Hi @Alex-Kushleyev ,
Has there been a tag created in https://gitlab.com/voxl-public/voxl-sdk/core-libs/libmodal-pipe that contains these changes? Please let me know when that happens, would would like to get these CPU gains moved into production ASAP!
All the best,
Rowan
Hi @Alex-Kushleyev,
I confirmed that the https://gitlab.com/voxl-public/voxl-sdk/core-libs/libmodal-pipe/-/tree/ion-32bit-test branch does indeed allow voxl-qvio-server to use ION pipes (using DEN_ION_BUF and DEN_32BIT_ION_BUF). I have not detected any other breakages of the Cleo SDK or performance degradations, but that's something we will monitor as we test more.
In CPU perf mode, average core util dropped from 31% to 22%, this is a huge help, thank you very much for your work.
We will be eagerly awaiting the release of a libmodal-pipe gitlab tag with these changes in it!
Cheers,
Rowan
Hi @Alex-Kushleyev,
I did a quick test. Can you please test to make sure the updated libmodal pipe works with your 32 bit application and also does not break the main 64-bit functionality? Then i think we can merge it to dev.
Yes absolutely, I can get that done today.
Thanks for your patience!
Thank you for taking this on!
Rowan
@Alex-Kushleyev Hi Alex, Happy New Years 
Is there an ETA on when the 32-bit ION buffer support in libmodal-pipe will be main-lined? For our internal time planning here at Cleo.
Thank you,
Rowan
@Moderator Hi Modal,
Does this mean you can create a custom data file partition that you can flash using fastboot?
Yes that is correct.
We have not explored this before but it has been asked a few times. This could be very helpful for other developers.
I'm certainly hopeful that it will be helpful here at Cleo Robotics! So far with my prototyping it works as expected and cuts flashing time of some large docker images we have here at Cleo down by a noticeable fraction (no file overhead via fastboot like with ADB).
I think the snippet I posted covers the baseline functionality of getting a custom "payload" into the data partition. However if there is more I can elaborate on in terms of the toolchain / what's in the payload, and if that elaboration will be helpful to other VOXL2 developers, I would be happy to elaborate Just let me know!
Other similar discussion points I tackled recently that I'm happy to talk about lessons of:
For anyone reading this in the future this works:
docker run --rm --privileged \
-v "$SYSTEM_IMAGE_DIR":/system-image:ro \
-v "$CLEO_IMAGE_DIR":/cleo-image \
-v "$CLEO_PROVISION_DIR":/cleo-provision:ro \
ubuntu:18.04 bash -c "
set -e
# Install tools
apt-get update -qq
apt-get install -y -qq android-tools-fsutils e2fsprogs >/dev/null
# Extract stock data partition contents
echo 'Extracting stock m0054-data-fs.ext4...'
mkdir -p /tmp/data_root
# Convert sparse to raw and mount
simg2img /system-image/m0054-data-fs.ext4 /tmp/stock_raw.ext4
mkdir -p /mnt/stock
mount -o loop,ro /tmp/stock_raw.ext4 /mnt/stock
# Copy all stock files
cp -a /mnt/stock/* /tmp/data_root/
umount /mnt/stock
rm /tmp/stock_raw.ext4
echo 'Stock contents copied.'
# Add cleo-provision directory
cp -a /cleo-provision /tmp/data_root/cleo-provision
echo 'Added cleo-provision directory'
# Create the ext4 image using same parameters as BitBake recipe:
# -s : sparse output
# -l SIZE : filesystem size in bytes
# -a /data : android mount point
# -b 4096 : block size
make_ext4fs -s -l $USERDATA_SIZE_EXT4 -a /data -b 4096 /cleo-image/m0054-data-fs.ext4 /tmp/data_root
echo ''
echo 'Created sparse ext4 image successfully'
"
You'll need to be on an arm64 architecture.
I found https://gitlab.com/voxl-public/system-image-build/meta-voxl2/-/blob/qrb5165-ubun1.0-14.1a/recipes-products/image/qti-ubuntu-robotics-image.bbappend#L148, which is in bitbake language (not familiar)?
