From: Richard Biener <rguenther@suse.de>
To: Kito Cheng <kito.cheng@gmail.com>
Cc: "juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai>,
"richard.sandiford" <richard.sandiford@arm.com>,
jeffreyalaw <jeffreyalaw@gmail.com>,
gcc-patches <gcc-patches@gcc.gnu.org>,
palmer <palmer@dabbelt.com>, jakub <jakub@redhat.com>
Subject: Re: Re: [PATCH] machine_mode type size: Extend enum size from 8-bit to 16-bit
Date: Wed, 12 Apr 2023 09:21:19 +0000 (UTC) [thread overview]
Message-ID: <nycvar.YFH.7.77.849.2304120920111.4466@jbgna.fhfr.qr> (raw)
In-Reply-To: <CA+yXCZCVd3c3c6KWf_t5x714T0brKu7Swh3AXXy+Z3PQ+1Y02A@mail.gmail.com>
On Wed, 12 Apr 2023, Kito Cheng wrote:
> Hi Richard:
>
> > > In order to model LMUL in backend, we have to the combination of
> > > scalar type and LMUL; possible LMUL is 1, 2, 4, 8, 1/2, 1/4, 1/8 - 8
> > > different types of LMUL, and we'll have QI, HI, SI, DI, HF, SF and DF,
> > > so basically we'll have 7 (LMUL type) * 7 (scalar type) here.
> >
> > Other archs have load/store-multiple instructions, IIRC those
> > are modeled with the appropriate set of operands. Do RVV LMUL
> > group inputs/outputs overlap with the non-LMUL grouped registers
> > and can they be used as aliases or is this supposed to be
> > implemented transparently on the register file level only?
>
> LMUL and non-LMUL (or LMUL=1) modes use the same vector register file.
>
> Reg for LMUL=1/2 : { {v0, v1, ...v31} }
> Reg for LMUL=1 : { {v0, v1, ...v31} }
> Reg for LMUL=2 : { {v0, v1}, {v2, v3}, ... {v30, v31} } // reg. must
> align to multiple of 2.
> Reg for LMUL=4 : { {v0, v1, v2, v3}, {v4, v5, v6, v7}, ... {v28, v29,
> v30, v31} } // reg. must align to multiple of 4.
> ..
> Reg for 2-tuples of LMUL=1 : { {v0, v1}, {v1, v2}, ... {v29, v30}, {v30, v31} }
> Reg for 2-tuples of LMUL=2 : { {v0, v1, v2, v3}, {v2, v3, v4, v5}, ...
> {v28, v29, v30, v31}, {v28, v29, v30, v31} } // reg. must align to
> multiple of 2.
> ...
>
> > But yes, implementing this as operations on multi-register
> > ops with large modes is probably the only sensible approach.
> >
> > I don't see how LMUL of 1/2, 1/4 or 1/8 is useful though? Can you
> > explain? Is that supposed to virtually increase the number of
> > registers? How do you represent r0:1/8:0 vs r0:1/8:3 (the first
> > and the third "virtual" register decomposed from r0) in GCC? To
> > me the natural way would be a subreg of r0?
> >
> > Somehow RVV seems to have more knobs than necessary for tuning
> > the actual vector register layout (aka N axes but only N-1 dimensions
> > thus the axes are
>
> The concept of fractional LMUL is the same as the concept of AArch64's
> partial SVE vectors,
> so they can only access the lowest part, like SVE's partial vector.
>
> We want to spill/restore the exact size of those modes (1/2, 1/4,
> 1/8), so adding dedicated modes for those partial vector modes should
> be unavoidable IMO.
>
> And even if we use sub-vector, we still need to define those partial
> vector types.
Could you use integer modes for the fractional vectors? For computation
you can always appropriately limit the LEN?
next prev parent reply other threads:[~2023-04-12 9:21 UTC|newest]
Thread overview: 63+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-10 14:48 juzhe.zhong
2023-04-10 14:54 ` Jeff Law
2023-04-10 15:02 ` juzhe.zhong
2023-04-10 15:14 ` juzhe.zhong
2023-04-11 9:16 ` Jakub Jelinek
2023-04-11 9:46 ` juzhe.zhong
2023-04-11 10:11 ` Jakub Jelinek
2023-04-11 10:25 ` juzhe.zhong
2023-04-11 10:52 ` Jakub Jelinek
2023-04-11 9:46 ` Richard Sandiford
2023-04-11 9:59 ` Jakub Jelinek
2023-04-11 10:11 ` juzhe.zhong
2023-04-11 10:05 ` Richard Earnshaw
2023-04-11 10:15 ` Richard Sandiford
2023-04-11 10:59 ` Richard Biener
2023-04-11 11:11 ` Richard Sandiford
2023-04-11 11:19 ` juzhe.zhong
2023-04-11 13:50 ` Kito Cheng
2023-04-12 7:53 ` Richard Biener
2023-04-12 9:06 ` Kito Cheng
2023-04-12 9:21 ` Richard Biener [this message]
2023-04-12 9:31 ` Kito Cheng
2023-04-12 23:22 ` 钟居哲
2023-04-13 13:06 ` Richard Sandiford
2023-04-13 14:02 ` Richard Biener
2023-04-15 2:58 ` Hans-Peter Nilsson
2023-04-17 6:38 ` Richard Biener
2023-04-20 5:37 ` Hans-Peter Nilsson
2023-05-05 1:43 ` Li, Pan2
2023-05-05 6:25 ` Richard Biener
2023-05-06 1:10 ` Li, Pan2
2023-05-06 1:53 ` Kito Cheng
2023-05-06 1:59 ` juzhe.zhong
2023-05-06 2:12 ` Li, Pan2
2023-05-06 2:18 ` Kito Cheng
2023-05-06 2:20 ` Li, Pan2
2023-05-06 2:48 ` Li, Pan2
2023-05-07 1:55 ` Li, Pan2
2023-05-07 15:23 ` Jeff Law
2023-05-08 1:07 ` Li, Pan2
2023-05-08 6:29 ` Richard Biener
2023-05-08 6:41 ` Li, Pan2
2023-05-08 6:59 ` Li, Pan2
2023-05-08 7:37 ` Richard Biener
2023-05-08 8:05 ` Li, Pan2
2023-05-09 6:13 ` Li, Pan2
2023-05-09 7:04 ` Richard Biener
2023-05-09 10:16 ` Richard Sandiford
2023-05-09 10:26 ` Richard Biener
2023-05-09 11:50 ` Li, Pan2
2023-05-10 5:09 ` Li, Pan2
2023-05-10 7:22 ` Li, Pan2
2023-05-08 1:35 ` Li, Pan2
2023-04-10 15:18 ` Jakub Jelinek
2023-04-10 15:22 ` juzhe.zhong
2023-04-10 20:42 ` Jeff Law
2023-04-10 23:03 ` juzhe.zhong
2023-04-11 1:36 ` juzhe.zhong
[not found] ` <20230410232205400970205@rivai.ai>
2023-04-10 15:33 ` juzhe.zhong
2023-04-10 20:39 ` Jeff Law
2023-04-10 20:36 ` Jeff Law
2023-04-10 22:53 ` juzhe.zhong
2023-04-10 15:10 ` Jakub Jelinek
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=nycvar.YFH.7.77.849.2304120920111.4466@jbgna.fhfr.qr \
--to=rguenther@suse.de \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=jeffreyalaw@gmail.com \
--cc=juzhe.zhong@rivai.ai \
--cc=kito.cheng@gmail.com \
--cc=palmer@dabbelt.com \
--cc=richard.sandiford@arm.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).