Raspberry Pi

Liam Proven wrote on Tue, 5 Mar 2013 00:16:44 +0000: And so they should be, and they should be proud of the Inmos Transputer too though that is no longer with us (except through its echo, the Xmos). The Raspberry Pi people actually tried to make the same deal with the BBC, but it was no longer possible because current laws don't allow that kind of partnership anymore. But the project did get mentioned in a BBC blog and that is how the world found out about it a bit too early. The first prototype used two Atmel AVR8 processors instead of an ARM. One generated video in software while the other ran the Python interpreter (but for some reason they decided to call it "Pi" instead of "Py"). The second prototype used a Broadcom chip which made it cheaper and the Python interpreter much faster, but it was only the third prototype that added the ARM and made it into a general purpose machine. The current board design was changed twice after that, but the main chip remained the same. This is way off topic for this list unless you consider the Smalltalk virtual machine to be a classic computer and what I am doing equivalent to someone putting a PDP-11 in a FPGA. In some ways FPGAs are so abstract that you might as well be doing just simulations, like in the famous "From NAND to Tetris" course. By adding a few steps before you get to the FPGA, however, it can be much more educational. Etienne Delacroix (from Belgium, but he spent years in South America working with children and university students) likes to start his "atelier labs" by building simple gates from transistors and resistors, then more interesting circuits with TTLs before repeating the projects with a Z80 and software. He didn't like that the Z80 is a black box, and so wanted to build his own processor from TTLs. He found a description from a one instruction machine (substract and branch if negative) on the Internet and I designed a board for him with 12 TTLs and did the macros in "as" so he could put his simple Z80 programs on it. I don't feel the result is very educational - the hardware is too simplistic and the complexity gets hidden elsewhere (the macro expansions generate huge binaries). An option is to have more layers, where the transition from one layer to the next isn't too radical. From the top down, we have: 1) very high level programming for games and 3D shared virtual worlds: Scratch, Etoys and OpenCobalt/OpenCroquet/OpenQwaq http://scratch.mit.edu/ http://www.squeakland.org/ http://www.opencobalt.org/ 2) Smalltalk text level Squeak Smalltalk, Pharo, Self and other Smalltalks www.squeak.org http://www.pharo-project.org/home http://selflanguage.org/ 3) Smalltalk bytecode level stack bytecodes, jump bytecodes, send bytecodes, special bytecodes https://github.com/4ji/Smalltalk-Bluebook-chapter-26-30 4) microcode level 16 bit instruction set described in: http://www.merlintec.com/download/jecelwplcr.pdf 5) cache level most of the registers visible in level 4 don't really exist and are instead mapped to the four caches 6) hardware level logic gates Some outdated information about the project can be found at: http://www.siliconsqueak.org/ That is great to hear (that someone is making use of the language). This morning I was thinking that I was too subtle - to get my reference you would have to know that some of the email applications on the Xerox Alto were "Laurel", "Hardy" and "Babar" and that Babar and Celeste were characters (talking elephants) from a 1930s series of children's stories. -- Jecel