This fixes the 20 second startup delay caused by tx/control threads
getting temporarily stuck while trying to set their own priority.
Apparently the only sane way for core affinity+priority is to set both
as attributes during thread creation using pthreads.
This switches the cmd queue to the timeout version, too, to ensure the
thread doesn't get stuck waiting for messages, and allows cleaner exits.
Change-Id: I7e2f83a9b9df024acaf9076c58189cb6b7bcc34b
The VA is already being used by the ms side and is part of the original
gsm design. It only works for gmsk, 4sps, and needs a bit of rx burst
scaling and burst shifting.
Change-Id: I9d7a4ff72e323832a94d885d5714fcde01ceeb3d
This commit adds support for rach bursts to the viterbi equalizer, which
is currently only being used by the ms side, so the equalizer can be used
by the osmo-trx network side, too. The difference is that rach bursts are
shorter than any other burst type (due to unknown TA) and start with
diffrent tail bits.
This drops the multiversioning which was only working for x86 anyway because
it can't be combined with no_ubsan.
Change-Id: I4a5cedc8c9a3289c75ce7b914eac286e601ebed0
After upgrading our CI environment to use Debian 12 with GCC 12, the
byteswap code fails the build with the following. I've talked to Eric
about this and he recommended to just remove the code as practically
nobody will use osmo-trx with a big endian system.
USRPDevice.cpp:591:30: error: 'data' is used uninitialized [-Werror=uninitialized]
591 | *wordPtr = host_to_usrp_u32(*wordPtr);
| ~~~~~~~~~~~~~~~~^~~~~~~~~~
Related: OS#6057
Change-Id: I806d8c1432cb20efca1830a2752a4cbc70384b54
Manual attempts to get the number of complex and single samples right
turned out to be a bit error prone at times...
Change-Id: I3c9953073555e3a7f70b78b0946dfdf949175a82
The timekeeper should never wait for lazy readers, because that causes
timekeeping and later usb transfers to fail.
Change-Id: Id0aad606a296b2885617013ce6637204357b13d7
One of the mystery bugs was that the blade ms needed two starts
after powercycling the bladerf due to transfer timeouts.
This is now fixed.
Change-Id: I1cd8790191790f4861a70bc55c8f4c9993fa10c8
* Do not hard-code '-O3' and '-g', let the user decide on them.
* Move '-lpthread' to LIBADD, it shall not be in CFLAGS.
* Also add '-Wall' to AM_CFLAGS.
Change-Id: Ia971fc4124b405514792e5309d42936aaebdc8c1
* AM_CPPFLAGS is for preprocessor flags like '-I' or '-D',
* AM_CFLAGS/AM_CXXFLAGS is for C/C++ compiler flags like '-Wall',
* AM_LDFLAGS is for linker flags like '-no-undefined', not libraries!
* Link ipc-driver-test against libdevice.la,
* Do not put $(UHD_CFLAGS) everywhere.
Change-Id: Iafd68974c9c613fb4e65a01d076b2c687b716c83
* AM_CPPFLAGS is for preprocessor flags like '-I' or '-D',
* AM_CXXFLAGS is for C++ compiler flags like '-Wall'.
Change-Id: I82fc1eb6796f5289f0feff7bee5264bfacd6e733
This option should be used for any executables which are used only
for testing, or for generating other files and are consequently never
installed. By specifying this option, we are telling Libtool that
the executable it links will only ever be executed from where it is
built in the build tree. Libtool is usually able to considerably
speed up the link process for such executables.
Change-Id: Ieb4ddc5799819c24ed357218c7b0197bcb6b5c91
* AM_CPPFLAGS is for preprocessor flags like '-I' or '-D',
* AM_CFLAGS is for C (not C++!) compiler flags like '-Wall'.
* AM_CXXFLAGS is for C++ specific compiler flags like '-Wall'.
By having the preprocessor flags in the proper variable we make
sure that they apply to both *.c and *.cpp files.
Change-Id: I7b3504a01e3350834b35c42d8d76d5d88d84a4b9
The "safe" scaling factor introduced in
7ac54b10d3 is too low and dates back to
the beginning and the move from usrp1->uhd, but the modulator will
exceed +-1 so "proper" scaling leads to overflows. Let's just do what
osmotrx has been doing for many years...
Change-Id: I75a2eba1f7f7b81249c06ce3fc9dfeee08878cb9
Given integral type A and non integral type B and depending on rounding
mode, optimization, compiler, and phase of the moon A(A)*B != A(A*B) so
split the two cases.
While at it, also make the template automagically work for complex types
instead of requiring manual casts, the general idea here is to allow
inlining and vectorization by treating all args as plain arrays, which is fine.
This works as expected with -tune=native, x64 implies sse2, and we do not
target any neon-less arm versions either.
Clang only array length hints can improve this even more.
Change-Id: I93f077f967daf2ed382d12cc20a54846b3688634
