The collected works of ivelegacy - Page 14

I have a portrait, a painting, which looks like a photograph of Heather Locklear
she is my "phenotype" about "blonde", and ... I got wicked threats like "remove it, or it will burn"
so it's hidden in the attic

hi
I am still blocked with 3Com 3C996B-T Gigabit Network Card, I got one but ...
I prefer to have an original card instead of hacking the flash (I do not have the equipment to fix the CRC)
which does not completely work, Irix claims errors, and I do not want to patch the driver

someone has one for sale?

if so, let me know!

I wish I was rich, Discreet-Logic-Flame-2008-2TB-Stones-Sapphire-Sparks-SGI-Tezro-maxed-out @ ~9K euro + shipping

Broadcom BCM5701TKHB chip, but SGI P/N

• SGI P/N 9210289 (3C996B-T-SGI1)

In the theory any Tigon3 gigabit cards should work with this hack, these include Compaq NC7770, 3Com 3c996B-T
but I prefer an original P/N if available under the definition of "reasonable price": no more than 50 euro

&& the solution is: to be on SunRay2's console
like a piece of cake on sunny summer party


money fault, but
I am going to invent a money printer machine

oh men, I guess he is exercising in xmass songs
(here I am so ahead that
I am focused on the Happy New Year Compilation)

here I am spending "a lot" of my free time writing strange things
(what the frog should I have to do with a BTREE? Who on the Earth knows?)

last but not the least: interviewing all the English girls I can meet
around the Linate Airport/Milan in order to "learn" English
(I have a few job tasks on my desk there, in the Theory )

so, no xmass-news for me ?

dear Sanda Claus

been an angel all year
* * * so hurry down * * *
the chimney on 25 night

ThkU

&& SantaBaby
under the tree
LoL

so if the worst comes to the worst: php a fractal of bad design , perhaps the worst ever.

hi
I wonder how fast a diskless setup goes, with a 10Mbit/sec built-in connection, and if I can make it "stripped" in order to reduce the data-transfer.

I also wonder how good is linux/X11 with a modern Xorg package, I played linux on IP22 in 2005 at my university unix lab (currently RIP, replace with linux/PC), there was a linux kernel 2.2, and completely different packages (Xorg has changed a lot)

In this case, I wonder how much smart is to look for a striped-ram-rootfs with an MIPS3-BE-softfloat-uclibc (the lighter ever) version of the whole rootfs (if my builder will build)


the purpose is: using IP22 as X11/Xterm (WM~=Fluxbox?)

just to know

it seems I am alone with IP30
I know only three guys with Gentoo/MIPS

• Kumba (gentoo/MIPS team)
• Ralf Baechle ( at linux-mips dot org, but … it's a bit difficult to get an email from him )
• Redhatter (Stuart Longland, australian dude, he was a gentoo dev/user, enjoying tinkering in the fields of embedded and RISC computing, and curing a blog, but I haven't read anything from him since a while)

any other ?

(nice to meet you, in advance)

nyef wrote: I have a couple of IP30 systems to install Linux on, I just haven't settled in to trying to get them to work

nice to hear

I am still currently running linux v2.6.17.4+hack on my IP30
but today I have evaluated v4.3, work in progress, it will take a while




I have my two ramrootfs, based on uclibc, soft float & hard float, mips3/be
so it's also good for

• RS/PRO, Atheros AR7161 CPU
• GL-iNet-6416, Atheros AR7240 CPU
• TL-WR703N, Atheros AR7240 CPU
• Fonera, FON2100A (hack SPI), FON2200, FON2202(built-in USB)

they are tiny routers

edit:
routers (MIPS32r2) vs SGI(MIPS4)

ivelegacy wrote: Interesting languages and books

• ADA & Protected Objects
• Modula3
• RUST
• Concurrent Euclid, Addison Wesley Computer-Science

let's add an other great coin to the list!

• Erlang!

