Archive for December, 2010
I saw many, many, many, many articles and presentations saying: "98% processors in the world are in embedded systems".
Where did they get this information from? Why people don’t cite original work and point to the source of data? Sad….
Anyway, after checking about 10 articles, at least one, published by ACM, pointed to the original source of this information:
I also found this in context, which is also very interesting:
Ten Lies About Microprocessors
ssh -NL 8118:localhost:8118 user@server
Configure your browser to use proxy – localhost:8118
Everyone knows (or not) that I am a big fan of ARM microcontrollers, as I already have a collection of them…of course, all running Linux.
But I wanted to know about the behavior of multi-core ARMs. As my PandaBoard did not arrive, I asked help from Raul Porcel, Gentoo developer (http://armin762.wordpress.com/) who kindly run some tests on his PandaBoard (Dual Core ARM Cortex A9) and his Nvidia Tegra 2 development kit from Nvidia (Dual Core also).
Well, the results are in the graph. They are compared to 2 different dual-core ARMs, and it is great to see that the speedup on ARM platform was roughly the same as in the PC platform. Note that by speedup I mean performance increase with more processors, and not execution time.
Also, I was curios about Hyper Threading. So, here it is another graph, now comparing x86 machines with multi-core architecture. Thanks here to my friend Alberto Daniel Junior (Sapo) who did a test on a Dell machine with 8 real cores (HT), which shows in Linux as 16 processors.
It is interesting to see that although Linux lists cores with HT as 2 cores, they do not behave as a real processor. Well, the numbers don’t lie…just take a look at the graph