Examining Soft Machines' Architecture: An Element of VISC to Improving IPC
by Ian Cutress on February 12, 2016 8:00 AM EST- Posted in
- CPUs
- Arm
- x86
- Architecture
- Soft Machines
- IPC
Dealing with Guest ISAs and a Translation Layer
Going back to this architecture diagram, everything up to the global front end is another interesting story as well.
Part of Soft Machines' product package is a low level virtual software layer that will translate a guest instruction set and convert it into the VISC ISA. This is to allow VISC to be used with existing software, and to more easily integrate into current environments rather than trying to establish an ecosystem for a new architecture in 2016. Soft Machines tells us that two instruction sets are supported, one of which will be ARMv8. It was implied that x86 would be the other, although they were reluctant to outright confirm it (ed: x86 translation is likely not to be looked upon fondly by Intel). Meanwhile we were told that writing additional translation layers, while not trivial, can be done and that they plan to support other guest ISAs in future.
So for all intents and purposes, this is a translation layer converting from ARMv8 to VISC. Many companies over the past couple of decades have tried with translation layers – Intel with Itanium, Transmeta to x86, and one of the latest was NVIDIA with Denver, which translated ARM to a custom ISA. Mentioning Itanium, Transmeta and Denver, for those who have followed the industry, might bring a chill down the spine given the very limited success each of these platforms have had. Soft Machines’ CEO was keen to point out that the purpose of the translation layer for VISC is very different to these previous attempts.
The VISC translation layer is designed to be a thin and lean implementation whose main role is to maintain compatibility to the VISC ISA, not to extract performance. Taking Denver as the most recent example, the translation layer there is designed to adjust the ARM instructions into Denver’s ISA and extract instruction level parallelism into the 7-wide design. For VISC, we are told, there is no need to go after performance at this level. The main point at which the VISC design increases performance is at threadlet generation, not in translation and making instruction sequences better fit the VISC hardware. This allows the ARM translation layer to have a less than 5% overhead, according to Soft Machines, and releases a point of contention with previous translation layer designs. As long as the translation layer is 100% compatible, the performance can in principle be extracted at the threadlet level.
This also means, again according to Soft Machines, that any specific compiler enhancement offered by others can also be used when translated. We put it to them that in the case of x86 certain codes are accelerated better on Intel’s compiler than say GCC (a question that arose out of the results we’ll go into later), and we were told that those instruction enhancements by ICC should translate well into the VISC ISA after going through the translation layer.
We asked about the VISC ISA, but were told that more information about this and the core design would be released at a later date as designs progress. We were told that it is a relatively small ISA (as to us sounds like a RISC, which is easier to extract ILP at lower power) with smaller instructions in comparison to ARM and x86. I would assume that this means they are fixed length, but this was not confirmed.
Facebook
Twitter
RSS
97 Comments
View All Comments
extide - Friday, February 12, 2016 - link
Because a compiler can only schedule instructions to the CPU's front end. This is scheduling of instructions to different ports on the back end of the cpu. The compiler can't tell the CPU what port an instruction goes down, the CPU picks that. THe compiler only gets to pick what instructions are issues, and in what order, and of course, modern CPU's can even change that order if they deem it faster to do so.Exophase - Friday, February 12, 2016 - link
To have any realistic chance of working the threading speculation/detection has to have a large dynamic component (detecting threadlets as they become desirable at runtime) and has to have architectural support for very lightweight thread splitting, merging, and inter-thread communication.That can't be provided by compilers targeting existing instruction sets.
AlexTi - Friday, February 12, 2016 - link
Thanks, I think I got the point finally. This looks similar to what instruction scheduler currently does for execution units in conventional CPU. Virtualization layer + CPUs will be a kind of very wide core. Right?It was already noted in the article, making curent CPU wider is problematic and not universally beneficial. So this new engine should be much more efficient than current implementations.
Good thing is that we'll see eventually :)
Senti - Friday, February 12, 2016 - link
Bullshit. I have no idea how technically incompetent writers should be to reprint that marketing nonsense again and again.First of all, this brings absolutely no advantages over existing fat core + SMT concept. More IPC per core with more pipelines is not done because it's hard to do without that 'virtual cores' nonsense, but simply because there are not enough actually independent instructions that can be automatically extracted from real code during parts where performance matters.
