Tag Archives: software development

Is the triumph of the GPU the failure of the CPU?

I’m at NVIDIA’s GPU tech conference in San Jose. The central theme of the conference is that the capabilities of modern GPUs enable substantial performance gains for general computing, not just for graphics, though most of the examples we have seen involve some element of graphical processing. The reason you should care about this is that the gains are huge.

Take Matlab for example, a popular language and IDE for algorithm development, data analysis and mathematical computation. We were told in the keynote here yesterday that Matlab is offering a parallel computing toolkit based on NVIDIA’s CUDA, with speed-ups from 10 to 40 times. Dramatic performance improvements opens up new possibilities in computing.

Why has GPU performance advanced so rapidly, whereas CPU performance has levelled off? The reason is that they use different computing models. CPUs are general-purpose. The focus is on fast serial computation, executing a single thread as rapidly as possible. Since many applications are largely single-thread, this is what we need, but there are technical barriers to increasing clock speed. Of course multi-core and multi-processor systems are now standard, so we have dual-core or quad-core machines, with big performance gains for multi-threaded applications.

By contrast, GPUs are designed to be massively parallel. A Tesla C1060 has not 2 or 4 or 8 cores, but 240; the C2050 has 448. These are not the same as CPU cores, but nevertheless do execute in parallel. The clock speed is only 1.3Ghz, whereas an Intel Core i7 Extreme is 3.3Ghz, but the Intel CPU has a mere 6 cores.  An Intel Xeon 7560 runs at 2.266 Ghz and has 8 cores.The lower clock speed in the GPU is one reason it is more power-efficient.

NVIDIA’s CUDA initiative is about making this capability available to any application. NVIDIA made changes to its hardware to make it more amenable to standard C code, and delivered CUDA C with extensions to support it. In essence it is pretty simple. The extensions let you specify functions to execute on the GPU, allocate memory for pointers on the GPU, and copy memory between the GPU (called the device) and the main memory on the PC (called the host). You can also synchronize threads and use shared memory between threads.

The reward is great performance, but there are several disadvantages. One is the challenge of concurrent programming and the subtle bugs it can introduce.

Another is the hassle of copying memory between host and device. The device is in effect a computer within a computer. Shifting data between the two is relatively show.

A third is that CUDA is proprietary to NVIDIA. If you want your code to work with ATI’s equivalent, called Streams, then you should use the OpenCL library, though I’ve noticed that most people here seem to use CUDA; I presume they are able to specify the hardware and would rather avoid the compromises of a cross-GPU library. In the worst case, if you need to support both CUDA and non-CUDA systems, you might need to support different code paths depending on what is detected at runtime.

It is all a bit messy, though there are tools and libraries to simplify the task. For example, this morning we heard about GMAC, which makes host and device appear to use a single address space, though I imagine there are performance implications.

NVIDIA says it is democratizing supercomputing, bringing high performance computing within reach for almost anyone. There is something in that; but at the same time as a developer I would rather not think about whether my code will execute on the CPU or the GPU. Viewed at the highest level, I find it disappointing that to get great performance I need to bolster the capabilities of the CPU with a specialist add-on. The triumph of the GPU is in a sense the failure of the CPU. Convergence in some form or other strikes me as inevitable.

NVIDIA talks up GPU computing, presents roadmap

At the NVIDIA GPU Technology Conference in San Jose CEO Jen-Hsun Huang talked up the company’s progress in GPU computing, showed some example applications, and announced a high-level roadmap for future graphics chip architectures. NVIDIA has three areas of focus, he said: the Quadro line for visualisation, Tesla for parallel computing, and GeForce/Tegra for personal computing. Tegra is a system on a chip aimed at mobile devices. Mobile, says Huang, is “a completely disruptive force to all of computing.”