Is
do_makeuserdata() {
make_ext4fs -s -l ${USERDATA_SIZE_EXT4} ${IMAGE_EXT4_SELINUX_OPTIONS} \
-a /data -b 4096 ${DEPLOY_DIR_IMAGE}/${OVERLAYIMAGE_TARGET} ${IMAGE_ROOTFS}/data
}
something I could do just on normal ubuntu in CI? I'm thinking to build a cleo-specific userdata in CI with all of our custom software.
What are ${IMAGE_EXT4_SELINUX_OPTIONS}, is there some secret sauce there?
Hi Modal,
I'm looking into adding some data in the m0054-data-fs.ext file system. I'm having some trouble getting the voxl2 to work correctly when I try to add the data and then create a new Android sparse image that I flash instead of m0054-data-fs.ext. Would it be possible to provide me with the tools used to create that m0054-data-fs.ext file system at ModalAI so that I can use the exact same tools to create the sparse android image after adding my files to it?
Thank you,
Rowan
@Alex-Kushleyev Sounds good! I did not make any changes since the revisions you shared 5 days ago.
sure, that would be helpful.
i will check with the team today to make sure we are all good, so perhaps you should wait for that.
Will do!
@Alex-Kushleyev Okay sounds like we're almost there in terms of readiness for mainline then. Would you like me to take that on and put up a gitlab MR from my fork into modalai's libmodal-pipe repo? I can schedule an hour or two to polish that up and get it ready sometime in the next few days.
@Alex-Kushleyev Hi Alex,
Success! I was able to run QVIO using the ION pipe and saw the resultant reduction in CPU util both on the camera server process as well as the QVIO process (small but noticeable reductions).
Note that I was not able to run the camera server with the 32-bit customized libmodal-pipe was installed. So the approach I took was run voxl-camera-server first with the unmodified libmodal-pipe, then install the 32-bit customized libmodal pipe, and run QVIO. Of course this would not be feasible in production. Here are the errors I'm getting when trying to run voxl-camera-server with the 32-bit customized libmodal-pipe installed:
=================================================================
thread is locked to cores: 4 5 6 7
connected to mavlink pipe
Connected to cpu-monitor
Starting Camera: back (id #0)
gbm_create_device(156): Info: backend name is: msm_drm
MISP Initializing for camera back
Detected 1 platform(s)
Detected 1 GPU device(s)
Estimated imu dt = 0.000977s
WARNING: OMX SetTargetBitrate: H265 CBR requires bps >= 3.0Mbit (1200000 bps provided). Using FPS hack. scale = 2.500000
WARNING: Encoder expecting(16) more buffers than module allocated(1)
ERROR: OMX Set config failed!
ERROR: Failed to initialize MISP encoder for camera: back
ERROR: Failed to start camera: back
Starting Camera: top (id #1)
MISP Initializing for camera top
WARNING: OMX SetTargetBitrate: H265 CBR requires bps >= 3.0Mbit (1200000 bps provided). Using FPS hack. scale = 2.500000
WARNING: Encoder expecting(16) more buffers than module allocated(1)
ERROR: OMX Set config failed!
ERROR: Failed to initialize MISP encoder for camera: top
ERROR: Failed to start camera: top
Starting Camera: tof (id #2)
Starting Camera: hires (id #3)
WARNING: OMX SetTargetBitrate: H265 CBR requires bps >= 3.0Mbit (2000000 bps provided). Using FPS hack. scale = 1.500000
WARNING: Encoder expecting(16) more buffers than module allocated(1)
ERROR: OMX Set config failed!
ERROR: Failed to initialize encoder for camera: hires
ERROR: Failed to start camera: hires
Starting Camera: bottom (id #4)
MISP Initializing for camera bottom
WARNING: OMX SetTargetBitrate: H265 CBR requires bps >= 3.0Mbit (1200000 bps provided). Using FPS hack. scale = 2.500000
WARNING: Encoder expecting(16) more buffers than module allocated(1)
ERROR: OMX Set config failed!
ERROR: Failed to initialize MISP encoder for camera: bottom
ERROR: Failed to start camera: bottom
What are the next steps towards mainlining this in a way that works "out of the box"? Anything else I should test or any way I can support that process?
