Project

General

Profile

OsmoTRX » History » Version 50

neels, 06/28/2016 02:02 PM

1 41 sylvain
{{>toc}}
2 1 ttsou
3 41 sylvain
h1. [[OsmoTRX]]
4 1 ttsou
5 1 ttsou
6 41 sylvain
[[OsmoTRX]] is a software-defined radio transceiver that implements the Layer 1 physical layer of a BTS comprising the following 3GPP specifications:
7 41 sylvain
* TS 05.01 "Physical layer on the radio path"
8 41 sylvain
* TS 05.02 "Multiplexing and Multiple Access on the Radio Path"
9 41 sylvain
* TS 05.04 "Modulation"
10 41 sylvain
* TS 05.10 "Radio subsystem synchronization"
11 1 ttsou
12 49 neels
[[OsmoTRX]] is based on the transceiver code from the [[OsmoBTS:OpenBTS]] project, but setup to operate independently with the purpose of using with non-OpenBTS software and projects, while still maintaining backwards compatibility with [[OsmoBTS:OpenBTS]]. Currently there are numerous features contained in [[OsmoTRX:]] that extend the functionality of the [[OsmoBTS:OpenBTS]] transceiver. These features include enhanced support for various embedded platforms - notably ARM - and dual channel diversity support for the Fairwaves [[umtrx:]].
13 41 sylvain
14 46 laforge
h2. OsmoTRX in the Osmocom GSM architecture
15 46 laforge
16 46 laforge
{{graphviz_link()
17 46 laforge
digraph G {
18 46 laforge
    rankdir = LR;
19 46 laforge
    SDR -> OsmoTRX [label="Raw Samples"];
20 46 laforge
    OsmoTRX -> OsmoBTS [label="bursts over UDP"];
21 46 laforge
    OsmoBTS -> OsmoNITB [label="Abis/IP"];
22 46 laforge
    OsmoBTS -> OsmoPCU [label="pcu_sock"];
23 46 laforge
    OsmoPCU -> OsmoSGSN [label="Gb/IP"];
24 46 laforge
    OsmoTRX [color=red];
25 46 laforge
}
26 46 laforge
}}
27 41 sylvain
28 41 sylvain
h2. Features
29 41 sylvain
30 41 sylvain
31 41 sylvain
*Intel SSE Support*
32 6 ttsou
* SSE3
33 6 ttsou
* SSE4.1
34 20 ttsou
35 41 sylvain
On Intel processors, [[OsmoTRX]] makes heavy use of the Streaming SIMD Extensions (SSE) instruction set. Accelerated operations include pulse shape filtering, resampling, sequence correlation, and many other signal processing operations. SSE3 is the minimum requirement for accelerated use.
36 1 ttsou
37 20 ttsou
SSE3 is present in the majority of Intel processors since later versions of the Pentium 4 architecture and is also present on low power Atom processors. Support is automatically detected at build time. For additional performance information, please see the performance and benchmarks section.
38 29 ttsou
39 41 sylvain
*ARM Support*
40 1 ttsou
* NEON
41 1 ttsou
* NEON-VFPv4
42 20 ttsou
43 41 sylvain
[[OsmoTRX]] runs on a variety of ARM processors with and without NEON coprocessors. Like SSE on Intel processors, NEON provides acceleration with SIMD vectorized instructions.
44 20 ttsou
45 1 ttsou
Tested popular architectures include ARM11 (Raspberry Pi), Cortex-A8 (!BeagleBoard), and Cortex-A15 (!ArndaleBoard). Loosely speaking, these platforms are representative of low cost embedded devices, mid-level handsets, and high-end smartphones respectively. Similarly, in order, these platforms include no NEON coprocessor, standard NEON, and NEON-VFPv4. The latter NEON variation, VFPv4, provides additional fused-multiply-accumulate (FMA) instructions useful for many DSP operations.
46 1 ttsou
47 26 ttsou
NEON support must be enabled by the user at build time. For additional information, please see the configuration and performance and benchmarks sections.