NVIDIA’s current chip architecture is called Fermi. The company is settling on a two-year product cycle and will deliver Kepler in 2011 with 3 to 4 times the performance (expressed as Gigaflops per watt) of Fermi. Maxwell in 2013 will have around 12 times the performance of Fermi. In between these architecture changes, NVIDIA will do “kicker” updates to refresh its products, with one for Fermi due soon.

The focus of the conference though is not on super-fast graphics cards in themselves, but rather on using the GPU for general purpose computing. GPUs are very, very good at doing mathematics fast and in parallel. If you have an application that does intensive calculations, then executing that part of the code on the GPU can offer impressive performance increases. NVIDIA’s CUDA library for C lets you do exactly that. Another option is OpenCL, a standard that works across GPUs from multiple vendors.

Adobe uses CUDA for the Mercury Playback engine in Creative Suite 5, greatly improving performance in After Effects, Premiere Pro and Photoshop, but with the annoyance that you have to use a compatible NVIDIA graphics card.

The performance gain from GPU programming is so great that it is unavoidable for applications in relevant areas, such as simulation or statistical analysis. Huang gave a compelling example during the keynote, bringing heart surgeon Dr Michael Black on stage to talk about his work. Operating on a beating heart is difficult because it presents a moving target. By combining robotic surgery with software that is able to predict the heart’s movement through simulation, he is researching how to operate on a heart almost as if it were stopped and with just a small incision.

Programming the GPU is compelling, but difficult. NVIDIA is keen to see it become part of mainstream programming, for obvious reasons, and there are new libraries and tools which help with this, like Parallel Nsight for Visual Studio 2010. Another interesting development, announced today, is CUDA for x86, being developed by PGI, which will let your CUDA code run even when an NVIDIA GPU is not present. Even if the performance gains are limited, it will mean developers who need to support diverse systems can run the same code, rather than having a different code path when no CUDA GPU is detected.

That said, GPU programming still has all the challenges of concurrent development, prone to race conditions and synchronization problems.

Stuffing a server full of GPUs is a cost-effective route to super-computing. I took a brief look at the exhibition, which includes this Colfax CXT8000 with 8 Tesla GPUs; it also has three 1200W power supplies. It may cost $25,000 but if you look at the performance you are getting for the price, machines like this are great value.

image

Delphi XE includes licenses for older versions back to Delphi 7

I’ve just picked up that Delphi XE, the latest RAD Windows development suite from Embarcadero, includes licenses for older versions going back to Delphi 7.

There’s an explanation and list of what’s on offer here. Delphi 7 was the last version to use the old fully native code IDE and is delightfully fast and lightweight by today’s standards. Delphi 2007 was the last version before big Unicode changes in Delphi 2009, which often broke code, so could be useful for older projects.

The FAQ includes a few points of interest. Embarcadero is dismissive of the old Delphi for .NET (before Prism) and will not supply it:

That is an old technology that was replaced by Delphi Prism and we don’t want to encourage use of that old product.

If you have purchased XE and want to take advantage of the offer, you must do so within 180 days.

Apple lifts restrictions on app development tools, publishes review guidelines

Apple has lifted its restrictions on the development tools used to create iOS (iPhone and iPad) apps, in a statement published today:

We have listened to our developers and taken much of their feedback to heart. Based on their input, today we are making some important changes to our iOS Developer Program license in sections 3.3.1, 3.3.2 and 3.3.9 to relax some restrictions we put in place earlier this year. In particular, we are relaxing all restrictions on the development tools used to create iOS apps, as long as the resulting apps do not download any code. This should give developers the flexibility they want, while preserving the security we need.

In addition, Apple says it is publishing the App Store Review Guidelines in the hope that this will make the approval process more transparent.

Good news I guess; but why? Maybe in part because the restrictions made little sense and were possibly unenforceable; and in part because Android’s popularity is putting pressure on Apple to be more developer-friendly. In practice, some apps that you would have thought breached the requirements apparently made it through the approval process; and those publishers of cross-platform tools which kept their nerve have their patience rewarded.

