Grappas/bugfixes #27
Grappas/bugfixes #27
Conversation
Less space taken
it was omitting every odd index of Frequency[i]
if none arguments were given gqrx-scanner performed sweep scan anyway
malloc() created single struct, sweep was performed and when
Frequencies[74] was touched it segfaulted into oblivion
|
Hi @grappas : can you please explain the inner "for loop" for i in Frequencies[] on the ScanFrequenciesInRange()? |
| @@ -1580,6 +1571,16 @@ int main(int argc, char **argv) { | |||
| print_usage(argv[0]); | |||
| } | |||
|
|
|||
| if (opt_scan_mode == sweep) | |||
| { | |||
| size_t freqeuencies_count = (size_t)(((opt_max_freq-opt_min_freq)/opt_scan_bw)+1); | |||
There was a problem hiding this comment.
This could grow very very big in case of large scan. Do you really need to store the noise floor for each step? I was even tempted to remove the GetSquelchLevel from the inner loop (to speed up a bit) and get a value before each round (the user will not change the setting during the scan, usually, so it is not optimized). Can we do something like that to avoid the fine-grained array?
There was a problem hiding this comment.
This could grow very very big in case of large scan. Do you really need to store the noise floor for each step? I was even tempted to remove the GetSquelchLevel from the inner loop (to speed up a bit) and get a value before each round (the user will not change the setting during the scan, usually, so it is not optimized). Can we do something like that to avoid the fine-grained array?
We need to know, for each frequency what noise floor we are dealing with and yeah - each one. If we have ie. UKF NFM 12500 Hz with bandwidth 8500 Hz or USB which can be even narrower and we would mesh it more loosely we would end up with omitting chatter.
In the other hand: lets say 2MHz scanning every 12500 Hz we are ending up with 160 elements, 1856 bytes each (if I remember correctly) we are ending up with 290 KiB of memory space. C'mon ( ͡° ͜ʖ ͡°) it's not that much.
If we go crazy and we wanted to cover whole 800MHz we would end up with 113 MiB.
If we could successfully cooperate with gqrx team and they would manage to implement array of frequencies only then we can discuss about optimization.
It has to be what it is for now.
It has to compare signal level to squelch each time and calculate noise_floor for each frequency. We could obviously put some of the functions outside for loop but due to lag (obviously because of querying for signal level each time we are changing frequency) it would be very annoying ie. adjusting squelch in gqrx in-flight. That and other stuff would be easier if we could access whole visible spectrum at once ( ͡° ͜ʖ ͡°) |
|
If you allow me I would consider - while optimizing the code - also make code a little bit more readable (make it more modular with inline funtions). |
|
Dude? Are you dead or somethin? |
I was busy with the job. |
To be perfectly safe I would say separate thread. |
Found few bugs. Plz see commit comments.