Thank you,
Rowan
@Alex-Kushleyev Here are the diffs:
Thanks for your help!
@Alex-Kushleyev Hi Alex, I compiled libmodal-pipe with ION client-only support in 32 bit builds. I then changed the voxl-mpa-tools toolchain to 32 bit and set the 32 bit toolchain to actually build the tool executables and removed the ones that didn't compile. After deploying those two modified packages to the VOXL2 and running the same repub command that I ran before, I am getting a runtime error while parsing the incoming ION buffers on the client side
voxl2:/$ voxl-image-repub top_misp_norm_ion -o top_misp_norm_ion_test
Subscribed to ION input pipe: top_misp_norm_ion
invalid metadata, magic number=85, expected 1229934146
invalid metadata, magic number=85, expected 1229934146
invalid metadata, magic number=85, expected 1229934146
Tracking it down, it looks like the call to pipe_client_bytes_in_pipe in the _ion_buf_cb is what's printing the error. If I remove that line (so not returning if pipe_client_bytes_in_pipe returns 0) it prints some garbage (incorrect) metadata and then segfaults:
voxl2:/$ voxl-image-repub top_misp_norm_ion -o top_misp_norm_ion_test
Subscribed to ION input pipe: top_misp_norm_ion
CH0 Got input ion frame size: 10330x-2843, format: STEREO_NV12, stride: 50834, size: 396 bytes
Segmentation fault:
Fault thread: voxl-image-repu(tid: 7900)
Fault address: 0x4
Address not mapped.
Segmentation fault
Note that I didn't recompile voxl-camera-server on-top of the modified libmodal-pipe, don't think that would change anything.
Let me know if you want me to send you my diffs of libmodal-pipe (w.r.t. master) and voxl-mpa-tools (w.r.t. add-new-image-tools).
P.S. I also tried voxl-inspect-ion-stream top_misp_norm_ion (that tool is also 32 bit now) and it's printing garbage too
| Pipe Name | fd | bytes | wide | hgt |exp(ms)| gain | frame id |latency(ms)| fps | mbps |t
| top_misp_norm_ion | 6 | 455 | 3237 |-29847 | 52.43 |-24576 |1448040524 |1955547.7 | 30.0 | 0.6
@Alex-Kushleyev Sounds good, I proceeded with resolving the basic typing issues related to 32 bit monotonic time. Next issue is a linker error that looks a bit more tricky:
-- Build files have been written to: /home/root/build32
[ 10%] Building CXX object library/CMakeFiles/modal_pipe.dir/src/buffers.cpp.o
[ 40%] Building C object library/CMakeFiles/modal_pipe.dir/src/server.c.o
[ 40%] Building C object library/CMakeFiles/modal_pipe.dir/src/common.c.o
[ 40%] Building C object library/CMakeFiles/modal_pipe.dir/src/misc.c.o
[ 50%] Building CXX object library/CMakeFiles/modal_pipe.dir/src/buffers/gbm.cpp.o
[ 60%] Building C object library/CMakeFiles/modal_pipe.dir/src/interfaces.c.o
[ 70%] Building C object library/CMakeFiles/modal_pipe.dir/src/client.c.o
[ 80%] Building C object library/CMakeFiles/modal_pipe.dir/src/sink.c.o
[ 90%] Building C object library/CMakeFiles/modal_pipe.dir/src/start_stop.c.o
[100%] Linking CXX shared library libmodal_pipe.so
/opt/sysroots/qrb5165_1/usr/lib/libgbm.so: file not recognized: file format not recognized
collect2: error: ld returned 1 exit status
make[2]: *** [library/CMakeFiles/modal_pipe.dir/build.make:228: library/libmodal_pipe.so] Error 1
make[1]: *** [CMakeFiles/Makefile2:106: library/CMakeFiles/modal_pipe.dir/all] Error 2
make: *** [Makefile:136: all] Error 2
@Alex-Kushleyev Hi Alex I'm running into compilation errors when building libmodal-pipe after turning the EN_ION_BUF flag ON (builds fine with the flag OFF)
Found voxl-cross version: 4.4
-- ---------------------------------------------------------
-- Using voxl-cross 32-bit toolchain for QRB5165
-- C Compiler : /usr/bin/arm-linux-gnueabi-gcc-7
-- C++ Compiler : /usr/bin/arm-linux-gnueabi-g++-7
-- Sysroot : /opt/sysroots/qrb5165_1