But what about Adobe? Apple’s development restrictions seems to trigger a significant change of direction, with work on the Packager for iPhone stopped, Android devices issued to employees, and evangelism for Android in Adobe blogs and tweets.

Since Apple is not changing its mind about runtimes, but only about development tools, this change of mind does not enable Flash on the iPhone; but I guess Adobe could now revive its cross-compilation work. On the other hand, Apple’s pronouncements have caused disruption for Adobe and perhaps served more as a wake-up call: this is a closed platform with one owner and therefore a risky target for investment.

Decompiling Silverlight

A Silverlight application is a .NET application. Most developers will be aware of this; but it is worth noting that whereas ASP.NET code executes on the server and is not normally available for download, Silverlight code is downloaded to the client and can easily be decompiled. It is almost as easy to view as JavaScript code in the browser.

If you want to investigate this, the first thing to do is to find the .xap file which contains the Silverlight application. You will likely find this in your browser cache, or you can download it directly from the web site hosting the application. If you have out-of-browser Silverlight apps, they are usually located at:

C:\Users\[username]\AppData\LocalLow\Microsoft\Silverlight\OutOfBrowser

Copy the .xap file somewhere convenient, and rename it to have a .zip extension. Then extract the files. The result looks something like this:

image

Next, you need a .NET decompiler such as Redgate .NET Reflector. Run Reflector and open a .dll file containing application code. Select a method, and Reflector does its best to show you the code. It does a good job too:

image

The purpose of this post is not to encourage decompiling other people’s code, but rather to make the point that even though Silverlight code is “compiled”, it is trivial to read it – just in case anyone thought it was a bright idea to store passwords or other authentication secrets there.

The solution is to never to put anything security-critical in client-side code. Second, you can use an obsfuscator such as dotfuscator to make the decompiled code harder to read.

Getting started with Android – a few hassles

I’m setting myself up for Android development, and my HTC Desire arrived today. It is a lot of fun, though I have had a few hassles, partly because my device is locked to Orange and of course I wanted to upgrade to Android 2.2 “Froyo” right away. There are also a few issues with the Orange branding on a Desire, including the absence of Google Talk and applications like Orange Maps that you probably will not use. I went through a series of steps to unbrand the phone and install Android 2.2.

It is an odd situation, where mobile operators work hard at adding their branding and applications to the device, only for canny users to work out how to remove them. Branding also delays updates. If Android is updated, the operator has to receive the updated operating system image from the phone manufacturer, HTC in this case, add its customizations, and then apparently sent it back to HTC for packaging.

My guess is that if an operator decided to ship the standard Android image it would be a selling point and attract more customers. I even have a catchy name to offer: “Naked Desire”. How about it?

Still, all went well with my hacks; and eventually I was up and running with Android 2.2 and of course Flash 10.0:

image

That said, I had an annoying issue with Android Market. I selected an application, it would appear to start installing, but stuck on “Starting download” or “Downloading” with no actual progress.

It turns out this is a common problem – see here for one of many discussions. A popular fix is to change your account from someone@gmail.com to someone@googlemail.com, but this did not apply to me. I discovered the reason, which is that my wifi is behind Microsoft’s ISA Server. Android Market requires TCP outbound connections on port 5228. I configured ISA accordingly, and after cancelling and restarting the downloads they succeeded.

Finally, I wanted a screen grab or two to decorate this post. I used the method described here, fine for developers but not ideal for casual users.

image

I’ll report more impressions shortly, including no doubt the inevitable iPhone comparison.

Review: Tapworthy – designing great iPhone Apps by Josh Clark

Developing for iPhone is a hot topic. Many developers are not only having to learn Apple’s Objective C and the Cocoa application framework, but are also new to mobile development. It is a big shift. Josh Clark is a iPhone designer, and his book Tapworthy is about how to design apps that people will enjoy using. It is not a programming book; there is not a single snippet of Objective C in it.

