On 12/05/2010 06:25 PM, Marcus D. Leech wrote:
Why should:
http://www.sbrac.org/files/testme.py
http://www.sbrac.org/files/testme.grc
Be causing Gnu Radio to attemp to allocate 16GB of virtual memory?
Under what drug-induced fantasy should a straight-line
FFT (admittedly a chunky FFT) buffer be ballooned out by a factor of
1000, when the FFT size is a power of 2!
I realize that only *half* of that space actually maps to physical
memory, but 8GB for an FFT of order 2**23 is decidedly piggy.
Furthermore, it always provokes a fatal error, no matter which
"factory" I use for buffer allocation:
gr_vmcircbuf_mmap_shm_open: invalid size = -795873280
gr_buffer::allocate_buffer: failed to allocate buffer of size 15999996 KB
terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc
Because it looks like that buffer size is 32-bits unsigned, regardless
of the underlying system architecture (I'm on an x86_64 machine).
Oh, never mind. Mixed-mode math in Python causing *non* power-of-2 FFT
size. Sorry guys. Never mind.
--
Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org
_______________________________________________
Discuss-gnuradio mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/discuss-gnuradio
