PowerPC 64-bit To Support Up To 512TB Virtual Address Space On Linux 4.12

[phoronix]

Phoronix: PowerPC 64-bit To Support Up To 512TB Virtual Address Space On Linux 4.12

While Intel is working on 5-level paging support to allow a virtual address space up to 128 PiB and physical address space of 4 PiB, the PowerPC guys are working on upping their address space capabilities too...

http://www.phoronix.com/scan.php?page=news_item&px=Power-Changes-Linux-4.12

[starshipeleven]

More RAM for the RAM god!

[jacob]

Now I'm confused, I thought that PowerPC was dead and that IBM's 64bit CPUs were based on its bigger brother POWER?

[starshipeleven]

the article says "POWER 64-bit server CPUs can now support up to 512TB of virtual address space compared to a previous limit of 128TB"

[Dawn]

Now I'm confused, I thought that PowerPC was dead and that IBM's 64bit CPUs were based on its bigger brother POWER?

Not that simple. POWER, the ISA, actually died in the late 90s; Power3 was mixed-mode PPC/POWER, while Power4 was PPC only. Subsequently, the PPC ISA got renamed to Power Architecture (not to be confused with the POWER ISA that Power1, Power2, and Power3 used) in the mid 2000s, but everyone still calls it PPC.

IBM Power server processors implement that ISA.

Confused yet?

[hax0r]

I wish you could buy a decent POWER9 workstation for home use these days akin to that of SGI workstations from 1990s, I'm tired of x86 garbage.

[davidbepo]

I wish you could buy a decent POWER9 workstation for home use these days akin to that of SGI workstations from 1990s, I'm tired of x86 garbage.

could you explain what's garbage about x86?
or you are just trolling?

[starshipeleven]

could you explain what's garbage about x86?
or you are just trolling?

I can say UEFI firmwares, for a start. If you thought BIOS firmwares were bad, UEFI allows the "bad" to reach a whole new level.
Mostly because they are more complex so OEMs manage to screw shit up much more.

Then there is all the ME/PSP stuff that was added because lulz and might or might not expose the system to super-duper rootkits.

[Dawn]

could you explain what's garbage about x86?
or you are just trolling?

There are plenty of valid criticisms of x86, IMO. Being two-operand doesn't exactly help alleviate register pressure, although this is far less of an issue in 64-bit mode (and is more of an annoyance than a serious problem anyway - in actual execution, on most uarches, false dependencies will be eliminated during rename.) Port-based I/O is pretty weird but doesn't matter unless you're porting drivers.

SSE was, in many ways, a steaming pile - but AVX went for the right approach ("clone VMX, and improve from there.")

[microcode]

could you explain what's garbage about x86?
or you are just trolling?

There are only two surviving vendors, and both of them have a questionable closed source blob of software running whether the machine is on or off, with full access to all the memory. The architecture is being crushed under its own weight, as they double down every year on the static-sized SIMD extensions (this affects POWER as well).

Granted, Intel's implementations of AMD64 still win for high single-thread performance, and might continue to for a while.

I think RISC-V probably has more promise than POWER, if you give it a few years. POWER has most of the same licensing issues as ARM, and none of the momentum.