Is openCL gpu drivers are supported?
I have included the User space Driver.
The binary library is /usr/lib/aarch64-linux-gnu/libmali.so
I will work on installing the OpenCl and report back.
YES —>>> openCL Platforms are available. See post below.
The included kernel has the Mali Driver and the user-space binary from Lebain with Tensorflow is included in the image.
To enable the opencl and see the platforms do the following.
sudo apt-get install mesa-utils
sudo apt-get install ocl-icd-* opencl-headers
Then run
---------------------------------------------------------------------------------------------
dpkg -s libglu1-mesa
You should see the following.
Package: libglu1-mesa
Status: install ok installed
Priority: optional
Section: libs
Installed-Size: 411
Maintainer: Ubuntu Developers <ubuntu-devel-discuss@lists.ubuntu.com>
Architecture: arm64
Multi-Arch: same
Source: libglu
Version: 9.0.0-2.1build1
Replaces: libglu1
Provides: libglu1
Depends: libc6 (>= 2.17), libgcc1 (>= 1:3.0), libgl1-mesa-glx | libgl1, libstdc+ +6 (>= 4.1.1)
Conflicts: libglu1, mesag3 (<< 5.0.0-1), xlibmesa3
Description: Mesa OpenGL utility library (GLU)
Finish things up by running.
----------------------------------------------------------------------------------------------------------
sudo mkdir -p /etc/OpenCL/vendors/
echo "libmali.so" | sudo tee /etc/OpenCL/vendors/mali.icd
Then you can run CLINFO and see the Mali Platforms.
sudo clinfo
AVAILABLE PLATFORMS
-----------------------------------------------------------------------------------------------------------
root@hikey970:/etc/OpenCL/vendors# cat mali.icd
libmali.so
root@hikey970:/etc/OpenCL/vendors# sudo clinfo
Number of platforms 1
Platform Name ARM Platform
Platform Vendor ARM
Platform Version OpenCL 2.0 v1.r10p0-01rel0.e99 0c3e3ae25bde6c6a1b96097209d52
Platform Profile FULL_PROFILE
Platform Extensions cl_khr_global_int32_base_atomi cs cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_l ocal_int32_extended_atomics cl_khr_byte_addressable_store cl_khr_3d_image_writes cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp16 cl_khr_icd cl_khr_egl_image cl_khr_image2d_from_buffer cl_khr_depth_images cl_arm_core_id c l_arm_printf cl_arm_thread_limit_hint cl_arm_non_uniform_work_group_size cl_arm_ import_memory cl_arm_shared_virtual_memory
Platform Extensions function suffix ARM
Platform Name ARM Platform
Number of devices 1
Device Name Mali-G72
Device Vendor ARM
Device Vendor ID 0x62210001
Device Version OpenCL 2.0 v1.r10p0-01rel0.e99 0c3e3ae25bde6c6a1b96097209d52
Driver Version 2.0
Device OpenCL C Version OpenCL C 2.0 v1.r10p0-01rel0.e 990c3e3ae25bde6c6a1b96097209d52
Device Type GPU
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 12
Max clock frequency 767MHz
Device Partition (core)
Max number of sub-devices 0
Supported partition types None
Max work item dimensions 3
Max work item sizes 384x384x384
Max work group size 384
Preferred work group size multiple 4
Preferred / native vector sizes
char 16 / 4
short 8 / 2
int 4 / 1
long 2 / 1
half 8 / 2 (cl_khr_fp 16)
float 4 / 1
double 0 / 0 (n/a)
Half-precision Floating-point support (cl_khr_fp16)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Single-precision Floating-point support (core)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations No
Double-precision Floating-point support (n/a)
Address bits 64, Little-Endian
Global memory size 4294967296 (4GiB)
Error Correction support No
Max memory allocation 1073741824 (1024MiB)
Unified memory for Host and Device Yes
Shared Virtual Memory (SVM) capabilities (core)
Coarse-grained buffer sharing Yes
Fine-grained buffer sharing Yes
Fine-grained system sharing No
Atomics Yes
Shared Virtual Memory (SVM) capabilities (ARM)
Coarse-grained buffer sharing Yes
Fine-grained buffer sharing Yes
Fine-grained system sharing No
Atomics Yes
Minimum alignment for any data type 128 bytes
Alignment of base address 1024 bits (128 bytes)
Preferred alignment for atomics
SVM 0 bytes
Global 0 bytes
Local 0 bytes
Max size for global variable 65536 (64KiB)
Preferred total size of global vars 0
Global Memory cache type Read/Write
