PoCL Installation on Jetson¶
Purpose & Motivation¶
PoCL is a portable open source (MIT-licensed) implementation of the OpenCL standard. In addition to being an easily portable/layered multi-device open-source OpenCL implementation, a major goal of this project is improving interoperability of diversity of OpenCL-capable devices by integrating them to a single centrally orchestrated platform. Also one of the key goals is to enhance performance portability of OpenCL programs across device types utilizing runtime and compiler techniques.
PoCL currently supports various CPU architectures (x86, ARM, RISC-V), NVIDIA GPUs via libCUDA, Intel GPUs via Level Zero and TCE ASIPs (OpenASIP) at different feature coverage levels. It also supports a remote backend for distributed OpenCL execution. PoCL is also known to have multiple (private) adaptations in active production use.
PoCL uses Clang as an OpenCL C frontend and LLVM for kernel compiler implementation, and as a portability layer. Thus, if your desired target has an LLVM backend, it should be able to get OpenCL support easily by using PoCL.
Warning
Nvidia Jetson boards do not support natively OpenCL, thus it is mandatory to install PoCL to run OpenCL applications on them.
In this page we will focus on installing PoCL version 5. This version supports CUDA 16-bit floats. The release note details the supported feature.
Note
On "old" Jetson boards (TX2, Xavier NX, AGX Xavier & Nano), it is not possible to install recent PoCL version 5 because the OS is too old (Ubuntu 18.04) and it is complicated to install a recent version of the required Clang compiler (version 17). This is why on these specific boards we will install PoCL version 3. One of the main drawback is that there is no GPU 16-bit float support in this version :-(.
Note
We suppose that on the Orin Nano, NX & AGX, the Jetpack 5.2 is installed.
Install LLVM¶
Download ARM64 build and extract it:
export LLVM_VERSION=10
sudo apt install -y build-essential ocl-icd-libopencl1 cmake git pkg-config libclang-${LLVM_VERSION}-dev clang-${LLVM_VERSION} llvm-${LLVM_VERSION} make ninja-build ocl-icd-libopencl1 ocl-icd-dev ocl-icd-opencl-dev libhwloc-dev zlib1g zlib1g-dev clinfo dialog apt-utils libxml2-dev libclang-cpp${LLVM_VERSION}-dev libclang-cpp${LLVM_VERSION} llvm-${LLVM_VERSION}-dev libncurses5
cd /opt
sudo wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.1.0/clang+llvm-11.1.0-aarch64-linux-gnu.tar.xz
sudo tar -xvvf clang+llvm-11.1.0-aarch64-linux-gnu.tar.xz
sudo mv clang+llvm-11.1.0-aarch64-linux-gnu llvm-11.1.0
sudo rm clang+llvm-11.1.0-aarch64-linux-gnu.tar.xz
export LLVM_VERSION=12
sudo apt install -y build-essential ocl-icd-libopencl1 cmake git pkg-config libclang-${LLVM_VERSION}-dev clang-${LLVM_VERSION} llvm-${LLVM_VERSION} make ninja-build ocl-icd-libopencl1 ocl-icd-dev ocl-icd-opencl-dev libhwloc-dev zlib1g zlib1g-dev clinfo dialog apt-utils libxml2-dev libclang-cpp${LLVM_VERSION}-dev libclang-cpp${LLVM_VERSION} llvm-${LLVM_VERSION}-dev libncurses5
cd /opt
sudo wget https://github.com/llvm/llvm-project/releases/download/llvmorg-17.0.6/clang+llvm-17.0.6-aarch64-linux-gnu.tar.xz .
sudo tar -xvvf clang+llvm-17.0.6-aarch64-linux-gnu.tar.xz
sudo mv clang+llvm-17.0.6-aarch64-linux-gnu llvm-17.0.6
sudo rm clang+llvm-17.0.6-aarch64-linux-gnu.tar.xz
Note
On the Jetson TX2, the /opt folder is replaced by /data/opt. This is
because the eMMC available space is low (~ 5 GB) and we preferred to install
software on the SSD which is mounted in /data.
