Our internal helper UHDBoost.cmake includes several hacks for Boost
version compatibility. Now that we require at least Boost 1.71, these
hacks are just clutter and can be removed.
Building static libraries with UHD was broken due to the way that
UHDBoost was structured. The following changes to UHDBoost.cmake were
applied:
- This file no longer observes ENABLE_STATIC_LIBS. That's because in UHD,
setting ENABLE_STATIC_LIBS is supposed to build both static and shared
libraries, but the way UHDBoost.cmake was written, it would apply
settings for static builds to all libraries. If ENABLE_STATIC_LIBS was
set, then this would cause linker errors.
- As a consequence, users of UHDBoost.cmake must specify
Boost_USE_STATIC_LIBS manually before including it.
- BUILD_SHARED_LIBS is no longer modified. UHD always specifies the
correct library modifier (SHARED, STATIC, etc.) directly in the
add_library() call, and that is good practice. By overriding
BUILD_SHARED_LIBS in UHDBoost.cmake, this file can make changes to
CMake files that call add_library() without a modifier.
In host/lib/CMakeLists.txt, we now call UHDBoost a second time for
static builds but set Boost_USE_STATIC_LIBS beforehand. This allows to
specify linker targets for the static library independently from the
shared library.
We also factored out some common library property manipulations that are
shared between uhd and uhd_static.
- For Ubuntu/Debian, the package lib dir (PKD_LIB_DIR), is set to
/usr/libexec/uhd. This is the path where examples, tests and
utils get installed.
- List usrp_hwd.py explicitly as not installed.
- Align the distribution of the files to packages (debian/*.install)
files to the Debian/Ubuntu official version from
https://salsa.debian.org/bottoms/pkg-uhd (branch experimental)
- Import rfnoc-blocks-track-runtime-abi patch from the official debian
sources at https://salsa.debian.org/bottoms/pkg-uhd.
This changes the UHD package data path to include the ABI version, e.g.:
old: /usr/share/uhd
new: /usr/share/uhd/4.9.0
Co-authored-by: A. Maitland Bottoms <bottoms@debian.org>
These sources are not required when building RFNoC bitfiles, and they
also reference csh, which causes some build systems to produce warnings.
As a consequence, we can also remove csh from the uhd-rfnoc-dev
dependency list.
Update Changelog
- Add mention of new RFNoC OOT block repository to changelog.
Add link to RFNoC OOT block github repository to README.
Update Readme with reference to new OOT block repo
Update Manifest
Co-authored-by: Niels Garibaldi <niels.steffen.garibaldi@emerson.com>
Co-authored-by: Marian Koop <marian.koop@emerson.com>
UHD_VERSION needs to be also checked against an empty string
since cmake 4.1 sets this variable during project setup, also
if it was undefined before. Furthermore we need to backup and
restore custom UHD_VERSION when defined for e.g. by using option
-DUHD_VERSION=<version>, because project() setup will reset the
version to an empty string.
- Adds calling pnpdriver util to apply all installed inf win drivers
- Change installer to be aware of the target proc architecture of
installed files. This reults into a 64-bit installer to default
installing registry keys into the 64-bit registry hive.
- Extended file info of installer executable to incl. publisher and
copyright and make installer binary version match version of
deployed binaries.
- Add custom icon to installer and uninstaller binaries as well as
ARP entry.
With this commit the host enables retranmit capabilities to tx
streamer. This can be used for host to FPGA transmission that
are real time depended but need absolut reliability. The main
use case will be transmissions from host to replay block. The
current implementation supports only single channel transmissions.
To enable retransmissions the streamer has to change the op mode
using the new STREAM_MODE_STOP_ON_SEQ_ERROR. The streamer will
check whether the FPGA reports that this mode was enabled before
proceeding. When enabled on the FPGA the FPGA will reject data
from the host when it detects a sequence error. Only after a new
initialize stream command is sent the host can transfer more data
to the FPGA. When the tx streamer detects a sequence error from the
FPGA it will stop the current send command and return the samples
sent so far. Note: This might be more than the FPGA has actually
processed. It is up to the user logic to determine if and where
the transmission has to restart (e.g. query the replay block for
arrived samples). If the tx streamer calls send again the
reinitialization is triggered automatically.
With -DENABLE_RFNOC_DEV=ON (the default is OFF), the HDL sources for
building RFNoC blocks and RFNoC images through rfnoc_image_builder are
now also installed into $PREFIX/share/uhd/rfnoc/fpga/usrp3. This re-uses
the existing share/uhd/rfnoc/fpga directory.
This commit also includes a Python script that copies the fpga/usrp3
directory into a staging directory inside the build dir. The
installation happens from the staging directory, not the fpga/usrp3
directory itself. This allows us to filter the usrp3 source dir without
having to use highly verbose CMake for that.
The uhd-host Debian install file is also modified so that the RFNoC
sources are not packaged into the uhd-host package.