Global Memory cache size 524288 (512KiB)
Global Memory cache line size 64 bytes
Image support Yes
Max number of samplers per kernel 16
Max size for 1D images from buffer 65536 pixels
Max 1D or 2D image array size 2048 images
Base address alignment for 2D image buffers 32 bytes
Pitch alignment for 2D image buffers 64 pixels
Max 2D image size 65536x65536 pixels
Max 3D image size 65536x65536x65536 pixels
Max number of read image args 128
Max number of write image args 64
Max number of read/write image args 64
Max number of pipe args 16
Max active pipe reservations 1
Max pipe packet size 1024
Local memory type Global
Local memory size 32768 (32KiB)
Max number of constant args 8
Max constant buffer size 65536 (64KiB)
Max size of kernel argument 1024
Queue properties (on host)
Out-of-order execution Yes
Profiling Yes
Queue properties (on device)
Out-of-order execution Yes
Profiling Yes
Preferred size 2097152 (2MiB)
Max size 16777216 (16MiB)
Max queues on device 1
Max events on device 1024
Prefer user sync for interop No
Profiling timer resolution 1000ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
printf() buffer size 1048576 (1024KiB)
Built-in kernels
Device Extensions cl_khr_global_int32_base_atomi cs cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_l ocal_int32_extended_atomics cl_khr_byte_addressable_store cl_khr_3d_image_writes cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp16 cl_khr_icd cl_khr_egl_image cl_khr_image2d_from_buffer cl_khr_depth_images cl_arm_core_id c l_arm_printf cl_arm_thread_limit_hint cl_arm_non_uniform_work_group_size cl_arm_ import_memory cl_arm_shared_virtual_memory
NULL platform behavior
clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...) ARM Platform
clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...) Success [ARM]
clCreateContext(NULL, ...) [default] Success [ARM]
clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT) Success (1)
Platform Name ARM Platform
Device Name Mali-G72
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU) No devices found in platfor m
clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU) Success (1)
Platform Name ARM Platform
Device Name Mali-G72
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM) No devices found in plat form
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL) Success (1)
Platform Name ARM Platform
Device Name Mali-G72
ICD loader properties
ICD loader Name OpenCL ICD Loader
ICD loader Vendor OCL Icd free software
ICD loader Version 2.2.11
ICD loader Profile OpenCL 2.1Never mind my bad …
Thank you for letting me know. I have not seen that happen.
If you are going to flash the ubuntu_bionic.hikey970.V-2.0.sparse.img there is nothing to worry about.
The image has the kernel installed prior to creating the sparse image. So everything that needs to be in the rootfs / system image is there.
The Kernel package is zipped to that file.
That kernel package is made in case you are just building the kernel by itself.
That way you can copy the gz to the hikey970 and install the kernel.
NATIVE KERNEL COMPILING
I will mention as well after you have flashed the image and run /etc/init.sh
everything that needs to be installed to natively compile the kernel on the hikey970 is installed.
So on the hikey you can clone the kernel tree and build on hikey970.
https://github.com/Bigcountry907/linux.git -b hikey970-v4.9-Debain-Working
Then run
export arch=arm64
make arch=arm64 hikey970_defconfig
make arch=arm64 -j8
make arch=arm64 modules_install
make arch=arm64 install
After that the kernel is installed on the board.
But since grub uses different names for Image.gz and the DTB please copy the Image.gz and the DTB to the directory /boot/.
The DTB has to be named kirin970-hikey970.dtb.
it is located in /arch/arm64/boot/dts/hisilicon/ in the kernel tree.
cp -rf /arch/arm64/boot/dts/hisilicon/kirin970-hikey970.dtb /boot/
The Kernel image.gz has to be named Image-hikey970-v4.9.gz.
it is located in /arch/arm64/boot/ in the kernel tree.
cp -rf /arch/arm64/boot/Image.gz /boot/
mv /boot/Image.gz Image-hikey970-v4.9.gz
Previous directions are only for building the kernel on the hikey970.
