From: "H.J. Lu" <hjl.tools@gmail.com>
To: Jan Beulich <jbeulich@suse.com>
Cc: Binutils <binutils@sourceware.org>,
"Jiang, Haochen" <haochen.jiang@intel.com>,
wwwhhhyyy <hongyu.wang@intel.com>
Subject: Re: [PATCH] x86: fold AVX512-VNNI disassembler entries with AVX-VNNI ones
Date: Fri, 14 Oct 2022 10:28:49 -0700 [thread overview]
Message-ID: <CAMe9rOoEjx8zzjQiuWZTm=WDk-ad5S+wryPbUneaTaTPQiZxSg@mail.gmail.com> (raw)
In-Reply-To: <7bac66be-535e-9051-d674-f2f5ba180e17@suse.com>
On Fri, Oct 14, 2022 at 3:22 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> Make %XV also print the separating blank in the VEX case, while making
> it do nothing for EVEX-encoded insns. This way the AVX-VNNI entries
> can be re-used for AVX512-VNNI, at the same time fixing the lack of
> EVEX.W decoding.
>
> For the AVX-VNNI ones further make sure only VEX.66 forms are actually
> decoded.
> ---
> Irrespective of this change I continue to disagree with the arbitrary
> printing of "{vex}" for the AVX-VNNI insns: If that's meant for
> disambiguation purposes, then EVEX-encoded insns not using EVEX-specific
> functionality by having VEX counterparts (vaddps %xmm0, %xmm0, %xmm0)
> should also be prefixed by "{evex}".
This is done to match the assembler. There are 3 kinds of VNNI processors:
1. AVX512-VNNI only.
2. AVX-VNNI only.
3. AVX512-VNNI and AVX-VNNI.
Since AVX512-VNNI came out first, all VNNI instructions without a prefix
are encoded as AVX512-VNNI. The existing VNNI instructions without
a prefix, generated by compiler or hand written, are encoded with EVEX.
If one needs VNNI with VEX encoding, the {vex} prefix should be used.
This applies to any AVX extensions which come after EVEX ones, including
AVX-IFMA.
> --- a/opcodes/i386-dis.c
> +++ b/opcodes/i386-dis.c
> @@ -1755,7 +1755,7 @@ struct dis386 {
> "XD" => print 'd' if !EVEX or EVEX.W=1, EVEX.W=0 is not a valid encoding
> "XH" => print 'h' if EVEX.W=0, EVEX.W=1 is not a valid encoding (for FP16)
> "XS" => print 's' if !EVEX or EVEX.W=0, EVEX.W=1 is not a valid encoding
> - "XV" => print "{vex3}" pseudo prefix
> + "XV" => print "{vex} " pseudo prefix
> "LQ" => print 'l' ('d' in Intel mode) or 'q' for memory operand, cond
> being false, or no operand at all in 64bit mode, or if suffix_always
> is true.
> @@ -7545,19 +7545,19 @@ static const struct dis386 vex_w_table[]
> },
> {
> /* VEX_W_0F3850 */
> - { "%XV vpdpbusd", { XM, Vex, EXx }, 0 },
> + { "%XVvpdpbusd", { XM, Vex, EXx }, PREFIX_DATA },
> },
> {
> /* VEX_W_0F3851 */
> - { "%XV vpdpbusds", { XM, Vex, EXx }, 0 },
> + { "%XVvpdpbusds", { XM, Vex, EXx }, PREFIX_DATA },
> },
> {
> /* VEX_W_0F3852 */
> - { "%XV vpdpwssd", { XM, Vex, EXx }, 0 },
> + { "%XVvpdpwssd", { XM, Vex, EXx }, PREFIX_DATA },
> },
> {
> /* VEX_W_0F3853 */
> - { "%XV vpdpwssds", { XM, Vex, EXx }, 0 },
> + { "%XVvpdpwssds", { XM, Vex, EXx }, PREFIX_DATA },
> },
> {
> /* VEX_W_0F3858 */
> @@ -10711,22 +10711,29 @@ putop (instr_info *ins, const char *in_t
> case 'V':
> if (l == 0)
> abort ();
> - else if (l == 1
> - && (last[0] == 'L' || last[0] == 'X'))
> + else if (l == 1)
> {
> - if (last[0] == 'X')
> + switch (last[0])
> {
> + case 'X':
> + if (ins->vex.evex)
> + break;
> *ins->obufp++ = '{';
> *ins->obufp++ = 'v';
> *ins->obufp++ = 'e';
> *ins->obufp++ = 'x';
> *ins->obufp++ = '}';
> - }
> - else if (ins->rex & REX_W)
> - {
> + *ins->obufp++ = ' ';
> + break;
> + case 'L':
> + if (!(ins->rex & REX_W))
> + break;
> *ins->obufp++ = 'a';
> *ins->obufp++ = 'b';
> *ins->obufp++ = 's';
> + break;
> + default:
> + abort ();
> }
> }
> else
> --- a/opcodes/i386-dis-evex.h
> +++ b/opcodes/i386-dis-evex.h
> @@ -383,8 +383,8 @@ static const struct dis386 evex_table[][
> { "vrsqrt14p%XW", { XM, EXx }, 0 },
> { "vrsqrt14s%XW", { XMScalar, VexScalar, EXdq }, PREFIX_DATA },
> /* 50 */
> - { "vpdpbusd", { XM, Vex, EXx }, PREFIX_DATA },
> - { "vpdpbusds", { XM, Vex, EXx }, PREFIX_DATA },
> + { VEX_W_TABLE (VEX_W_0F3850) },
> + { VEX_W_TABLE (VEX_W_0F3851) },
> { PREFIX_TABLE (PREFIX_EVEX_0F3852) },
> { PREFIX_TABLE (PREFIX_EVEX_0F3853) },
> { "vpopcnt%BW", { XM, EXx }, PREFIX_DATA },
> --- a/opcodes/i386-dis-evex-prefix.h
> +++ b/opcodes/i386-dis-evex-prefix.h
> @@ -233,14 +233,14 @@
> {
> { Bad_Opcode },
> { "vdpbf16p%XS", { XM, Vex, EXx }, 0 },
> - { "vpdpwssd", { XM, Vex, EXx }, 0 },
> + { VEX_W_TABLE (VEX_W_0F3852) },
> { "vp4dpwssd", { XM, Vex, EXxmm }, 0 },
> },
> /* PREFIX_EVEX_0F3853 */
> {
> { Bad_Opcode },
> { Bad_Opcode },
> - { "vpdpwssds", { XM, Vex, EXx }, 0 },
> + { VEX_W_TABLE (VEX_W_0F3853) },
> { "vp4dpwssds", { XM, Vex, EXxmm }, 0 },
> },
> /* PREFIX_EVEX_0F3868 */
OK.
--
H.J.
next prev parent reply other threads:[~2022-10-14 17:29 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-14 10:22 Jan Beulich
2022-10-14 17:28 ` H.J. Lu [this message]
2022-10-17 6:39 ` Jan Beulich
2022-10-17 21:10 ` H.J. Lu
2022-10-18 5:41 ` Jan Beulich
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='CAMe9rOoEjx8zzjQiuWZTm=WDk-ad5S+wryPbUneaTaTPQiZxSg@mail.gmail.com' \
--to=hjl.tools@gmail.com \
--cc=binutils@sourceware.org \
--cc=haochen.jiang@intel.com \
--cc=hongyu.wang@intel.com \
--cc=jbeulich@suse.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).