Age | Commit message (Collapse) | Author |
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
While build aarch64 with gcc <= 8.2, there is a ice failure[1]:
...
|external/eigen_archive/unsupported/Eigen/CXX11/../../../Eigen/src/Core/
products/GeneralBlockPanelKernel.h:1879:3: internal compiler error:
in emit_move_insn, at expr.c:3722
...
Backport a patch[2] from upstream eigen to workaround the issue
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89752
[2] https://gitlab.com/libeigen/eigen/-/commit/0d906378385a9f94f612528523ba2ad9a6d1e265
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
There is gcc Killed failure:
...
|tmp-glibc/work/corei7-32-wrsmllib32-linux/lib32-tensorflow/
1.13.0-r0/git/tensorflow/core/kernels/BUILD:3158:1: C++ compilation
of rule '//tensorflow/core/kernels:cwise_op' failed (Killed): gcc
failed: error executing command
...
From Bazel User Manual
--local_resources availableRAM,availableCPU,availableIO
This option, which takes three comma-separated floating point arguments,
specifies the amount of local resources that Bazel can take into
consideration when scheduling build and test activities. Option expects
amount of available RAM (in MB), number of CPU cores (with 1.0 representing
single full core) and workstation I/O capability (with 1.0 representing
average workstation). [1]
According to [2][3], explicitly set the amount of local host RAM
4096 MB available to Bazel, the number of local CPU 4 cores
available to Bazel, the number of jobs is 4 which should be
executed concurrently.
Refer:
[1] https://stackoverflow.com/questions/34756370/is-there-a-way-to-limit-the-number-of-cpu-cores-bazel-uses
[2] https://github.com/tensorflow/tensorflow/issues/7723
[3] https://github.com/tensorflow/tensorflow/issues/24318
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
The python3-core provides /usr/bin/python3 which required by
python3 scripts of tensorboard/tensorflow
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
The python3 script contains build path:
...
'''exec' tmp-glibc/work/corei7-64-wrs-linux/tensorflow/1.13.0-r0/
recipe-sysroot-native/usr/bin/nativepython3 "$0" "$@"
...
Improve sed operation to match nativepython3
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
If [batch mode is] set, Bazel will be run as just a client process
without a server, instead of in the standard client/server mode. [1]
It is not necessary to shutdown a bazel server in batch mode.
[1] https://docs.bazel.build/versions/master/user-manual.html
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Move the duplicated part of tensorflow and tensorflow-native to
tensorflow.inc
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Since Yocto has gcc9, the ice fix is not necessary, drop it
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Support to override bazel jobs, then user could set it in local.conf
according to the host performance
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
7.3.1
Previously, we workaround tensorflow-native ice failure while building
from source with host gcc 7.3.1, and tensorflow build requires the same
fix since bazel cross compiling will do the native build at the same time.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Do not support libn32-tensorflow on qemumips64, it introduces
many build failures, we do not plan to maintain them.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
On opensuse151_64, the gcc lib include dir is /usr/lib64/gcc/ rather than
/usr/lib/gcc/, which caused failure
...
|tensorflow/1.13.0-r0/bazel/output_base/external/nasm/BUILD.bazel:8:1:
undeclared inclusion(s) in rule '@nasm//:nasm':
|this rule is missing dependency declarations for the following
files included by 'external/nasm/x86/regs.c':
| '/usr/lib64/gcc/x86_64-suse-linux/7/include-fixed/limits.h'
| '/usr/lib64/gcc/x86_64-suse-linux/7/include-fixed/syslimits.h'
| '/usr/lib64/gcc/x86_64-suse-linux/7/include/stddef.h'
| '/usr/lib64/gcc/x86_64-suse-linux/7/include/stdarg.h'
...
Add /usr/lib64/gcc/ to cxx_builtin_include_directory
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Tensorflow does not work on some arch, add bbclass to skip
all involved recipes.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Previously we removed -Wl,-latomic to fix native build failure,
but some arch's gcc (such as qemumips) still requires this link option.
Then add -Wl,-latomic back for target build only
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
gcc 7.3.1
The root cause is gcc ice issue:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89752
But we could not patch on native gcc,so try to workaround,
Refer: https://github.com/tensorflow/tensorflow/issues/25323
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
While enabling ccache, build bazel-native failed:
...
|Cannot find gcc or CC (ccache gcc); either correct your path
or set the CC environment variable
...
So disable ccache to workaround the failure.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Since convert python2 to python3, do not need to depend on meta-python2
Revert "layer.conf: depend on meta-python2"
This reverts commit af3f5917ebd83754b69b2b7f02af0718b7a6950b.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Drop python2, use python3 to build bazel
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Since Yocto has upgraded gcc to 9, it does not need to explicitly
link to libatomic.so to avoid failure of libatomic.so missing.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
If host does not provide unzip, build bazel will fail even though
Yocto native sysroot does have unzip.
The reason is var-PATH was not passed to bazel build in some cases.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
The build of tensorflow-for-poets will download model tarball from internet,
add it to SRC_URI to follow Yocto download mirror.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Backport patch to fix grpc conflicts issue
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
glibc 2.30 (ubuntu1904_64) will declare its own gettid; see
https://sourceware.org/git/?p=glibc.git;a=commit;h=1d0fc213824eaa2a8f8c4385daaa698ee8fb7c92.
Rename the grpc versions to avoid naming conflicts
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
We met the following error with python 3.8.
TypeError: _logger_find_caller() takes from 0 to 1 positional arguments but 2 were given
Backport a patch to fix this issue.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
|
|
Fix the following compliation error.
error: cannot convert 'std::nullptr_t' to 'Py_ssize_t {aka long int}' in initialization
This patch references: https://github.com/tensorflow/tensorflow/issues/34197
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
|
|
bazel-native needs python2-native, so depend on meta-python2.
Also update the README file.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
|
|
The compilation process needs python (python2) to be there, so
fix it by making bazel-native inherit pythonnative.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
|
|
The issue was caused by commit [bad2757 meta-demo/conf/layer.conf:
do not support to save avi video], clean it.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
...
|ERROR: tensorflow-1.13.0-r0 do_package: tensorflow: Multiple shlib
providers for libtensorflow_framework.so: tensorflow, tensorflow (used
by files: tmp-glibc/work/core2-64-wrs-linux/tensorflow/1.13.0-r0/packa
ges-split/tensorflow/usr/lib64/python3.7/site-packages/tensorflow/contrib/
image/python/ops/_distort_image_ops.so)
...
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
...
| requires /usr/bin/python3, but no providers found in RDEPENDS_tensorflow-for-poets? [file-rdeps]
...
Add python3-core to RDEPENDS
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|
|
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
|
|
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
|