Be careful - Baikal-T1! Or the story of one project with an attempt to apply "import substitution"

Effective managers and component suppliers can safely ignore this material. This note is for hardware developers who are already looking, or are only going to look in the direction of the Baikal-T1 processor. So to say, “food for thought” about the real history of communication with the firm Baikal Electronics (BE) and what the result was from an applied project with an attempt to use “import substitution”.

So, everything is in order ...

Background


Our company has been developing and producing electronic equipment since 1993. and we are well known in our market. What is the element base with which we haven’t reworked for these 25 years and haven’t altered anything! Although, the truth is that they definitely didn’t work with the domestic base during this time. The last time in my memory this was back in 1987, long before the formation of our company - the best memories of Soviet analog and digital components from Kiev Kristall, Novosibirsk, Tallinn, Riga, Zelenograd, Chisinau, Leningrad, and others.

So, at the end of 2016. We conceived one product, which, among other components, should include a controlling controller, more precisely SoC (System-on-Chip). No specific requirements for speed and performance to the controller (400 MHz-1 GHz is quite enough). It is only necessary to have standard ports (1GbE, I2C, SPI, SATA, etc.) and external DDR memory up to 1GB, and also the presence of two or more processor cores is desirable. The type of processor core is not important in principle. ARM core is desirable but not necessary. Temperature range - commercial and industrial. Consumption is also not important (again, within reasonable limits), because Compared to how much the main components consume in this product, so this, as mathematicians say, is “o” -small, with a third and higher order. The main requirement, as usual, is a well-established environment for designing and debugging software under Windows, a JTAG emulator, “basic” libraries for working with a processor and peripherals in the “bare-metal” mode, and compatibility with any of the standard RTOS supplied in the source code (ThreadX from ExpressLogic, FreeRTOS, etc.) with support for IP protocols and file system. Linux is not suitable in principle and is not considered. The hardware part of the device as a whole is not very complicated, but the software is “very, very” and will take at least a year before the first release and with the documentation. The device is likely subject to further certification. The quantity is small - about 100pcs per year for the first few years.

Our application experience and software developed for Freescale / NXP processors with PPC and ARM cores, TI and Atmel processors, allowed us, in principle, not to think about choosing a controlling controller and eliminate the entire “hemorrhoids” with “basic” software, RTOS and software development environment. But the devil pulled me and pecked me then in the back some rooster ... I decided to see - “Is there anything domestic?” To the place of the controlling controller. Just at that time, there was a “ringing” about import substitution, and, just in time, I read about the first domestic Baikal-T1 processor. And spun ...

First impression


I went to the BE site, found a Baikal-T1 processor, downloaded a flyer with a block diagram. The first impression is quite suitable for the parameters. Nothing extra ordinary for 2016, but quite well for a wide range of embedded applications and without any claims to a “superfood.” Well done! If everything works as intended, great work! Good hit in the market.

I filled out an “on-line” application from the site with an indication of “who is this”, from where, and with a request to send more detailed technical information on the hardware and software parts. And, as usual, switched to routine.

It takes a week, second, third ... Silence. No letters and calls from the BE. Remembering the BE. Call yourself. It turns out that there are no chips (at the end of 2016) and they are planned only by the end of 2017. No technical documentation is available yet.

Well, let's wait ... Our product was still only at the stage of reflection. There is time.

The second impression is the last


February 2018 Finally, we have approached. The concept of the product is ready and the market is clear and ready. It's time to start designing. The idea of ​​using Baikal-T1 is still in force. I go to the site BE. Today the site is beautiful, everything blinks, moves ... The font is large. Cool! Again I filled out an “on-line” application from the site with an indication of “who is this”, from where, and with a request to send more detailed technical information on the hardware and software parts. And, as usual, again switched to the routine in anticipation of an early response.

However, a week passes, a second, a third begins ... Silence again. No letters and calls from the BE. The steepness of the site somehow begins to not correspond to the true state of affairs.

Again, recall the BE. We call. We find out with whom to talk about technical issues. All the same people - Mr. R.S. We communicate.

And here it turns out ...

CPU and demo board


The price of the chip is about $ 65 (3.7 tr.). In stock is in stock, take it even now. Fine.
I set a number of specific technical questions about the processor. I get generally adequate answers, although in some places a person “floats”, but within reason (the questions are very specific, and there is a lack of experience). In any case, it is clear that full technical documentation will be required for the work.

I ask, "Is there a test board for testing software and hardware?". The answer is, but there are some slight delays with the new revision. No question, wait a bit. But the price of the board ~ 150t.r. - this is really cool for a 3.7t.p. chip! There is a common question “Why would it?”, While test boards from Freescale / NXP, TI, Atmel in the price range of ~ $ 60 (3.6t.r.) to ~ $ 400 (23t.r.) from the manufacturer . A reasonable answer for the price, as always, no. Do not want - do not take it! I conclude that things are not going well with chip sales, so they are focusing on test boards and state orders. Indeed, in their right mind no one (or very few) will buy this fee for the price and for their hard-earned money. By the way, just the other day I learned that a training class in Baikal-T1 was opened at Moscow State University. This is of course great! But only money is again state and they have a limit and there are a lot of those who want it!

The situation is clear. Not scary yet. It will be necessary - and buy a test board. Go ahead ...

Soft


I ask "how are things now with Baikal-T1 software?" In accordance with the requirements of our task?

