355個大學正在傳授CUDA高速平行運算 -- 找出最近的地點進行學習

355個大學正在傳授CUDA高速平行運算 -- 找出最近的地點進行學習

CUDA and GPU Computing Courses

355 universities are teaching CUDA and GPU Computing courses.

See the list: http://www.nvidia.com/object/cuda_courses_and_map.html

CUDA University Courses

The NVIDIA CUDA™ technology is being taught at universities throughout the world, helping the development of innovative solutions to some of the most complex computation-intensive challenges.
We encourage academic institutions to apply to our professor partnership program. Click here for more information.

355 Universities Teaching the CUDA Parallel Programming Model

Are you teaching CUDA now?
If you are teaching a class in CUDA submit your information and link to the coursework by email.

Do you want to teach a CUDA class?
Apply to our professor partnership program here.

What do we recommend as CUDA capable GPUs for students?
See the CUDA-enabled Products here.

StreamBase – 用CUDA把華爾街提升到另一個境界

StreamBase – 用CUDA把華爾街提升到另一個境界

StreamBase – Taking Wall Street to the Next Level with CUDA

StreamBase系統公司本週在紐約宣佈, 他們正在與NVIDIA合作進行財務領域”最優秀品種”(best-of-breed)的計算方式.

當交易策略需要改變時, 進行計量與交易的財務公司們, 會需要將新策略套用於大量的歷史交易紀錄, 以驗證新的交易策略能否奏效.

運用NVIDIA Tesla GPU高速運算圖形處理器, 加上StreamBase公司”最優秀品種”(best-of-breed)的計算方式, 可以讓財務公司在瞬間部署有效的交易策略.

StreamBase Systems announced this week in New York that it is working with NVIDIA to offer best-of-breed computing capability for the capital markets. Firms whose traders and quants want to rapidly test out new trading strategies against large amounts of historical data can now combine StreamBase technology with NVIDIA Tesla GPUs to accelerate time to deployment.

StreamBase的技術長, Richard Tibbetts, 說道: “先進技術(CUDA)將華爾街提升到另一個境界”, “財務運算與NVIDIA GPU高速運算圖形處理器双劍合璧, 為財務產業帶來了革命”.
"Advanced technologies are taking Wall Street to the next level," said Richard Tibbetts, CTO at StreamBase. "This partnership underscores the technical revolution in the financial industry."

美國著名諮詢公司TABB集團的資深分析師, Kevin McPartland, 說道: “當你在百萬分之一秒內做出的交易決策具有成與敗如此巨大的差異時, 快那麼一點點就顯得極為重要.” “我們不只看到GPU高速圖形運算處理器在選擇權領域發光發熱(Davis註: 我已經在六月翻譯過幾篇衍生性金融商品用GPU獲得50x~300x倍快的文獻), 現在我們還看到GPU高速圖形運算處理器在多重資產分析上的最新應用”.
"When you're talking about trading where microseconds make a difference, the extra hops within the server can make a difference," said Kevin McPartland, Senior Analyst at TABB group. "We are seeing renewed interest in GPU technology not only for options, but also for multi-asset analytics."

[Davis延伸閱讀]:

百萬分之一秒的勝負

2010年五月六日, 紐約股市大跌千點, Accenture (NYSE:ACN)股價跌到一美分(5/7收盤價$41.09), 如果當時花一千美元買了十萬股, 現在已經市值$4109000(台幣一億三千一百九十萬), 這與中彩劵無異!

雖然SEC與NASDAQ都決定取消這20分鐘的異常交易, 不過對於周邊的交易, 像是指數型期貨與選擇權交易則不受影響, 因此承做這些週邊交易的人真的會發大財呢.

Referenced blog: http://blog.pchome.com.tw/prudencio/post/1320938021

GPU高速財務模擬既然能在百萬分之一秒改變策略, 就能把握與這20分鐘異常交易類似的投資機會 – 當然GPU高速運算並非只是用在這個方面的啦.

醫療影像判斷加快 30X倍速– 用CUDA高速運算以相基礎(phase based)為根本的影像體積套合(Volume Registration)技術

醫療影像判斷加快 30X倍速– 用CUDA高速運算以相基礎(phase based)為根本的影像體積套合(Volume Registration)技術

Phase Based Volume Registration Using CUDA – 30x 倍速

在醫療影像判斷上, 我們剛剛用GPU完成了最新的, 以相基礎(phase based)為根本的影像體積套合(Volume Registration)技術 – 30x倍速快.

對比於其他已經用CUDA GPU高速運算基於影像強度(Image Intensity)的影像套合技術, 我們剛剛完成的新運算法則, 能讓CUDA GPU高速運算更上一層樓, 更加可靠的判斷出具有多重影像強度的同體積物質.

這個新技術剛剛在2010年達拉斯州(Dallas)的IEEE ICASSP研討會上面發表.