48 37 ttsou
49 41 sylvain
*Dual Channel (UmTRX and B210)*
50 7 ttsou
51 1 ttsou
Two dual channel modes are available: standard dual channel mode and diversity. In standard dual channel mode, each RF
52 28 ttsou
path of the dual channel device supports a different ARFCN. Each path operates independently a
53 1 ttsou
nd operates similarly to two separate devices. GSM channel capacity in this mode is doubled. This option can be configured at run time from the command line.
54 1 ttsou
55 41 sylvain
*Dual Channel Diversity (UmTRX, experimental)*
56 1 ttsou
57 28 ttsou
Diversity mode is similar to the standard dual channel mode except each antenna supports both ARFCN channels. In this case, the receiver sample bandwidth is widened to handle both ARFCN's and subsequently converted and demultiplexed into separate sample streams. Each GSM receive path is fed dual signals, where antenna selection diversity is performed by taking the stronger signal on a burst-by-burst basis. This diversity setup improves uplink reception performance in multipath fading environments.
58 16 ttsou
59 28 ttsou
Limitations are increased CPU utilization and that ARFCN spacing is restricted (currently at 400 kHz) by the receiver sampling bandwidth. Setting the ARFCN spacing beyond the sampling limit will disable the diversity path and operate in standard dual channel mode. This options can be configured at run time from the command line.
60 20 ttsou
61 41 sylvain
*Uplink Burst Detection*
62 39 ttsou
63 49 neels
[[OsmoTRX]] utilizes an updated receive burst detection algorithm that provides greater sensitivity and reliability than the original [[OsmoBTS:OpenBTS]] approach, which relied on energy detection for the initial stage of burst acquisition.
64 39 ttsou
65 1 ttsou
The limitation of the previous approach was that it was slow to adapt to highly transient power levels and false burst detection in challenging situations such as receiver saturation, which may occur in close range lab testing. The other issue was that a high degree of level tuning was often necessary to operate reliably.
66 1 ttsou
67 1 ttsou
The current receiver code addressed those limitations for improved performance in a wider variety of environments.
68 1 ttsou
69 41 sylvain
*Low Phase Error Modulator*
70 16 ttsou
71 16 ttsou
The default GSM downlink signal is configured for low distortion using a linearized GMSK modulator. The implementation is based on a two pulse Laurent approximation of continuous phase modulated (CPM) signals. The baseband output signal measures with very low phase error and is capable of passing industry spectrum mask requirements. Please note that actual performance will depend strongly on the particular device in use.
72 1 ttsou
73 41 sylvain
Theoretical details can be found in the report on "GMSK":http://tsou.cc/gsm/report_gmsk.pdf. Octave / Matlab code for "pulse generation":http://tsou.cc/gsm/laurent.m is also available.
74 1 ttsou
75 1 ttsou
This option can be enabled or disabled at run time from the command line.
76 28 ttsou
77 28 ttsou
Very Low Phase Error (Ettus Research N200)
78 16 ttsou
79 42 laforge
!http://tsou.cc/gsm/osmo-trx-phase75.gif!
80 1 ttsou
81 21 ttsou
Spectrum Mask (Ettus Research N200)
82 1 ttsou
83 42 laforge
!http://tsou.cc/gsm/osmo-trx-spectrum75.gif!
84 1 ttsou
85 41 sylvain
h2. RF Hardware support
86 1 ttsou
87 1 ttsou
88 50 neels
Multiple RF devices are currently supported. These include USRP family products from Ettus Research, and the [[UmTRX:]] from Fairwaves.
89 41 sylvain
90 41 sylvain
||*Fairwaves*||*Notes*||
91 20 ttsou
||UmTRX||Dual channel||
92 20 ttsou
93 20 ttsou
All Ettus Research devices are supported.
94 1 ttsou
95 41 sylvain
||*Ettus Research*||*Notes*||
96 1 ttsou
||USRP1||Requires legacy libusrp driver and clocking modification||
97 1 ttsou
||USRP2||10 MHz external reference required||
98 1 ttsou
||B100||
99 1 ttsou
||B110||
100 1 ttsou
||B200||GPSDO or 10 MHz external reference recommended||
101 1 ttsou
||B210||Dual channel, 10 MHz external reference recommended||
102 1 ttsou
||N200||
103 20 ttsou
||N210||
104 1 ttsou
||E100||
105 1 ttsou
||E110||
106 1 ttsou
107 41 sylvain
h2. Embedded Platform Support
108 1 ttsou
109 41 sylvain
110 41 sylvain
[[OsmoTRX]] has been tested on the multiple embedded platforms representing a wide range of device types. Low cost ARM devices are generally limited by memory and I/O as much CPU utilization.
