Squashed commit of the following:
commit d12066467c99bdbb7048285c0616db70ff383e2f
Merge: 3493a3b3c 2770a187a
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Tue Sep 19 13:56:34 2017 -0400
Merge branch 'master' into rtucker-host-capabilities
commit 3493a3b3c1ab17ea4a616bbf8b543108e5defeca
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Thu Sep 14 13:40:33 2017 -0400
host: libbladeRF: add bladerf_get_channel_count call
This will return the number of channels available on a given hardware.
commit 609bfc7d4bc4bb1fea09164de184b94fe9f691da
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Thu Sep 14 13:37:04 2017 -0400
host: libbladeRF/board: add FW capabilities for bladeRF 2
Not entirely certain we need a capability flag for bladeRF 2 support in
the firmware, but now there is one.
commit 12f6d61c0c04df1a52d7552ff3ee45f6faaaf6b1
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Thu Sep 14 13:35:11 2017 -0400
host: libbladeRF/board: move capabilities to board.h
Keeping them common between boards seems prudent for now.
expansion_get_attached should only report that a board is attached if it
has already gone through expansion_attached, to avoid skipping the
initialization
This includes everything but AGC-related stuff.
Squashed commit of the following:
commit ab00f2a68353024c5faf14dedbb5fd05a8438f26
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 9 10:58:08 2017 -0400
libbladeRF: add compilation hint to boilerplate.c
h/t owen on IRC
commit 3cbeaf72dbe22525a5c907dfb9643db81c7aca85
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 12:27:21 2017 -0400
host/CMakeLists: dequote variables in ifs (CMP0054) #511
Per CMake Policy CMP0054, the meaning of constructs like
if("${FOOBAR}" STREQUAL "quux")
is ambigious, and the old meaning is deprecated.
https://cmake.org/cmake/help/git-stage/policy/CMP0054.html
commit 73376b7aac65d945c5399a5130aca33265080653
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 13:40:37 2017 -0400
host/README.md: add pthreads-win32 build instructions (#447)
On recent versions of Windows / Visual Studio, struct timespec is
already defined, and the definition is subtly different than what
pthreads-win32 is built with.
It may be possible to handle this in a more automatic fashion via CMake,
but this is TBD.
commit b90ff648e5cf754a29264f2a2c2c19d924c5c070
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 18:09:30 2017 -0400
host/cmake/CyAPI: define Windows 10 support
This requires that the user get special additional drivers in addition
to the ones provided by the usual SDK; a URL will point in the right
direction.
Note that this does not remove the requirement for driver signing. Some
work will be needed to document how to do this.
commit 888254b975f3b64ec787c741b937a561a08ccc44
Author: Jon Szymaniak <jon.szymaniak@nuand.com>
Date: Thu Jun 9 17:47:21 2016 -0400
libbladeRF: Treat incompatible OS version as "not found" in FindCyAPI.cmake
commit a9e97b2de6c22539266447f755e5a42cdf9a4f70
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 11:35:16 2017 -0400
host/libbladeRF/cyapi: ReleaseMutex before CloseHandle
When working on mutexes for the libusb backend, I noticed that I needed
to ReleaseMutex() when closing the device, or else that device would not
be usable again until all of the mutexes were released.
I have not tested this with CyAPI as yet, but I believe the same problem
will occur, and I also believe this may fix it.
commit b897475aa1a5b80358fcf2c8902089740d04582f
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 18:00:34 2017 -0400
host/libusb.c: properly release/close handles
commit f9b99ee7715af1ccb64d41d3363a3cc8e66aa063
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 17:17:29 2017 -0400
host/libusb.c: fix create_device_mutex errors (#484)
This builds and correctly enforces exclusive access under Windows 10.
commit 6e9c048ff2e8321052c3e72e14f51eb2aa66b5dc
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Thu Aug 3 13:14:51 2017 -0400
host/backend/libusb: enforce exclusive access (#484)
Windows seems to allow multiple instances of libbladeRF to open the same
device, which is problematic in many ways. This borrows the mutex
approach from cyapi.c to enforce exclusivity.
commit 519f26b6b88d1cbdfec7138fe9a0fc13715c7f93
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 16:07:49 2017 -0400
host/cmake: search for pthreadVC2.dll more vigorously (#517)
With the pre-built binary pack, pthreadVC2.dll lives in a
dll/${ARCHITECTURE}/ directory. However, when compiling from scratch,
it just lives in the source code's root directory. Look in both places
and set a LIBPTHREADSWIN32_DLL variable accordingly.
commit b7617cda8e87396da19101f10de02619d0fd8e1d
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 18:05:33 2017 -0400
CONTRIBUTORS: self-updating
commit 3a4e57b3c2f2b9ac27db832de0bbc945c3217dd7
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Tue Jul 25 15:07:54 2017 -0400
debian: chmod 0444 downloaded fpga/fw images
The mktemp/wget workflow creates the data files chmod 0600, which breaks
loading of firmware for non-root users. Explicitly chmod 0444 to allow
all users to read.
