Huawei Linux Laptop Driver Improvements On The Way

[oiaohm]

And now you are talking about something completely different! SiFive is an American company that designs custom computer chips for others. That's right, theirs core designs are not free. You have to pay for the license. And Huawei is unable to buy it, because of U.S. sanctions. Moreover, they don't even offer cores that would be comparable to high-performance ARM cores, like ARM Cortex-A76 or upcoming ARM Cortex-A77.

I never said the core designs are free. But the cores SiFive offer in their custom chips are not all theirs. Its a good place if you have a login to see what is on offer.

Of course, there are royalty-free RISC-V cores, but they are totally crap, at least for smartphones.

You cannot be sure of that. RISC-V royallty free have not been made at 7nm mostly 22nm or larger. At the same nm running the same workloads risc-v benchmarks out using 40% less power than the arm chip. So not total crap for smartphones as swapping to risc-v would increase battery life by quite a bit.

The biggest barrier is the software stack. Followed by lack of parties interested to invest the money to make risc-v work at the high end nm. This is why arm position is not as solid as it first sounds. There is a unrestricted alternative that can perform very well.

This GPU is totally crap, completely unsuitable for smartphones. Do you even read what you refer to?! [email protected] does not even provide smooth video playback (BTW: you can forget about H.265 hardware decoding!), not to mention the fact that most smartphones provide HiDPI display! And performance is just terrible. 5~6 GFLOPs - Do you even know what it means?! Adreno 640 has 898.56 GFLOPS in FP16 and 449.28 GFLOPS in FP32! Maybe 10 years ago it wouldn't look so bad, but today it only makes people laugh.

Maybe you miss the basic problem here. You say 10 years ago it would not look so bad. Risc-v gpu will be 28nm so 2011 8+ year old nm of production. Even so it power usage per GFLOPS is better than the Adreno. That GPU is not totally crap.

Here is a horrible question how fast is Adreno 640 when its doing FP64 that right it does not support that. This risc-v gpu design is a 64FP and 64bit Integer design. That 5-6 GFLOPs on a single processing core of a GPU is not bad once you 1 wake up its only a single processing core 2 its old nm and it is FP64.

Adreno 405 is what you would have to compare that risc-v prototype against and then allow that this has more silicon are and cores and is only FP32.

Big thing about this first chip is to provide the RISC-V graphics ISA with a test platform to develop drivers against and validate the ISA design. Once that is done how fast it can be scaled up will be a interesting question mostly likely totally limited by budget/investment.

So please, stop wasting people's time with these nonsense statements.

No I told you to get the specifications of those base stations not marketing stuff. There is a reason why Huawei has been maintaining their own OS risc-v compatible.

Did you miss that Arm and Risc-v can use compatible busses. The question is being answered very carefully because that chip was not just arm its arm/risc-v hybrid.

Risc-v is something interesting in this regard with arm it does not have to be all or nothing. Yes it possible to do like arm big/little but instead arm be big and risc-v be the little or reversed. This makes that basestation chip of Huawei very interesting.

[the_scx]

I do know what I am talking about. https://www.sifive.com/ These guys have been doing custom risc-v for a while.

No, you don't. You said that most things could be transferred from their current SoC ("more of their existing soc design parts could be transferred to risc-v cores."), which is totally bullshit. First of all, the core design belongs to Arm Holding, not Huawei. Secondly, the most important part of their SoCs are Cortex-A cores with Mali GPU.
And now you are talking about something completely different! SiFive is an American company that designs custom computer chips for others. That's right, theirs core designs are not free. You have to pay for the license. And Huawei is unable to buy it, because of U.S. sanctions. Moreover, they don't even offer cores that would be comparable to high-performance ARM cores, like ARM Cortex-A76 or upcoming ARM Cortex-A77.

Of course, there are royalty-free RISC-V cores, but they are totally crap, at least for smartphones.

https://riscv.org/2019/02/phoronix-a...being-plotted/
2020 we should start seeing the Risc-V low power GPU start turning up.

So in 2020 Cortex-A core designs, as well Mali GPU get swapped with a full feature Risc-v and Custom feature Risc-v GPU comes highly likely this would be a core using open source hardware that is not restricted by government sanctions.