"Alternatively, if multiple programs or threads want to use the hardware, then a single core is inaccessible to additional threads while the first thread is still in use (though this can be avoided somewhat by simultaneous multithreading or SMT which will let another thread have access when the first has encountered a stall such as waiting for L1/L2 memory)." - total lies. That describes coarse-grained multithreading which is not very popular atm. For example, Intel HT allows usually 2 threads to execute simultaneously dynamically sharing pipelines of the same core all the time. POWER8 uses 8 'virtual' threads per core.
Why no one splits instructions from the same thread over several cores (other than the obvious reason that there are not enough independent instructions to split)? Almost quote from the text: "cross-core communication adds latency and reduces performance".
Instruction set emulation? Far from new concept. Why not popular? Reason is very simple: significant overhead. Try translating something non-trivial like AVX/NEON instructions to some generic internal instruction set.
Finally, the last point: everyone can draw cute performance graphs and huge numbers in marketing presentations, but how about giving actually working chips for independent reviews of performance and power efficiency on real code?
vladx - Sunday, February 14, 2016 - link
Skim the article again, there's a roadmap so let's see how things will go from here.Exophase - Friday, February 12, 2016 - link
There's another big question with their power measurements. They take differences between idle and 1C and 1C and 2C to cancel out the static contribution of other peripherals. But this still ignores the dynamic contribution.For example, we can look at Cortex-A72, where ARM claims that one core at 2.5GHz on the TSMC 16FF+ process will consume about 750 mW. In Kirin 950, the power consumption appears to be about 900 mW at 2.3GHz. Is ARM exaggerating or is Huawei's implementation inferior to ARM's expectations? The discrepancy can actually be pretty easily explained by losses in the PMIC/VRMs, the SoC's memory controller, and the DRAM - all components which use more power the more the CPU load increases.
This is especially a factor for wall measurements because they take into effect an additional AC/DC convertor. While it's possible that Soft Machines included these figures in their power estimations I doubt it since they didn't mention it, and like ARM it's more practical and beneficial for them to work with core power estimations only.
So there could easily be another 25+% that the non-VISC platforms are being penalized.
Something else that raises a red flag to me is the 16FF+ test chip. There are only 100 pins. When you take out power, ground, and various control signal are accounted for that leaves a very small interface either to a memory controller or memory (if the controller is integrated). Even a single channel 32-bit interface would be a hard fit. So does this chip really represent both realistic power consumption or realistic performance? I think they're trading one for the other on this one and that makes me question the applicability of the power numbers they've given for it.
Arnulf - Friday, February 12, 2016 - link
Since one cannot buy these "scaled" chips, IMHO it'd make more sense for SMI to publish performance per watt figures of real hardware and let the market decide whether their concept is attractive enough. Yes, Intel may have process node advantage, yes, different CPUs are targetting different performance and power profiles but at least it's a straight comparison and if VISC doesn't beat its entire competition at at least one metric then it's destined to fail anyway.Oh and the remark in the article regarding "VISC advantage" because of it using twice the number of cores while running a single thread in tests - who cares as long as it comes out on top in performance per watt? If they can beat other CPUs by using more cores, kudos to them!
ppi - Saturday, February 13, 2016 - link
Regarding core count, I would direct you to recent AT article on Android usage of multiple cores. Simplified conclusion may be, that Android tends to utilise 4 cores pretty well.In real world, this significantly reduces impact of distributing single thread over multiple cores.
kgardas - Friday, February 12, 2016 - link
Interesting stuff, but to be honest, combining "simpler" cores into more complex is also done by software on SPARC64. At least Fujitsu mentions this on some of their hotchip presentation for SPARC64 VII. So you have 4 cores CPU with 4-wide core and you can combine this by software (compiler) into 8-wide or more depending on your needs for instruction parallelism.Another thing is that something like that is IIRC also supported by POWER8 where you do have a lot of duplicated resources, but not enough so in case 1 thread is able to consume all core resources you may switch-off 7 others. IIRC IBM's compilers contains some optimizations for this too.
Pity think you have mentioned Itanium only in this negative way. Honestly speaking Itanium design was really great and really pity that Intel stopped developing it and not provided any OoO designs on this architecture. If Denver will be successful we will see, but NVidia still counts with it for some designs which may be interesting in a light that they are using ARM's core (A57) for some time now and don't need Denver that much. Also automotive does not care if Denver is there or not yet nVidia pushes it there so I would bet they needs to have really good reason for it. Perhaps their VLIV is good for some special tasks...
So to me whole this looks like they are on another round for money.
Oxford Guy - Friday, February 12, 2016 - link
"Honestly speaking Itanium design was really great and really pity that Intel stopped developing it"The market disagreed so, if you're right, it's a pity the market dictates product success to such a degree.