commit 24594d3e34eca05c2fc738dc2fa5d1fcc9ab0bda
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 17:34:11 2017 -0400
host: add note about Fedora and /usr/local (#490)
commit 6219e15a38c8dfe957e12ce38fbd76e0bf40a177
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 15:35:27 2017 -0400
nios: fix si5338_write typo (#491)
Unnecessary bit is being set when writing address.
commit 0d14e6ec0a3d438c206f4d02a95aa4b250ec3674
Author: Oliver Jowett <oliver@mutability.co.uk>
Date: Thu Apr 27 21:03:12 2017 +0100
Document why postinst should succeed if the FPGA download fails
commit 84042b6aa5bcc761c631a06687dc690343f4d952
Author: Oliver Jowett <oliver@mutability.co.uk>
Date: Thu Apr 27 13:20:25 2017 +0100
Don't make postinst fail entirely (producing half-configured packages and
making the overall install operation fail) if the FPGA download fails.
(cherry picked from commit 13befc071b3cddb913ce961c3cd7b08216bba2e5)
commit b203fbca8c789d8a99bbef34e20feda9d1a6e940
Author: ifrasch <fraschi17@gmail.com>
Date: Sat Feb 18 17:28:59 2017 -0500
hdl: Fixed simulation files
-Added missing ports to nios_system.vhd
-Added missing files to nuand.do
-Compile.do will now compile successfully in Modelsim
This includes everything but AGC-related stuff.
Squashed commit of the following:
commit ab00f2a68353024c5faf14dedbb5fd05a8438f26
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 9 10:58:08 2017 -0400
libbladeRF: add compilation hint to boilerplate.c
h/t owen on IRC
commit 3cbeaf72dbe22525a5c907dfb9643db81c7aca85
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 12:27:21 2017 -0400
host/CMakeLists: dequote variables in ifs (CMP0054) #511
Per CMake Policy CMP0054, the meaning of constructs like
if("${FOOBAR}" STREQUAL "quux")
is ambigious, and the old meaning is deprecated.
https://cmake.org/cmake/help/git-stage/policy/CMP0054.html
commit 73376b7aac65d945c5399a5130aca33265080653
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 13:40:37 2017 -0400
host/README.md: add pthreads-win32 build instructions (#447)
On recent versions of Windows / Visual Studio, struct timespec is
already defined, and the definition is subtly different than what
pthreads-win32 is built with.
It may be possible to handle this in a more automatic fashion via CMake,
but this is TBD.
commit b90ff648e5cf754a29264f2a2c2c19d924c5c070
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 18:09:30 2017 -0400
host/cmake/CyAPI: define Windows 10 support
This requires that the user get special additional drivers in addition
to the ones provided by the usual SDK; a URL will point in the right
direction.
Note that this does not remove the requirement for driver signing. Some
work will be needed to document how to do this.
commit 888254b975f3b64ec787c741b937a561a08ccc44
Author: Jon Szymaniak <jon.szymaniak@nuand.com>
Date: Thu Jun 9 17:47:21 2016 -0400
libbladeRF: Treat incompatible OS version as "not found" in FindCyAPI.cmake
commit a9e97b2de6c22539266447f755e5a42cdf9a4f70
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Mon Aug 7 11:35:16 2017 -0400
host/libbladeRF/cyapi: ReleaseMutex before CloseHandle
When working on mutexes for the libusb backend, I noticed that I needed
to ReleaseMutex() when closing the device, or else that device would not
be usable again until all of the mutexes were released.
I have not tested this with CyAPI as yet, but I believe the same problem
will occur, and I also believe this may fix it.
commit b897475aa1a5b80358fcf2c8902089740d04582f
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 18:00:34 2017 -0400
host/libusb.c: properly release/close handles
commit f9b99ee7715af1ccb64d41d3363a3cc8e66aa063
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 17:17:29 2017 -0400
host/libusb.c: fix create_device_mutex errors (#484)
This builds and correctly enforces exclusive access under Windows 10.
commit 6e9c048ff2e8321052c3e72e14f51eb2aa66b5dc
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Thu Aug 3 13:14:51 2017 -0400
host/backend/libusb: enforce exclusive access (#484)
Windows seems to allow multiple instances of libbladeRF to open the same
device, which is problematic in many ways. This borrows the mutex
approach from cyapi.c to enforce exclusivity.
commit 519f26b6b88d1cbdfec7138fe9a0fc13715c7f93
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Aug 4 16:07:49 2017 -0400
host/cmake: search for pthreadVC2.dll more vigorously (#517)
With the pre-built binary pack, pthreadVC2.dll lives in a
dll/${ARCHITECTURE}/ directory. However, when compiling from scratch,
it just lives in the source code's root directory. Look in both places
and set a LIBPTHREADSWIN32_DLL variable accordingly.