Compile and Install PoCL¶
Clone the repository, compile the code and install it:
cd ~/
mkdir softwares
cd softwares
git clone -b release_3_0 https://github.com/pocl/pocl.git pocl_3.0
cd pocl_3.0
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/opt/pocl-3.0 -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-funroll-loops -march=native" -DCMAKE_C_FLAGS="-funroll-loops -march=native" -DWITH_LLVM_CONFIG=/opt/llvm-11.1.0/bin/llvm-config -DSTATIC_LLVM=ON -DENABLE_CUDA=ON ..
make -j6
sudo make install
sudo mkdir -p /etc/OpenCL/vendors/
sudo touch /etc/OpenCL/vendors/pocl.icd
echo "/opt/pocl-3.0/lib/libpocl.so" | sudo tee --append /etc/OpenCL/vendors/pocl.icd
cd ~/
mkdir softwares
cd softwares
git clone -b release_5_0 https://github.com/pocl/pocl.git pocl_5.0
cd pocl_5.0
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/opt/pocl-5.0 -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-funroll-loops -march=native" -DCMAKE_C_FLAGS="-funroll-loops -march=native" -DWITH_LLVM_CONFIG=/opt/llvm-17.0.6/bin/llvm-config -DSTATIC_LLVM=ON -DENABLE_CUDA=ON .. -DLLC_HOST_CPU=cortex-a78
make -j6
sudo make install
sudo mkdir -p /etc/OpenCL/vendors/
sudo touch /etc/OpenCL/vendors/pocl.icd
echo "/opt/pocl-5.0/lib/libpocl.so" | sudo tee --append /etc/OpenCL/vendors/pocl.icd
Now OpenCL should be successfully installed on the system. You can check if it works with the following command:
It should display something like this (example below has be taken from the Jetson TX2):
Number of platforms 1
Platform Name Portable Computing Language
Platform Vendor The pocl project
Platform Version OpenCL 3.0 PoCL 3.0-rc2 Linux, Release, RELOC, LLVM 11.1.0, SLEEF, FP16, CUDA, POCL_DEBUG
Platform Profile FULL_PROFILE
Platform Extensions cl_khr_icd cl_pocl_content_size
Platform Host timer resolution 0ns
Platform Extensions function suffix POCL
Platform Name Portable Computing Language
Number of devices 2
Device Name pthread-cortex-a57
Device Vendor ARM
Device Vendor ID 0x13b5
Device Version OpenCL 1.2 PoCL HSTR: pthread-aarch64-unknown-linux-gnu-cortex-a57
Driver Version 3.0-rc2
Device OpenCL C Version OpenCL C 1.2 PoCL
Device Type CPU
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 6
Max clock frequency 2035MHz
Device Partition (core)
Max number of sub-devices 6
Supported partition types equally, by counts
Max work item dimensions 3
Max work item sizes 4096x4096x4096
Max work group size 4096
Preferred work group size multiple 8
Preferred / native vector sizes
char 16 / 16
short 8 / 8
int 4 / 4
long 2 / 2
half 8 / 8 (cl_khr_fp16)
float 4 / 4
double 2 / 2 (cl_khr_fp64)
Half-precision Floating-point support (cl_khr_fp16)
Denormals No
Infinity and NANs No
Round to nearest No
Round to zero No
Round to infinity No
IEEE754-2008 fused multiply-add No
Support is emulated in software No
Single-precision Floating-point support (core)
Denormals No
Infinity and NANs Yes
Round to nearest Yes
Round to zero No
Round to infinity No
IEEE754-2008 fused multiply-add No
Support is emulated in software No
Correctly-rounded divide and sqrt operations No
Double-precision Floating-point support (cl_khr_fp64)
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
Address bits 64, Little-Endian
Global memory size 6094360576 (5.676GiB)
Error Correction support No
Max memory allocation 2147483648 (2GiB)
Unified memory for Host and Device Yes
Minimum alignment for any data type 128 bytes
Alignment of base address 1024 bits (128 bytes)
Global Memory cache type Read/Write
Global Memory cache size 2097152 (2MiB)
Global Memory cache line size 64 bytes
Image support Yes
Max number of samplers per kernel 16
Max size for 1D images from buffer 134217728 pixels
Max 1D or 2D image array size 2048 images
Max 2D image size 8192x8192 pixels
Max 3D image size 2048x2048x2048 pixels
Max number of read image args 128
Max number of write image args 128
Local memory type Global
Local memory size 32768 (32KiB)
Max number of constant args 8
Max constant buffer size 32768 (32KiB)
Max size of kernel argument 1024
Queue properties
Out-of-order execution Yes
Profiling Yes