Github is partially implemented in Erlang && The "great strange thing" about Erlang is the message passing model for parallelism and the "let it fail" philosophy

I guess Erlang will become very relevant in the near future with cpus with hundreds of cores, also I think it's a good language for fpga soft cores, this because in Erlang you build lots of independent processes that communicate sending messages. No locks, no shared memory, no semaphores, no contention. That's also how hardware works or should work. There is no globally shared data. If you want/need globally shared data, give it to a process and communicate with it with messages. Make a server, and when a process starts another, it can "ask the system" for a message when the created process "crashes". So you can have very easily an hierarchy of "supervisor" processes watching "workers" and restarting them when one "dies".

As far as I understood, It may not suitable for the "embedded" side near the sensors, but if you need to connect hundreds or thousands of embedded systems, or clients or whatever, you better look into Erlang!

Nice to know




~~~~~~~~ going back to the C language ~~~~~~~~

well … until GCC 4.6.3, compiler built-ins for atomic operations were a pure compiler extension, and in GCC they were grouped into the __sync_* family of functions (it means problems, a lot!), while as of version 4.7.0, both the new C++11 and the C11 standards had been finalized, and GCC updated their atomic built-ins to better reflect the new memory model of those two new language revisions. The new functions are grouped into the __atomic_* family.

(the older built-ins are still available, for legacy purposes)

Hopefully I'll be able to get you a working beta-netboot within the next 1-2 weeks for the Octane,

already done for 2.6.17, I can re-use the ramrootfs with kernel 4.*

p.s.
is there a Gigabit/sec PCI card that is known to work with XIO_PCI (ShoeHorn or ShoeBox) ?

Kumba wrote: You can, as I sometimes use that old ~2007 netboot for fixing things

I have already built mine, also used for my routers, as I do not like the OpenWRT rootfs, but in case, I can re-use the OpenWRT builder

All the routers in the above list are MIPS3/be-uslibc-softfloat profiled, good enough for SGI R1*K


FON2100A (hack SPI), FON2200, FON2202(built-in USB)



I have already shown you a screenshot, the above ramrootfs runs on IP30 too

(about good test-code, I am using nbench)


IP30/SMP (2.6.17.4.hack), Toshiba RISC 2xR12000 @ 400Mhz with FPU -> BogoMIPS 598.01+600.00



IP30 SMP (2.6.17.4.hack), Toshiba RISC 2xR14000 @ 600Mhz with FPU -> BogoMIPS 897.02+897.02




So I can measure & compare "progresses" from 2.6.17 to 4.*

Guys, don't worry about kernel crashes, since I have implemented this hw hack, I am now able to reboot the machine remotely! So I can test whatever you want to be tested, including unsafe kernel code

which stress & test apps do you run ? && what do you think about "nbench" ?

unfortunately all my routers are FPU_LESS, so I am using two solutions

• soft float profile, the floating point is provided in userspace by an external emulation library, directly linked to the App, and gcc is patched and configured to use it.
• kernel support, trapping all the exceptions in kernel space, so the Application believes that the hardware owns a floating point unit

currently I am using the first one, as it's the default choice used by OpenWRT (and I am recycling a part of their code/builder)

I'd like to have "something" (an application, a piece of C code, whatever) to measures the fine grane scheduling speedup, as my 2.6.17 was still using big-kernel-locks. Today I have completed the v4.3 build up, I am going to give it a full try, planned for this WE

A gentleman has warned me about : don't do it, you are still in time, stop your maleficent plan … too late, sir




I am coding a similC interpreter & I have redefined the grammar in a strictly-schoolmarm way
look at the above code, I haven't implemented the semantical part, just (a part of) the syntax tree

Schoolmarm' rules

• a type MUST ends with "_t", otherwise a syntax error is issued
• a pointer MUST begins with "p_", otherwise a syntax error is issued
• a function MUST have its parameters in the form TYPE DATADIR VARNAME, where DATADIR = { In, Out, InOut }, otherwise a syntax error is issues
• comments MUST be in the form /* … */, the // form is banned
• every while && if/else statement MUST be followed by a block { … } otherwise a syntax error is issues
• everything MUST be explicitly declared "private" or "public", there are NO includes (#include ".."), everything is automatically built using a db, and this is a part of the language (auto interfaces)

that's all, for the moment, but … the SafeC Schoolmarm used to drink Conium Maculatum (Aka Cicuta) for her tea break
don't call her "poison-schoolmarm"

oh men, I am afraid that I have been writing worst wicked witch ever

skywriter wrote: Bummer, I love //

Geany (the IDE) also loves //, but it's banned by MISRA-C-95

skywriter wrote: btw is this just for fun, or are you getting paid to do this?

Yes, I am payed(2) to develop a C checker = { MISRA-95 + MIL+STD + DO178B}(3) & an a few other tools(1), and in order to develop such a tool I need to learn a bit of things

I am studying a few books, and exercising with a few toys, but I can recycle the code
