[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YQ75wVUXPO6yRx7T@casper.infradead.org>
Date: Sat, 7 Aug 2021 22:23:13 +0100
From: Matthew Wilcox <willy@...radead.org>
To: Hugh Dickins <hughd@...gle.com>
Cc: Zi Yan <ziy@...dia.com>, Vlastimil Babka <vbabka@...e.cz>,
David Hildenbrand <david@...hat.com>, linux-mm@...ck.org,
"Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
Mike Kravetz <mike.kravetz@...cle.com>,
Michal Hocko <mhocko@...nel.org>,
John Hubbard <jhubbard@...dia.com>,
linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 00/15] Make MAX_ORDER adjustable as a kernel boot
time parameter.
On Sat, Aug 07, 2021 at 02:10:55AM +0100, Matthew Wilcox wrote:
> This hasn't been helped by the scarce number of 1GB TLB entries in Intel
> CPUs until very recently (and even those are hard to come by today).
Minor correction to this. I just fixed x86info to work on my Core
i7-1165G7 (Tiger Lake, launched about a year ago) and discovered it has:
L1 Store Only TLB: 1GB/4MB/2MB/4KB pages, fully associative, 16 entries
L1 Load Only TLB: 1GB pages, fully associative, 8 entries
L2 Unified TLB: 1GB/4KB pages, 8-way associative, 1024 entries
My prior laptop (i7-7500U, Kaby Lake, 2016) has only 4x 1GB TLB entries at
the L1 level, and no support for L2 1GB pages. So this speaks to Intel
finally taking performance of 1GB TLB entries seriously. Perhaps more
seriously than they need to for a laptop with 16GB of memory! There are
Xeon-W versions of this CPU, so I imagine that there are versions which
can support 1TB or more memory.
I still think that 1GB pages are too big for anything but specialist
use cases, but those do exist.
Powered by blists - more mailing lists