Zhaoxin submitted a support patch for merging the "Lujiazui" CPU to the upstream of the GCC compiler again last week, because the previous submission was too late and missed the merge window of GCC 12.
In 2019, Zhaoxin, a joint venture between Taiwan's VIA (VIA) and the Shanghai municipal government, launched the ZX-E / KX-6000 series of x86_64 micro-architecture processors, code-named "Lujiazui", for use in laptops and servers. The KX-6000 is manufactured with a 16nm process and has a frequency of 3.0GHz. It is an eight-core x86-64 processor. Its performance is 50% higher than that of the previous generation KX-5000 series, reaching the level of the Intel i5-7400.
▲ Zhaoxin "Lujiazui" CPU
When the "Lujiazui" CPU was first released in 2019, the Linux kernel upstream added support for it, but the widely used GCC compiler has not correctly identified the Lujiazui processor, but incorrectly identified it as Intel's Core 2 or i386 CPUs. Zhaoxin has been trying to submit upstream detection and support for "Lujiazui" in the GCC compiler in recent months.
According to the GCC mailing list, Zhaoxin submitted a patch supporting "Lujiazui" to the upstream again last week.
This patch fixes the problem of Zhaoxin's CPU vendor ID detection, and adds support for Zhaoxin's "Lujiazui" processor. Currently gcc cannot identify Zhaoxin CPUs (vendor IDs "CentaurHauls" and "Shanghai"). Users get confused if they use the -march=native option. This patch enables -march=native and -march/-mtune=lujiazui in Zhaoxin family 7th processors, setting cost and tuning according to the processor's characteristics. We have added a new md file to describe the Lujiazui pipeline.