e.g. the lib_tokener is almost completed, and will be used in the production tool, while the SchoolMarm is just a "learning toy"

(1) dead code checker & inter module-dependencies
(2) unfortunately not as "consultant", so I will be paid for the final product, without caring about the effort in term of how many hours
(3) the reason why is the funniest part of this story: my tool can't be used to validate anything (because nobody has validated it), but! The official tool has just 2 licenses, and the dev squad is composed by a lot of guys, that means they looks like the "Dining philosophers problem" by Edsger Dijkstra where there are N users with M resources, with N >> M, so in order to solve it, my boss has asked me to develop a tool to let them to check their code without the need of accessing the official-C-checker, so they will use my tool during the development phase, and then to validate everything in one shot with the official one (2 guys per time). Funny Avionics World

edit:
syntax error fixed

robespierre wrote: Ada is rather verbose with many "noise chars" that seem redundant. It does have a nicer type system that can explicitly define upper and lower bounds. VHDL was based on Ada.

yeah, and let me say: I like VHDL

(also my poor soft core likes the VHDL, whenever I promise to complete it …
… can I invent a cloning machine in order to get a few copies of me
in order to complete all my wicked plans ? seriously tempted)

skywriter wrote: those are beyond a parser

every trip begins with a foot step, I am currently involved into a syntax checker, while the next step will involve a full parser
e.g. no multi return in a function, no casting, check if a data respects its DataDir { In, Out, InOut }, etc etc etc

The rabbit hole goes deeper than you can imagine: just a few days ago (and confirmed by email 5 minutes ago) I was told that there is step3 of the wicked plan.

So everything begins with a simple syntax checker, and now it ought implement a real simil-C interpreter for an Ethernet debugger
oh man, the story goes funnier!

There is a similar product made by Windriver, and (probably due to the fact that it costs more than I can imagine, which is really too much) my boss wants to clone in order to have the full control. Windriver && Green Hills dudes sell these kins of debuggers, but you have what you pay, they sell just the bare debugger, if you want the scriptable engine (which is physically an add on) you have to pay as extra feature, and it costs a lot, and if you need to customize it will double the cost.

Hell knows why this must be the case: he has explained that a customization is needed, and they need a lot of these toys for the production, while they do not need a certified for such a purpose. Funny guys, they want to save a part of their money, and here we go.

So, since then, my boss has been thinking about my simile-C interpreter, pointing out all the good things out of the box IF it should run into an embedded box, arm-m4-driven and bare metal (developped with Keil ucVision), with a bandwidth of 20Mbyte/sec through a DMA & usb bulk driven combined methods (probably with an external USB 2 chip, I have to check their project files) and able to follow the debugger tap instruction (this part is implemented as TAP inside the PowerPC chip) in order to handle the low level of the debug interface

practically
- code injection
- flash programming
- flash reading
- hw breakpoints
- hw watchpoints
- stack trackers
- ram viewer
- bus error trappers
- not invasive

all of these "features" are planned for a PPC4xx line (which already implement a jtag port), already developed by a squad they have (lucky I am, all those parts would be too complex for me in term of several order of magnitude), but there is no product because they miss the user interface, and here we go trying to export to the user through a scriptable engine which looks like a C interpreter: "it's good when you have to write testing scripts" - he said. Which is ok, alright, and why on the why do you also want to put the "poison-schoolmarm" into a debugger ?

The science doesn't know What the Frog is beyond the motivation

I guess the problem is the cost, plus other problem with the EXEC manager, for example the QaC-tool costs too much (in several orders of magnitude), and they do not want to buy all the licenses for all the squad dudes (~ 40 dudes), so, they have bought just a pair of licenses. I am bit shocked, but the price is really too high, not for the final customer (which has enoght money to buy an aircraft), but too high for a consultant agency, and the problem is about communication, we are an external agency, and my boss should understand that we'd better use the customer's tools (forcing the customer to buy them for us) instead of being independent (which practically means not enough money to buy those tools for all the dudes in the squad, plus all the dudes in the production).

I don't have an idea about that (hidden business? hidden reasons? who knows), I am too young, 2 years in avionics, 4 years in automotive, and they are completely different worlds, like the Earth planet (automotive) vs Mars (avionics)

but I have understood that I'd better take secret everything I do develop in private (e.g. the SafeC-Schoolmarm) because it's "dangerous" if my boss hears about (and in second place because I am not paid more), and I guess I will leave avionics, automotive is better.


