The generation Alder Lake of Intel processors introduced a radical change that did not sit well with Linux. For those who are lost, Alder Lake released a hybrid architecture that has mixed high and low performance cores in the same package. The chip giant focused its work on Windows and left Linux as a second course, which made Microsoft’s system clearly superior in initial comparisons over Alder Lake.
The news, obviously, did not go unnoticed by Linux kernel developers, who got to work to at least close the gap. The first blow came with Linux 5.16, which included a patch that greatly improved performance through more intelligent use of high-performance kernels. The situation improved a little more in Linux 5.18, although without completely surpassing Windows, with the introduction of features and support for technologies included specifically for Alder Lake.
Many months after the release of Alder Lake processors, Intel continues to contribute to Linux to improve performance. Ricardo Neri, engineer of Intel, dedicated to Linux, has released a new patch that allegedly improves task scheduling between P (high-performance) and E (low-performance) cores..
Neri explained that “Intel processors that support Intel Turbo Boost Max 3.0 use ‘asym_packing’ to assign higher priorities to higher clocked CPUs. However, he artificially assigns a lower priority to higher numbered SMT siblings to ensure they are used last. This results in unnecessary task migrations within SMT domains.”
“On processors with a mix of higher-frequency SMT cores (the P cores) and lower-frequency non-SMT cores (the E cores), a lower-priority CPU offloads tasks from the higher-priority cores if there is more than one SMT brother busy.”
“Don’t use different priorities for each SMT sibling. Instead, modify the ‘asym_packing’ load balancer to recognize SMT cores with more than one busy sibling and allow lower priority CPUs to perform tasks. Removing these artificial priorities prevents unnecessary migrations and allows lower priority kernels to inspect all SMT siblings for the busiest queue.”.
Despite his status as an engineer, Neri’s explanation is simple enough to understand that the purpose of the patch is to make smarter use of low and high performance kernels when handling tasks, making low-performing cores get more share when a high-performing one is detected to be busy. The expected consequence of applying the patch is that Linux will be able to take better advantage of Alder Lake processors and possibly future Raptor Lake processors as well.
That Intel contributes directly to Linux is not news, since it has been one of the top contributors to the kernel for a very long time. In other words, Intel has a lot of weight in Linux development, although the relationship with the hard core of the kernel developers has not always been idyllic.
-
-
-
-
-
-
-
-
