> [coined in a paper by T.H. Myer and I.E. Sutherland On the Design of Display Processors, Comm. ACM, Vol. 11, no. 6, June 1968)]
> Term used to refer to a well-known effect whereby function in a computing system family is migrated out to special-purpose peripheral hardware for speed, then the peripheral evolves toward more computing power as it does its job, then somebody notices that it is inefficient to support two asymmetrical processors in the architecture and folds the function back into the main CPU, at which point the cycle begins again.
> Several iterations of this cycle have been observed in graphics-processor design, and at least one or two in communications and floating-point processors. Also known as the Wheel of Life, the Wheel of Samsara, and other variations of the basic Hindu/Buddhist theological idea.
This applies to all types of coprocessors and hardware offloading. But it's pretty funny to think that the effect is first observed in GPUs, and today the main player is still the GPU. The history is already established, 50 years ago.
In the 90s Auspex sold fileservers that had lots of specialty offload processors. This was before modern GPUs. Then the advantage of functional processors disappeared, and Auspex with it. Later came RDMA, which essentially is akin to having some functional processing in NICs. Anyways. This has been happening for a long time. But CPUs are maxed out for now, I think, so I'd expect that the current cycle will be a long one.
> But CPUs are maxed out for now, I think, so I'd expect that the current cycle will be a long one.
CPUs in the future may be a mixed bag, there may be different coprocessors, but integrated in the same physical chip - more advanced SoC. At least this picture is what I find to be the most convincing after the end of exponential scaling by Moore's Law.
Yes. It is already an issue of AVX2, which has the hilarious problem that the user must try finding a net performance gain between speeding things up and heating the CPU into thermal-throttle... But I think if an on-chip coprocessor can offload a common task and making it more energy efficient, the thermal issue can be avoided. We've already seen successful applications of cryptographic instruction sets / coprocessors, video transcoding, and the number can only get bigger.
”As we have said, we know of no remote display in which the computer and display channel are integrated into one machine, i.e. exactly one turn around the wheel.”
The later ZX80 was an extreme example, I think. It also supported their claim that ”Having one processor would be cheaper”, but that partly was because it was such a poor display processor.
What's interesting is how fast the evolution is compared to the decades in the Jargon file. The other interesting bit is how bandwidth and RAM costs would change the architecture today.
http://catb.org/jargon/html/W/wheel-of-reincarnation.html
> wheel of reincarnation
> [coined in a paper by T.H. Myer and I.E. Sutherland On the Design of Display Processors, Comm. ACM, Vol. 11, no. 6, June 1968)]
> Term used to refer to a well-known effect whereby function in a computing system family is migrated out to special-purpose peripheral hardware for speed, then the peripheral evolves toward more computing power as it does its job, then somebody notices that it is inefficient to support two asymmetrical processors in the architecture and folds the function back into the main CPU, at which point the cycle begins again.
> Several iterations of this cycle have been observed in graphics-processor design, and at least one or two in communications and floating-point processors. Also known as the Wheel of Life, the Wheel of Samsara, and other variations of the basic Hindu/Buddhist theological idea.
This applies to all types of coprocessors and hardware offloading. But it's pretty funny to think that the effect is first observed in GPUs, and today the main player is still the GPU. The history is already established, 50 years ago.