#cortex a15

Rivales Raspberry Pi: ODROID-HC1

Otra vuelta de tuerca a una de las arquitecturas de más éxito de Hardkernel y sus placas ODROID. Se trata del mismo hardware base de la ODROID-XU3, la ODROID-XU3Lite y sobretodo la ODROID-XU4. Los cambios son para convertirla en una base para un NAS casero:

Cambios:

Aún más barata, unos 60€ incluyendo disipador con ventilador pero sin gastos de envío ni fuente de alimentación incluidos.

Tamaño…

Rivales Raspberry Pi : BeagleBoard-X15

Las BeagleBone eran antes de las Raspberry Pi las placas de este tipo más populares, soportaban Ubuntu y tenían una comunidad respetable en el mundo de la electrónica y la robótica. Su precio hizo que nunca alcanzara los niveles de ventas de la Raspberry y cuando esta llegó trataron de reaccionar con la BeagleBone Black, pero ya era tarde. Consiguieron mantener a su comunidad anterior, que no es…

New Post has been published on http://www.phonerumors.net/evolved-samsung-exynos-5-octa-coming-week/

Evolved Samsung Exynos 5 Octa coming next week

The Samsung Exynos Twitter account crackled to life on Thursday morning with a tweet hinting that an evolved Exynos 5 Octa is coming. The attachment to the tweet states that a more powerful and enhanced Exynos 5 Octa will be introduced next week. Right now, the most current version of the Samsung Exynos 5 Octa is found on certain versions of the Samsung Galaxy S4.

Speculation calls for the new chip to be the Samsung Exynos 5420 and will feature a higher clock speed along with improved optimization for the 8 integrated cores. Not technically a true octa-core chip, the Exynos 5 Octa is comprised of two sets of quad-cores with one set of Cortex-A15 chips doing the heavy lifting while the other set of quad-core Cortex-A7 chips takes care of the light housekeeping. Not all 8 cores can be used at once. When one set of quad-core chips are being used, the other cannot.

Recently, a road map from low cost chip producer MediaTek showed off a true octa-core chip coming, manufactured using TSMC’s 28nm process. Mass production of this chip, the MT6592, is expected to start in November with the first handsets powered by the silicon to launch early next year. All 8 cores will be available for use at once.

Turning back to the tweet about the Samsung Exynos 5 Octa, the Korean tech titan said that it will be holding a Twitter based contest based on the evolved chip and the rules will be released next week. According to Samsung, the prizes will consist of Exynos powered devices.

There has been some speculation on the net about what sort of processor is in the A6 SoC found in the iPhone 5. The two dominant choices are either a quad core version of the Cortex A9 found in previous versions of the Apple SoCs or a new dual core version of the Cortex A15, which until now has not been found in any shipping product. Without access to actual devices, it would seem the odds are stacked against us in determining what the actual CPU is.

Except that with the latest Xcode now has support for building for a new architecture -- armv7s. There is little to go on about what the differences are between armv7s and armv7, so I decided to look in the opposite direction. I went with the assumption that if Apple is using an Cortex A15, they would want to support any new instructions provided. And therefore, I went and found out what was added in the A15 vs the A9.

It turns out there are a handful of new instructions. Most of the features seem to be oriented around virtualization support, which wouldn't be of any use to an iOS developer. But there were two instructions added that could be very useful -- a "fused" multiply and add instruction (MAC) and more importantly, an integer divide instruction. This later one is interesting because currently if division is required on the armv7 platform, the compiler inserts a call to a helper function to carry out the division in software.

My plan was then as follows. I would build a test application for both armv7 and armv7s. In this application, I made a test function that was as follows:

int my_divide(int a, int b) { return a / b; }

I built the project and then when diving into ~/Library/Developer/Xcode looking for the .o files for both the arm7 and arm7s version. Since I didn't have an arm7s disassembler handy, I chose to use nm instead. nm will list all of the unresolved external symbols used in the .o file. Since the software division routine would need to be linked in, my expectation was that on the armv7 version, I would see an unresolved symbol for that routine and on the armv7s version, there would be no unresolved symbol.

Here's what I got when I ran it on the armv7 version:

$ nm File.o U ___divsi3 00000000 T _my_divide

Here is what I got when I rant it on the armv7s version:

$ nm File.o 00000000 T _my_divide