Intel releases CPU optimization code examples
Intel has released the latest CPU optimization reference manual and uploaded relevant sample codes on GitHub. Intel has always maintained a lengthy optimization reference manual to help developers understand how to optimize code for their latest CPU micro-architecture.
Recently, Intel not only updated the manual but also open-sourced many related code examples on GitHub to better help developers learn Intel optimization technology, so as to make full use of its latest processors. These code examples include Intel 64-bit and IA-32 architectures, mainly involving AVX/AVX2/FMA optimization, INT8 deep learning inference, and the use of AVX-512 to target the latest Intel CPUs (such as Xeon Scalable Ice Lake, Tiger Lake, and Rocket Lake) for best performance.
In addition, these code examples need to be run on a Haswell processor or higher, and no runtime checks are performed on the instructions in Haswell (such as FMA or AVX-2). Therefore, if you run some code samples on a device that does not support these instructions, a crash may occur. However, these code samples will check the Haswell instruction set before running, so it will skip some tests when the CPU does not support the relevant instructions. GCC 8.1 or higher is required to build unit tests.