From: "Cui, Lili" <lili.cui@intel.com>
To: "Beulich, Jan" <JBeulich@suse.com>
Cc: "Lu, Hongjiu" <hongjiu.lu@intel.com>,
"binutils@sourceware.org" <binutils@sourceware.org>
Subject: RE: [PATCH v4 3/9] Support APX GPR32 with extend evex prefix
Date: Tue, 26 Dec 2023 07:00:51 +0000 [thread overview]
Message-ID: <SJ0PR11MB56000519C69C31D771FB32209E98A@SJ0PR11MB5600.namprd11.prod.outlook.com> (raw)
In-Reply-To: <4c64e863-22a3-43e4-a566-e05c7fb909bd@suse.com>
> > --- a/opcodes/i386-dis-evex-prefix.h
> > +++ b/opcodes/i386-dis-evex-prefix.h
> > @@ -285,6 +285,14 @@
> > { "%XEvfmsub213s%XW", { XMScalar, VexScalar, EXdq, EXxEVexR }, 0 },
> > { "v4fnmadds%XS", { XMScalar, VexScalar, Mxmm }, 0 },
> > },
> > + /* PREFIX_EVEX_0F38F2_L_0 */
> > + {
> > + { "andnS", { Gdq, VexGdq, Edq }, 0 },
> > + },
>
> So not being able to re-use the VEX entry for this and ...
>
> > --- a/opcodes/i386-dis-evex-reg.h
> > +++ b/opcodes/i386-dis-evex-reg.h
> > @@ -49,3 +49,10 @@
> > { "vscatterpf0qp%XW", { MVexVSIBQWpX }, PREFIX_DATA },
> > { "vscatterpf1qp%XW", { MVexVSIBQWpX }, PREFIX_DATA },
> > },
> > + /* REG_EVEX_0F38F3_L_0_P_0 */
> > + {
> > + { Bad_Opcode },
> > + { "blsrS", { VexGdq, Edq }, 0 },
> > + { "blsmskS", { VexGdq, Edq }, 0 },
> > + { "blsiS", { VexGdq, Edq }, 0 },
> > + },
>
> ... this was due to the VEX entries having PREFIX_OPCODE, which would be
> getting in the way? This is the sort of thing that would be useful to have in the
> description, to avoid raising the same question again that (I think) was raised
> before.
>
> Yet then - why do you strip PREFIX_OPCODE from the VEX entries? If you do
> that (as iirc I did suggest), there's no need for having separate EVEX ones (the
> suggestion, after all, was to be able to re-use the VEX entries). That re-work of
> existing VEX encodings could, btw, also have been split out quite easily. That
> way this huge patch would have further shrunk a little.
>
I re-used the VEX entry, but forgot to remove the redundant, removed.
> > --- /dev/null
> > +++ b/opcodes/i386-dis-evex-x86-64.h
> > @@ -0,0 +1,50 @@
> > + /* X86_64_EVEX_0F90 */
> > + {
> > + { Bad_Opcode },
> > + { VEX_W_TABLE (VEX_W_0F90_L_0) }, },
> > + /* X86_64_EVEX_0F91 */
> > + {
> > + { Bad_Opcode },
> > + { VEX_W_TABLE (VEX_W_0F91_L_0) }, },
> > + /* X86_64_EVEX_0F92 */
> > + {
> > + { Bad_Opcode },
> > + { VEX_W_TABLE (VEX_W_0F92_L_0) }, },
> > + /* X86_64_EVEX_0F93 */
> > + {
> > + { Bad_Opcode },
> > + { VEX_W_TABLE (VEX_W_0F93_L_0) }, },
> > + /* X86_64_EVEX_0F38F2 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE (PREFIX_VEX_0F38F2_L_0) }, },
> > + /* X86_64_EVEX_0F38F3 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE (PREFIX_VEX_0F38F3_L_0) }, },
> > + /* X86_64_EVEX_0F38F5 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE (PREFIX_VEX_0F38F5_L_0) }, },
> > + /* X86_64_EVEX_0F38F6 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE(PREFIX_VEX_0F38F6_L_0) }, },
> > + /* X86_64_EVEX_0F38F7 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE(PREFIX_VEX_0F38F7_L_0) }, },
> > + /* X86_64_EVEX_0F3AF0 */
> > + {
> > + { Bad_Opcode },
> > + { PREFIX_TABLE (PREFIX_VEX_0F3AF0_L_0) }, },
>
> Am I misremembering that we had agreed that this new file isn't necessary, by
> having USE_X86_64_EVEX_FROM_VEX_TABLE handle the non-64-bit case? At
> least I couldn't find a mail from you saying this isn't possible (and why).
