永慶房產集團房產事業群總經理葉凌棋表示,影響第4季房市的因素有三,第一是美國即將縮減QE規模、第二是奢侈稅修法、第三是油、電、基本工資調漲。美QE縮減規模,台灣有雄厚的外匯存底可調節,影響房市的關鍵還是在利率,若利率仍低則無妨。

永慶房產集團房產事業群總經理葉凌棋表示,影響第4季房市的因素有三,第一是美國即將縮減QE規模、第二是奢侈稅修法、第三是油、電、基本工資調漲。美QE縮減規模,台灣有雄厚的外匯存底可調節,影響房市的關鍵還是在利率,若利率仍低則無妨。

繼續閲讀

Hyperscale Cloud Datacenters Redefined

Hyperscale Cloud Datacenters Redefined
Low-cost systems-on-chip (SoC), light-workload micro-servers, specialized controllers for “cold” storage, entry-level networking, optical-interconnected racks and with a software-defined infrastructure (SDI) are...
Overclocking, Asus x79-Deluxe Spotlighted at Intel Developer Forum
Overclocking—the process of pushing your computer components harder and faster than the manufacturer designed them to go—took center stage during the...
Master Template Functions in C++
Mastery of Threading Building Blocks requires a mastery of certain aspects of C++, such as template functions. In the first post...
Exascale Supercomputer Expected by 2018
An exscale supercomputer that’s 1,000 times faster than today’s petascale systems will become a reality within the decade, Intel fellow Shekhar...
Intel’s Parallel Programming Tools Updated
Intel’s suite of parallel software development tools has just been updated to add support for all Xeon and third- generation Core...
IDF 2013 Promises Chip News, HTML5 Tools, Parallel Computing
About 40 billion devices will be connected to the “Internet of Things” by 2020, Cisco executive Rob Soderbury predicted in a.... Video Tutorial: Testing Multicore Code on a High Performance AWS Server. Understanding Mutexes in Threading Building Blocks. Intel’s James Reinders Talks Xeon Phi, MICs and the Idea of Supercomputers on Chips. New Directions for Embedded. Perceptual Computing & The Future of User Interaction. Inside Intel® Parallel Studio XE 2013.See all Videos.
Case StudiesRADVISION* leverages Intel® Parallel Studio to dramatically accelerate the performance of IP-based applications on multicore platformsEvaluation GuidesImprove C++ Code Quality with Static Security Analysis (SSA)Tech DocumentsEstimating FLOPS using Event Based Sampling (EBS)TutorialsHow to manage the analysis overhead of Correctness Modeling in Intel® Parallel Advisor 2011VideosThree Things You must Teach: Module 3. Programming with OpenMP (pt.1)White PapersIntroduction to Multi-threaded Debugging Techniques.
Multicore Chips As ‘Mini-Internets’Intel To Ship Xeon Phi For “Exascale” Computing This YearIntel Ivy Bridge Processor Hits 7GHz Overclock RecordIntel Releases Ivy Bridge Programming Docs Under CC LicenseGeekbench Confirms Ivy Bridge MacBook Pro and iMacIntel Launches Z77 Motherboards, Preparing For Ivy Bridge.

Six-Step Path to Parallel Programming Share your comment!Share on linkedinShare on redditShare on emailMore Sharing ServicesIt’s no secret that learning parallel programming is no small task, but the payoff can be enormous. To help ease the move to parallel is a six-step plan for optimizing your code, both for small quad-core chips up to the massively-parallel Xeon Phi coprocessor. The plan is outlined in Jim Jeffers and James Reinders book on high-performance programming. Jeff Cogswell explores these steps and offers some additional suggestions.

Six-Step Path to Parallel Programming Share your comment!Share on linkedinShare on redditShare on emailMore Sharing ServicesIt’s no secret that learning parallel programming is no small task, but the payoff can be enormous. To help ease the move to parallel is a six-step plan for optimizing your code, both for small quad-core chips up to the massively-parallel Xeon Phi coprocessor. The plan is outlined in Jim Jeffers and James Reinders book on high-performance programming. Jeff Cogswell explores these steps and offers some additional suggestions.
While experimenting is one way to learn parallel programming, it’s best to make sure you first fully understand what is going on under the hood so that you can get the most out what you’re doing. Otherwise, you could end up with your code looking right, but running slower than serial code. Also, it’s always a good idea to consider the advice of the experts before trying to re-invent the wheel.