The YAML files that get installed as part of UHD in order to enable
RFNoC image building support are no longer in the 'header'component, but
are moved to a newly created 'rfnoc' component. This allows separate
rules for installing YAML files vs. C/C++ header files.
When ENABLE_LIBUHD is OFF, headers are no longer installed. This is
relevant for the use case where only Python utilities (like the RFNoC
tooling) is installed.
Modernized USRP inf files by:
- Removing using co-installers and depend instead on WINUSB.NT
install service (provided by Windows OS itself)
- Defined version string in addition to date string
- Switched to using unique USRPs specific ClassGuid, instead of
standard Windows defined ClassGuid to avoid infver (WDK) errors
- ERROR(1280) Class name and ClassGuid mismatch, expecting Class
"Sample" for ClassGuid "{78A1C341-4539-11d3-B88D-00C04FAD5171}"
- ERROR(1285) Cannot specify [ClassInstall32] section for
Microsoft-defined class
- Verified inf files with WDK infver utility.
Added new inf file for B206mini.
Extended erllc_uhd_b200_reinit.inf file to support initializing
unprogrammed FX3 based USRPs (B200 Series).
This adds rfnoc-gainConfig.cmake, which can be used by other projects to
find this OOT after installation by calling
```cmake
find_package(rfnoc-gain)
```
For identifying the export type of symbols in rfnoc-gain, the OOT
previously used UHD's config.hpp. However, that file requires defining
a specific preprocessor macro for setting the export types correctly.
Now, every OOT should carry their own config.hpp file, which means the
export-API macros need to be renamed (for rfnoc-gain, instead of
recycling UHD_API, it is now rfnoc_gain_API).
The corresponding template for rfnoc_modtool has also been adapted.
This is a bugfix for 8684d3c1:
- rfnoc_modtool create would generate an rfnoc/CMakeLists.txt that said
'gain' instead of 'newmod'.
- rfnoc_modtool add would put the gateware files in the wrong directory
(fpga/<modulename>/<blockname> instead of (rfnoc/fpga/...).
This is a vestage from the UHD-3 era, where RFNoC was optional. To
reduce confusion, we remove the CMake variable UHD_RFNOC_FOUND, although
we leave it in UHDConfig.cmake for not breaking 3rd party applications.
During redesign of the OOT module directory layout, the block YAMLs and
dts files were moved under `rfnoc/`, but the gateware was not. This
causes an inconsistency in the directory layout before and after
installation, making it hard to switch between installed and
not-installed versions of RFNoC modules.
For example, block YAMLs are in `./rfnoc/blocks/*.yml`, and gateware is
in `fpga/gain/rfnoc_block_gain/*`. After installation, block YAMLs are
in `$PREFIX/share/uhd/rfnoc/blocks/*.yml`, and gateware is in
`$PREFIX/share/uhd/rfnoc/fpga/gain/rfnoc_block_gain/*` -- note the
discrepancy.
This fixes two immediate issues:
- When building the X410 example in-tree, the DTS file was not being
found because it used the same finding logic as the gateware source files.
- When building any image core file post-installation, none of the
gateware could be found because it used the same logic for finding
gateware inside the OOT, which is in a different location relative to
the block YAMLs.
By removing the inconsistencies, both these use cases work.
A note on backward compatibility:
- Existing OOT modules already have the issue that building bitfiles
post-install does not work, and that will continue to not work.
However, otherwise, they will keep working as before.
- Because people may have gotten accustomed to specifying the RFNoC OOT
module as the include directory (e.g., `rfnoc_image_builder -I
/path/to/rfnoc-gain`), we allow for this use case, and auto-correct
the include path in this case with some backward compatibility code.
- Building the rfnoc-gain/icores/x410_XG_200_rfnoc_image_core.yml
example image core file now also works, even with older versions of
rfnoc-gain.
This line is redundant (we call CMakeRC.cmake from other cmake files,
and we have our own cmake_minimum_required() call in
host/CMakeLists.txt), and it is also less strict than our own
requirements (we currently require at least 3.12 for all of UHD). With
recent CMake versions, this causes warnings so we remove this line.
This is the warning that is now removed:
```
CMake Deprecation Warning at cmake/Modules/CMakeRC.cmake:63
(cmake_minimum_required):
Compatibility with CMake < 3.5 will be removed from a future version of
CMake.
Update the VERSION argument <min> value or use a ...<max> suffix to tell
CMake that the project does not need compatibility with older versions.
```
- Newer versions of cmake do not set install_name, so we need to do it
ourselves.
- Made Boost a system dependency to suppress compiler warnings from it.
In UHDPython.cmake, when calling find_package(Python3), we now add the
NumPy component. However, because this component was added in CMake 3.14,
and our minimum version is 3.12, we do this conditionally. For older
CMake versions, we rely on UHD_PYTHON_CHECK_MODULE_VERSION() to detect
numpy, which can fail to properly find the include headers in some
corner cases.
Add the patch version number to poetry version. Because UHD uses
the patch number for adding branch information on development branches
we need to save the patch number in some different variable for later
access.
