Open loewal opened 4 years ago
https://github.com/simonyiszk/csdr/issues/11 Orange pi pc-2 uses Cortex-A53, you need to manually modify the Makefile when compiling csdr…
And how....? Thnx!
PARAMS_NEON = -O3 -mcpu=cortex-a53 -march=armv8-a+crc+crypto -fomit-frame-pointer -funroll-loops -Wformat=0
PARAMS_RASPI = -mcpu=cortex-a53 -mtune=cortex-a53 -funsafe-math-optimizations -Wformat=0
it compiles, but the result....the same
PARAMS_NEON = -O3 -mcpu=cortex-a53 -march=armv8-a+crc+crypto -fomit-frame-pointer -funroll-loops -Wformat=0
PARAMS_NEON = -mfloat-abi=hard -march=armv7-a -mtune=cortex-a8 -mfpu=neon -mvectorize-with-neon-quad -funsafe-math-optimizations -Wformat=0 -DNEON_OPTS
tnx Jan Szumiec for the Raspberry Pi support
PARAMS_RASPI = -mfloat-abi=hard -mcpu=arm1176jzf-s -mfpu=vfp -funsafe-math-optimizations -Wformat=0
PARAMS_RASPI = -mcpu=cortex-a53 -mtune=cortex-a53 -funsafe-math-optimizations -Wformat=0
Keep PARAMS_RASPI
unchanged, It should like this:
PARAMS_NEON = -O3 -mcpu=cortex-a53 -march=armv8-a+crc+crypto -fomit-frame-pointer -funroll-loops -Wformat=0
#tnx Jan Szumiec for the Raspberry Pi support
PARAMS_RASPI = -mfloat-abi=hard -mcpu=arm1176jzf-s -mfpu=vfp -funsafe-math-optimizations -Wformat=0
Still the same: Part of the output... libcsdr.c:123:5: note: LOOP VECTORIZED normalize_fir_f: normalize pass 2 libcsdr.c:121:5: note: LOOP VECTORIZED normalize_fir_f: normalize pass 1 libcsdr.c:136:5: note: not vectorized: loop contains function calls or data references that cannot be analyzed firdes_lowpass_f: calculate taps libcsdr.c:156:5: note: not vectorized: multiple nested loops. firdes_bandpass_c libcsdr.c:162:14: note: not vectorized: number of iterations cannot be computed. firdes_bandpass_c libcsdr.c:161:14: note: not vectorized: number of iterations cannot be computed. firdes_bandpass_c libcsdr.c:193:5: note: not vectorized: multiple nested loops. shift_math_cc libcsdr.c:204:14: note: not vectorized: number of iterations cannot be computed. shift_math_cc: normalize phase libcsdr.c:203:14: note: not vectorized: number of iterations cannot be computed. shift_math_cc: normalize phase libcsdr.c:217:5: note: not vectorized: relevant stmt not supported: _34 = __builtin_sinf (_30); libcsdr.c:217:5: note: not vectorized: relevant stmt not supported: _34 = __builtin_sinf (_30); libcsdr.c:237:5: note: not vectorized: multiple nested loops. shift_math_cc libcsdr.c:262:14: note: not vectorized: number of iterations cannot be computed. shift_math_cc: normalize phase libcsdr.c:261:14: note: not vectorized: number of iterations cannot be computed. shift_math_cc: normalize phase libcsdr.c:276:5: note: not vectorized: multiple nested loops.
It takes 35 seconds before the waterfall starts to move and another 5 for hearing sound. Soundbuffer around 2.2 seconds. Strange because i use a value of 1 (to test) in config_webrx.py.
More strange, because when running under Jessie last week I didn't need to change anything to compile....
Now running Buster
Perhaps not the fault of csdr, do you mind pasting the OpenWebRX configure file here and let me have a look?
Changed it to .txt....... config_webrx.txt
This configuration is the same, except buffervalue, which was 6. Changed it to check if it made any difference. After 45 seconds it runs fine. OrangePi load is about 40%.
It runs on an Orange Pi pc2 and a Cubietruck 3. Both ran well last week under Jessie.
This configuration is the same, except buffervalue, which was 6. Changed it to check if it made any difference. After 45 seconds it runs fine. OrangePi load is about 40%.
Try to decrease the value of fft_size & fft_voverlap_factor and see if there is any improvement.
fft_size = 4096
fft_voverlap_factor = 0.2
I am sorry, but I think you attack this problem the wrong way (sorry!). Only the startup is slow, after that its allright. I used the same settings last week under Jessie, so it is not logical to change them. I changed it, but startup still the same and audibuffer 2.1 secs.
These are some general solutions... I'm sorry for leading you into a wrong way : (
It seems you're using SDRPlay, you can change the value of nmux_memory
to 128
or higher, which can increase the size of the circular buffer used by nmux.
Other than that, I can't remind a better way.
I also use Debian Buster on Orange Pi Lite, RTL-SDR, up to now, everything goes well...
Again, sorry to bring you so much trouble.
Installed OpenWebRX following the instructions of Andras Retzler (https://blog.sdr.hu/2015/06/30/quick-setup-openwebrx.html) on my new Raspberry Pi 4. Exactly the same slow start as the other two. Help anyone??
Are you viewing the web interface on a browser on the (Orange)Pi itself, or on a different computer, like a laptop with a better CPU?
On my iMac. I installed it on 3 mini’s including Pi4. On Jessie it was fine. Buster? Sloooooow
Since I installed Debian Buster on my OrangePi PC-2, it takes about hals a minute to display a running waterfall. Changing all kinds of parameters give no result. Even changing buffersize from 6 to 1 still shows an audiobuffer of 2.1 seconds. What is happening? After one minute it runs quite well, but that slow startup.....?