20.22.16952
·
16619 commits
to master
since this release
Packages included in the release
- intel-gmmlib (https://github.com/intel/gmmlib)
- intel-igc-core and intel-igc-opencl (https://github.com/intel/intel-graphics-compiler)
- intel-opencl, intel-ocloc, intel-level-zero-gpu (https://github.com/intel/compute-runtime)
Components revisions included in the release
Additional components revisions used in build
- intel/libva@c9bb65b (Compatible with va_api_major_version = 1)
- intel/llvm-patches@595c1e3
- intel/[email protected]
- KhronosGroup/SPIRV-LLVM-Translator@424e375
- llvm/[email protected]
- oneapi-src/[email protected] (Compatible with oneAPI Level Zero API Spec v0.91)
Installation procedure on Ubuntu 18.04
- Create temporary directory
Example:
mkdir neo
- Download all *.deb packages
Example:
cd neo
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-gmmlib_20.1.1_amd64.deb
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-igc-core_1.0.4062_amd64.deb
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-igc-opencl_1.0.4062_amd64.deb
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-opencl_20.22.16952_amd64.deb
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-ocloc_20.22.16952_amd64.deb
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/intel-level-zero-gpu_0.8.16952_amd64.deb
- Verify sha256 sums for packages
Example:
wget https://github.com/intel/compute-runtime/releases/download/20.22.16952/ww22.sum
sha256sum -c ww22.sum
- Install all packages as root
Example:
sudo dpkg -i *.deb
sha256 sums for packages
134e2199bb93b3fb38414b6f1d265bfd331fdec967382fd8f7e9d2e87838f3ad intel-gmmlib_20.1.1_amd64.deb
83fe23d2715be3161b70ed4bbeb2868c0cfd03312b278c6b0d87c52783767ed4 intel-igc-core_1.0.4062_amd64.deb
66cd371eca0797111eacda3cb44eb2e19265d3c7b9692bd8e5d8dab0f234c466 intel-igc-opencl_1.0.4062_amd64.deb
de4bd81f2eb2b8b264edce4938ab558b5501c037f9a78da9e5480ffe55d7bde8 intel-level-zero-gpu_0.8.16952_amd64.deb
56633882e93a8a4e7fc12d79c7eefa7d83aef5518b044e2e52c2bf33dee36a7f intel-ocloc_20.22.16952_amd64.deb
3af27dc996518436b3d4fa0938b54beabc40d33384b40549cc57370ffb0b004b intel-opencl_20.22.16952_amd64.deb
Other installation procedures are described here.
Quality expectations
| Platform | OpenCL | Level Zero |
|---|---|---|
| Broadwell | 2.1 - Production | - |
| Skylake | 2.1 - Production | 0.8 - Pre-Release |
| Kaby Lake | 2.1 - Production | 0.8 - Pre-Release |
| Coffee Lake | 2.1 - Production | 0.8 - Pre-Release |
| Ice Lake* | 2.1 - Production | 0.8 - Pre-Release |
| Tiger Lake* | 3.0* - Early support | 0.8 - Early support |
| Apollo Lake | 1.2 - Production | - |
| Gemini Lake | 1.2 - Production | - |
Unless noted otherwise, the driver was validated on Ubuntu 18.04 LTS with stock kernel.
Moving to Ubuntu 20.04 for testing soon.
(*) Product is based on a published Khronos specification and is expected to pass the Khronos Conformance Process when available. Current conformance status can be found at https://www.khronos.org/conformance.
(*) For ICL and TGL platforms kernel from drm-tip repository was used:
- 5.6.0-rc7 commit 257af0ddcea3a234dcb79579600f971edd47353f 2020-03-23_15-07-40 drm-tip: 2020y-03m-23d-15h-06m-46s UTC integration manifest
Quality levels
- Experimental - no quality expectations
- Early support - platform may not be available on the market yet
- Pre-Release - suitable for integration and testing, gathering customer feedback
- Beta - suitable for integration and broad testing
- Production - Beta + meets API-specific conformance requirements; suitable for production
Issues & regressions of note
- For ICL platform i915.alpha_support=1 kernel parameter is required for kernel before 5.2
Important changes
- Updated IGC
- Enabled OpenCL 3.0 support on TGLLP
- Added support for new device info queries
- Added new clGetDeviceInfo queries
- Added new API properties queries
- Added new function setFilterMode
- Disabled Device Enqueue and Pipes support on TGLLP