~~~~~~~~~~

when I was a student, I developed this interpreter for an homework examination. It's written in C++, but it's possible to port it to C89 with a minimal effort (a few hours). It's a toy, a minimal implementation which looks like pascal, if someone wants to play, here it is on google code, with a few examples, while calc_host_v2 is a tiny integer calculator, written in C, and ables to evaluate an expression. I have coded in order to "test" a soft core (mips2), but both of them are stalled project, so Enjoy

gentoo-4.3.0.20151126: the first try … fails



I am switching back to "ip30_defconfig-4.3.0", then I will try to add features and options




the the second try fails funnier



emerged & switched to sys-devel/kgcc64-v4.9.3!
(64bit kernel compiler)

Kumba wrote: That said, at the above point you indicated, did it simply freeze up or auto-reboot itself?

auto-reboot itself!

You have dual R14K/600MHz in your Octane, right? What graphics card?

my IP30 is SMP2xR12K@400Mhz, my friend machine is 2XR14K@600Mhz
both machines do not have a GfX installed, mine has a MENET and ShoeHorn
his machine has a empty carrier (just the plastics without anything installed)

the above screenshot is from his machine, mine auto-reboot itself
so we get the same reaction from both of them

Kumba wrote: gcc-5.3.0

my production kernel, 2.6.17-rc4-hack, gets compiled by kgcc-v4.3.5 and boots fine
I have switched to the last kgcc in the portage

gcc-5.3.0, as far as I see from the portage (last --sync, yesterday) it's currently ~mips
but I can force it, just give me the time, my Atheros9 takes 9 hours to compile gcc

Atheros9 = mips32r2/be/32bit, funny 900Mhz machine

The first head-node in Cerberus, 24 June 2009
It's still running kernel 2.6.15 on a single pATA

Kumba wrote: Still need to get you off of 2.6.x entirely. I can only wonder about the security flaws that cumulatively exist in such an old release.

oh well, let me say a few words about:

Cerberus (click on the link in my signature to see it) is a NAS system, build on PowerPC technology, composed by 4 nodes, and loaded with different fs services (FUSE/SSHFS, NFS, SAMBA, …), and repo services (GIT, SVN, MERCURIAL, ...), ables to performs soft-mirroring (softRAID). A pretty old Kernels 2.6.15 and 2.6.24 are running on such a system since the beginning with no troubles in years of continuously working. Cerberus was born in single node configuration, then I added more nodes, all my source code archives are stored by that multi-heads-monster & in all of his heads I trust

All my routers (RS/PRO, TL703) are 2.6.* are 2.6.39.4 and no troubles with them, even if recent routers they are migrating to 3.3.* (e.g. in TL703 the RF wifi module needs a kernel workaround to fix an hardware bug, and this "patch" is available only for kernel 3.*, so I'd best move up)

Acme System is still supporting NetusG20 (arm5/embedded) with 2.6.39.4, no modern kernel are available from them without troubles, so, for the production I am using 2.6.39.4.

My PowerPC laptop is 2.6.26, my PDA C1K is 2.6.23 and it's the last kernel available really usable, so I have back ported an USB ethernet module from 2.6.26 (which sucks greatly about the battery charge and the suspend feature, which are the most important things in a portable device)

In short, everything I have developed for hobby & job is based on 2.6, this implies that my kernel modules (e.g. USB_LCD, used in shipbuilding, job purpose) are based on 2.6, and in order to "port them" to kernel 4.* I just need effort: time, a lot of time, in first place!