111 41 sylvain
112 1 ttsou
Running a full or near full ARFCN configuration (7 simultaneous TCH channels with Combination V) may require running the GSM stack remotely, which can be configured at runtime on the command line. This limitation appears to be scheduling related more so than lack of CPU resources, and may be resolved at a later time.
113 1 ttsou
114 43 laforge
|_.Platform|_.SoC*|_.Processor|_.SIMD/FPU|_.Testing Notes|
115 43 laforge
|ArndaleBoard|Samsung Exynos 5250|ARM Cortex-A15|NEON-VFPv4|7 TCH|
116 43 laforge
|BeagleBoard-xM|Texas Instruments OMAP3|ARM Cortex-A8|NEON|7 TCH, remote [[osmobts:]] stack|
117 43 laforge
|Ettus E100|Texas Instruments OMAP3|ARM Cortex-A8|NEON|7 TCH, remote [[osmobts:]] stack|
118 43 laforge
|Raspberry Pi|Broadcom BCM2835|ARM11|VFP|2 TCH, remote [[osmobts:]] stack|
119 43 laforge
|Shuttle PC|NA|Intel Atom D2550|SSE3|Dual channel, 15 TCH|
120 1 ttsou
121 1 ttsou
All embedded plaforms were tested with low-phase error modulator disabled. Use of the more accurate modulator on embedded platforms has not been extensively tested.
122 19 ttsou
123 41 sylvain
h2. Mailing List
124 22 ttsou
125 41 sylvain
126 49 neels
For development purposes, [[OsmoTRX:]] is discussed on both [[OsmoBTS:OpenBTS]] and [[OpenBSC:]] mailing lists at openbts-discuss@lists.sourceforge.net and openbsc@lists.osmocom.org respectively.
127 41 sylvain
128 1 ttsou
Please direct questions and bug reports to the list appropriate for the GSM stack being used.
129 41 sylvain
130 47 laforge
Subscription information is available at "and [http://lists.osmocom.org/mailman/listinfo/openbsc/":https://lists.sourceforge.net/lists/listinfo/openbts-discuss].  Please make sure to read our [[cellular-infrastructure:MailingListRules]] before posting.
131 1 ttsou
132 41 sylvain
h2. GPRS support
133 1 ttsou
134 1 ttsou
135 44 laforge
[[OsmoTRX]] supports GPRS through [[osmobts:]].
136 1 ttsou
137 49 neels
For GPRS support with [[OsmoBTS:OpenBTS]], please use the transceiver supplied with [[OsmoBTS:OpenBTS]].
138 41 sylvain
139 41 sylvain
140 41 sylvain
h2. Source code
141 41 sylvain
142 41 sylvain
143 1 ttsou
The source code is available from git.osmocom.org (module osmo-trx).
144 18 ttsou
145 18 ttsou
Public read-only access is available via
146 41 sylvain
<pre>
147 19 ttsou
$ git clone git://git.osmocom.org/osmo-trx
148 41 sylvain
</pre>
149 1 ttsou
You can browse it via cgit: http://cgit.osmocom.org/cgit/osmo-trx/
150 1 ttsou
151 48 neels
h2. Dependencies
152 1 ttsou
153 48 neels
Install libusb-1.0 and libbost dev packages. On debian 8.4:
154 1 ttsou
155 48 neels
<pre>
156 48 neels
sudo apt-get install --no-install-recommends libusb-1.0-0-dev libboost-dev
157 48 neels
</pre>
158 41 sylvain
159 48 neels
*UHD*
160 1 ttsou
161 48 neels
Unless using USRP1, you will need the Universal Hardware Driver (UHD),
162 48 neels
which is available from Ettus Research or Fairwaves; the UHD implementation
163 48 neels
must match your hardware:
164 48 neels
165 48 neels
* Ettus Research UHD for USRP devices
166 48 neels
* Fairwaves UHD with [[UmTRX]]
167 48 neels
* USRP1 does not use the UHD driver, it is supported through the legacy libusrp driver provided in GNU Radio 3.4.2.
