Intel 5-level paging: Difference between revisions

Content deleted Content added
x86 processors have supported multi-level page tables since the 80386. What the 5-level paging extension does is add support for one *more* level in the page table.
GreenC bot (talk | contribs)
Rescued 1 archive link. Wayback Medic 2.5 per WP:URLREQ#anandtech.com
 
(6 intermediate revisions by 5 users not shown)
Line 2:
{{Use dmy dates|date=August 2018}}
 
'''Intel 5-level paging''', referred to simply as ''5-level paging'' in [[Intel]] documents, is a processor extension for the [[x86-64]] line of processors.<ref name="intel-white-paper">{{Cite web|url=https://www.intel.com/content/www/us/en/content-details/671442/5-level-paging-and-5-level-ept-white-paper.html|title=5-Level Paging and 5-Level EPT|publisher=Intel Corporation|date=May 2017}}</ref>{{Rp|11}} It extends the size of [[virtual address]]es from 48&nbsp;bits to 57&nbsp;bits by adding an additional level to x86-64's [[Page table#Multilevel page tables|multilevel page tables]], increasing the addressable [[virtual memory]] from 256&nbsp;[[terabytetebibyte|TBTiB]] to 128&nbsp;[[petabytepebibyte|PBPiB]]. The extension was first implemented in the [[Ice Lake (microprocessor)|Ice Lake]] processors.<ref name="anandtech-13699">{{Cite web|url=https://www.anandtech.com/show/13699/intel-architecture-day-2018-core-future-hybrid-x86/2|archive-url=https://web.archive.org/web/20190502215444/https://www.anandtech.com/show/13699/intel-architecture-day-2018-core-future-hybrid-x86/2|url-status=dead|archive-date=2 May 2019|title=Sunny Cove Microarchitecture: A Peek At the Back End|work=Intel's Architecture Day 2018: The Future of Core, Intel GPUs, 10nm, and Hybrid x86|last=Cutress|first=Ian|access-date=2019-10-15}}</ref>
 
== Technology ==
[[File:X86 Paging 64bit.svg|thumb|right|555px|4-level paging of the 64-bit mode]]
In the 4-level paging scheme (previously known as [[IA-32e]] paging), the 64-bit virtual memory address is divided into five parts. The lowest 12 bits contain the offset within the 4 KiB memory page, and the following 36 bits are evenly divided between the four 9 bit descriptors, each linking to a 64-bit [[X86-64#Page table structure|page table entry]] in a 512-entry page table for each of the four paging levels. This makes it possible to use bits 0 through 47 in the virtual address, for a total of 256&nbsp;TBTiB.<ref name="x86-software-developers-manual" />{{Rp|page=4{{hyp}}2}}
 
[[File:Page Tables (5 levels).svg|thumb|A diagram of five levels of paging]]
5-level paging adds another 9 bit page table descriptor, making it possible to use bits&nbsp;0 through&nbsp;56. This multiplies the address space by 512 and increases the limit to 128&nbsp;PBPiB.
 
With 5-level paging enabled, bits&nbsp;57 through&nbsp;63 must be copies of bit&nbsp;56.<ref name="intel-white-paper" />{{Rp|17}} This is the same as with 4-level paging, where the high-order bits of a virtual address that do not participate in address translation must be the same as the most significant implemented bit.
Line 18:
 
== Drawbacks ==
Adding another level of indirection makes [[page table]] "walks" longer.<ref>{{Cite conference|title=CSALT: Context Switch Aware Large TLB|book-title=MICRO-50: the 50th Annual IEEE/ACM International Symposium on Microarchitecture : proceedings |___location=Cambridge, MA|publisher=Institute of Electrical and Electronics Engineers., IEEE Computer Society., ACM Special Interest Group on Microprogramming|doi=10.1145/3123939.3124549|page=450|isbn=978-1-4503-4952-9|oclc=1032337814|date = 14 October 2017}}</ref> A page table walk occurs when either the processor's [[memory management unit]] or the memory management code in the operating system navigates the tree of page tables to find the [[page table entry]] corresponding to a virtual address.<ref>{{Cite web|url=http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0301h/I1026235.html|title=ARM Information Center|website=infocenter.arm.com|access-date=2018-04-26}}</ref><ref name="x86-software-developers-manual">{{Cite book|url=https://www.intel.com/content/www/us/en/developer/articles/technical/intel-sdm.html|title=Intel® 64 and IA-32 Architectures Software Developer's Manual|volume=3A|publisher=[[Intel Corporation]]}}</ref>{{Rp|page=4{{hyp}}22}} This means that, in the worst case, the processor or the memory manager has to access physical memory six times for a single virtual memory access, rather than five for the previous iteration of x86-64 processors. This results in slightly reduced memory access speed.<ref name="cse-451-paging-tlbs-slides" /> In practice this cost is greatly mitigated by caches such as the [[translation lookaside buffer]] (TLB).<ref name="cse-451-paging-tlbs-slides">{{Cite web|url=https://courses.cs.washington.edu/courses/cse451/08au/lectures/10-paging_TLBs.pdf|title=CSE 451: Operating Systems: Paging & TLBs|last=Levy|first=Hank|author-link=Hank Levy (computer scientist)|date=Autumn 2008|website=[[University of Washington]]|access-date=26 April 2018}}</ref> Future extensions may reduce page walks by limiting virtual address space per application, with dedicated hardware flags in an extended 128 bit page table entry, and allowing a larger 64&nbsp;KiB or 2&nbsp;MiB [[Page (computer memory)|page size]]s and backward compatibility with 4&nbsp;KBKiB page operations.<Ref name=VA64>{{cite patent | country = US | number = 9858198 | status = patent | title = 64KB64KiB page system that supports 4KB4KiB page operation | pridate = 2015-06-26 | fdate = 2015-06-26 | pubdate = 2016-12-29 | gdate = 2018-01-02 | invent1 = Larry Seiler | assign1 = Intel Corp.}} </ref>
 