I am planning to attach my USB_LCD to ip30 through a NEC USB/OHCI chip installed inside the ShoeHorn, it's currently working fine with 2.6.17.4, but it's time to move , no doubt about that, and no doubt kernels 4.* are really supporting Btrfs, the new copy on write CoW filesystem for Linux aimed at implementing advanced features while focusing on fault tolerance, repair and easy administration.

It's time to move from ext3, I want the Btrfs filesystem

however, consider that moving kernel modules from 2.6 to 4.* requires a lot of effort in first place, and consider that I have other pending tasks in my queues (e.g. VHDL, soft core & TAP, planned to be completed the sooner )


OK, let's get rock & thank you for your help

(start--stop - who - where - what - result)

so, gcc took longer to compile

I guess you won't find one on ebay, all of these boards are gone.

I was told they came from business failure, I bought a stock of 20 boards from Mexico in 2009

I sold a lot of them, and I am using 3-4 nodes, how many boards do you need ?
I might provide one (offered to a friend of mine), but I have to check if it's possible (if he still needs it, or not)

I also have a PPC440 and a PPC460 board for sale, full ATX, even if not so well supported by linux
(bought for avionics purpose, because Avionics uses PPC440/460, Windriver and Green Hills do support them)

and I have two Sandpoint/7410(=G4) boards for sale, full ATX

contact me in private

p.s.
these nodes do not have a great firmware support , you can't bootstrap a kernel linux which requires dev-tree
also, bootstrapping from pATA is not supported by Uboot, and my patch is too old, and no time to port it
('cause a lot of things changed since 2007, which is the uboot version I used)

p.s.2.
I am working around "kexec", it allows an old kernel to load and boot a new kernel.

p.s.3
I have opened a topic in the WTS section of this forum, in case you want to purchase one of these PowerPC boards!

@kumba



is there any problem if the kernel image is 22Mbyte (kernel + ramrootfs) ?
just to know, I currently also have a smaller 9Mbyte file (kernel + uclibc ramrootfs)
is your testing machine a dual R12K ? or R14K ?

I have

• forced R10K (with R12K, R14K, R16K it hangs up)
• force early serial
• removed gfx kprintf (we do not have a gfx installed, so it's no sense)
• attached a compressed ramrootfs

&& it boots !!!!!!!!!!!!!!!!!



hApPyhApPy
Celebration








p.s.
the SMP is not enabled (cat /proc/cpuinfo says 1 CPU), the Menet has 2 NICs instead of 4

OMG
impressed

hAllo, it happens I have one of my hands in avionics, sometimes carry out immoral acts(1) with their software

(1)we, norm-life people, are used to call "patches" and to apply them (especially to the kernel linux, or generally, it's a common practice with open sources), but you cannot say "patches" with those dudes because Avionics MUST avoid to have any evidence of mistakes, especially when a project has been already approved (from a draft/proof) and committed: if you find then a mistake, it means that someone (going to be get fired after his/her 45 minutes of public shame) did not do his/her job properly, and a lot of bloody money will flow, because you have to call back all the aircrafts in order to "patch" them

Avionics does not apply patches, it applies "changes" due to the failures in obtaining hardware parts ( )

oh well, but I want to say: never seen an SGI/IRIX@MIPS machine there, they use PC/WindowsXP & Sun-SunOS@Sparc, software like Office, Doors, QaC, vectorcast, stood, AAL, etc

Windrivers and Green Hills are the most involved into the Avionics OS, more specifically, things like VxWorks/DO178B, Vel/OS, Integrity, … && DevChains like AdaMulti, Tornado, Workbench, plus specific ICE/PPC4xx from Green Hills and Windriver



But, as far as your request, You can enjoy A/UX, announced at the February 1988 Uniforum conference, seven months behind schedule, based on AT&T's Unix System V.2.2 with additional features from BSD Unix, and sold bundled with a Macintosh II for ~ $8500, even if the the most interesting part of this story came in 1991, when Apple formed a new business division for enterprise systems to serve "large businesses, government, and higher education".

