Intel has released version 2.34.4 of the Intel Graphics Compiler — an update to the open shader and compute kernel compiler used in Intel's graphics stack for Linux and Windows.
The Intel Graphics Compiler (IGC) is one of the key components of Intel's GPU driver infrastructure and works in conjunction with OpenCL, Level Zero, and Vulkan drivers. The project is responsible for compiling shaders, SPIR-V, and compute code into machine instructions for Intel graphics processors.
Release 2.34.4 mainly focuses on bug fixes, code generation optimizations, and improving compatibility with the new Xe2 and Battlemage GPU architectures. The developers also noted:
- optimizations for SPIR-V compilation;
- fixes in SIMD handling;
- improvements in register allocation;
- updating LLVM dependencies;
- resolving issues that caused OpenCL applications to crash.
- support for SPV_INTEL_bfloat16_arithmetic for Xe3P;
- optimizations for Panther Lake;
- improvements in code scheduling;
- enabling 2D load splitting pass by default;
- support for WideMulMad;
- enhancements for Crescent Island.
Part of the changes aims to improve performance for compute workloads and gaming applications on Linux, particularly in conjunction with Mesa ANV and Compute Runtime.
The project is developed as open source and is distributed under the MIT license. The source code is published on GitHub, and new versions are usually integrated quickly into Compute Runtime and the oneAPI stack.
The source texts and changelog are available on the Intel Graphics Compiler GitHub.
Source: linux.org.ru
