Date: Fri, 25 May 2018 16:41:52 -0700 From: Maxim Sobolev <sobomax@freebsd.org> To: David Chisnall <theraven@freebsd.org> Cc: Mark Linimon <linimon@lonesome.com>, Matthew Macy <mat.macy@gmail.com>, Andrew Gallatin <gallatin@cs.duke.edu>, Pedro Giffuni <pfg@freebsd.org>, Cy Schubert <Cy.Schubert@cschubert.com>, Brooks Davis <brooks@freebsd.org>, Eugene Grosbein <eugen@grosbein.net>, Stefan Esser <se@freebsd.org>, "rgrimes@freebsd.org" <rgrimes@freebsd.org>, Gleb Smirnoff <glebius@freebsd.org>, Sean Bruno <sbruno@freebsd.org>, "svn-src-head@freebsd.org" <svn-src-head@freebsd.org>, "svn-src-all@freebsd.org" <svn-src-all@freebsd.org>, "src-committers@freebsd.org" <src-committers@freebsd.org> Subject: Re: Deorbiting i386 Message-ID: <CAH7qZfs6CGVjRK4jvRaVJ6aDoanxgSmwjiS34zn7gFYNdhKYSg@mail.gmail.com> In-Reply-To: <05C5BD86-70D0-4B02-AC29-36E68B3602AE@FreeBSD.org> References: <201805232218.w4NMIxMA067892@slippy.cwsent.com> <e4311612-d1c4-b118-187b-7086945a312d@FreeBSD.org> <18a87d6d-14af-ef9d-80ff-403039e36692@cs.duke.edu> <CAPrugNomum%2BDO7M3GET3y0DrFse7jy1PmSUwnXGU5Sm6DXRrVg@mail.gmail.com> <20180525003949.GA710@lonesome.com> <CAH7qZfsbGheNqnwNmkP5jYiE=FXzc65yZSBoX_mM%2BuNce9rhyQ@mail.gmail.com> <05C5BD86-70D0-4B02-AC29-36E68B3602AE@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
That again is very subjective view, David. Sorry. Arm32 is kinda kind of the hill these days in the low-power/low cost space, but arm as a company is much more interested it seems in going into server / mobile device space rather than solidifying it's current de-facto kingdom. Those platforms based on their IP are very short lived and tightly coupled to a particular vendor with zillions busses, various kinds of weird quirks, vendor-maintained bootloaders etc. On the other hand, Intel is quickly closing the gap. If you've seen any of the atom bay trail systems in action you may understand what I mean. You get full blown x64 system with four cores and it takes only 2W of power. This is roughly equivalent of ARM8 system with a single core @ only 900MHz. So my prediction is in the 32-bit land arm will fade out as a platform and be replaced with RiscV in matter of few years and i386 will probably continue to be the platform of choice for many if Intel/Amd play that card right. -Max On Fri, May 25, 2018, 12:27 AM David Chisnall <theraven@freebsd.org> wrote: > On 25 May 2018, at 05:27, Maxim Sobolev <sobomax@freebsd.org> wrote: > > > > The idea looks very inmature and short-sighted to me. i386 is here to > stay not as a server/desktop platform but as an embedded/low power/low co= st > platform for at least 5-10 years to come. There are plenty of application= s > in the world that don't need > 3gb of memory space and have no use for > extra bits (and extra silicon) to function. > > This argument seems very odd to me. If you are targeting the embedded > space, it is far easier to build a low-power chip that targets the x86-64 > ISA than the x86-32 ISA. You can move all of the 80-bit floating point > stuff into microcode. You can put anything using pair-of-32-bit-register > 64-bit operations into slow microcode. You can skimp on store forwarding > for stack addresses. You actually need fewer rename registers (one of th= e > biggest consumers of power), because x86-64 code needs to do less registe= r > juggling to fit in the architectural register space. All of these things > are big consumers of power and area and are far less necessary when runni= ng > code compiled for x86-64. You can also do tricks like the one that Intel > did on the early Atoms, where the SSE ALUs are actually only 64 bits wide > and the 128-bit ops are cracked into pairs of 64-bit micro-ops. > > As to =E2=80=98not needing more than 3GB of memory space=E2=80=99, that= =E2=80=99s what the x32 ABI > is for. This lets you get all of the advantages of the x86-64 ISA (of > which there are very many, in comparison to x86-32), without needing 64-b= it > pointers. You get the instruction density of x86-64 combined with the da= ta > density of x86-32. This is what Intel and Centaur have been pushing in t= he > embedded space for several years. > > You do pay a slight hardware cost from supporting a 48-bit virtual addres= s > space, though with superpages that=E2=80=99s negligible and the hardware = targeted > at these applications often doesn=E2=80=99t support more than a 32-bit vi= rtual > address space. > > And this completely ignores the fact that Intel has almost no presence in > the low-end embedded space. AArch32 is vastly more important there and i= f > we dropped x86-32 and shifted that effort to AArch32 then I think we=E2= =80=99d see > a lot more adoption. > > David >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAH7qZfs6CGVjRK4jvRaVJ6aDoanxgSmwjiS34zn7gFYNdhKYSg>