This GPU is totally crap, completely unsuitable for smartphones. Do you even read what you refer to?! [email protected] does not even provide smooth video playback (BTW: you can forget about H.265 hardware decoding!), not to mention the fact that most smartphones provide HiDPI display! And performance is just terrible. 5~6 GFLOPs - Do you even know what it means?! Adreno 640 has 898.56 GFLOPS in FP16 and 449.28 GFLOPS in FP32! Maybe 10 years ago it wouldn't look so bad, but today it only makes people laugh.

Also it gets worse I pointed to Huawei involvement with Risc-v. If you get a chance to see the features of the Huawei made 5G towers you will notice something interesting custom risc-v SOC chips inside. So Huawei already has teams in different department making risc-v based hardware so making a Risc-V SOC chip for a phone they would not be starting off with zero experience as they would be able to start with the personal working on the 5G towers

https://www.capacitymedia.com/articl...-base-stations

Hence today’s event, at which Huawei launched what it calls the industry’s first 5G core chip, called Huawei Tiangang, claiming breakthroughs in integration, computing power, and spectral bandwidth.

In its presentation in Beijing today, Huawei suggested that the Tiangang chip uses intellectual property from UK-based design company ARM, now owned by SoftBank.

These [ARM chips] are used in the Chinese company's 5G base stations and computer servers in addition to its smartphones.

So please, stop wasting people's time with these nonsense statements.

[oiaohm]

Tell me one thing: does someone pay you for trolling or are you just doing it for fun?


Again, you have absolutely no idea what you're talking about.

I do know what I am talking about. https://www.sifive.com/ These guys have been doing custom risc-v for a while.

memory (DRAM) controller, point-to-point high-speed processor interconnect or another type of fast link, GPU/DSP and audio processor, IO extension

Restricting to non USA sources. Only bit in that list that is any form of problem is the GPU/DSP and its not absolute stuck you have horrible PowerVR. That is without using the Huawei IP there cpu already have their own custom memory controller.... Yes and a design for arm cores and risc cores use the same on chip interconnect standard to those parts as well.

And to make it competitive with hi-end ARM, you have to develop your own core design and instruction set extensions (SIMD extensions, like SSE/AVX in x86, AltiVec in Power or NEON in ARM, virtualization extensions like VT-x or AMD-V, security/crypto extensions for RNG, MD5, SHA/SHA2/SHA3, SM3/SM4, AES, etc.).

Again you don't need to design your own core there is a core that is already designed that is open source hardware.
1) SIMD extensions, like SSE/AVX in x86, AltiVec in Power or NEON in ARM=Risc-v Vector.
https://content.riscv.org/wp-content...uctions-v2.pdf
Yes the bugger is very competitive.

2) Hypervisor instructions are part of risc-v instruction set. So VT-x and AMD-v again are covered.

3)security/crypto extensions for RNG, MD5, SHA/SHA2/SHA3, SM3/SM4, AES,
https://fuse.wikichip.org/news/1068/...iot-processor/

There are already open source IP for all that stuff for risc-v. Yes this stuff plugs into the risc-v chisel generation process. Reality here it possible to design and make a high end risc-v chip with luck in under 2 months yes my 6 months is to allow for a few failures.

Reality here as a instruction set risc-v is competitive with high and arm and x86.

Basically if you were going off the shelf today.
This Cortex-A core designs, as well Mali GPU get swapped with a full feature Risc-v and PowerVR GPU feature wise there is no difference.any IP part that you use to connect to a Cortex-A design will connect to a full feature risc-v core.because the full feature ones are designed as drop in for arm cores for people who start off with a soc designed for arm and then find they cannot pay the extra IP cost.

Arm location is not as solid as you are making out or as hard to replace. The biggest problems will be software side not hardware side. Forcing a company with size to fix up the software side will not do Arm any good long term.

The reason why I said they had to do it this year is not the CPU the CPU of arm was replaceable by Risc-v designs Last year..
https://riscv.org/2019/02/phoronix-a...being-plotted/
2020 we should start seeing the Risc-V low power GPU start turning up.