commit b7617cda8e87396da19101f10de02619d0fd8e1d
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 18:05:33 2017 -0400
CONTRIBUTORS: self-updating
commit 3a4e57b3c2f2b9ac27db832de0bbc945c3217dd7
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Tue Jul 25 15:07:54 2017 -0400
debian: chmod 0444 downloaded fpga/fw images
The mktemp/wget workflow creates the data files chmod 0600, which breaks
loading of firmware for non-root users. Explicitly chmod 0444 to allow
all users to read.
commit 24594d3e34eca05c2fc738dc2fa5d1fcc9ab0bda
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 17:34:11 2017 -0400
host: add note about Fedora and /usr/local (#490)
commit 6219e15a38c8dfe957e12ce38fbd76e0bf40a177
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Wed Aug 2 15:35:27 2017 -0400
nios: fix si5338_write typo (#491)
Unnecessary bit is being set when writing address.
commit 0d14e6ec0a3d438c206f4d02a95aa4b250ec3674
Author: Oliver Jowett <oliver@mutability.co.uk>
Date: Thu Apr 27 21:03:12 2017 +0100
Document why postinst should succeed if the FPGA download fails
commit 84042b6aa5bcc761c631a06687dc690343f4d952
Author: Oliver Jowett <oliver@mutability.co.uk>
Date: Thu Apr 27 13:20:25 2017 +0100
Don't make postinst fail entirely (producing half-configured packages and
making the overall install operation fail) if the FPGA download fails.
(cherry picked from commit 13befc071b3cddb913ce961c3cd7b08216bba2e5)
commit b203fbca8c789d8a99bbef34e20feda9d1a6e940
Author: ifrasch <fraschi17@gmail.com>
Date: Sat Feb 18 17:28:59 2017 -0500
hdl: Fixed simulation files
-Added missing ports to nios_system.vhd
-Added missing files to nuand.do
-Compile.do will now compile successfully in Modelsim
Squashed commit of the following:
commit b06ffe841dc5e8cc91cfa743213979be49843b83
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Sep 15 13:46:59 2017 -0400
host: libbladeRF: forgot bladerf_get_gain_modes implementation
commit ef115599d0baea3ae27136059aedadaaf67f5f43
Author: Rey Tucker <rey.tucker@nuand.com>
Date: Fri Sep 15 12:15:07 2017 -0400
libbladeRF: add bladerf_get_gain_modes
bladerf_get_gain_modes provides a pointer to an array of gain modes
supported by the device, including human-readable names.
If the system is configured to require more than 80 Msps total
throughput, warn the user that they will likely lose some samples.
The maximum possible throughput between the FPGA and the FX3 is
3.2 Gbps (32 bits wide, 100 MHz), and the practical limit is closer to
about 2.7 Gbps once overhead is accounted for. The AD9361 can do 2x2
MIMO at 61.44 Msps, which turns out to be 7.8 Gbps (32 bits per sample,
4 channels, 61.44 Msps/channel). Thus, the warning.
The channels are enabled by the software one at a time, so enable will
go high before the second channel is enabled in a MIMO scenario. This
resulted in samples_left_init always being 1, dropping half of the
received samples in a 2xRX configuration.
When we change freq for one {RX,TX} port, it also applies indirectly to
the other {RX,TX} port, as they share the same LO. This behavior should
follow through to the RF switches etc.
Change bladerf_enable_module to expect a bladerf_channel instead of a
bladerf_direction. This should have no impact on existing applications:
BLADERF_RX and BLADERF_TX map to BLADERF_CHANNEL_RX(0) and
BLADERF_CHANNEL_TX(0) respectively.
Modify bladerf2_enable_module to set the expected RFFE control reg bits.
Removed the channel_sel from the "Nios GPIO".
Added 4 outputs from the "RFFE GPIO" to control RX1_EN, TX1_EN, RX2_EN,
and TX2_EN. The Nios writes these bits to control which MIMO port(s) are
used/ignored. This also effects how the sample buffers are
packed/unpacked between the host and FPGA.
The host tells the FPGA which channel to listen to using the
"channel_select" GPIO from the Nios. This update makes this branch
backwards-compatible with the host software.
TODO: update libbladeRF to process packed MIMO samples.
This update allows for unpacking of packed MIMO stream data from the
tx_fifo when zero or more streams are disabled. Without it, a disabled
stream will submit zeroes for transmit over the USB bus from the host to
the bladeRF, reducing bandwidth that could have otherwise been used to
transmit useful samples.
Also made some changes to rudimentarily support different platforms in
nuand.do and sample_stream_tb.vhd.
Fixed a range bug (typo?) in synthesis/fifo_reader.vhd
Separate the ranges for {dac,adc}_{controls,streams}... there are the
same numbers of DAC and ADC channels for now, but removing the
dependency on the two seems prudent.
Also a spacing tweak.