168 48 neels
169 48 neels
*Debian*
170 48 neels
171 48 neels
At time of writing, the debian 8.4 packages for UHD are sufficient for running osmo-trx and osmo-bts-trx.
172 48 neels
here are some of the packages that need to be installed:
173 48 neels
174 48 neels
<pre>
175 48 neels
sudo apt-get install --no-install-recommends libuhd-dev uhd-host
176 48 neels
</pre>
177 48 neels
178 48 neels
It may be necessary to use the pothos PPA instead:
179 48 neels
180 48 neels
<pre>
181 48 neels
sudo add-apt-repository ppa:guruofquality/pothos
182 48 neels
sudo apt-get update
183 48 neels
sudo apt install libboost-dev uhd
184 48 neels
</pre>
185 48 neels
186 48 neels
*Firmware*
187 48 neels
188 48 neels
You also need to download the firmware using a script provided by the UHD package.
189 48 neels
Instructions suggest running the script as root, but this way is less dangerous:
190 48 neels
191 48 neels
<pre>
192 48 neels
sudo mkdir /usr/share/uhd
193 48 neels
sudo chown $USER: /usr/share/uhd
194 48 neels
/usr/lib/uhd/utils/uhd_images_downloader.py
195 48 neels
</pre>
196 48 neels
197 48 neels
*Group*
198 48 neels
199 48 neels
You may need to add yourself to the usrp group:
200 48 neels
201 48 neels
<pre>
202 48 neels
sudo gpasswd -a $USER usrp
203 48 neels
# and re-login to acquire the group
204 48 neels
</pre>
205 48 neels
206 48 neels
*Verify*
207 48 neels
208 48 neels
run uhd_find_devices to make sure b200 is available:
209 48 neels
210 48 neels
<pre>
211 48 neels
$ uhd_find_devices 
212 48 neels
linux; GNU C++ version 4.9.1; Boost_105500; UHD_003.007.003-0-unknown
213 48 neels
214 48 neels
--------------------------------------------------
215 48 neels
-- UHD Device 0
216 48 neels
--------------------------------------------------
217 48 neels
Device Address:
218 48 neels
    type: b200
219 48 neels
    name: MyB210
220 48 neels
    serial: 1C0FFEE
221 48 neels
    product: B210
222 48 neels
</pre>
223 48 neels
224 48 neels
h2. Configuration and Build
225 48 neels
226 41 sylvain
First, run autoreconf to remake the build system files.
227 1 ttsou
<pre>
228 18 ttsou
$ autoreconf -i
229 41 sylvain
...
230 18 ttsou
</pre>
231 41 sylvain
232 18 ttsou
*Intel Platforms (All)*
233 1 ttsou
234 41 sylvain
Intel SSE support is automatically detected on Intel x86 platforms. No user intervention is necessary. The general configuration defaults to the low phase error modulator. Atom users may wish to use the low-CPU utilization modulator, which can be later enabled from the command line at runtime.
235 18 ttsou
<pre>
236 1 ttsou
$ ./configure
237 1 ttsou
...
238 19 ttsou
checking whether mmx is supported... yes
239 18 ttsou
checking whether sse is supported... yes
240 18 ttsou
checking whether sse2 is supported... yes
241 18 ttsou
checking whether sse3 is supported... yes
242 18 ttsou
checking whether ssse3 is supported... yes
243 18 ttsou
checking whether sse4.1 is supported... yes
244 18 ttsou
checking whether sse4.2 is supported... yes
245 41 sylvain
...
246 18 ttsou
</pre>
247 41 sylvain
248 18 ttsou
*ARM Platforms with NEON*
249 41 sylvain
250 41 sylvain
Many popular ARM development boards fall under this category including BeagleBoard, PandaBoard, and Ettus E100 USRP. This option will disable the low phase error modulator, which can be re-enabled at runtime. NEON support must be manually enabled.