Prefer user sync for interop Yes
Profiling timer resolution 1ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels Yes
printf() buffer size 16777216 (16MiB)
Built-in kernels
Device Extensions cl_khr_byte_addressable_store cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_fp16 cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp64
Device Name NVIDIA Tegra X2
Device Vendor NVIDIA Corporation
Device Vendor ID 0x10de
Device Version OpenCL 1.2 PoCL HSTR: CUDA-sm_62
Driver Version 3.0-rc2
Device OpenCL C Version OpenCL C 1.2 PoCL
Device Type GPU
Device Topology (NV) PCI-E, 00:00.0
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 2
Max clock frequency 1020MHz
Compute Capability (NV) 6.2
Device Partition (core)
Max number of sub-devices 1
Supported partition types None
Max work item dimensions 3
Max work item sizes 1024x1024x64
Max work group size 1024
Preferred work group size multiple 32
Warp size (NV) 32
Preferred / native vector sizes
char 1 / 1
short 1 / 1
int 1 / 1
long 1 / 1
half 0 / 0 (n/a)
float 1 / 1
double 1 / 1 (cl_khr_fp64)
Half-precision Floating-point support (n/a)
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 (cl_khr_fp64)
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
Address bits 64, Little-Endian
Global memory size 8241844224 (7.676GiB)
Error Correction support No
Max memory allocation 2060461056 (1.919GiB)
Unified memory for Host and Device Yes
Integrated memory (NV) Yes
Minimum alignment for any data type 128 bytes
Alignment of base address 4096 bits (512 bytes)
Global Memory cache type None
Image support No
Local memory type Local
Local memory size 49152 (48KiB)
Registers per block (NV) 32768
Max number of constant args 8
Max constant buffer size 65536 (64KiB)
Max size of kernel argument 4352 (4.25KiB)
Queue properties
Out-of-order execution No
Profiling Yes
Prefer user sync for interop Yes
Profiling timer resolution 1ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
Kernel execution timeout (NV) No
Concurrent copy and kernel execution (NV) Yes
Number of async copy engines 1
printf() buffer size 16777216 (16MiB)
Built-in kernels
Device Extensions cl_khr_byte_addressable_store cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_nv_device_attribute_query
NULL platform behavior
clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...) Portable Computing Language
clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...) Success [POCL]
clCreateContext(NULL, ...) [default] Success [POCL]
clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT) Success (1)
Platform Name Portable Computing Language
Device Name pthread-cortex-a57
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU) Success (1)
Platform Name Portable Computing Language
Device Name pthread-cortex-a57
clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU) Success (1)
Platform Name Portable Computing Language
Device Name NVIDIA Tegra X2
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL) Success (2)
Platform Name Portable Computing Language
Device Name pthread-cortex-a57
Device Name NVIDIA Tegra X2
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 2.1
NOTE: your OpenCL library only supports OpenCL 2.1,
but some installed platforms support OpenCL 3.0.
Programs using 3.0 features may crash
or behave unexepectedly
Warning
If clinfo is returning Number of platforms 0, it means that the current
installation is not working :-(.
Info
On some boards, clinfo can show the following error message:
video group like
this:
Run clpeak Benchmark¶
cd ~/
mkdir workspace
cd workspace
git clone https://github.com/krrishnarraj/clpeak.git
cd clpeak
git submodule update --init --recursive --remote
mkdir build
cd build
cmake ..
make -j6
./clpeak
Sources¶
- https://github.com/pocl/pocl
- https://yunusmuhammad007.medium.com/build-and-install-opencl-on-jetson-nano-10bf4a7f0e65
- https://forums.developer.nvidia.com/t/can-the-xavier-run-opencl-applications/70262/7
- https://forums.developer.nvidia.com/t/jetpack-5-0-dp-nvrmmeminitnvmap-failed-with-permission-denied/219804
- https://github.com/krrishnarraj/clpeak