A little while ago, ARM Holdings (ARMH) put out a slide claiming that a large portion of the applications available on the Google (GOOG 0.39%) Play store require some degree of "binary translation" in order to run properly on Intel (INTC 0.29%) hardware. Binary translation essentially means that compiled applications originally intended for the ARM architecture need to be "converted" into something that Intel's X86 processors can run.
Remembering ARM's claims
According to ARM Holdings, a significant chunk of the most popular applications running on Android do have some level of ARM-specific code, thus necessitating some level of binary translation.
Source: ARM Holdings.
At first glance, these numbers are downright worrisome -- how can Intel really gain a foothold in the Android space if its chips have serious compatibility issues? Of course, Intel has done a lot of work with its binary translation scheme, which allows these applications to work, which puts it far ahead of ARM's efforts to penetrate the Windows PC world; but for very performance-sensitive applications, like 3D games, binary translation does come with a non-trivial penalty, as demonstrated here.
That said, at Intel's recent London Analyst Summit, Intel VP Bill Savage gave a very illuminating presentation with respect to these binary translation claims, and how not all is as it appears.
Intel's great Android adventure
First, Intel is working diligently to get 80% of the top 100 free and paid applications ported to X86. Given Intel's willingness to push into the tablet/mobile markets, there's little doubt that the company will hit this goal sooner rather than later.
Source: Intel.
Further, in Savage's presentation, he noted that, while many applications on the Android market do require binary translation, only relatively small portions of those applications are native and require such translation in most cases. Savage also claimed in his presentation that the graphics portion of these applications don't require translation either. This is only partially true -- the pixel/vertex shader programs that run on the GPU do not require translation, but the actual game engine code would, so there will be a performance impact in a real-world game.
At any rate, the idea here is that, for most applications, the amount of native code is negligible, and thus the translation overhead is lost in the noise of the full program execution. However, this doesn't mean that Intel can afford to not get as many of the major applications ported, nor does it mean that Intel shouldn't be aggressive in trying to get developers to target Intel as a first-class citizen on Android.
What is Intel doing to help that transition?
In order to promote this transition, Intel is leveraging its extensive software and services group to get developers on board. For example, the company is promoting its Intel Integrated Native Developer Experience product, which allows developers to easily build native programs across both Android and Windows:
Source: Intel.
However, while making easy-to-use tools for developers is nice, not all developers are going to adopt them, so Intel needs to do a lot of work with developers to make sure that, going forward, X86 is a prime target. Intel has indicated that it is doing so and, hopefully, as more Intel-powered Android devices roll out this year, developers will be increasingly attracted to the platform.
Foolish bottom line
Let's be realistic -- ARM has the first-mover advantage, and many powerful players have invested large sums of money in the ARM software ecosystem. While the nature of Android makes it so that Intel architecture has a pretty good shot at becoming an equal partner to the ARM architecture in terms of compatibility long term, it is going to be neither easy nor cheap. However, Intel seems to recognize the importance of this, and is going to fight tooth and nail for the hearts and minds of developers.