251 24 ttsou
<pre>
252 41 sylvain
$ ./configure --with-neon
253 1 ttsou
</pre>
254 41 sylvain
255 1 ttsou
*ARM Platforms with NEON-VFPv4*
256 41 sylvain
257 41 sylvain
Currently very few development platforms support this instruction set, which is seen mainly in high end smartphones and tablets. Available development boards are ArndaleBoard and ODROID-XU. This option will disable the low phase error modulator, which can be re-enabled at runtime. NEON-VFPv4 support must be manually enabled.
258 1 ttsou
<pre>
259 41 sylvain
$ ./configure --with-neon-vfpv4
260 1 ttsou
</pre>
261 41 sylvain
262 1 ttsou
*ARM Platforms without NEON*
263 41 sylvain
264 1 ttsou
This configuration mainly targets the Raspberry Pi. ARM platforms without NEON vector units are almost always very slow processors, and generally not very suitable for running [[OsmoTRX]]. Running [[OsmoTRX]] on a Raspberry Pi, however, is possible along with limited TCH (voice) channel support. Currently this configuration requires minor code changes.
265 1 ttsou
266 1 ttsou
Coming soon...
267 41 sylvain
268 1 ttsou
*Build and Install*
269 16 ttsou
270 16 ttsou
After configuration, installation is simple.
271 41 sylvain
272 16 ttsou
<pre>
273 16 ttsou
$ make
274 41 sylvain
$ sudo make install
275 16 ttsou
</pre>
276 16 ttsou
277 41 sylvain
h2. Running
278 16 ttsou
279 41 sylvain
280 41 sylvain
[[OsmoTRX]] can be configured with a variety of options on the command line. In most cases, the default settings will suffice. Notable options include UHD device argument passing, which is often useful for using network based devices with firewalls, and external 10 MHz reference support.
281 41 sylvain
282 41 sylvain
<pre>
283 16 ttsou
$ osmo-trx -h
284 16 ttsou
linux; GNU C++ version 4.8.1 20130603 (Red Hat 4.8.1-1); Boost_105300; UHD_003.005.004-140-gfb32ed16
285 16 ttsou
286 16 ttsou
Options:
287 16 ttsou
  -h    This text
288 1 ttsou
  -a    UHD device args
289 16 ttsou
  -l    Logging level (EMERG, ALERT, CRT, ERR, WARNING, NOTICE, INFO, DEBUG)
290 16 ttsou
  -i    IP address of GSM core
291 1 ttsou
  -p    Base port number
292 1 ttsou
  -d    Enable dual channel diversity receiver
293 16 ttsou
  -x    Enable external 10 MHz reference
294 16 ttsou
  -s    Samples-per-symbol (1 or 4)
295 38 ttsou
  -c    Number of ARFCN channels (default=1)
296 38 ttsou
  -f    Enable C0 filler table
297 16 ttsou
  -o    Set baseband frequency offset (default=auto)
298 41 sylvain
</pre>
299 16 ttsou
300 41 sylvain
<pre>
301 1 ttsou
$ osmo-trx -a "addr=192.168.10.2"
302 16 ttsou
linux; GNU C++ version 4.8.1 20130603 (Red Hat 4.8.1-1); Boost_105300; UHD_003.004.000-b14cde5
303 16 ttsou
304 16 ttsou
Config Settings
305 16 ttsou
   Log Level............... INFO
306 1 ttsou
   Device args............. addr=192.168.10.2
307 16 ttsou
   TRX Base Port........... 5700
308 1 ttsou
   TRX Address............. 127.0.0.1
309 16 ttsou
   Channels................ 1
310 16 ttsou
   Samples-per-Symbol...... 4
311 16 ttsou
   External Reference...... Disabled
312 16 ttsou
   Diversity............... Disabled
313 16 ttsou
314 41 sylvain
-- Opening a [[UmTRX]] device...
315 13 ttsou
-- Current recv frame size: 1472 bytes
316 38 ttsou
-- Current send frame size: 1472 bytes
317 41 sylvain
-- Setting [[UmTRX]] 4 SPS
318 38 ttsou
-- Transceiver active with 1 channel(s)
319 41 sylvain
</pre>
320 38 ttsou
321 1 ttsou
322 49 neels
h2. [[OsmoTRX]] with [[OsmoBTS:OpenBTS]]
323 38 ttsou
324 38 ttsou
325 49 neels
[[OsmoTRX]] is fully compatible with [[OsmoBTS:OpenBTS]] for voice and SMS services. Due to differences in handing of GPRS, [[OsmoTRX]] does not support GPRS when used with [[OsmoBTS:OpenBTS]], however, GPRS with the Osmocom stack is supported.
