HPL 浮点数理论性能与实际性能相差6倍问题

HPL

为什么是6

file

dispatch

https://www.nextplatform.com/2021/03/26/deep-dive-into-amds-milan-epyc-7003-architecture/

https://www.nextplatform.com/wp-content/uploads/2021/03/amd-milan-epyc-zen2-versus-zen3.jpg

AMD Zen4 发布会:https://www.bilibili.com/video/BV1B84y1v7nW/?spm_id_from=333.337.search-card.all.click&vd_source=4871cfa497362c1a843af2ecff18ab7f

http://www.nextplatform.com/wp-content/uploads/2021/03/amd-milan-epyc-ipc-versus-intel-xeon-sp.jpg

超标量处理器指令发射的基本逻辑 | Sherlock's blog

3000页长文
AMD64 Architecture Programmer’s Manual, Volume 2: System Programming

FPU VP8 | Aida64