His book illustrates the power shift that has taken place in computing. In the early days, it was the developer’s task to make an application that worked, and the user’s task to understand how to use it, through manuals, training courses, or whatever it took.

There are still traces of this approach in the software industry, but when it comes to iPhone apps it has reversed completely. The app creator has to build an app that the user will find intuitive, useful and fun; otherwise – no sale.

An early heading reads “Bored, Fickle and Disloyal”. That’s the target user for your app.

Clark’s point is valid, and he does hammer it home page after page. You will get the message; but it can get tiresome. His style is frank and conversational: some readers will love it, others will find it grating after a chapter or two.

Even if you are one of the latter group, it is worth persevering, because there is a ton of good content here. There are also numerous short interviews with developers of actual apps, many of them well-known, discussing the issues they faced. The persistent issue: we’ve got a complex app, a small screen, and intolerant users, how on earth do we make this seem simple and intuitive?

Constraints like these can actually improve applications. We saw this on the web, as the enforced statelessness and page model of web applications forced developers to simplify the user interface. It is the same with mobile. Joe Hewitt, author of the first generations of Facebook for iPhone:

There is so much stuff that is actually better on the small screen because it requires designers to focus on what’s really important.

So what’s in the book? After a couple of scene-setting chapters, Clark drills down into how to design for a tiny touchscreen. Be a scroll sceptic, he says. Chapter 4 then looks at app structure and navigation. Chapter 5 takes you blow by blow through the iPhone controls and visual elements. Then we get a chapter on making your app distinctive, a chapter on the all-important start-up sequence and how to make seem instantaneous, and a chapter on touch gestures.

The last three chapters cover portrait to landscape flipping, alerts, and finally inter-app communication and integration.

Throughout the book is illustrated in full colour, and the book itself is a pleasure to read with high quality paper and typography. 300 pages that will probably improve your app design and increase its sales; a bargain.

 

Enterprise app development on Apple iPhone and iPad

Apple’s iPhone is still perceived as primarily marketed to individuals rather than corporates. However, I was interested to see how much Apple is doing to attract corporate developers. First, Apple now supports some basic enterprise-friendly features, such as Microsoft Exchange (with a few caveats), VPN, remote wipe, and the ability to lock down iTunes to some extent. Without these capabilities, the devices would not be acceptable in many environments, making it pointless to consider them for custom applications.

Unfortunately iTunes is still needed for activation, deploying software updates, and installing applications. It is silly that Apple requires business users to install a music library to use its phone, I guess reflecting the device’s history as a music player. It is also a somewhat intrusive application especially on Windows.

If you then want to develop internal applications, you sign up for the iPhone Developer Enterprise Program. At $299 per year this is more expensive than the more general equivalent, but no big deal. Then you have to get a digital certificate from Apple. Next, create one or more “provisioning profiles” that install onto the device and authorize it to run your applications. Applications you create must be signed with your digital certificate. Finally, you can add the signed applications to an iTunes library, and users can then drag them to their iPhone or iPad. It will only run on devices that have the matching provisioning profile installed. Organisations can also revoke applications by revoking the identity used to sign the provisioning profile.

As Adobe pointed out to me, since these apps do not go through Apple’s approval process, there is nothing to stop corporate developers using the Flash Packager for iPhone that is available in Creative Suite 5.

There is more detail on Apple’s iPhone in Business page.

Google App Inventor – another go at visual programming

Google has put App Inventor for Android on Google Labs:

To use App Inventor, you do not need to be a developer. App Inventor requires NO programming knowledge. This is because instead of writing code, you visually design the way the app looks and use blocks to specify the app’s behavior.

Sharon Machlis at Computerworld says it is a breakthrough:

App Inventor has the potential to do for mobile app creation what VisiCalc did for computations — move it out of the exclusive realm of specialists in glassed-in data centers (or, in the case of mobile apps, programmers who can use a conventional SDK) into the hands of power users as well as make it easier for IT departments to create corporate apps.

