4 ms·
> "Microsoft, at least, has Windows running on ARM, but it doesn't have Rosetta" I thought that Microsoft's emulation efforts were hampered by Intel making thr
by ThrowawayR2 6y ago
> "Microsoft, at least, has Windows running on ARM, but it doesn't have Rosetta"
I thought that Microsoft's emulation efforts were hampered by Intel making threats, e.g. "Intel recently made an unprecedented public challenge to Microsoft and Qualcomm that basically told the latter two companies: if you ship an x86 instruction set architecture (ISA) emulator, we’re coming after you.", from https://www.forbes.com/sites/tiriasresearch/2017/06/16/intel-threatens-microsoft-and-qualcomm-over-x86-emulation/?sh=eeca82454f43 https://www.forbes.com/sites/tiriasresearch/2017/06/16/intel...
If so, the blame for that really lies at Intel's doorstep as well.
- usefulcat 6y agoDid Intel similarly threaten Apple? If not, why not? Even if they didn't know for sure exactly what Apple was up to with the M1, they surely could have speculated, as so many others did.
- shaggyfrog 6y agoThe threat was regarding emulation. Rosetta 2 isn't emulation, it's translation.
- erik 6y agoRosetta 2 includes both ahead-of-time translation and run-time emulation. It can run x64 jit software for example.
- coldtea 6y agoThe jit is run-time translation still - not run-time emulation.
- dtech 6y agoCan you provide a definition that specifies the difference? The end result is identical: being able to run software compiled for foreign ISA.
- coldtea 6y agoIt doesn't emulate a whole CPU and has the simulated CPU run the code. It translates instructions directly to the different ISA.
- saagarjha 6y agoThat is literally how performant emulators work.
- buzzerbetrayed 6y agoSo why does Microsoft go the translation route instead of emulation?
- leoc 6y agoFWIW, the twentieth anniversary of the publication of the x86-64 specification was apparently August of this year (https://web.archive.org/web/20120308025559/http://www.amd.com/us/press-releases/Pages/Press_Release_715.aspx https://web.archive.org/web/20120308025559/http://www.amd.co... , via WP https://en.wikipedia.org/wiki/X86-64#History https://en.wikipedia.org/wiki/X86-64#History ), while MS announced x86-64 emulation support for ARM W10 at the end of September https://uk.pcmag.com/migrated-3765-windows-10/128922/microsoft-to-bring-64-bit-x86-windows-apps-to-arm-powered-pcs-with-new-emulator https://uk.pcmag.com/migrated-3765-windows-10/128922/microso... https://blogs.windows.com/windowsexperience/2020/09/30/now-more-essential-than-ever-the-role-of-the-windows-pc-has-changed/ https://blogs.windows.com/windowsexperience/2020/09/30/now-m... .
- xxpor 6y agoDoesn't the force behind that threat depend heavily on the outcome of Oracle v Google?
- Spooky23 6y agoWho knows what kind of weird deals Microsoft and Intel have. Also, Microsoft historically is awful at transitions. Look at the 64 but print driver transition - complete shitshow.
- fomine3 6y agoI don't think it was awful. It's not failure to 64bit transition but they simultaneously force drivers to be signed.
- dwaite 6y agoIntel has patents in things like some of the newer vector instructions still - Apple AFAIK will not translate those, and will not report them as available. Microsoft has a performance disadvantage in that the memory consistency model of ARM is much weaker than x86 sequential ordering or x86_64s total store ordering. Apple’s chips (at least back to the A12Z) support enabling TSO per thread on the performance cores, so the translated code does not need as much fencing.