![]() |
![]() OCAU News - Wiki - QuickLinks - Pix - Sponsors |
|
|||||||
| Notices |
|
Sign up for a free OCAU account and this ad will go away! Search our forums with Google: |
![]() |
|
|
Thread Tools |
|
|
#1 | |
|
Member
Join Date: Jul 2001
Location: Sydney
Posts: 8,889
|
Real World Tech discusses ARM's transition to 64bit. Aka: ARMv8 architecture.
=> http://www.realworldtech.com/arm64/ Its a technical read of the ARMv8. From the Conclusion... Quote:
__________________
"Haters gonna hate." => "I have no logical rebuttal to your criticism so I'm going to say something simple and meaningless to make myself feel better."-crabjokeman from Techreport.com forums. "The politicians don’t just want your money. They want your soul. They want you to be worn down by taxes until you are dependent and helpless."-James Dale Davidson. |
|
|
|
|
| Join OCAU to remove this ad! |
|
|
#2 |
|
SLATYE, not SLAYTE
Join Date: Nov 2002
Location: Canberra
Posts: 25,832
|
I'm curious about how they'll be maintaining backward-compatibility while removing so many of the ARMv7 features. Emulation? I'm sure that it's a good thing from a chip design point of view, but if Intel had been able to get rid of all the old x86 features and still maintain compatibility then they would have done that years ago.
__________________
Main system: Phenom II X4 920 | 8GB (4x 2GB) DDR2-800 | Gigabyte M57SLI-S4 v2.0 | Leadtek Geforce 9600GSO 384MB | Enermax Modu82+ 525W | 1TB Hitachi HDD | 3.5" + 5.25" FDD Laptop: Compal EL80 | C2D T7200 | 320GB Fujistu HDD | 2GB DDR2-667 | GF Go 7600 |
|
|
|
|
|
#3 |
|
Member
Join Date: Aug 2009
Posts: 2,649
|
(note: having problems loading the article here)
Compatible at what level, direct machine code or at the compile stage? I have a feeling that they actually mean the latter, because most ARM integrators have full control over the direct software that interfaces with the hardware - and those who are going to be worried about machine code level changes are unlikely to have any reason to skip to ARMv8 until it's mature anyway - or indefinitely if the costs of v7 remain lower than v8. In that respect x86 is much a different beast because full backwards compatibility is the status quo there, and quite a lot of the market is going to baulk when you change something. Plus only recently has low power computing become an issue - and a lot of the 'legacy' stuff in the grand scheme of things only takes a tiny portion of a high performance x86 processor, emulated or otherwise so it probably hasn't been a problem until now. |
|
|
|
|
|
#4 |
|
SLATYE, not SLAYTE
Join Date: Nov 2002
Location: Canberra
Posts: 25,832
|
Right, that would make sense.
That could make it a bit hard for ARM to get a foothold in the desktop/laptop market; people are used to the x86 system where a modern CPU will run any software ever written for an x86 CPU without any changes at all. It might also open the door a bit for ARM's competitors. If ARM stuck with a single instruction set and vendors became accustomed to compiling software for that instruction set and nothing else, then it'd be a major change for them to support alternatives (eg. MIPS). If they're already compiling for many different CPUs, then adding one more (as long as it doesn't require extensive code changes) is not such a big deal.
__________________
Main system: Phenom II X4 920 | 8GB (4x 2GB) DDR2-800 | Gigabyte M57SLI-S4 v2.0 | Leadtek Geforce 9600GSO 384MB | Enermax Modu82+ 525W | 1TB Hitachi HDD | 3.5" + 5.25" FDD Laptop: Compal EL80 | C2D T7200 | 320GB Fujistu HDD | 2GB DDR2-667 | GF Go 7600 |
|
|
|
|
|
#5 |
|
Member
Join Date: Aug 2009
Posts: 2,649
|
I would imagine the ARM consortium made an assessment on that front and decided that this is the way they're going to pace their architecture changes - and it does play into the direction userland software is heading down as well.
