Age | Commit message (Collapse) | Author |
|
Users may use "sumo" or previous release branches to build for this
machine.
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
This adds LAYERSERIES_COMPAT warrior to common and all BSP layers, i.e.
* common
* v1000
* r1000
* snowyowl
* amdfalconx86
* steppeeagle
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
This removes the BSP and PATCH versions for MEL releases as these are
not applicable anymore.
This is being done in preparation for yocto based warrior branch.
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
To use a single description for all INCLUDE_<components> alongwith
the supported <component> summary in each BSP's local.conf.
This also helps to have a single place for all the configurable
<components> in the local.conf instead of having huge paragraph
for each <component> and having to scroll to find each <component>.
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
The new release process and numbering requires these
versions being defined by each BSP.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
We deprecated the LunarG SDK in one of the previous updates
but missed one instance where these packages were being
explicitly installed on the target, this generates a build
time error. Fix this by dropping the deprecated packages from
the installation list.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
We're now deprecating the LunarG SDK support but we'll still
keep basic Vulkan support in. The directory structure has
also been changed just to make it clear.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
MEL is now basing its features in a way that disregard
which _particular image_ is being built and gives this
flexibility to user where they can decide which extra
features should be populated. This is manipulated in the
local.conf using EXTRA_IMAGE_FEATURES.
This aligns the AMD features accordingly and basis all
functionality on EXTRA_IMAGE_FEATURES. This would help
minimize the risk of going off track with feature manipulation
while doing open as well as MEL based builds.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
All the GPU based BSPs use mesa as their GL. The newer
versions of mesa require python-mako on host (native) to
generate some build files and python-mako-native is only
available through meta-python at this time. The core
layer only provides python3-mako-native which is not
usable here.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
MEL provides an additional graphics feature which is
required in order to support different types of images
under MEL.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
All the currently supported BSPs are 64bit so use the
appropriate m64 TUNE_FEATURE in order to get the build
environment evaluated properly. Otherwise the basic -m64
switch which is required for proper linker output
specification and other such things do not end up in
the correct form and the build fails in case a toolchain
that supports 32 and 64bit builds is used.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This makes configurable feature component variables to use a standardized
naming convention i.e. should end with _PKGS instead of _COMPONENTS or
_PACKAGES
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
This defines AMD Features that can be added as EXTRA_IMAGE_FEATURES
to different machines based on what is supported on a machine. This adds
a broad flexibility and control over packages being installed on a
machine for any image without needing to create bbappends for all the
images that are to be supported.
> Each feature can contain packages and packagegroups as its components.
> Feature components can be dependent on DISTRO, IMAGE_FEATURE, or any
variable in general. e.g.: Components of "amd-feature-graphics" may be
dependent on "x11-base" as an IMAGE_FEATURE. Some packages may only be
included for a specific DISTRO. Some packages may only be installed if
user allows them in local.conf etc.
> Each machine must add the required features to EXTRA_IMAGE_FEATURES in
its own machine config file.
> All required features must be added to a machine regardless of the
image being built, but make sure that feature components are included
based on dependency conditions. e.g. say "amd-feature-graphics" was
added to a machine that supported graphics, but components of this
feature must not be installed on an image that is only console based
such as "core-image-base", therefore such components must depend on an
IMAGE_FEATURE that is based on graphics such as "x11-base".
> Each machine can also override feature components in its own machine
config when adding the feature to EXTRA_IMAGE_FEATURES. e.g.: a
feature may be added to a specific machine with minimal (or extended)
packages based on requirement.
Features are classified as:
* amd-common-pkgs : Common pkgs to be added to all machines
* amd-feature-multimedia : Multimedia packages (it does not depend
on graphics because a machine may not
have a GUI but could play videos and
sounds from console)
* amd-feature-graphics : Graphics packages
* amd-feature-networking : Networking packages
* amd-feature-debug-profile : Debugging and Profiling tools
* More features may be added later as needed
Signed-off-by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
|
|
Drop layer specific recipe and leverage the upstream
one which is more recent. This would also make the
maintenance feasible.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This is now an upstream requirement for compliance
and throws warnings on the console if not handled
appropriately.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
The correct way of using this is through the meta/lib/oe
so drop the deprecated calls.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
|
|
The vulkan-tools configuration process checks for
Qt5 _and_ Qt5Svg availability before enabling the
traceviewer but our enablement of the feature only
checks for Qt5 layer which creates a discrepancy.
This ends up in an installation failure when we
try to manually install traceviewer binary which
is skipped during the build process due to the
above reason.