>
I Prefer not to change the current implement, we need a table that all instructions must go through, it can be x86-64 or len_table,
but I think x86-64 is better. It can reuse more old parts of x86-64 (for example X86_64_VEX_0F38E*) than len_table. If we use len_table instead, we need to let another 18 instructions through the len_table, this will also add the number of entries.
18 instructions are:
X86_64_VEX_0F38E0~ X86_64_VEX_0F38EF, X86_64_VEX_0F3849, and X86_64_VEX_0F384B.
Thanks,
Lili.
next prev parent reply other threads:[~2023-12-26 7:01 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-19 12:12 [PATCH v4 0/9] Support Intel APX EGPR Cui, Lili
2023-12-19 12:12 ` [PATCH v4 1/9] Support APX GPR32 with rex2 prefix Cui, Lili
2023-12-22 13:08 ` Jan Beulich
2023-12-25 6:14 ` Cui, Lili
2024-01-04 8:57 ` Jan Beulich
2023-12-19 12:12 ` [PATCH v4 2/9] Created an empty EVEX_MAP4_ sub-table for EVEX instructions Cui, Lili
2023-12-19 12:12 ` [PATCH v4 3/9] Support APX GPR32 with extend evex prefix Cui, Lili
2023-12-22 13:49 ` Jan Beulich
2023-12-25 12:23 ` Cui, Lili
2024-01-04 9:08 ` Jan Beulich
2024-01-04 12:32 ` Cui, Lili
2024-01-04 12:55 ` Jan Beulich
2023-12-22 14:19 ` Jan Beulich
2023-12-26 7:00 ` Cui, Lili [this message]
2024-01-04 9:01 ` Jan Beulich
2024-01-04 12:47 ` Cui, Lili
2023-12-19 12:12 ` [PATCH v4 4/9] Add tests for " Cui, Lili
2023-12-22 14:41 ` Jan Beulich
2023-12-25 13:40 ` Cui, Lili
2024-01-04 9:16 ` Jan Beulich
2024-01-05 6:58 ` Cui, Lili
2023-12-19 12:12 ` [PATCH v4 5/9] Support APX NDD Cui, Lili
2023-12-19 12:12 ` [PATCH v4 6/9] Support APX Push2/Pop2 Cui, Lili
2023-12-19 12:12 ` [PATCH v4 7/9] Support APX PUSHP/POPP Cui, Lili
2023-12-19 12:12 ` [PATCH v4 `8/9] Support APX NDD optimized encoding Cui, Lili
2023-12-19 12:12 ` [PATCH v4 9/9] Support APX JMPABS for disassembler Cui, Lili
2023-12-19 12:35 ` [PATCH v4 0/9] Support Intel APX EGPR Jan Beulich
2023-12-20 8:50 ` Cui, Lili
2023-12-20 8:57 ` Jan Beulich
2023-12-20 10:42 ` Cui, Lili
2023-12-20 11:00 ` Jan Beulich
2023-12-20 11:50 ` Cui, Lili
2023-12-20 12:01 ` Jan Beulich
2023-12-20 12:16 ` Cui, Lili
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=SJ0PR11MB56000519C69C31D771FB32209E98A@SJ0PR11MB5600.namprd11.prod.outlook.com \
--to=lili.cui@intel.com \
--cc=JBeulich@suse.com \
--cc=binutils@sourceware.org \
--cc=hongjiu.lu@intel.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).