That’s where Intel’s Jim Jeffers and James Reinders come in. In their excellent book “Intel Xeon Phi Coprocessor High Performance Programming,” the authors provide an insider look at how to get the most out of parallel programming, for C++ as well as Fortran compiling. (And as I’ve mentioned in previous blogs, a lot of the information in the book applies not just to Xeon Phi Coprocessor programming, but general multi-core programming as well.)

In the book, Jeffers and Reinders provide a methodology for programming with vectorization. I’ve been covering vectorization and Cilk Plus programming lately in reference to scientific programming, particularly with linear algebra and matrix operations. Let’s review what these experts have to say about a proper vectorization methodology.

Six-Step Methodology

The methodology they present consists of six steps using Intel Parallel Studio. While I encourage you to get the book and read the actual text (starting on page 110), I want to offer my own additional thoughts and notes. Here are the steps, with my thoughts:

1. Measure baseline release build performance.

The authors point out the importance of using a release version of the build. This is vital, because the debug version introduces extra parameters that could ultimately decrease your performance and give you an inaccurate measurement of the final results, both before tuning and after tuning. But on the other hand, the release build also introduces extra parameters in the way of optimization. The optimization could potentially optimize out code that is a good opportunity for vectorization, as the authors point out. But that doesn’t mean you want to turn off optimization; in fact, you want it on. But you want to first find out what the baseline performance is for the release build before you add vectorization for a fair comparison to the final product.

2. Use Intel VTune Amplifier to locate hotspots

This is rather self-explanatory; use the tools for performance profiling. But note that the authors recommend using the full hotspot analysis, not the “Lightweight Hostspots” analysis.

3. Determine loop candidates

The authors suggest using the vectorization report from the compiler, and determine if there are loops that are hotspots that are not auto-vectorized. The loops didn’t auto-vectorize for whatever reason; and you can take this time to determine if they could be re-coded slightly to allow for vectorization. Remember, not everything is automatic, and there are times you might want to re-think an algorithm to look for opportunities to make it parallel. That brings us to the next step.

4. Use the Guided Auto-Parallelization (GAP)

This part is also self-explanatory; run the tool and see what it says, making note of its suggestions.

5. Follow the suggestions from the GAP

This is the main part that the previous steps were leading up to. The idea is that you may want to consider recoding in the parts that the tool suggests. But remember, the tool is only an automatic tool and doesn’t have the full analysis abilities of the human brain! In other words, the advice the tool gives you could be wrong in that you might recode the algorithm so it’s in parallel and faster, but doesn’t do exactly what it’s supposed to. Or, there might be a way to re-code it, but you simply code it wrong.

The authors bring up an interesting point, which I’ll quote here directly from the book, page 112:

“One way to ensure that the loop has no dependencies that may be affected is to consider if executing the loop in backwards order would change the results.”

The reason for this seemly bizarre but good advice is that parallel loops should be able to run independently of each other. For example, if each iteration of the loop relies on the results of the previous iteration, then the loop as defined can’t be run in parallel, and clearly running it in reverse would not work. But that’s not to say it can’t be done; perhaps you need to rethink the loop. Here’s a simple example: Suppose each iteration relies on a sum calculated in the previous iterations. In such cases you may be able to divide up the algorithm into blocks such that each block can build a partial sum, and then you finally combine the sums together. This is the concept of a reducer, and reducers can be coded to function in parallel. So don’t give up! And remember, there are large libraries at your disposal, including reducers in both Cilk Plus and the Threading Building Blocks thread library.

6. Rinse and Repeat

Then you repeat the steps from the start until you reach a point where you get the performance you’re looking for.

Conclusion

Parallel programming—both multicore and vectorized—is no small task, but the rewards can be huge, especially if you’re taking your code to a many-core chip like the Xeon Phi coprocessor. But even if you’re just sticking to a run-of-the-mill quad-core processor, you can still make big strides in maximizing your performance. Use the tools, and keep studying, keep learning, and soon you’ll be mastering parallel programming. (And seriously—get the Jeffers and Reinders book.)