This moves the calibration Python utilities into the
uhd Python module (because they have dependencies to
uhd). Also the poetry script adds them as scripts to
the uhd module.
The install Python module is not executed on Windows
systems.
usrpctl, rfnoc image builder and rfnoc modtool are
not affected.
This sets classifiers in pyproject.toml. Most of them
are static except for the development status. The status
is set to 2 (pre alpha) as a default. Only if UHD_RELEASE_MODE
is set it changes to 4 (beta) if we are on a release canditate tag
or 5 (release) if we are on release tag. If we aren't on any tag
or the tag does not meet the correct formatting we stay on 2
and a warning is issued.
Simplifying the Windows installer experience by reducing the
number of individual downloads from 4 (for USB devices) to 2.
Including the libusb.dll into the installer will remove a common
onboarding error.
This change introduces a now, optional cmake feature called
-DENABLE_EXTEND_WIN_PACKAGING=ON, that defaults to off.
Enabling this feature will cause a number of convenience components
to be included into the windows deployment package (aka. installer).
Enabling this feature has no effect on other platforms.
The changes to the NSIS itself are trivial and limited to another
license dialog (for libusb redistribution). Most changes were in
the cmake code used by cpack (for which NSIS is the utilized Windows
cpack generator option).
These additional components now controlled via the feature
component are currently:
- incl. of MSVC Runtime binaries into ${RUNTIME_DIR}
- incl. of libusb.dll into ${RUNTIME_DIR}
- incl. of erllc_uhd_winusb_driver into ${PKG_DATA_DIR}/usbdriver
- icores/CMakeLists.txt now contains the correct macro example
- The CMakeLists.txt file for the gateware itself is now created, and
will dynamically contain the correct argument for
RFNOC_REGISTER_BLOCK() depending on --skip-testbench
- To enable this, the fpga/<module_name>/CMakeLists.txt is now also
properly updated on `rfnoc_modtool add`
- This was facilitated by adding the `append` step option
- The autogenerated Makefile.srcs has the correct file endings now (.sv
instead of .v)
- There was a bug in the autogenerated Python block controller wrapper,
which referenced the wrong namespace
- Fix more missing commas in generated HDL.
This adds a pyproject.toml to generate Python wheels using poetry.
The toml file is generated by cmake to add build dependend information to the
file. On Windows the build process invokes poetry using the configured
Python executable. The existing distribution way (copy python data
to site-packages folder of UHD installation) is not changed yet.
Replaced instances of email address packages@ettus.com with new
corporate email address uhd.maintainer@emerson.com.
For some yet undefined transition time emails to the old address
are being forwarded to the new one.
Previously, the content of /usr/share was packaged with the
`libuhd4.x.y` package. However, this was both problematic and incorrect:
- The problem is that the intention of the `libuhd.4.x.y` packages is
that multiple packages with different versions can be installed
side-by-side, which doesn't work if different versions contain the
same file paths.
- The files that were shipped in the `libuhd` package are not actually
required by the UHD library (DLL).
- Specifically, they include:
- The .fbs files used for calibration data, only required when
manually inspecting and modifying cal data using flatbuffer, see the
section "Modify Calibration Data" in the manual.
- The rfnoc-newmod directory, which is required by rfnoc_modtool,
which itself is packaged in `uhd-host`.
- The rfnoc/ subdirectory which contains the block and device BSP
YAMLs. These are currently only used by `rfnoc_image_builder`, which
is also packaged in `uhd-host`.
Because the UHD library itself requires none of these files, and the
consumers that actually require these files are packaged in `uhd-host`,
we move them into `uhd-host` as well.
The parameters for compiling and linking against DPDK (DPDK_CFLAGS
and DPDK_LDFLAGS or DPDK_INCLUDE_DIRS, DPDK_LIBRARIES and
DPDK_LIBRARY_DIRS) are not trivial to determine and can be best
provided by pkg-config. Therefore, use pkg-config to find DPDK and
do not allow user provided DPDK variables.
To compile UHD against DPDK which was installed to a custom prefix,
e.g. /opt/dpdk-22.11, provide this path as parameter
CMAKE_PREFIX_PATH:
cmake .. -DCMAKE_PREFIX_PATH=/opt/dpdk-22.11
This replaces rfnoc_modtool that was previously shipped with gr-ettus.
It is fully RFNoC/UHD4.x compatible. There is no shared code between
this version and the old one; this is a clean rewrite.
As of now, it lacks many of the features that were available with the
previous version, most importantly, the ability to generate GNU Radio
bindings.
This installs UHDPython.cmake to lib/uhd/, so that OOT RFNoC modules can
use it (and don't have to use a version that ships with the OOT). This
makes maintanance of that file a lot easier.
Now that this file can (and should) be used outside of UHD, we
"namespace" the macros and functions by prefacing them with 'UHD_'
(e.g., there's now UHD_PYTHON_CHECK_MODULE).
This file requires creating UHDMinDepVersions.cmake, which also gets
installed.