I’d like to believe this but I do not. It is visual programming; it is interesting; but it is similar to other visual programming tools that we’ve seen in the past. These tools have their place for learning, and there is probably some small sub-section of programming tasks for which they are ideally suited, and some small sub-section of developers for whom they work better than text-based tools, but for most of us textual code is easier and more productive when we are coding the logic rather than the user interface of an application.

I took a look at the Quiz Me tutorial. Here’s a code snippet – it is a click event handler:

image

and here is the complete application. Note the navigator at top right, which would be vital for finding your way around a more complex app:

image

It is often a problem with visual programming tools: scaling an app beyond a few simple blocks introduces difficulties with navigation and project management. Our text-based tools are highly evolved for managing large projects with thousands of lines of code.

What about democratisation of programming through visual tools like this, coding without coding, that will allow domain specialists to develop apps without involving developers? Is visual programming really easier for the non-specialist than textual programming? I’m not convinced. It should be easier to get started: no syntax errors, no language reference to consult. In the end though, is a purple “if” block with jigsaw connections for “test” and “then-do” much easier than typing if (test) {code block}?

It is just a different way of presenting the same thing, but less concise and less flexible. I suspect the domain specialist who can succeed with App Inventor will also succeed with code; and that people who struggle with code-based programming in an accessible language like Basic will also struggle with visual programming.

Where this gets interesting is when you have powerful components that expose a simple interface. A high-level non-specialist programmer can drag a component onto a design palette and do amazing things, because of the smarts that are hidden inside. Components do democratise development. One reason for the success of Microsoft’s development platform is that from Visual Basic through COM and then .NET, the company has always made it easy to use components and fostered a strong third-party market for them. If App Inventor provides a great way to deliver components to high-level developers, it could succeed.

That said, components do not require visual programming. Microsoft has flirted with visual programming – like the abandoned PopFly – but despite using the name “visual” everywhere, Microsoft has never delivered a mainstream visual programming tool.

DevExpress merges its Silverlight and WPF UI controls, says VS 2010 is light years ahead

Developer Express is a component vendor with add-ons for Visual Studio and Delphi. It has offered a library of components for Silverlight for some time, and a separate set for WPF (Windows Presentation Foundation), but now says that Silverlight and WPF are close enough that it can merge the two into a single codebase to be called XPF (Express Presentation Framework). CTO Julian Bucknall says:

Silverlight in v4 has the ability to create desktop applications that aren’t sandboxed into triviality. In fact, Silverlight, more than ever, resembles a WPF-lite on the desktop side, to the extent of pundits considering their eventual merging. At long last it is possible to write one set of non-trivial code and compile it both for Silverlight and for WPF without having to reinvent so many wheels on the Silverlight side (and to a much lesser extent on the WPF side).

Even though Visual Studio 2010 is only just released, DevExpress is focusing all its new Silverlight and WPF development on the new platform and IDE:

The Silverlight and WPF controls in DXperience v2010.1 will require .NET 4 and VS2010. In particular, you must use the new Silverlight 4 and WPF 4; the controls will not function with the previous versions of WPF and Silverlight, such as Silverlight 3. Similarly, you cannot use VS2008 or earlier, but must use VS2010. To my mind this isn’t that much of a downside: VS2010 is light years ahead of its earlier brethren in terms of user experience and its use is de rigueur if you are creating applications with either Silverlight or WPF.

Of course it’s in Bucknall’s interests to move developers on; he’s keen to sell upgrades. I still find this interesting. Like him, I find Visual Studio 2010 a major advance on earlier versions. More significant though is the idea of a common WPF and Silverlight codebase, though presumably still with added capabilities when running on WPF. I don’t think Windows-only development is dead; the success of Windows 7 may even stimulate the market for applications that take advantage of its new features. That said, for the large subset of applications where cross-platform is desirable, Silverlight seems to me a better choice than WPF.