So in 2020 Cortex-A core designs, as well Mali GPU get swapped with a full feature Risc-v and Custom feature Risc-v GPU comes highly likely this would be a core using open source hardware that is not restricted by government sanctions.

Also it gets worse I pointed to Huawei involvement with Risc-v. If you get a chance to see the features of the Huawei made 5G towers you will notice something interesting custom risc-v SOC chips inside. So Huawei already has teams in different department making risc-v based hardware so making a Risc-V SOC chip for a phone they would not be starting off with zero experience as they would be able to start with the personal working on the 5G towers

Basically the current barriers has been software stack and gpu. In a year the problem caused by blocking Huawei would have been a software stack only problem.

Remember my reason for suspecting Huawei will look risc-v is that this means they will not be starting from scratch in fact able to reused what they already have developed for other departments.

You can say that it should be possible to transfer experiences from developing other CPUs, like MIPS (Loongson), x86 (Zhaoxin) or Shenwei (Sunway).

I would say you are right for the ones you listed. As Loongson, Zhaoxin and Shenwei are not design to interface with IP designed for Arm core designs or even each others. Risc-v and Arm SoC designs can use the same third party IP blocks. So at the silicon design level would require a lot more work having to test out new IP parts going to any one of those 3 you listed. If Huawei goes Risc-v they can use what they are using in the 5G towers and any third party IP blocks they were using with the arm cores in phones that are not effected by the USA blockade. Yes the 5G towers are requiring some very fast chips and highly power effective.

Its simple to forgot Huawei is diversified in CPU usage. Huawei has been using 3 different cpu types. Arm, x86(from intel) and risc-v. If they go bugger it you will expect focus on risc-v at least in the short term as this requires the least amount of staff training due to existing experience in other effected departments.

Think about it how effect will sanctions be if they end up fully on Risc-v and other license fee free technology. Yes this could make Huawei products more cost effective long term due to not having to pay as many license fees. So short term pain long term gain for Huawei is totally possible.

[the_scx]

I don't want something that is Android based because it's a unnecessary fork of GNU/Linux.

Android is definitely not a fork of GNU/Linux, whatever it means (the "fork" part). It doesn't even have glibc, nor does it support the full set of standard GNU libraries. Android's standard C library, Bionic, was developed by Google specifically for Android.

Please keep in mind that AOSP is an open source project, so according to your words, you have to love it, unless you are freedom hater.

For many People Google play service is a big anti-service

The market has already verified that there are not too many of them. Currently, nobody has a chance to become a leader in the smartphone market without Android with Google Play services (well, Apple would have a chance when it comes to profits, but that's not the point, because nobody but apple can sell smartphones with iOS).

Also could you explain me why they could not design new socs? Isn't that only a legal problem? Are they technically not able to do it? Does Arm have some special computers / Programs that are the only one on the world with what you can design socs? Please educate me, but as far as I understand the only thing that hinders them to make new design is intellectual property laws?

1. This is a legal problem, because Arm Holdings sells licenses for both ISAs (e.g. ARMv8-A, ARMv8.1-A, ..., ARMv8.5-A) and core desings (i.e. Cortex-A w/ Mali).

ARM instructed employees to halt "all active contracts, support entitlements, and any pending engagements” with Huawei and its subsidiaries to comply with a recent US trade clampdown.

2. This is also a technical problem, because without access to new core designs (i.e. Cortex-A77 & Mali-G77 Valhall), Huawei will be unable to put them in their new SoCs.

In other cases, manufacturers license ARM's processor core designs - which describes how the chips' transistors should be arranged. These blueprints still need to be combined with other elements - such as memory and radios - to create what is referred to as a system-on-chip.

Please note that Huawei has literally zero experience in creating their own ARM cores. This apply as well to the GPUs.

Please keep in mind that currently HiSilicon uses the Cortex-A core designs, as well Mali GPUs, licensed from ARM Holdings. They didn't developed their own core designs, nor GPUs for their SoCs. It applies as well to any Chinese company on ARM market.

3. Because of U.S. sanctions, they will be unable to buy chips from others (i.e. Qualcomm).

Please educate me, but as far as I understand the only thing that hinders them to make new design is intellectual property laws? Which china could declare void for Arm as part of the retribution of this trade war that the US started.

