From: Mike Stump <mikestump@comcast.net>
To: liuhongt <hongtao.liu@intel.com>
Cc: gcc-patches@gcc.gnu.org, crazylht@gmail.com, hjl.tools@gmail.com,
ubizjak@gmail.com, jakub@redhat.com
Subject: Re: [PATCH] [i386] Support type _Float16/__bf16 independent of SSE2.
Date: Wed, 19 Apr 2023 09:43:02 -0700 [thread overview]
Message-ID: <03F85853-4291-437A-A725-5C260227716E@comcast.net> (raw)
In-Reply-To: <20230419071551.3478647-1-hongtao.liu@intel.com>
LLM, machine learning and AI likes coding with data types that are weird, float16, bf16, 8 bit float and 4 bit floats. Longer term, would be nice to natively support these everywhere. Would be nice to trial run them in the compiler, sort it all out, so that the implementation experience can drive language adoption. A little speculative and a little narrow focus for the field, but, AI isn't going away in the next 20 years I don't think. Anyway, I like the direction.
On Apr 19, 2023, at 12:15 AM, liuhongt via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
> That said, these fundamental types whose presence/absence depends on ISA flags
> are quite problematic IMHO, as they are incompatible with the target
> attribute/pragmas. Whether they are available or not available depends on
> whether in this case SSE2 is enabled during compiler initialization (aka after
> parsing command line options) and then they are available or unavailable to
> everything else based on that.
> -----comments end----------
>
> Enable _Float16 and __bf16 all the time but issue errors when the
next prev parent reply other threads:[~2023-04-19 16:43 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-19 7:15 liuhongt
2023-04-19 16:43 ` Mike Stump [this message]
2023-04-20 12:18 ` Jakub Jelinek
2023-04-21 13:53 ` [PATCH 1/2] " liuhongt
2023-04-21 13:53 ` [PATCH 2/2] [i386] def_or_undef __STDCPP_FLOAT16_T__ and __STDCPP_BFLOAT16_T__ for target attribute/pragmas liuhongt
2023-05-15 1:21 ` Hongtao Liu
2023-05-15 1:20 ` [PATCH 1/2] [i386] Support type _Float16/__bf16 independent of SSE2 Hongtao Liu
2023-07-17 8:35 ` Hongtao Liu
2023-07-17 11:38 ` Uros Bizjak
2023-07-19 5:58 ` Hongtao Liu
2023-07-19 10:51 ` Jakub Jelinek
2023-07-20 4:50 ` [PATCH] Fix fp16 related testcase failure for i686 liuhongt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=03F85853-4291-437A-A725-5C260227716E@comcast.net \
--to=mikestump@comcast.net \
--cc=crazylht@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=hjl.tools@gmail.com \
--cc=hongtao.liu@intel.com \
--cc=jakub@redhat.com \
--cc=ubizjak@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).