No way! Linux only. There is no support for either RTOS or bare-metal, nothing more. As if someone is porting QNX there to Baikal-T1, but we don’t know for sure.

Something reminds me very much of the “Soviet” times, when the processors released, gave documentation describing the system of processor commands and their codes, and then everyone else, who else can ... However, there was no competition. It was just a task to do it by any means!

In general, for us it is not a question. To which processors only, we did not transfer different RTOSs during our work. Well, they would also make a port for Baikal-T1. Would also write “base” libraries for “bare-metal” applications.

That's just a pity to your time. And even more pity to observe the flawed approach of the BE to this issue. After all, "silicone" engineers did their job well. But the chiefs, apparently, do not understand at all that the software should be dealt with in various directions (and not just Linux), and precisely in the BE itself. And it should be started much earlier than the release of the silicone itself.

In fact, all these benchmarks, the comparison of processors among themselves, the presence of one or another hardware port and other architectural things here and their absence there does not matter in principle for the vast majority of applications. It will be necessary - we will add what is missing. What is the difference between 1000MIPS and 800MIPS, or 10,000 CoreMarks and 7000 CoreMarks for a control controller? Yes, no! The real importance is only the availability of adequate supported software and development tools, as is the case for our task. And to make it quite real to the processor manufacturer. We just need to understand that this is the key point for the successful sale of processors. By the way, then the price of the test board will be quite adequate, because priorities will shift in the right direction. And here and new buyers will catch up.

Sadly But nevertheless, not critical. Go ahead ...

Software Design and Debugging Tools


I ask "how about debugging tools?". Do not believe it, but in 2018 all too: Yes, no way! Neither the design and debugging environment, "sharpened" for a specific chip with its periphery, nor the JTAG emulator from the EB is not and is not expected. They said only that somewhere, as if the environment project on Eclipse is lying, you can download it. But it is not supported, so compile and use yourself at your own risk! This is the approach! Great!

JTAG emulator as if available from the company OLIMEX (Bulgaria!). Gave a link - there is not a word about MIPS, only emulators for ARM. And here is ARM, when is the core of MIPS? Gave a link to JTAG emulators on the MIPS website. There is something like the truth. But what am I supposed to take the JTAG emulator from the USA?

I ask the question: “Where is the debugger for the JTAG emulator?” The answer is worthy - “Only GDB remains.” I didn’t understand something. There that, in the BE all buyers for morons that they think? Lower case debugger in 2018?

And I need it? Moreover, for an unknown MIPS P5600 (Baykal-T1) core.

I can only repeat the same thing as in the previous paragraph. “Effective” managers from the EB should be understood that only the availability of adequate supported software and development tools is real, and directly from the chip manufacturer itself. It is the BE company itself that needs to deliver the tested JTAG emulators and its own chip-optimized design and debugging environment, as all processor manufacturers do.

The situation becomes critical. However, we go further ...

Worthy finale


Although it was already clear that the situation was “joints”, I would probably have to completely give up the idea of ​​Baikal-T1, asking about obtaining detailed technical documentation. You need to look at a complete description of the ports and chip specifications. I especially want to take a look at the “Silicon Errata” list, which contains a list of all errors in silicone and the presence of software “patches” for them, since This is what determines the possibility of normal use of the chip for a specific task.

In order of information: most chip manufacturers have detailed technical documentation and Silicon Errata either immediately available for download from the site, or after a simple registration on the site (mainly for market research).

BE is asked to sign an NDA (non disclosure agreement, non-disclosure agreement). No question - they signed, and more than once, with different chip manufacturers. This is usually required for very complex and special chips, for which there is intense competition in terms of functionality and IP (intellectual property). Only here with a manufacturer of general-purpose processor chips - this is the first time in my memory. Nevertheless, I was sure that we would sign it without any problems and asked to send a file to the mail.

And here the ambush waited ...

In a day I receive a letter in the mail with the NDA file and the accompanying text. And it turns out that in order to sign the NDA in order to get only a description of the processor (!), The following documents must be submitted to the company in EB (hereinafter I give an exact quotation from the letter):

"...
1. A copy of the certificate of state registration.
2. A copy of the certificate of tax registration.
3. A copy of the charter, certified by the head.
4. Protocol or decision on the appointment of a director.
5. A copy of the notice of the application of the simplified taxation system (if applicable).
6. Power of Attorney and Representative Contact Information.
7. Company card.
... "

I remember this unforgettable moment when I read it. Just like in the BMW advertising slogan - “Delight ...”. Highest "Respect" to the company BE with its legal service! In this, she went too far and surpassed all who could and could not, in the matter of signing the NDA to obtain documentation for a general-purpose processor. Apparently, the lawyer from the bank was taken from the banker and he was obviously confused that we (the buyers) want to get something adequate for their money from the banker, and not vice versa. But life certainly will fix it, and I hope that soon.

All that I was capable of at this unforgettable moment is to write a response letter to the BE with the modest question “Can you still have the keys to the safe where the money is?” And with the proposal to forget each other like a bad dream.

“Dry” residue


Full "paragraph"!

What comforted?


I had to return to the good old ARM-platform hardware, for which we have everything for a long time and works great. Application software for our product is a key point, and there is a lot of it. And it is completely useless to take on solutions to minor and other people's problems with development and debugging tools, and possibly with unknown hardware.

I just wanted the extreme with the use of "import substitution" ...!

P.A. Semenov, Ph.D., "Microlab Systems" (Moscow)

Source: https://habr.com/ru/post/412077/


All Articles