Skip to main content

theory-driven-pipeline-optimization

当需要为计算密集型算法建立理论性能上限时使用:先识别核心指令和依赖链,编写微基准脚本测量这些指令的时延、吞吐和混合执行成本,再根据算法的指令计数、依赖图和数据移动推算每个处理单位需要多少 cycles,得到理论 cycles 下界和吞吐上限。适用于 C/C++、汇编、SIMD、加密、压缩、字符串处理、数学库等 kernel 的理论性能建模、headroom 判断和优化方向选择。

Jump to install

Source facts

Repository
kunpengcompute/skills
Last source activity
June 12, 2026 at 02:43
Detected SKILL.md language
Chinese
Stars
0
Forks
0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.