== Implementation ==
5-level paging is implemented by the [[Ice Lake (microprocessor)|Ice Lake]] [[microarchitecture]],<ref name="anandtech-13699"/> [[Epyc#Fourth_generation_Epyc_(Genoa,_Bergamo_and_Siena)|EPYC 9004 and 8004 Series Processors]]<ref>{{Cite web|url=https://www.amd.com/content/dam/amd/en/documents/epyc-technical-docs/tuning-guides/58020-epyc-9004-tg-high-perf-toolchain.pdf|title=Tuning Guide for AMD EPYC™ 9004 Processors|publisher=[[AMD]]|date=September 2023}}</ref><ref>{{Cite web|url=https://www.amd.com/content/dam/amd/en/documents/products/epyc/4th-gen-amd-epyc-processor-architecture-whitepaper.pdf|title=4TH GEN AMD EPYC™ PROCESSOR ARCHITECTURE|publisher=[[AMD]]|date=May 2024}}</ref> and [[List_of_AMD_Ryzen_processors#Storm_Peak_desktop|Storm peak]] [[Ryzen#Threadripper_series|Ryzen Threadripper]] PRO 7900WX series.<ref>{{Cite web|url=https://github.com/InstLatx64/InstLatx64/blob/37bd7b1ac29d95ddcf1fed90efaf7e82c89ab65d/AuthenticAMD/AuthenticAMD0A10F81_K19_StormPeak_01_CPUID.txt#L81C17-L81C25|title=CPUID dump for 96-Core AMD Ryzen Threadripper PRO 7995WX (Storm Peak) Zen4|website=[[GitHub]] |date=October 19, 2023}}</ref>
 
The 4.14 [[Linux kernel]] adds support for it.<ref>{{Cite news|url=https://www.zdnet.com/article/first-linux-4-14-release-adds-very-core-features-arrives-in-time-for-kernels-26th-birthday/|title=First Linux 4.14 release adds "very core" features, arrives in time for kernel's 26th birthday {{!}} ZDNet|last=Tung|first=Liam|work=ZDNet|access-date=2018-04-25|language=en}}</ref>
Support for the extension was submitted as a set of patches to the [[Linux kernel]] on 8 December 2016.<ref name="phoronix">{{Cite web|url=https://www.phoronix.com/scan.php?page=news_item&px=Intel-5-Level-Paging|title=Intel Working On 5-Level Paging To Increase Linux Virtual/Physical Address Space - Phoronix|author=Michael Larabel|date=9 December 2016|website=[[Phoronix]]|language=en|access-date=2018-04-26}}</ref> As was reported on the [[Linux kernel mailing list]], it consisted of extending the Linux memory model to use five levels rather than four.<ref>{{Cite mailing list|url=http://lkml.iu.edu/hypermail/linux/kernel/1612.1/00383.html|title=[RFC, PATCHv1 00/28] 5-level paging|last=Shutemov|first=Kirill A.|mailing-list=[[Linux kernel mailing list]]|date=December 8, 2016|access-date=2018-04-26}}</ref> This is because, although Linux [[Abstraction (software engineering)|abstracts]] the details of the page tables, it still depends on having a number of levels in its own representation. When an [[Instruction set architecture|architecture]] supports fewer levels, Linux emulates extra levels that do nothing.<ref>{{Cite web|url=https://www.kernel.org/doc/gorman/html/understand/understand006.html|title=Page Table Management|website=www.kernel.org|access-date=2018-04-26}}</ref> A similar change was previously made to extend from three levels to four.<ref>{{Cite web|url=https://lwn.net/Articles/106177/|title=Four-level page tables [LWN.net]|date=October 12, 2004|website=lwn.net|access-date=2018-04-26}}</ref>