[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200605234523.GU31009@gate.crashing.org>
Date: Fri, 5 Jun 2020 18:45:23 -0500
From: Segher Boessenkool <segher@...nel.crashing.org>
To: Rich Felker <dalias@...c.org>
Cc: Daniel Kolesa <daniel@...aforge.org>, musl@...ts.openwall.com,
Michal Suchánek <msuchanek@...e.de>,
Joseph Myers <joseph@...esourcery.com>, libc-alpha@...rceware.org,
eery@...erfox.es, Will Springer <skirmisher@...tonmail.com>,
Palmer Dabbelt via binutils <binutils@...rceware.org>,
via libc-dev <libc-dev@...ts.llvm.org>, linuxppc-dev@...ts.ozlabs.org
Subject: Re: Re: ppc64le and 32-bit LE userland compatibility
Hi!
On Fri, Jun 05, 2020 at 01:50:46PM -0400, Rich Felker wrote:
> On Fri, Jun 05, 2020 at 12:27:02PM -0500, Segher Boessenkool wrote:
> > > I'm also not really all that convinced that vectors make a huge difference in non-specialized code (autovectorization still has a way to go)
> >
> > They do make a huge difference, depending on the application of course.
> > But VSX is not just vectors even: it also gives you twice as many
> > floating point scalars (64 now), and in newer versions of the ISA it can
> > be beneficially used for integer scalars even.
>
> Vectorization is useful for a lot of things, and I'm sure there are
> specialized workloads that benefit from 64 scalars, but I've never
> encountered a place where having more than 16 registers made a
> practical difference.
20 years ago 32 FP registers was already often a limitation, making FFT
and similar kernels almost twice slower than they could otherwise be.
Things are only *worse* with short vectors, not better. In general with
floating point data you need more registers (because you have more state
to look at concurrently) than with integer data.
*Of course* having 64 floating point registers does not matter if your
whole program only ever uses three floating point values, total, let
alone concurrently.
> The fact that there are specialized areas where this stuff matters
> does not imply there aren't huge domains where it's completely
> irrelevant.
There are very few domains where ISA 2.07 does not have significant
advantages over ISA 2.01. That is Power8 vs. Power4.
> > No, that is exactly the point of requiring ISA 2.07. Anything can use
> > ISA 2.07 (incl. VSX) without checking first, and without having a
> > fallback to some other implementation. Going from ISA 2.01 to 2.07 is
> > more than a decade of improvements, it is not trivial at all.
>
> This only affects code that's non-portable and PPC-specific, which a
No, it does not. It is not only about vector registers, either.
> I think a lot of the unnecessary fighting on this topic is arising
> from differences of opinion over what an ABI entails. I would call
> what you're talking about a "platform" and more of a platform-specific
> *API* than an ABI -- it's about guarantees of interfaces available to
> the programmer, not implementation details of linkage.
No, this is very much about the ABI. The B stands for Binary. Which
is what this is about.
Segher
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