326 41 sylvain
327 49 neels
For use with [[OsmoBTS:OpenBTS]], enable the filler table option "Enable C0 filler table", which enables [[OsmoBTS:OpenBTS]] style idle bursts and retransmissions.
328 41 sylvain
329 41 sylvain
<pre>
330 1 ttsou
$ osmo-trx -f
331 41 sylvain
</pre>
332 17 ttsou
333 49 neels
The [[OsmoTRX]] transceiver should be started before running [[OsmoBTS:OpenBTS]]. No symbolic link to './transceiver' should exist in the [[OsmoBTS:OpenBTS]] directory. This prevents [[OsmoBTS:OpenBTS]] from starting its own transceiver instance.
334 35 ttsou
335 1 ttsou
336 41 sylvain
h2. Benchmarks
337 1 ttsou
338 35 ttsou
339 49 neels
A variety of performance benchmarks are available for various code optimizations. These include floating point - integer conversions, convolution, and convolutional decoding. Note that convolutional decoding does not take place in [[OsmoTRX]], but one stop higher in the Layer 1 stack - either in [[osmobts:]] or [[OsmoBTS:OpenBTS]] core.
340 35 ttsou
341 41 sylvain
*Repository*
342 41 sylvain
343 41 sylvain
Currently the trx-bench repository holds the test files and contains the same NEON and SSE code as [[OsmoTRX]]. The test code may be merged into [[OsmoTRX]] at a later time, but, for now, it exists as a separate repository. NEON configure options are the same as [[OsmoTRX]].
344 41 sylvain
345 41 sylvain
<pre>
346 35 ttsou
$ git clone https://github.com/ttsou/trx-bench.git
347 35 ttsou
348 35 ttsou
$ cd trx-bench
349 35 ttsou
$ autoreconf -i
350 35 ttsou
$ ./configure [--with-neon] [--with-neon-vfp4]
351 1 ttsou
$ make
352 1 ttsou
$ src/conv_test
353 35 ttsou
$ src/convert_test
354 35 ttsou
$ src/convolve_test
355 41 sylvain
</pre>
356 35 ttsou
357 35 ttsou
The convolutional decoding test includes command options including experimental support for benchmarking with multiple threads.
358 35 ttsou
359 41 sylvain
<pre>
360 35 ttsou
$ ./conv_test -h
361 35 ttsou
Options:
362 35 ttsou
  -h    This text
363 1 ttsou
  -i    Number of iterations
364 1 ttsou
  -j    Number of threads for benchmark (1 to 32)
365 13 ttsou
  -b    Run benchmark tests
366 13 ttsou
  -a    Run validity checks
367 13 ttsou
  -e    Run bit-error-rate tests
368 41 sylvain
</pre>
369 10 ttsou
370 1 ttsou
Selected benchmark results are provided below. All tests are run on a single core only.