Government OS, oh my pants

I guess, the most interesting part comes with mission tactical computers, also known as "MTC"
they are used in war, so they are formally war-machines, built on "strange things" like the ARINC technology

edit:
fixed

ATX form factor, they work with a common ATX PSU

• PowerPC440, codename "Ebony"

contact me in private by email
thanks





edit:
sold

• PowerPC7410, codename "Sandpoint3", made by Motorola/Freescale
• PowerPC7445, codename "Sandpoint3", made by Motorola/Freescale

let me know if you are interested, I have worked only on Sandpoint3, and I have a working linux kernel for this board

my personal experience is limited to ARINC 429, more specifically as alternative to MIL-STD-1553. Both of them ARINC 429 and 1553B are "communication standards" for wiring and protocols widely used in civil and miltary aircraft applications.

• The 1553 standard uses a command-response protocol along a time-division multiplexing, so within multiple nodes, it means - say sensors in the aircraft- can send their data "at the same time" along the same wire, up to 31 at data speed up to 1M.
• The ARINC 429 standard requires separate wires for transmit and receive at nodes, up to 20 data speed 100k, which means it's not a bus in the conventional sense as it requires more than one cable, and it's more robust and hard to be destroyed

coming back to "software", well Avionics uses special software & equipment, e.g. the "APX1553 MILScope", an equipment which integrates a unique Digital Scope function with an on board ADC digitizing the MIL-STD-1553 waveform for real time measurement & storage which supports automatic test functions to meet the "Electrical Test" requirements defined in "AS4112" RT Production Test Plan.

"Test Plan", it's a common "term" in Avionics, because everything MUST be tested, and reported (Test Plan ---> Test Report ---> Product approved ), so there are special Software Applications for these purposes, one of these (perhaps the most used) is called "Doors", it's a special "committer", even if it's used like … the big brother of GIT/Subversion/Mercurial.

I have forced cpu_type=R10000 && Multi-Processing support && Preemption Model=No Forced Preemption
and … something bad is happening now

jan-jaap wrote: I work for a company which specializes in military and aerospace applications of IEEE-1394 (FireWire). IRIX is one of our supported platforms.

That's a news for me, never seen Irix in my Avionics experience, only SunOS on Sparc machines && Wintel. Interesting news

Can I ask why Irix (on on which system ? Ip30 ? Ip35 ? …) ? I wonder where Irix is involved in the process

FlasBurn wrote: how much do you want for one board and can you provide pictures?

which of them ? Sandpoint-X3? Ebony?





Sandpoint-X3-mobo, with a 74xx CPU module


Sandpoint-X3-block

Features

• integrated host processor
• JTAG connector on PMC module
• One PMC slot on motherboard to accept processor PMC modules
• Switch-selectable modes
• Four 5V 32-bit PCI slots at 33 MHz
• Two standard 16650-compatible serial ports
• IEEE 1284 parallel port
• NVRAM and real-time clock (RTC)
• Two IDE ports
• LED monitors critical functions
• Advanced Power Controller ("soft on/off")
• PCI slot attached on Winbond W83C553F

&& 2.6.19 ramrootfs/uclibc microroot screenshot



I also have a NIC for sale, the Sandpoint X3 board needs a specific NIC because DINK32 has a restricted compatibility list


the ebony board looks similar, its firmware has a few bugs

• ram0 is working, ram1 is not working and (if inserted) it causes uboot-1.3 to freeze
• tftpboot has a buffer overflow, that limits the max amount of download to 2Mb

I do not have a jtag equipment on hand, and no time to fix these issues

I can also try to provide a little PPC405GP


*** give me an offer ***




also you can consider these auctions from ebay

• AMCC-POWERPC-440SPE-Evaluation-Board
• EST-SBC8240-Wind-River-Systems-Power-PC-Dev-Board-Evaluation
• Wind River CPCI evaluation Boards MIPS ppmc4KC & PPMC7400

Can You Ride Your Meta Compiler ?

According to TypeScript Language Specification, Version 1.6, August, 2015, in Microsoft they guess that you can
and TypeScript actually aims to fix issues in JavaScript by a meta-compiling technique