]}
.RS
Threadpool composition, either in a separate Json file, or inline as arguments
(see COMPOSITION SYNTAX).
.RE
.TP
\-\-designFrequency [NUM]
.RS
Target design frequency (PE clock) in MHz.
.RE
.TP
\-\-implementation [NAME]
Composer implementation (default: Vivado)
.PP
.B Options
.TP
\-\-architectures|\-a [NAME] (\',\' [NAME])*
.RS
Filter for Architectures, restrict set of Architectures to names.
.RE
.TP
\-\-platforms|-p [NAME] (\',\' [NAME])*
.RS
Filter for Platforms, restrict set of Platforms to names.
.RE
.TP
\-\-debugMode [NAME]
.RS
Activates the given debug mode; currently implemented modes are:
.RS
.PP
\'r\' generates random results, ranging from Success, PlacerErrors to
TimingFailures
.PP
\'p\' always generate PlacerError
.PP
\'t\' always generate TimingFailure and
.PP
\'s\' always generate a Success, w/o reports.
.RE
.RE

." -----------------------------------------------------------------------------
.SS corestats [options]
Collects results for all Cores in in coreDir from different reports (synthesis,
utilization, power, ...) and stores them in one CSV file per Target, i.e.,
Architecture + Platform combination.
.PP
.B Options
.TP
\-\-prefix [PREFIX]
.RS
Prefix file names with [PREFIX]. Files will be called
"[PREFIX][ARCH]@[PLATFORM].csv".
.RE
.TP
\-\-architectures|\-a [NAME] (\',\' [NAME])*
.RS
Filter for Architectures, restrict set of Architectures to names.
.RE
.TP
\-\-platforms|-p [NAME] (\',\' [NAME])*
.RS
Filter for Platforms, restrict set of Platforms to names.
.RE

." -----------------------------------------------------------------------------
.SS dse [composition] [dimensions] [batch size] [options]*
Perform automatic
.I design space exploration (DSE)
for the given Composition, Targets (Architecture + Platform) and dimensions.
.PP
.B Mandatory Arguments
.TP
\-\-composition {[FILE] | [DEF]}
.RS
Threadpool composition, either in a separate Json file, or inline as arguments
(see COMPOSITION SYNTAX).
.RE
.TP
\-\-dimensions [dimension] [\',\' [dimension]]* where [dimension] is one of
.RS
.I area
activate utilization variation, i.e., linear scaling of PE instance counts
.PP
.I frequency
activate frequency variation, i.e., automatic selection of PE clock frequency
.PP
.I alternatives
activate alternative implementations, i.e., allow TaPaSCo to replace any PE with
a PE with the same ID.
.RS
.PP
Example:
Let "counter" and "countdown" both have ID 14; then a Composition
.RS
[counter x 12]
.RE
may be changed by the DSE to be
.RS
[countdown x 12]
.RE
.RE
.RE
.TP
\-\-batchSize [NUM]
.RS
Excute [NUM] runs in parallel in each batch; should not exceed the number of
available processors and licences.
.RE

.PP
.B Options
.TP
\-\-basePath [PATH]
.RS
Output base directory for all Compositions etc. (default: $TAPASCO_HOME/DSE_[TIMESTAMP]).
.RE
.TP
\-\-heuristic [heuristic]
Select design space ordering heuristic (default: job throughput).
.TP
\-\-frequency [NUM]
.RS
Initial target design frequency (PE clock) in MHz.
.RE
.TP
\-\-implementation [NAME]
.RS
Composer implementation (default: Vivado)
.RE
.TP
\-\-architectures|\-a [NAME] (\',\' [NAME])*
.RS
Filter for Architectures, restrict set of Architectures to names.
.RE
.TP
\-\-platforms|-p [NAME] (\',\' [NAME])*
.RS
Filter for Platforms, restrict set of Platforms to names.
.RE
.TP
\-\-debugMode [NAME]
.RS
Activates the given debug mode; see compose.
.RE

." -----------------------------------------------------------------------------
.SS hls [options]*
Execute High-Level Synthesis to turn Kernel into Core instances for the
specified Targets (Architecture + Platform).
.PP
.B Options
.TP
\-\-implementation [NAME]
.RS
Composer implementation (default: VivadoHLS)
.RE
.TP
\-\-kernels|-k [NAME] (\',\' [NAME])*
.RS
Filter for Kernels, restrict set of Kernels to names.
.RE
.TP
\-\-architectures|\-a [NAME] (\',\' [NAME])*
.RS
Filter for Architectures, restrict set of Architectures to names.
.RE
.TP
\-\-platforms|-p [NAME] (\',\' [NAME])*
.RS
Filter for Platforms, restrict set of Platforms to names.
.RE

." -----------------------------------------------------------------------------
.SS import [options]*
Imports a existing IP-XACT core archive into the core library of TaPaSCo for the
specified Targets (Architecture + Platform). Will perform
.I out-of-context (OOC) synthesis
if existing reports cannot be found, to generate estimates for area utilization,
max. operating frequency and power consumption.
.PP
.B Options
.TP
\-\-zip [FILE]
.RS
Path to .zip file containing IP-XACT core (sources + component.xml).
.RE
.TP
\-\-id [NUM]
.RS
Kernel ID (> 0) for use in TaPaSCO (used by APIs).
.RE
.TP
\-\-averageClockCycles [NUM]
.RS
Avg. clock cycles per execution (optional).
.RE
.TP
\-\-description "[TEXT]"
.RS
Kernel description text (optional).
.RE
.TP
\-\-architectures|\-a [NAME] (\',\' [NAME])*
.RS
Filter for Architectures, restrict set of Architectures to names.
.RE
.TP
\-\-platforms|-p [NAME] (\',\' [NAME])*
.RS
Filter for Platforms, restrict set of Platforms to names.
.RE
.SH ENVIRONMENT
.TP
.B TAPASCO_HOME
.RS
Path to TaPaSCo root installation folder (where setup.sh is located).
.RE

.SH SEE ALSO
itapasco(1), tapasco-build-libs(1), tapasco-logviewer(1),
tapasco-reportviewer(1), tapasco-load-bitstream(8), libtapasco(3),
libplatform(3), tapasco-api.h(3), tapasco-api.hpp(3), platform-api.h(3)