371 1 ttsou
372 41 sylvain
*Intel Haswell (i7 4770K 3.5 GHz)*
373 1 ttsou
374 41 sylvain
<pre>
375 1 ttsou
--- Floating point to integer conversions
376 1 ttsou
-- Testing 40000 iterations of 3120 values
377 1 ttsou
- Measuring conversion time
378 1 ttsou
- Elapsed time base...                  0.065508 secs
379 1 ttsou
- Validating SIMD conversion results... PASS
380 3 ttsou
- Measuring conversion time
381 3 ttsou
- Elapsed time SIMD ...                 0.011424 secs
382 3 ttsou
- Speedup...                            5.734244
383 41 sylvain
</pre>
384 1 ttsou
385 41 sylvain
<pre>
386 3 ttsou
[+] Testing: GSM TCH/AFS 7.95 (recursive, flushed, punctured)
387 3 ttsou
[.] Input length  : ret = 165  exp = 165 -> OK
388 3 ttsou
[.] Output length : ret = 448  exp = 448 -> OK
389 3 ttsou
[.] Pre computed vector checks:
390 3 ttsou
[..] Encoding: OK
391 3 ttsou
[..] Decoding base: 
392 3 ttsou
[..] Decoding SIMD: 
393 1 ttsou
[..] Code N 3
394 1 ttsou
[..] Code K 7
395 1 ttsou
OK
396 1 ttsou
[.] Random vector checks:
397 1 ttsou
[.] Testing baseline:
398 17 ttsou
[..] Encoding / Decoding 10000 cycles:
399 17 ttsou
[.] Elapsed time........................ 1.435066 secs
400 17 ttsou
[.] Rate................................ 3.121808 Mbps
401 17 ttsou
[.] Testing SIMD:
402 17 ttsou
[..] Encoding / Decoding 10000 cycles:
403 17 ttsou
[.] Elapsed time........................ 0.073524 secs
404 17 ttsou
[.] Rate................................ 60.932485 Mbps
405 17 ttsou
[.] Speedup............................. 19.518334
406 41 sylvain
</pre>
407 17 ttsou
408 41 sylvain
*Intel Atom (D2500 1.86 GHz)*
409 41 sylvain
<pre>
410 17 ttsou
--- Floating point to integer conversions
411 17 ttsou
-- Testing 40000 iterations of 3120 values
412 17 ttsou
- Measuring conversion time
413 17 ttsou
- Elapsed time base...                 1.147449 secs
414 1 ttsou
- Validating SSE conversion results... PASS
415 17 ttsou
- Measuring conversion time
416 1 ttsou
- Elapsed time SSE ...                 0.347838 secs
417 1 ttsou
- Quotient...                          3.298803
418 41 sylvain
</pre>
419 17 ttsou
420 41 sylvain
<pre>
421 1 ttsou
[+] Testing: GSM TCH/AFS 7.95 (recursive, flushed, punctured)
422 1 ttsou
[.] Input length  : ret = 165  exp = 165 -> OK
423 1 ttsou
[.] Output length : ret = 448  exp = 448 -> OK
424 1 ttsou
[.] Pre computed vector checks:
425 17 ttsou
[..] Encoding: OK
426 17 ttsou
[..] Decoding base: 
427 17 ttsou
[..] Decoding SIMD: 
428 17 ttsou
[..] Code N 3
429 17 ttsou
[..] Code K 7
430 17 ttsou
OK
431 17 ttsou
[.] Random vector checks:
432 19 ttsou
[.] Testing baseline:
433 19 ttsou
[..] Encoding / Decoding 10000 cycles:
434 19 ttsou
[.] Elapsed time........................ 11.822688 secs
435 17 ttsou
[.] Rate................................ 0.378932 Mbps
436 17 ttsou
[.] Testing SIMD:
437 17 ttsou
[..] Encoding / Decoding 10000 cycles:
438 17 ttsou
[.] Elapsed time........................ 0.550423 secs
439 17 ttsou
[.] Rate................................ 8.139195 Mbps
440 17 ttsou
[.] Speedup............................. 21.479277
441 41 sylvain
</pre>
442 17 ttsou
443 41 sylvain
*!ArndaleBoard (ARM Cortex-A15 1.7 GHz)*
444 17 ttsou
445 17 ttsou
Please note that the Viterbi implementations on ARM is largely C based with speedup generated primarily through algorithm changes. In comparison, vector optimization on Intel platforms with SSE is currently much more aggressive, which explains the disparity on decoding performance.