We have implemented phase based volume registration using CUDA, in contrast to all other GPU based image registration implementations that are based on the image intensity. Our registration algorithm is more robust for volumes that differ significantly in intensity. This work was presented at the IEEE conference ICASSP in Dallas 2010.

[Davis延伸閱讀]:

· 醫療影像判斷, 尤其是腫瘤判斷, 需要很多的學問.

· 無論是以影像強度(Image Intensity)判斷, 以level-Set, 或是以相基礎(Phase Based)判斷, 都是為了精準標示身體組織的尺寸與估計其組成材質.

· 一般超音波掃描器在身體內約每10秒創建一張2D超音波成像, 醫師再將超音波掃描器轉向不同角度成像, 交由電腦將連續2D影像裁切為3D立體影像.

· 醫療上的2D-to-3D成像, 除了能取得的2D影像張數非常有限(這已經夠困難的), 以及其他可以想像的困難之外, 還有非常多的因素要考慮.

· 比方說, 超音波探頭在掃描的過程會壓迫腫瘤, 使腫瘤組織形變. 而良性腫瘤受壓迫時的形變比惡性腫瘤來的大, 這是計算上要考量的地方.

· 另外, 超音波探頭在掃描的過程也會使腫瘤位移, 而良性腫瘤因為組織較軟, 受壓迫時的位移比惡性腫瘤來的大, 這是計算上另一個要考量的地方.

· 無論採用哪些技術, 醫療影像判斷都是件耗時費力的事, 而且還有病患療程時間的迫切性, 既能判斷並3D繪製病理組織的大小, 還能估計病理組織屬於良性或惡性, 與獲得可能的其餘資訊.

· 30x倍速的運算, 同時也表示相同運算時間下具有30x倍的準確度.

· NVIDIA CUDA GPU高速運算, 肯定是現在人類醫療領域的一盞名燈.

地震解答者Seismic Solvers – 200x倍速 加速石油天然氣探勘

(CUDA軟體) 地震解答者Seismic Solvers – 200x 倍速 加速石油天然氣探勘之多皺折岩層震波地質研究

Acceleware是高速震波地質模擬與研究領域的領導廠商. Acceleware的震波地質學專家群, 使用NVIDIA的GPU高速平行運算來加速Acceleware的震波地質研究軟體, NVIDIA GPU大幅減少多皺折岩層地質模擬傳統CPU所需要的運算時間達200x倍速.

Acceleware is leading the market in providing acceleration solutions for seismic data processing and reservoir simulation. By combining our core knowledge in parallelization with an in-house team of seismic industry experts, Acceleware provides software solutions for seismic data processors, which access the massively parallel capabilities of compute GPUs. The Acceleware seismic processing solutions provide multi-fold performance increases to reduce lengthy processing times and deliver faster business decisions for the seismic industry. By harnessing the parallel processing power of GPU accelerators to dramatically increase the computation power of data centers, seismic jobs are processed faster and with a reduced total cost of IT ownership.

以GPU高速平行運算為核心的雲端運算服務中心 – Hoopoe (快來Hoopoe挖寶吧)

以GPU高速平行運算為核心的雲端運算服務中心 – Hoopoe (快來Hoopoe挖寶吧)

Hoopoe – Cloud Services for GPU Computing

http://www.hoopoe-cloud.com/About.aspx

什麼是Hoopoe?

What is Hoopoe?
Hoopoe是一個致力於提供雲端運算服務的計畫, 其與眾不同的地方是, Hoopoe是以GPU高速平行運算為核心的雲端運算服務計畫, 使用許多最新的技術與工具.

Hoopoe提供一個使用GPU圖形處理器硬體高速運算的, 支持任何model的通用環境.

它讓使用者可以指定運算過程中想要使用的硬體, 或者很有彈性的讓系統自我選擇並非忙碌中的資源, 讓運算以最快的速度完成.

Hoopoe is a project that aims to provide cloud computing services, the same as in general computing, but for GPUs, using various technologies and tools.
It provides a general environment for running computations, of any kind and model, on GPU hardware.
Users may specify the type of hardware they wish to use during the computation or the system will use the available resources to complete the computation as fast as possible.

Hoopoe計畫名稱的由來

Origin of the name
Hoopoe是一種鳥類的名稱, 也是以色列建國60週年國鳥. 在希伯來文中, 這個名字念起來像是”Dukhifat”, 我們可以在許多希伯來文的文獻與聖經中看到Hoopoe的蹤影. 以Hoopoe鳥為計畫名稱, 約略反映出如”鳥”般”靈巧”的雲端計算.

Hoopoe is a bird, which was chosen to officially represent Israel at the 60's birthday celebrations. In Hebrew, the name is pronounced: "Dukhifat" and found in many literature sources, including the bible itself.
As being a bird, it reflects very well the idea of using services in a "cloud" oriented model.