..澳幣愈貶國人愈愛 愈買愈多

..澳幣愈貶國人愈愛 愈買愈多
.-字+字.作者: 記者黃惠聆╱台北報導 | 中時電子報 – 2013年9月2日 上午5:30.
.
...工商時報【記者黃惠聆╱台北報導】

澳幣今年來大貶值,不少投資人匯兌虧了不少,但澳幣愈貶,台灣投資人卻愈買愈多,根據投信投顧公會統計,今年前上半年澳幣計價債券基金金額成長了42%,已超越歐元計價債券基金,成為第二大外幣債券基金。

觀察目前國內持有澳幣資產變化,澳幣仍相當受到國人偏愛。投信投顧公會統計,6月國人持有澳幣計價的境外基金金額達1,710億元,已超越歐元計價的1,565億元,今年成長幅度更達42%,顯見資金湧向澳幣等高息資產趨勢明顯。

雖然有不少投資人澳幣資產因匯損讓資產縮水了1成,但也正因澳幣今年貶很多以及投資人認同澳幣還有再起之時,只要保險公司或者基金公司推出澳幣商品,銷售成績依然不錯,壽險銷售員表示,因為現在價格比年初價格至少便宜了1成以上,有的甚至打到85折,當然很具吸引力。

德銀遠東DWS澳幣7年期保本型基金經理人楊斯淵指出,澳洲出口至大陸占總出口的35%,大陸下半年起重拾經濟動能,提升原物料中長線需求成長;另外,澳洲為金屬礦出口大國,澳幣走勢與金屬礦價格走勢相關度高,近期鐵礦砂價格已出現上漲走勢,金屬預估價格長線走多,也會對澳幣帶來支撐。

摩根投信副總經理謝瑞妍也認為,短線澳幣走勢偏弱,但投資人不需急於殺出澳幣資產,反而應該著眼於長線升值趨勢,長期持有。

至於紐幣今年貶值幅度不如澳幣大,法人對紐幣後市也不悲觀。國泰投信基金經理人廖維苡表示,因為紐西蘭在十大工業國當中,利率水準相對較高,對紐幣匯率具有明顯的支撐效果。

題目 : 日常
部落格分类 : 網路部落

王者再臨 – 移動與桌面開發的真正王者

王者再臨 – 移動與桌面開發的真正王者



您還在為需要同時開發 Windows, Mac, iOS 和 Android 而焦頭爛額嗎? 您還在花費無數寶貴時間搜尋和評估跨平 台的快速開發工具嗎? 您還受挫於虛擬執行環境和解釋型語言緩慢的執行效率嗎? 那為什麼不來看看最新世代的Delphi 呢?

Delphi 從面市值到現在都被業界公認為是Windows 平台最佳的原生快速開發工具, ,而目前最新的 Delphi 更允許您開發原生的iOS/Android App. 很驚訝嗎? 藉由新一代的 LLVM 編譯器技術,Delphi 更能讓您藉由一份程式碼就能同時開發Windows, Mac, iOS 和Android 四個平台的App, 而且開發完成的App是四個平台真正原生的App, 擁有最高的執行效率.

如同Delphi 強調的RAD 精神, 全新的Delphi 將向您展示什麼
才是真正的快速Android/iOS App 開發. 新世代的移動設備, 多平台開發將由 Delphi 來領導和定義.

讓您一睹從 Windows,iOS 到 Android 一氣呵成且一碼到底的美夢成真!敬邀您免費參加!

研討會內容

歡迎和簡介
真正的多平台, 多設備,原生App開發工具
全新登場! Android快速App 開發工具
中場休息
一份程式碼同時開發移動和桌面應用程式
開發結合資料和服務的新世代Apps
誰適合參加:
專業程式設計師
專案團隊主管
軟體開發經理
技術開發部門主管

題目 : 公告文章
部落格分类 : 網路部落

自我介紹

v2oip,arith.info and yunoo.info and 房市分析.投資分析,房地產分析 賣book

Author:v2oip,arith.info and yunoo.info and 房市分析.投資分析,房地產分析 賣book
新年快樂














target="_blank" style="font-family: Geneva, Arial, Helvetica, sans-serif;
font-size: 9px; color: #908C86; text-decoration: underline ;">How Many People Visit


MANSION OF THE GLORY 5price



















最新文章
最新留言
最新引用
月份存檔
類別
搜尋欄
RSS連結
連結
加為好友

和此人成爲好友