From: Stefan Monnier <monnier@iro.umontreal.ca>
To: Spencer Baugh <sbaugh@catern.com>
Cc: Eli Zaretskii <eliz@gnu.org>,
dgutov@yandex.ru, arnold@tdrhq.com, rms@gnu.org,
emacs-devel@gnu.org
Subject: Re: [PATCH v2 00/16] Speeding up DEFVAR_PER_BUFFER
Date: Mon, 30 Nov 2020 17:10:20 -0500 [thread overview]
Message-ID: <jwvh7p6v7m5.fsf-monnier+emacs@gnu.org> (raw)
In-Reply-To: <87y2iihafg.fsf@catern.com> (Spencer Baugh's message of "Mon, 30 Nov 2020 15:11:31 -0500")
> Yes, it's probably a trade-off. My guess is that the extra overhead of
> BVAR will not be significant, but benchmarking is the only way to know
> for sure. (Indeed, it's possible things will speed up by removing the
> metadata, which reduces the size of the working set and lets more things
> stay in cache)
Maybe one way to figure it out is to magnify the impact: add extra
slowdown code to BVAR to make sure the slowdown is noticeable.
This might point to the place where a performance impact may be
measurable, or it might make it obvious that the impact is negligible.
> Well, with a completely naive static branch-predictor which assumes the
> Qunbound branch will never be taken, there will be no overhead for
> fields which are never Qunbound, namely non-DEFVAR_PER_BUFFER fields and
> permanent-buffer-locals. We could encourage that prediction by marking
> the Qunbound branch with an "unlikely()" macro using GCC's
> __builtin_expect.
While this might work for some vars, I suspect that it could be
detrimental for some.
> It's probably prematurely low-level for me to talk about such things
> right now though - benchmarking is the only way to know.
100% agreement. And modern branch predictors are quite sophisticated so
I expect they'll do a much better job than any hint we may be able
to provide.
> A DEFVAR_PER_BUFFER variable is normally marked as
> permanent-buffer-local by:
Indeed, there are two concepts: one if the concept that corresponds to
the `permanent-local` symbol property (that's implemented via
buffer_permanent_local_flags) and the other is the concept of
buffer-local variables which are always buffer-local and don't even have
a default value (that's the one your patch modifies, tho it only
modifies the implementation, not the semantics).
> - in the old version, setting -1 in the corresponding field in
> buffer_local_flags (which is a `struct buffer' distinct from
> buffer_permanent_local_flags which is an array of chars)
> - in the new version, setting the corresponding field in buffer_defaults
> to Qunbound (indicating there's no default value)
BTW, one potentially unintended side effect of your change is that
`makunbound` on PER_BUFFER vars won't work quite as usual. I guess we
could address that by using some other special value than Qunbound, tho
I'm not sure how important this is.
> All this is rather confusing and inconsistent; I don't think we should
> add more of these "pseudo-permanent" variables.
Agreed: I think we should get rid of this special category of "always
local" variables. This is a concept that only applies to the following
variables:
bset_filename (&buffer_local_flags, make_fixnum (-1));
bset_directory (&buffer_local_flags, make_fixnum (-1));
bset_backed_up (&buffer_local_flags, make_fixnum (-1));
bset_save_length (&buffer_local_flags, make_fixnum (-1));
bset_auto_save_file_name (&buffer_local_flags, make_fixnum (-1));
bset_read_only (&buffer_local_flags, make_fixnum (-1));
bset_major_mode (&buffer_local_flags, make_fixnum (-1));
bset_mode_name (&buffer_local_flags, make_fixnum (-1));
bset_undo_list (&buffer_local_flags, make_fixnum (-1));
bset_mark_active (&buffer_local_flags, make_fixnum (-1));
bset_point_before_scroll (&buffer_local_flags, make_fixnum (-1));
bset_file_truename (&buffer_local_flags, make_fixnum (-1));
bset_invisibility_spec (&buffer_local_flags, make_fixnum (-1));
bset_file_format (&buffer_local_flags, make_fixnum (-1));
bset_auto_save_file_format (&buffer_local_flags, make_fixnum (-1));
bset_display_count (&buffer_local_flags, make_fixnum (-1));
bset_display_time (&buffer_local_flags, make_fixnum (-1));
bset_enable_multibyte_characters (&buffer_local_flags, make_fixnum (-1));
I don't see the benefit of forcing those vars to always be buffer-local.
