Alibaba T-Head TH1520 RISC-V CPU & A Few New Arm SoCs Ready For Linux 6.5

[phoronix]

Phoronix: Alibaba T-Head TH1520 RISC-V CPU & A Few New Arm SoCs Ready For Linux 6.5

The Arm (and RISC-V) SoC updates have been submitted for merging to the Linux 6.5 kernel. Additions this cycle include an exciting RISC-V processor now supported, NVIDIA Tegra234 "Orin" upstream additions, and other new SoCs and devices/boards being upstreamed...

https://www.phoronix.com/news/Linux-6.5-SoCs

[morydris]

[QUOTE][Alibaba T-Head TH1520 support is arguably most exciting of these new pull requests, but it remains to be seen if this higher-performing RISC-V SoC will work its way to western markets./QUOTE]

Eh what? I’ve had it running for months as have many in North America. It’s not exactly difficult to get.

Also, it runs at 1.85 GHz. 2.5 might be possible but I don’t know anyone running more than stock.

[coder]

Alibaba T-Head TH1520 support is arguably most exciting of these new pull requests, but it remains to be seen if this higher-performing RISC-V SoC will work its way to western markets.

Eh what? I’ve had it running for months as have many in North America. It’s not exactly difficult to get.

Also, it runs at 1.85 GHz. 2.5 might be possible but I don’t know anyone running more than stock.

And how does it perform? Based on the specs, I'm going to guess it's a little slower than a Pi 4. Maybe on par with the quad Cortex-A57 in a Jetson TX1?

https://www.t-head.cn/product/c910?lang=en

[tuxd3v]

Cool..

I suspect it to be a bit slower than Cortex A72, clock per clock, but at 2.5Ghz it will be maybe faster.
However...building it for that specification is more costly, and probably it will not be built for it..so expect only 2Ghz..in my opinion.

I have some questions

The floating point unit says that its fully in compliance with IEEE-754, but it doesn't support exceptions
What's happen when I divide a number by zero?
It should launch an exception and terminate the program..

morydris
can you code a sample program with something like this:

#include <stdio.h>
int main( int argc, char** argv )
{
        double a = 1.0 / 0;
        double b = -1.0 / 0;
        double c = 0.0 / 0;​

        return 0;
}

then compile it, and run it
a should be +inf
b should be -inf
c should be NaN

But my point is, how the fpu detects that is dividing by zero, if it doesn't do exceptions?
Its a curiosity, for sure they found a way to solve it.

One other question I have is, about the V extension, does it has one?if so, its the ratified version?

For what I read about the ROMA laptop, it will have 2 versions, one less expensive, and a more performant one..
I presume Xiangshan 2.0 will be used for that...they are taping out now...so soon we will have some news about that opensource core( with better performance than cortex A76 ).

[coder]

The floating point unit says that its fully in compliance with IEEE-754, but it doesn't support exceptions
What's happen when I divide a number by zero?
It should launch an exception and terminate the program..

You can get a QNaN, instead.

https://en.wikipedia.org/wiki/NaN#Quiet_NaN

But my point is, how the fpu detects that is dividing by zero, if it doesn't do exceptions?

Didn't you already answer that? A FPU exception interrupts program flow, whereas a QNaN just affects dataflow. At some point, you have to check for any NaNs via isnan(), but you can defer that until several operations after the NaN is first introduced.

https://linux.die.net/man/3/isnan

Not supporting FP exceptions has an obvious benefit for complexity and efficiency. GPUs don't support them, for instance.

[tuxd3v]

Not supporting FP exceptions has an obvious benefit for complexity and efficiency. GPUs don't support them, for instance.

I agree that for complexity yes its easier, smaller footprint for the die and so on, and RISCV has a long road to go..

I think its better to have exceptions implemented in the FPU... its a security mechanism.. amd64 and afaik arm implement them too.
I didn't knew GPUs skipped them..but for the price they cost...they shouldn't!

[jacob]

What's exciting about the T-Head TH1520? Another hilariously overpriced chip with performance that would have been competitive 20 years ago?

[ayumu]

What's exciting about the T-Head TH1520? Another hilariously overpriced chip with performance that would have been competitive 20 years ago?

That chip has boards that are cheaper than Raspberry Pi 4 or 400, and is faster than them.

Regardless, it isn't meant for you, but for developers who want to help get the RISC-V software ecosystem ready for the masses.

[GreatLord]

I did check the spec from aliexpress it say it is BXM-4-64 GPU, and was dispointed over the gpu speed only 256 GFLOPS and 16 gigapixels per second.
The speed info are from theregister ( https://www.theregister.com/2020/10/...ation_b_series ).

so the gpu is very week and can not be use for heavy calc, But it is faster what rasbery pi gpu rpi 4 gpu is 13.5 GFLOPS

[coder]

I did check the spec from aliexpress it say it is BXM-4-64 GPU, and was dispointed over the gpu speed only 256 GFLOPS and 16 gigapixels per second.
The speed info are from theregister ( https://www.theregister.com/2020/10/...ation_b_series ).

so the gpu is very week and can not be use for heavy calc, But it is faster what rasbery pi gpu rpi 4 gpu is 13.5 GFLOPS

Good info, but I don't really see the problem. Yes, speed is nice & important, but it really isn't the reason to be buying this SoC. The main reason is as a development vehicle, and for that it's fast enough.