-- C flags : -isystem=/usr/arm-linux-gnueabi/include -isystem=/usr/include -idirafter /usr/include -fno-stack-protector -march=armv7-a -mfloat-abi=softfp -mfpu=vfpv4
-- CXX flags : -isystem=/usr/arm-linux-gnueabi/include -isystem=/usr/include -idirafter /usr/include -fno-stack-protector -march=armv7-a -mfloat-abi=softfp -mfpu=vfpv4
-- EXE Link Flags : -L/opt/sysroots/qrb5165_1/usr/lib32 -L/opt/sysroots/qrb5165_1/lib -L/opt/sysroots/qrb5165_1/usr/lib -L/opt/sysroots/qrb5165_1/usr/arm-linux-gnueabi -L/opt/sysroots/qrb5165_1/usr/lib/gcc-cross/arm-linux-gnueabi/7 -L/usr/lib
-- SO Link Flags : -L/opt/sysroots/qrb5165_1/usr/lib32 -L/opt/sysroots/qrb5165_1/lib -L/opt/sysroots/qrb5165_1/usr/lib -L/opt/sysroots/qrb5165_1/usr/arm-linux-gnueabi -L/opt/sysroots/qrb5165_1/usr/lib/gcc-cross/arm-linux-gnueabi/7 /opt/sysroots/qrb5165_1/usr/lib/gcc-cross/arm-linux-gnueabi/7/libgcc.a /opt/sysroots/qrb5165_1/usr/lib/gcc-cross/arm-linux-gnueabi/7/libgcc_eh.a /opt/sysroots/qrb5165_1/usr/lib/gcc-cross/arm-linux-gnueabi/7/libssp_nonshared.a /opt/sysroots/qrb5165_1/usr/arm-linux-gnueabi/lib/libc_nonshared.a -L/usr/lib
-- The C compiler identification is GNU 7.3.0
-- The CXX compiler identification is GNU 7.3.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/arm-linux-gnueabi-gcc-7 - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/arm-linux-gnueabi-g++-7 - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- INFO: building with ion buf support
-- INFO: building with mavlink support
INFO: Skipping Building Python Bindings
-- INFO: Skipping examples and tools
-- Configuring done (1.8s)
-- Generating done (0.0s)
-- Build files have been written to: /home/root/build32
[ 10%] Building C object library/CMakeFiles/modal_pipe.dir/src/start_stop.c.o
[ 30%] Building C object library/CMakeFiles/modal_pipe.dir/src/client.c.o
[ 30%] Building CXX object library/CMakeFiles/modal_pipe.dir/src/buffers.cpp.o
[ 50%] Building C object library/CMakeFiles/modal_pipe.dir/src/interfaces.c.o
[ 50%] Building C object library/CMakeFiles/modal_pipe.dir/src/sink.c.o
[ 60%] Building C object library/CMakeFiles/modal_pipe.dir/src/common.c.o
[ 70%] Building C object library/CMakeFiles/modal_pipe.dir/src/misc.c.o
[ 80%] Building C object library/CMakeFiles/modal_pipe.dir/src/server.c.o
[ 90%] Building CXX object library/CMakeFiles/modal_pipe.dir/src/buffers/gbm.cpp.o
/home/root/library/src/server.c: In function 'pipe_server_write_ion_buffer':
/home/root/library/src/server.c:1757:102: error: format '%ld' expects argument of type 'long int', but argument 5 has type 'int64_t {aka long long int}' [-Werror=format=]
printf("server preparing to send ion buffer id %d (fd %d) to clients, n_clients: %d, time: %ld\n",
~~^
%lld
ion_buf->buffer_id, fd, c[ch].n_clients, _time_monotonic_ns());
~~~~~~~~~~~~~~~~~~~~
/home/root/library/src/client.c: In function '_stop_helper_and_remove_pipe':
/home/root/library/src/client.c:1353:56: error: format '%ld' expects argument of type 'long int', but argument 3 has type 'int64_t {aka long long int}' [-Werror=format=]
if(en_debug) printf("ch: %d, shutdown socket %ld ns\n", ch, _time_monotonic_ns());
~~^ ~~~~~~~~~~~~~~~~~~~~
%lld
/home/root/library/src/client.c: In function 'pipe_client_close':
/home/root/library/src/client.c:1475:43: error: format '%ld' expects argument of type 'long int', but argument 3 has type 'int64_t {aka long long int}' [-Werror=format=]
printf("ch: %d, shutdown socket %ld ms\n", ch, _time_monotonic_ns());
~~^ ~~~~~~~~~~~~~~~~~~~~
%lld
cc1: all warnings being treated as errors
make[2]: *** [library/CMakeFiles/modal_pipe.dir/build.make:93: library/CMakeFiles/modal_pipe.dir/src/client.c.o] Error 1
make[2]: *** Waiting for unfinished jobs....
