Found it:
Created an C++ app that called that function with the same parameter values as with python for this flow graph. Witch I was able to debug normally.
In window.cc line 265, with i = ntaps - 1, temp = 1.0000000000000000002 that cause sqrt (0) witch return "-nan" on the next line and screw all the rest of the calculations.
This only happens when compiled with "CFLAGS=-march=native -O2", if I don't specify the march it's working correctly. The function is called on my system with taps=787 and beta = 7.
Regards: Cor
On Wed, 2017-06-21 at 15:13 +0200, Cor Legemaat wrote: Hi:
Sorry for the late replay... The intel pc call filter.firdes.low_pass with the same values but return 768 proper float values, not like the <nan>'s on the AMD pc.
Tried to debug with "nemiver /usr/bin/python2.7 -u <path>/fm_receiver.py" and the breakpoint at firdes.cc line 100 witch get triggered and I can read the function parameters but when I try to step true the function it jumps to the assembly of pthread. If I put more breakpoints in firdes.cc I get back to the function but cant read any variables any more. Also tried exporting "export GR_SCHEDULER=STS" but the same symptoms.
Don't know if Ubuntu will trigger the bug it's probably compiled more generic...
Regards: Cor
On Wed, 2017-06-07 at 04:26 -0400, Anon Lister wrote: I have an AMD system with the same chip running Ubuntu 16.xx. I can probably try to duplicate this weekend, if Cor doesn't get to it, as another data point.
_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
|