What is the point of producing something that you can not sell (at least outside of China)?

[blackiwid]

We already have Tizen, Sailfish OS, Ubuntu Touch, KaiOS, etc. And PureOS will join this group pretty soon. What is more, we also have AOSP and several systems based on it, such as LineageOS and its forks. So tell me, why do you need yet another mobile Linux platform or Android distribution? Why these systems are not enough for you?

I don't want something that is Android based because it's a unnecessary fork of GNU/Linux. If they then take Pure os and rebrand it or they create their own PureOS I don't care to much like I don't care which of the desktop GNU/Linux distros you use because they are are all app compatible, not package format compatible but that is no real big issue, at least for the opensource world and I don't care about proprietary Software developers if they get some android compatibility running fine if not not... my only point is A they would likely be able to do something them self if it's android without play store or something else is a different question and B that I hate android so there are people that would want a big alternative to Android ideally they would rebrand PureOS but heck even some sort of Google free Android would be a big selling point. For many People Google play service is a big anti-service not to mention that it eats 80% battery live if installed but also nonstop tries to force you to send google all data. Asks you 1 trillion times if you want to activate it and if you mistakenly press one time yes then all your data is send to this motherfuckers.

Also could you explain me why they could not design new socs? Isn't that only a legal problem? Are they technically not able to do it? Does Arm have some special computers / Programs that are the only one on the world with what you can design socs? Please educate me, but as far as I understand the only thing that hinders them to make new design is intellectual property laws? Which china could declare void for Arm as part of the retribution of this trade war that the US started. Not that they would have to give a good reason as dictatorship for it, but they also could call that a nationally security issue because else they can't supply themself with hardware/software that has no american spyware on them.

[Weasel]

Tell me one thing: does someone pay you for trolling or are you just doing it for fun?

He's obviously paid by Red Hat, his bias is too clear here.

[Weasel]

Canonical with Ubuntu Core is a snap store. There is a problem. We are talking IOT snaps are not known for their disk and memory effectiveness.

Man, you are such a Red Hat shill, my god. Flatpak, Wayland, GNOME, systemd, freedesktop, etc etc... all of them you praise and knock down the competition. It's laughable how visible your bias is.

How much do they pay you?

[the_scx]

Tell me one thing: does someone pay you for trolling or are you just doing it for fun?

Not at all more of their existing soc design parts could be transferred to risc-v cores. So would not be a new SoC from scratch. In fact it possible todo a new risc-v SoC from base parts in about 4 months. So with what Huawei has it could be a little as 6 months before chips start appearing.

Again, you have absolutely no idea what you're talking about.

You get free ISA, but everything else you have to develop literally from scratch: memory (DRAM) controller, point-to-point high-speed processor interconnect or another type of fast link, GPU/DSP and audio processor, IO extension, etc. And to make it competitive with hi-end ARM, you have to develop your own core design and instruction set extensions (SIMD extensions, like SSE/AVX in x86, AltiVec in Power or NEON in ARM, virtualization extensions like VT-x or AMD-V, security/crypto extensions for RNG, MD5, SHA/SHA2/SHA3, SM3/SM4, AES, etc.).

Please keep in mind that currently HiSilicon uses the Cortex-A core designs, as well Mali GPUs, licensed from ARM Holdings. They didn't developed their own core designs, nor GPUs for their SoCs. It applies as well to any Chinese company on ARM market. So you have literally nothing here and expecting some miracle. Don't you thing that this is impossible, at least in near future?

You can say that it should be possible to transfer experiences from developing other CPUs, like MIPS (Loongson), x86 (Zhaoxin) or Shenwei (Sunway). However, I don't believe that. It just doesn't work like that. If that were the case, Intel and IBM would have created their own ARM chips a long time ago.

I'm not going to respond to your pointless statements anymore, so don't even try to answer.

[the_scx]

1. I want another mobile platform as developer and user so your assumption about "nobody" is wrong and I am not the only one.