cc1: all warnings being treated as errors
make[2]: *** [library/CMakeFiles/modal_pipe.dir/build.make:149: library/CMakeFiles/modal_pipe.dir/src/server.c.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:106: library/CMakeFiles/modal_pipe.dir/all] Error 2
make: *** [Makefile:136: all] Error 2
Seems like I'm going down an uncharted path here. I'm okay with spending my time charting it out since getting QVIO to work with ION pipes will save us critical CPU util. Does ModalAI have plans to switch the official SDK QVIO server release (for qrb5165) to using ION pipes (I saw ModalAI switched to using the MISP pipes for QVIO a few months ago, taking a CPU hit)?
@Alex-Kushleyev Actually had some time today to do the profiling and explore the ION pipes.
I see the same improvements that you quoted:
On 2.2.17
Normal pipes: 45%
MISP pipes: 66%
MISP pipes + 4x tracking pipes: 160%
On perf improvements
Normal pipes: 45%
MISP pipes: 52%
MISP pipes + 4x tracking pipes: 71%
On perf improvements + ION pipes
MISP pipes: 48%
MISP pipes + 4x tracking pipes: 52%
I started looking into using the ION pipe in the QVIO server code but saw this in the voxl-mpa-tools build script:
qrb5165)
check_docker "4.3"
mkdir -p build32
cd build32
cmake -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_QRB5165_1_32} \
-DBUILD_TOOLS=OFF \
-DEN_ION_BUF=OFF ../
make -j$(nproc)
cd ../
mkdir -p build64
cd build64
cmake -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_QRB5165_1_64} \
-DBUILD_TOOLS=ON \
-DEN_ION_BUF=ON ../
make -j$(nproc)
cd ../
;;
qrb5165-2)
check_docker "4.3"
mkdir -p build
cd build
cmake -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_QRB5165_2_64} \
-DBUILD_TOOLS=ON \
-DEN_ION_BUF=OFF ../
make -j$(nproc)
cd ../
;;
Since we're using the qrb5165 (not the qrb5165-2) we have to use the 32bit toolchain to build QVIO (right?):
qrb5165)
check_docker "4.4"
mkdir -p build
cd build
cmake -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_QRB5165_1_32} ../
make -j$(nproc)
cd ../
;;
Does this mean that on the qrb5165 (using the 32 bit toolchain), the QVIO server cannot be a client to the ION pipes?
Thank you,
Rowan
@Alex-Kushleyev Hi Alex, just acknowledging your messages, thanks for looking into this and providing us with a path forward. I will have time over the next 2-3 days to test the perf-optimizations branch and familiarize myself the ION based clients. I will update you then.
Hi Modal,
Just following up here, I am going to put our migration to OpenVINS onto the backlog for the time being. Please let me know when the SDK is ready for QVIO -> OpenVINS migration, and when there is bandwidth to assist us with the migration. You can reach me at rowan.dempster@cleorobotics.com.
@Alex-Kushleyev Hi Alex, just following up on any update regarding the shared ION buffers or other methods to work around the CPU hit taken by having many clients to misp based image pipes. Happy to try out any methods/suggestions with our specific clients!
Rowan