【Hacker News搬运】告诉HN:GpuOwl/PRPLL，用于查找最大素数的GPU软件

[hackernews]

Title: Tell HN: GpuOwl/PRPLL, GPU software used to find the largest prime number

告诉HN:GpuOwl/PRPLL，用于查找最大素数的GPU软件

Text: Hi, I'm Mihai Preda the author of GpuOwl/PRPLL [1], an OpenCL software used by Luke Durant for his recent discovery of the largest prime number know, the 52nd Mersenne prime 2^136279841 - 1 [2].<p>Feel free to ask questions about technical aspects of the GpuOwl implementation, about optimizations, tricks, efficient FFT implementation on GPUs etc. Or anything else.<p>[1] GpuOwl: <a href="https://github.com/preda/gpuowl">https://github.com/preda/gpuowl</a>

[2] GIMPS: <a href="https://www.mersenne.org/" rel="nofollow">https://www.mersenne.org/</a>

嗨，我；m Mihai Preda，《GpuOwl》的作者；PRPLL[1]，卢克·杜兰特最近发现了已知的最大素数，即第52个梅森素数2^136279841-1[2]，他使用的OpenCL软件<p> 欢迎就GpuOwl实现的技术方面、优化、技巧、GPU上的高效FFT实现等提出问题。或者别的什么<p> [1]GpuOwl:<a href=“https:&#x2F；&#x2F; github.com&#x2F-preda&#x2Gpowl”>https:&quot&#x2F；github.com；preda；GPU猫头鹰</a>[2] GIMPS:<a href=“https:”www.mersenne.org.“rel=”nofollow“>https:”&#x2F；www.mersenne.org</一

hn link

Url:

Post by: mpreda

Comments:

mpreda: Some topic ideas:<p><pre><code> - Why use OpenCL when implementing GPU software

• Does it run on AMD or on Nvidia GPUs?
• How does the primality test implemented in GpuOwl work?
• How fast is it to test a Mersenne candidate?
• Why use FFTs? how large are the FFTs?
• What do you use for sin/cos?</code></pre>

mpreda: 一些主题想法：<p><pre><code>-为什么在实现GPU软件时使用OpenCL-它是在AMD还是Nvidia GPU上运行？-GpuOwl中实现的素数测试是如何工作的？-测试梅森候选人的速度有多快？-为什么要使用FFT？FFT有多大？-你用什么来赎罪；因为</代码></pre>

motorolnik: Hi,
I've got few questions:<p>1). What profiling tools do you use for GPU code?<p>2). Where one would start, in terms of learning resources, about coding using inline GPU assembler?<p>3). Do you verify GPU assembler generated by a compiler from C/C++ code, in terms of effectiveness? If so, which tools do you use for that?<p>4). Is SIMD on GPUs a thing?<p>5). What are the primary factors being taken into account by you (cache sizes, microoptimizations, etc.) when you write code for a tool like gpuowl/prpll? Which factor is the most important?
Thanks!

motorolnik: 你好，我；我有几个问题：<p>1）。您对GPU代码使用哪些分析工具<p> 2）。在学习资源方面，从哪里开始使用内联GPU汇编器进行编码<p> 3）。您是否验证了编译器从C&#x2F生成的GPU汇编器；C++代码，在有效性方面？如果是这样，你使用哪些工具<p> 4）。GPU上的SIMD是一种东西吗<p> 5）。当您为gpuowl这样的工具编写代码时，您考虑的主要因素是什么（缓存大小、微优化等）；prpll？哪个因素最重要？谢谢！

dgacmu: First, congrats! Awesome work and appreciate you sharing more.<p>Second: I'm confused by something in your readme. It says:<p>> For Mersenne primes search, the PRP test is by far preferred over LL, such that LL is not used anymore for search.<p>But later notes that PRP is computationally nearly identical to LL. Was that sentence supposed to say TF and P-1 instead of PRP or am I misunderstanding something about the actual computational cost of PRP?

dgacmu: 首先，恭喜！很棒的工作，谢谢你分享更多<p> 第二：我；我对你自述中的某些内容感到困惑。上面写着：<p>&gt；对于梅森素数搜索，到目前为止，PRP测试比LL更受欢迎，因此LL不再用于搜索<p> 但后来注意到，PRP在计算上几乎与LL相同。这句话应该说TF和P-1而不是PRP，还是我误解了PRP的实际计算成本？

primecurious: I'd also like to draw attention that a lot of this work was sponsored by IMC the market maker, Mihai's employer.

primecurious: 我；d还想提请注意，这项工作的大部分是由做市商IMC Mihai赞助的；的雇主。

iyn: Wow, congrats!<p>Indeed, I’m curious why you’ve used OpenCL. And what was the hardware/general setup used for finding the prime?<p>What was your motivation behind building this software?

iyn: 哇，恭喜你<p> 事实上，我很好奇你为什么使用OpenCL。硬件是什么；用于查找素数的一般设置<p> 你开发这个软件的动机是什么？