Hoopoe如何能帶來好處?

Why is it good?
從資訊管理人員的角度來看, 建立並維持一個GPU圖形處理器的運算叢集不是一件簡單的事. 尤其是新的, 更快速的GPU圖形處理器不斷推陳出新的今日, 我們很難取捨於, 該繼續維持使用現有的設備, 還是大刀闊斧的更新為新一代的硬體?

Hoopoe可以為資訊管理人員解決以上的問題.

短期間使用或開發GPU高速平行運算的人有福了. 既然Hoopoe提供的是雲端運算服務, 那就表示沒有大筆的設備投入與維持的開銷, 取而代之的是, “用多少付多少”的費用計算. 當你暫時不使用GPU高速運算時, 你甚至不用花一毛錢.

Hoopoe讓需要圖形運算能力的團體, 於彈指之間獲得數千顆GPU圖形處理器的高速運算雲端服務, 而不需要冗長的硬體設施規劃/驗證/組裝/維護的時間與金錢成本.

Hoopoe不是簡單的雲端服務, 倚靠即時分散運算引擎的Hoopoe, 能迅速統整百萬顆GPU圖形的高速平行運算能力, 沒有絲毫損耗.

(註: 傳統CPU是序列運算架構, 當它要進行叢集多核運算時, 所耗損的運算能力太多, 所以CPU沒有競爭力.)

Maintaining a GPU cluster is not always simple, from IT aspects to general maintenance services. The same problem comes to play when a new generation of hardware is available. Should we leave the current investment and buy new hardware?

Hoopoe can answer most of these questions and problems.
As a cloud solution, it hides the maintenance costs from organizations, providing them with a Pay-Per-Use solution, if you don't use it, you don't pay for it.
It allows organizations get the power of thounsands of GPUs with a single click, without taking care of operational costs, general maintenance issues and distribution problems.

Because Hoopoe is not just a cloud service, it relies on a real-time distributing engine that can harness the power of millions of GPUs with no performance loss.
The distribution model used with Hoopoe, can help every organization fit the requested computation inside, and enjoy the flexibility of the distributed environment without spending too much effort on the distribution.

Hoopoe提供的服務

Provided services
Hoopoe提供以下兩種服務介面:

(1) 網頁服務的API(應用程序接口) – 允許各種技術與平台透過Hoopoe進行無阻礙的聯繫

(2) 網頁為基礎的GUI(圖形使用者介面) – 以圖形使用者介面提供與(1)相同功能的服務.

As part of Hoopoe, users are provided with two interfaces to communicate with the service:
(1) Web Service based API - To allow every technology and platform communicate with Hoopoe transparently
(2) Web based GUI - Which provides the same set of features for users as the Web Service API
運用以上兩種介面, 使用者可以:

* 管理Hoopoe上的工作 – 提出/取消/暫停/繼續/.. GPU高速平行運算工作.

* 監控已提出的工作排程, 包括排程狀態, 以及各種不同的資訊.

* 管理工作所需要的檔案與權限.
With these interfaces, users can:
* Manage tasks in Hoopoe - Submit, cancel, pause resume and more
* Monitor submitted tasks, including status information and various statistics
* Manage files to be used by tasks
The interfaces provided by Hoopoe can assist both automatic systems and individuals communicate with the system through the internet.

註: 以後若有機會, 我將繼續為各位介紹Hoopoe的各種GPU高速運算服務方案細節:

(1). CUDA.NET http://www.hoopoe-cloud.com/Solutions/CUDA.NET/Default.aspx è 已被下載逾兩萬次的CUDA.NET免費服務.

(2). OpenCL.NET http://www.hoopoe-cloud.com/Solutions/OpenCL.NET/Default.aspx è 與CUDA.NET類似, 只是改用OpenCL為基礎的, GPU高速運算.NET服務.

(3). Encodeo™ http://www.hoopoe-cloud.com/Apps/Encodeo.aspx è 轉檔在雲端? x10倍快的雲端影片轉檔服務.

(4). Vicodeo™ http://www.hoopoe-cloud.com/Solutions/VideoDecoding/Default.aspx è GPU加速的雲端影片解碼技術 (廣泛支持.NET, Java, Python各種語法).

(5). jCUDA http://www.hoopoe-cloud.com/Solutions/jCUDA/Default.aspx è 免費提供JAVA下的CUDA服務.

(6). Fortran CUDA http://www.hoopoe-cloud.com/Solutions/Fortran/Default.aspx è 免費提供Fortran下的CUDA服務.

(7). CUFFT Apps http://www.hoopoe-cloud.com/Apps/Default.aspx è 透過NVIDIA官方CUFFT Library的支持, Hoopoe持續延展並提供支持的應用程式.

看到很多寶貴的GPU高速運算資源, Hoopoe竟然都有提供免費服務嗎?

趕緊來Hoopoe挖寶吧…