The state of the game is JIT compilation, so they only need to remain largely compatible with with JIT compilers (in this case we expect the JIT provider to only need to recompile their binaries to achieve this) in order to retain a consistent experience between architecture changes. I don't think that the ARM consortium themselves are that worried about competing architectures as such as well - which of course JIT compatibility would also allow for (a given JIT just needs to fully function on another arch for the userland experience to be consistent). Whilst you could foresee that MIPS or whatever architecture might move in on their territory, the members still have a combined expertise on ARM to support downstream integrators of the processor products - affecting the fairly important time to market - as well as supporting the microcontroller market which is their traditional cornerstone for a lot of the larger ARM partners. The JIT compilation movement is largely outside of the ARM consortium's hands anyway, so from that angle they also have no choice but to go with the flow. The other thing is that I don't think ARM is aiming for the unified HPC space yet, that is still the domain of other processor technologies - they're obviously aiming for the celluar (or 'cloud') style computing angle to get a foothold into the server and cluster computing market - where there are many comparatively lower performing modules operating on distributed tasks. Given this form of structure, even with the forseeable short term change of expanding the memory space to the full 64 bits (or perhaps beyond) that still sets them up for quite some time till the next arch jump. |
|
|
|
|
|
#6 |
|
Member
Join Date: Jun 2006
Location: Daejeon - South Korea
Posts: 966
|
I think this is a great move and the architecture improvements will outweigh any potential issues from removing legacy stuff in the core. I can see ARM positioning itself to take on intel and AMD as a desktop/laptop contender (in the long term) and it is better that they clean up the core to remove legacy crap sooner rather than later.
|
|
|
|
|
|
#7 |
|
Member
Join Date: May 2002
Location: Griffith
Posts: 357
|
Will this be the first ARM architecture with Virtualisation Extensions?
HP Redstone should have used ARMv8 might make interesting web server cluster (or sickbeard cluster haha) might see a vcenter for ARM one day yet |
|
|
|
|
|
#8 |
|
Member
Join Date: Aug 2009
Posts: 2,649
|
Are not ARM and virtualisation residents of conjugate paradigms?
I don't see how the combination of them really presents the advantages of either technology to be honest. |
|
|
|
|
|
#9 |
|
Member
Join Date: Jul 2001
Location: Sydney
Posts: 8,889
|
Today, ARM announces their first two processors on 64bit.
ARM's Cortex A57 and Cortex A53: The First 64-bit ARMv8 CPU Cores => http://www.anandtech.com/show/6420/a...rmv8-cpu-cores Official press release... ARM Launches Cortex-A50 Series, the World’s Most Energy-Efficient 64-bit Processors => http://www.arm.com/about/newsroom/ar...processors.php ARMv8 Architecture => http://www.arm.com/products/processo...chitecture.php From ARM's Youtube channel...
__________________
"Haters gonna hate." => "I have no logical rebuttal to your criticism so I'm going to say something simple and meaningless to make myself feel better."-crabjokeman from Techreport.com forums. "The politicians don’t just want your money. They want your soul. They want you to be worn down by taxes until you are dependent and helpless."-James Dale Davidson. |
|
|
|
|
|
#10 |
|
Member
Join Date: Jul 2001
Location: Sydney
Posts: 8,889
|
Cortex A57 overview/highlights...
__________________
"Haters gonna hate." => "I have no logical rebuttal to your criticism so I'm going to say something simple and meaningless to make myself feel better."-crabjokeman from Techreport.com forums. "The politicians don’t just want your money. They want your soul. They want you to be worn down by taxes until you are dependent and helpless."-James Dale Davidson. |
|
|
|
![]() |
| Bookmarks |
|
Sign up for a free OCAU account and this ad will go away! |
| Thread Tools | |
|
|