[WIP] Draft benchmark roadmap for Apache Mahout QDP

[Copilot]

Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date as I form a plan and make progress.

Original prompt

please help draft a benchmark roadmap(RFC) for Apache Mahout QDP(in qdp/), it should focus on 1. fairness (implement warmup, and others tbd)
2. create nice graph for blogposts or academic papers
you should evaluate existing benchmarks, and think about it, there's a person's suggestions:要注意的應該就是 clear cache 和 event + synchronize 的設置，沒有 gradient 的問題就可以忽略下面實作中的那部分
可以用 torch.cuda 的 event 來記錄時間點，最好多跑幾次(記得 clear cache)算出很多個時間，然後算出統計值 (mean, std, err, percentiles, max, median, min 等)，有那些統計值 matplotlib 就很好畫出來了
benchmark 架構大概都長得差不多
https://github.com/pytorch/helion/blob/main/helion/autotuner/benchmarking.py
https://github.com/triton-lang/triton/blob/2c3afac48287af8ea5c270722124da9ec1cba2a1/python/triton/testing.py#L127
統計值
https://github.com/triton-lang/triton/blob/2c3afac48287af8ea5c270722124da9ec1cba2a1/python/triton/testing.py#L42
https://github.com/pytorch/helion/blob/main/helion/autotuner/base_search.py#L974
https://github.com/gpu-mode/reference-kernels/blob/main/problems/nvidia/eval_better_bench.py#L109

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.