446 17 ttsou
447 41 sylvain
<pre>
448 17 ttsou
--- Floating point to integer conversions
449 17 ttsou
-- Testing 40000 iterations of 3120 values
450 17 ttsou
- Measuring conversion time
451 17 ttsou
- Elapsed time base...                 0.384097 secs
452 17 ttsou
- Validating SSE conversion results... PASS
453 17 ttsou
- Measuring conversion time
454 17 ttsou
- Elapsed time SSE ...                 0.100877 secs
455 17 ttsou
- Quotient...                          3.807578
456 41 sylvain
</pre>
457 17 ttsou
458 41 sylvain
<pre>
459 17 ttsou
[+] Testing: GSM TCH/AFS 7.95 (recursive, flushed, punctured)
460 17 ttsou
[.] Input length  : ret = 165  exp = 165 -> OK
461 17 ttsou
[.] Output length : ret = 448  exp = 448 -> OK
462 17 ttsou
[.] Pre computed vector checks:
463 17 ttsou
[..] Encoding: OK
464 17 ttsou
[..] Decoding base: 
465 17 ttsou
[..] Decoding SIMD: 
466 17 ttsou
[..] Code N 3
467 17 ttsou
[..] Code K 7
468 17 ttsou
OK
469 17 ttsou
[.] Random vector checks:
470 17 ttsou
[.] Testing baseline:
471 17 ttsou
[..] Encoding / Decoding 10000 cycles:
472 17 ttsou
[.] Elapsed time........................ 5.371288 secs
473 17 ttsou
[.] Rate................................ 0.834064 Mbps
474 3 ttsou
[.] Testing SIMD:
475 3 ttsou
[..] Encoding / Decoding 10000 cycles:
476 3 ttsou
[.] Elapsed time........................ 1.016621 secs
477 3 ttsou
[.] Rate................................ 4.406755 Mbps
478 3 ttsou
[.] Speedup............................. 5.283471
479 41 sylvain
</pre>
480 3 ttsou
481 41 sylvain
*!BeagleBoard-xM (ARM Cortex-A8 800 MHz)*
482 41 sylvain
<pre>
483 5 ttsou
--- Floating point to integer conversions
484 3 ttsou
-- Testing 40000 iterations of 3120 values
485 3 ttsou
- Measuring conversion time
486 3 ttsou
- Elapsed time base...                  6.292542 secs
487 4 ttsou
- Validating SIMD conversion results... PASS
488 3 ttsou
- Measuring conversion time
489 3 ttsou
- Elapsed time SIMD ...                 0.839081 secs
490 3 ttsou
- Quotient...                           7.499326
491 41 sylvain
</pre>
492 1 ttsou
493 41 sylvain
<pre>
494 31 ttsou
[+] Testing: GSM TCH/AFS 7.95 (recursive, flushed, punctured)
495 31 ttsou
[.] Input length  : ret = 165  exp = 165 -> OK
496 31 ttsou
[.] Output length : ret = 448  exp = 448 -> OK
497 31 ttsou
[.] Pre computed vector checks:
498 31 ttsou
[..] Encoding: OK
499 31 ttsou
[..] Decoding base: 
500 31 ttsou
[..] Decoding SIMD: 
501 31 ttsou
[..] Code N 3
502 31 ttsou
[..] Code K 7
503 1 ttsou
OK
504 32 ttsou
[.] Random vector checks:
505 32 ttsou
[.] Testing baseline:
506 32 ttsou
[..] Encoding / Decoding 10000 cycles:
507 32 ttsou
[.] Elapsed time........................ 21.963257 secs
508 32 ttsou
[.] Rate................................ 0.203977 Mbps
509 32 ttsou
[.] Testing SIMD:
510 32 ttsou
[..] Encoding / Decoding 10000 cycles:
511 32 ttsou
[.] Elapsed time........................ 3.083282 secs
512 32 ttsou
[.] Rate................................ 1.452997 Mbps
513 32 ttsou
[.] Speedup............................. 7.123337
514 41 sylvain
</pre>
515 32 ttsou
516 32 ttsou
517 41 sylvain
*Full Results*
518 32 ttsou
519 41 sylvain
"[http://tsou.cc/gsm/shuttle.txt":http://tsou.cc/gsm/haswell.txt]
520 31 ttsou
521 41 sylvain
"[http://tsou.cc/gsm/beagle.txt":http://tsou.cc/gsm/arndale.txt]
522 1 ttsou
523 30 ttsou
524 1 ttsou
525 41 sylvain
h2. Authors
526 1 ttsou
527 1 ttsou
528 49 neels
[[OsmoTRX]] is currently developed and maintained by Thomas Tsou with generous support from Fairwaves, the Open Technology Institute, and Ettus Research. The code is derived from the [[OsmoBTS:OpenBTS]] project, which was originally developed by David Burgess and Harvind Samra at Range Networks.
Add picture from clipboard (Maximum size: 48.8 MB)