If anyone has any questions at all feel free to ask.
I will do my best to help.
Thanks… as I said I goofed. I forgot to do ‘sudo -s’ before running ./BB.sh
I’m glad it worked out good for you.
If anyone has any recommendations or would like to see other functions added please let me know.
For anyone building the kernel only to use with any image other than the rootfs created by the builder please remember to copy the 1920x1080.bin from the ~/Bionic-Builder/binaries/ folder to /lib/firmware/edid on the hikey970 board.
Otherwise the display may not work.
In the future i will build the firmware into the kernel.
Truly awesome tool. Everything went smoothly except for few self inflicted injuries. The only problem I have is my Ethernet connection behaves erratically. Sometimes it comes up at startup and sometime it doesn’t. If it doesn’t come up during start the manual start also does not work. I have to keep restating until it comes up. I spent lot of time online and fiddling with netplan config file ‘01-dhcp.yaml’. But so far no luck.
Lastly kudos to ‘BigCountry907’ for all the hard work to make this happen.
Thank you very much for your kind words.
I have not experienced the issue with the Ethernet but i don’t use it a lot.
I can tell you that 01-dhcp.yaml is a very very touchy file because of the .yaml format.
The spacing used has to be 100 % correct or you will get errors.
This is good page to reference for netplan configurations.
When i get a chance i will take a look and see if i can get the ethernet to have any problems.
The Ethernet controller is run by one lane of the pcie switch.
https://www.broadcom.com/products/pcie-switches-bridges/pcie-switches/pex8606
Are you running any cards in the Mini Pcie slot or the M.2 Key m slot?
I have seen reports that having a card in both slots has caused at least 1 person problems.
Soon as i get a ATX power supply I will be testing some pcie 1x to 16x risers in the mini and m.2 slots at which time if there are any problems with the switch I should find out pretty quickly.
Thanks… and no I am not running any cards in Mini Pcie slot or the M.2. Yes I was very careful with config file as I read it could spell havoc if not done properly. Meanwhile I will just keep looking for solution online and trying it.
On the board i have if I connect the Ethernet prior to booting up it works almost every time. If i boot first then connect the Ethernet sometimes it will work and sometimes it will not.
Okay after spend lot of time on this issue I am convinced that it is hardware issue. My RJ45 jack seems to be the culprit. If I push the cable hard (even when it is locked in the jack) before startup, it works fine. I tried several different cables with same result. Anyway thanks for all the help.
Sorry to hear about your bad luck.
I have a wireless router that acts the same way.
If you really need the Ethernet to work I would recommend adding a card to the mini pcie slot.
Other than that all i can think is push the cable in tight to where it works and then hot glue it. At least with the hot glue you can always peel the glue off and remove it easily. Maybe the place you purchased the board from will exchange it for you.
Thanks… I will give it a shot.
I thought tesorflow is builtin in this image but I can’t seem to find it. Anyone else tried it ?
Open Cl is included in this image.
You should be able to take the tensorflow binaries from the LeMaker image and use them. Otherwise find a arm64 build of tensorflow or compile tensorflow from source.
Thanks… how do I install headers for kernel 4.9.78-147560-g37a1403604e8 ? Or 4.15.0-50-generic build will do ?
The kernel source is located in the ~/Bionic-Builder/Kernel-SRC/linux/ directory.
Usually after i have ubuntu up and running on the Hikey 970 I clone the kernel src tree and build the kernel on the hikey970.
git clone https://github.com/Bigcountry907/linux.git -b hikey970-v4.9-Debain-Working
Then i will run
make arch=arm64 mrproper
make arch=arm64 hikey970_defconfig
make arch=arm64 -j8
make arch=arm64 install
make arch=arm64 modules_install
I have not checked if the kernel headers are installed using that process or not. But you can check that. If the headers are not installed you have to install the headers that match the kernel version. 4.9.78
The 4.9.78 is the only portion of the version that needs to match.
You can not install headers for 4.15.0-50 if you are running the 4.9.78 kernel.
Here is a tutorial on installing the headers.
Oh wow… thanks for the detailed explanation. I appreciate it. Will give it a try…
Hello, BigCountry907, Is there a URL for ubuntu_bionic.hikey970.V-2.0.sparse.img?