We fix this by depending upon the qtsvg package
as well which is checked for during the configuration
process otherwise a race condition would cause
undesireable/unknown condition.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
|
|
Increases maintainability and readability by using a single
file with the actual license text being tested.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
PV is extracted from the filename and PR is
managed through the PR service so no need to
define them explicitly.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
meta-steppeeagle and meta-seattle do not have a vulkan-tools
recipe, so move the relevant changes to the original recipes
for now.
Signed-off-by: Ahsan Hussain <ahsan_hussain@mentor.com>
|
|
This tool was not enabled and would cause configuration issue when Qt layers
are configured in the build.
Signed-off-by: Ahsan Hussain <ahsan_hussain@mentor.com>
|
|
We now default to the upstream version of libdrm which
is 2.4.81 and hence is much stable than the one we
were using earlier for AMD BSPs.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
V1000 release
|
|
Signed-off-by: Sanjay R Mehta <sanju.mehta@amd.com>
|
|
This is a workaround for hang issues with compute shaders.
Signed-off-by: Ahsan Hussain <ahsan_hussain@mentor.com>
|
|
The Vulkan loader looks for pre-installed layers in
various hardcoded paths on the system and broadcasts
this information to the application when requested.
This now corrects the path for these layers and moves
the layers to one of the loader defined paths so they
are found automatically when searched.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This enables the ability to get max. and used UVD handles from
userspace.
JIRA Ticket: INTAMDDET-1890
Signed-off-by: Ahsan Hussain <ahsan_hussain@mentor.com>
|
|
These configurations conflict with KGDB debugging in
some cases and generate stack dumps in various other
situations.
The stack dumps related to mwaitx are properly fixed
through
https://patchwork.kernel.org/patch/8549611/
but the mentioned patch is only available for Yocto
kernel 4.8 and we don't use these so it is much
better to avoid the problem with this for the time
being rather than backporting the patch.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
Setting PN and FILES inside the recipe was conflicting with the
module.bbclass bits. SPI driver made it into the RFS because of
different module package name (due to source files) than what was
specified in the recipe. This is also why we were seeing two packages
for it, namely kernel-module-spi-amd and kernel-module-amd-spi.
Signed-off-by: Ahsan Hussain <ahsan_hussain@mentor.com>
|
|
OGL 4.3 and kernel bug fixes
|
|
|
|
Signed-off-by: Sanjay R Mehta <sanju.mehta@amd.com>
|
|
The Vulkan integration mainly targets the MEL Commercial
configurations so the required tools should only be
populated in that particular case.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
We have analyzed that the configuration variable INCLUDE_GAMING_COMPONENTS,
is no longer valid as the components it include are not particularly
specific to gaming. So we have splitted it into two configuration variables
INCLUDE_VULKAN and INCLUDE_CODEXL.
Signed-off-by: Adeel Arshad <adeel_arshad@mentor.com>
|
|
The vulkaninfo command segfaults if the DISPLAY
environment variable is not set because it uses xlib and
xcb interfaces which in a way that assumes availability
of the said variable.
We now explicitly check such scenarios to avoid
segfaulting and provide user with a bit of detail.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
The machine configuration now includes a config fragment
that can be provided through customer layers. This will
help in situations where a variable can only be touched
through conf files rather than recipes one such example is
the SERIAL_CONSOLES variable.
Also pick up APPEND_ADDITIONAL which can be passed through
the same customer layer for flexibility.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
Fixes warnings printed out during kernel_configcheck
phase regarding CONFIG_BT_6LOWPAN and
CONFIG_DRM_I915_PRELIMINARY_HW_SUPPORT.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
libdrm: Cleanup libdrm recipes
|
|
Move libdrm recipe into the common layer.
Signed-off-by: Drew Moseley <drew_moseley@mentor.com>
|
|
The AMD_GNB_BUS was available under kernel 4.1 which
has been dropped during the upgrade to 4.4 so the
auto loading mechnism is not required anymore.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
The glslangValidator is required for application
development on the host through CodeBench.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This updates all the components of LunarG SDK to version
1.0.26 from 1.0.21 and adjusts patches/configurations
to cater the updated bits.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
As a best practise the serial consoles should be
provided earlier on the kernel command line than
the vga consoles because the last console is used
for taking user input and is mapped to /dev/console.
Using it differently can break things at places
where conflicting/incorrect serial drivers can
generate a kernel panic in such a scenario.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
Same name of demos are available under mesa-demos
package so we rename here just to avoid any conflicts
and make both set of demos available on hte target.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This consolidates common settings between standard and
RT type kernel across the BSPs.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
This commit does a few things which cannot be separated easily
so it cannot be split in to separate commits.
* Drops all 4.1 kernel bits.
* Moves all common patches from meta-amdfalconx86 to common.
* Moves SE/BE builds to 4.4 kernel.
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
Signed-off-by: Awais Belal <awais_belal@mentor.com>
|
|
|