We already have Tizen, Sailfish OS, Ubuntu Touch, KaiOS, etc. And PureOS will join this group pretty soon. What is more, we also have AOSP and several systems based on it, such as LineageOS and its forks. So tell me, why do you need yet another mobile Linux platform or Android distribution? Why these systems are not enough for you?

Everybody that likes free software must want this because Android is only 99.9% proprietary apps that are therefor useless garbage.

It isn't true. The fact that someone likes free/open software does not necessarily mean that he is eagerly waiting for yet another Linux distribution.

why would Huawei have to care about a license from Arm?

Because without it they will be unable to design new SoCs. And because of the U.S. ban they are unable to buy chips from i.e. Qualcomm. Anyway, if sanctions persist, in the next 2-3 years they will not be able to sell their products in the Western world (United States, Mercosur, Commonwealth, European Union). This may apply as well to markets of Japan, South Korea, and ASEAN.
What is worse, the Trump's ban means that you can either make business with American companies or with companies on the U.S. blacklist (if it concerns US origin technology), but not with both of them. That's why it has a global reach. Almost nobody wants to make deals with Huawei, because it means that they will be unable cooperate with U.S. companies (or at least there is such a risk).
Please keep in mind that Arm Holding stated that their core designs (Cortex-A) contained "US origin technology", so even MediaTek shouldn't be able to sell chips to Huawei, at least in theory.

A similar situation took place in the case of sanctions against Iran. European companies wanted to do business with Iran, but they had to withdraw, because otherwise U.S. sanctions would affect them too.

As you said they can use google and that's all you need, blackberry did it pretty well for a long time and that's absurd small company that had 20 Billion dollar turnover

BlackBerry was doing well before iOS and Android appeared on the market. Since then, it was just a slippery slope.

If Samsung can do that again why would Huawei not be able to do just what Samsung did? They replaced the store and other apps?

You didn't listen carefully. This is not about some pre-installed apps. This is about hundreds of thousands or maybe even millions of apps that depend on Google Play Services.
Samsung has never got rid of Google Play Services on its own Android devices. Amazon tried to do this and failed so badly that they never tried this again on the smartphone market.

But that said I would also not wonder if Trump again just pretended to do something strong.

It is obvious to me that Trump simply wants to make a new trade deal between the US and China. This deal will be more beneficial to the US than China. Unfortunately, China has no good options here. Either they accept this, or Huawei will be out of business. They can also escalate this conflict, threatening to cut off supplies of rare earths to U.S. companies, but no one knows how it will end.

[oiaohm]

The reality here is that you have no idea what you're talking about. You still don't distinguish between basic embedded systems and complex IoT solutions. Do you want to create an app store for all your devices? If you go into Canonical ecosystem, you could easily create your own branded IoT app store, hosted by Canonical. With Debian, you have to build everything on your own - the entire infrastructure have to be created from scratch.

Canonical with Ubuntu Core is a snap store. There is a problem. We are talking IOT snaps are not known for their disk and memory effectiveness.

https://endlessos.com/ There is the endlessos solution they are a debian base. Yep flatpak instead of snap. Nothing really special to Ubuntu core.

Because they are magicians and will develop a new SoC literally from scratch in one year!

Not at all more of their existing soc design parts could be transferred to risc-v cores. So would not be a new SoC from scratch. In fact it possible todo a new risc-v SoC from base parts in about 4 months. So with what Huawei has it could be a little as 6 months before chips start appearing.

Cyrix went into "military/government" solutions under the leadership of National Semiconductor. Please remind us where they are now...

Do more research they got acquired by VIA because they were successfully still making new product. Not all those who go after the "military/government" when they are in trouble die. IBM with power platform is good example of go after "military/government" and stay in the game. This could give decades to refactor.

But developers can sell games on their own, right? This is exactly the same difference as between Debian and Ubuntu Core in IoT.

That is not a difference. Ubuntu Core Store can be selling to Debian devices with snapd installed. So the most important fact is Debian. If you have your own copy protection system you can sell your game in flatpaks as well.

And you can get the same Steam client from Arch or Gentoo repo as well. It doesn't mean that Steam games are developed for Arch or published on Arch.

If you have not noticed that Microsoft has started releasing games they make on Steam. You have to have people trust you store for them to buy product from it.