3 ms·

> How would hardware emulation even work? Like this: https://en.wikipedia.org/wiki/Transmeta_Crusoe https://en.wikipedia.org/wiki/Transmeta_Crusoe In fact, mo

by pawadu 9y ago

> How would hardware emulation even work? Like this: https://en.wikipedia.org/wiki/Transmeta_Crusoe https://en.wikipedia.org/wiki/Transmeta_Crusoe In fact, most high-performance CPUs be it ARM or AMD or Intel do instruction translation internally.

DCKing 9y ago
Nvidia makes an architecture (Denver [1]) that is made by former Transmeta people and is in many ways a direct successor to this chip. In its current forms, Denver is implemented to execute ARMv8 instructions, but it could in theory execute x86 just the same. Now it's performing its internal exotic VLIW instructions by performing ARM "hardware emulation", but it could do x86 "hardware emulation" just the same. Qualcomm uses a less exotic architecture in their chips so that's not an option for them. [1]: https://en.wikipedia.org/wiki/Project_Denver https://en.wikipedia.org/wiki/Project_Denver
- ethbro 9y ago
  Do you have any links offhand about the internal arch? Curious about it being VLIW and couldn't dig up anything on a quick trawl. Good to hear that idea's still kicking around though. Diversity in CPU design is a good thing!
  - Symmetry 9y ago
    Yeah, it's still a 7-wide VLIW like the Efficeon but it also has a unit to run native arm code in hardware emulation at a lower IPC for code that hasn't been translated. The idea is to run code that's only used once in hardware emulation but to recompile all the hot code to native VLIW execution. https://www.hotchips.org/wp-content/uploads/hc_archives/hc26/HC26-11-day1-epub/HC26.11-2-Mobile-Processors-epub/HC26.11.234-Denver-Darrell.Boggs-NVIDIA-rev4.pdf https://www.hotchips.org/wp-content/uploads/hc_archives/hc26...
- lloeki 9y ago
  This is way over my head, but IIUC practically this would be both an arm CPU and an x86 CPU at the same time? Can it switch "on the fly" in a practical manner, and if so what kind of context switch are we talking about? Would it be viable to multitask between the two archs?

by pawadu 9y ago

DCKing 9y ago

Nvidia makes an architecture (Denver [1]) that is made by former Transmeta people and is in many ways a direct successor to this chip. In its current forms, Denver is implemented to execute ARMv8 instructions, but it could in theory execute x86 just the same. Now it's performing its internal exotic VLIW instructions by performing ARM "hardware emulation", but it could do x86 "hardware emulation" just the same. Qualcomm uses a less exotic architecture in their chips so that's not an option for them. [1]: https://en.wikipedia.org/wiki/Project_Denver https://en.wikipedia.org/wiki/Project_Denver

ethbro 9y ago
Do you have any links offhand about the internal arch? Curious about it being VLIW and couldn't dig up anything on a quick trawl. Good to hear that idea's still kicking around though. Diversity in CPU design is a good thing!
- Symmetry 9y ago
  Yeah, it's still a 7-wide VLIW like the Efficeon but it also has a unit to run native arm code in hardware emulation at a lower IPC for code that hasn't been translated. The idea is to run code that's only used once in hardware emulation but to recompile all the hot code to native VLIW execution. https://www.hotchips.org/wp-content/uploads/hc_archives/hc26/HC26-11-day1-epub/HC26.11-2-Mobile-Processors-epub/HC26.11.234-Denver-Darrell.Boggs-NVIDIA-rev4.pdf https://www.hotchips.org/wp-content/uploads/hc_archives/hc26...
lloeki 9y ago
This is way over my head, but IIUC practically this would be both an arm CPU and an x86 CPU at the same time? Can it switch "on the fly" in a practical manner, and if so what kind of context switch are we talking about? Would it be viable to multitask between the two archs?