We could just as well make them "normal" and they'll still have
a buffer-local value pretty much all the time anyway.
Stefan
next prev parent reply other threads:[~2020-11-30 22:10 UTC|newest]
Thread overview: 80+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-11-14 16:34 ido-switch-buffer is slow with many buffers; others are fast catern
2020-11-14 18:22 ` Stefan Monnier
2020-11-14 20:06 ` sbaugh
2020-11-14 23:00 ` Stefan Monnier
2020-11-14 23:16 ` Dmitry Gutov
2020-11-15 0:19 ` Spencer Baugh
2020-11-15 15:15 ` Stefan Monnier
2020-11-15 20:49 ` Spencer Baugh
2020-11-15 23:58 ` Arnold Noronha
2020-11-19 15:38 ` [PATCH 00/10] Speeding up DEFVAR_PER_BUFFER (Was: ido-switch-buffer is slow) Spencer Baugh
2020-11-19 17:29 ` [PATCH 00/10] Speeding up DEFVAR_PER_BUFFER Stefan Monnier
2020-11-22 2:34 ` [PATCH v2 00/16] " Spencer Baugh
2020-11-22 2:34 ` [PATCH v2 01/16] Add a test for let-binding unwinding Spencer Baugh
2020-11-25 20:53 ` Stefan Monnier
2020-11-30 17:31 ` Spencer Baugh
2020-12-01 16:44 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 02/16] Assert not local-variable-p after setq in let_default binding Spencer Baugh
2020-11-25 20:54 ` Stefan Monnier
2020-12-01 16:45 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 03/16] Stop checking the constant default for enable_multibyte_characters Spencer Baugh
2020-11-25 20:57 ` Stefan Monnier
2020-12-01 16:52 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 04/16] Take buffer field name in DEFVAR_PER_BUFFER Spencer Baugh
2020-12-01 16:56 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 05/16] Add BVAR_DEFAULT for access to buffer defaults Spencer Baugh
2020-12-01 17:00 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 06/16] Use bset_* functions instead of BVAR Spencer Baugh
2020-12-01 17:12 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 07/16] Take offset not idx in PER_BUFFER_VALUE_P Spencer Baugh
2020-12-01 17:20 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 08/16] Combine unnecessarily separate loops in buffer.c Spencer Baugh
2020-12-01 17:22 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 09/16] Add and use BUFFER_DEFAULT_VALUE_P Spencer Baugh
2020-12-01 17:24 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 10/16] Add and use KILL_PER_BUFFER_VALUE Spencer Baugh
2020-12-01 17:26 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 11/16] Assert that PER_BUFFER_IDX for Lisp variables is not 0 Spencer Baugh
2020-12-01 17:32 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 12/16] Rearrange set_internal for buffer forwarded symbols Spencer Baugh
2020-12-01 17:35 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 13/16] Get rid of buffer_permanent_local_flags array Spencer Baugh
2020-11-22 16:16 ` Eli Zaretskii
2020-12-01 17:39 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 14/16] Remove unnecessary Qunbound check Spencer Baugh
2020-12-01 17:40 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 15/16] Remove local_flags array in struct buffer Spencer Baugh
2020-11-22 20:04 ` Stefan Monnier
2020-12-01 17:57 ` Eli Zaretskii
2020-11-22 2:34 ` [PATCH v2 16/16] Remove usage of buffer_local_flags Spencer Baugh
2020-12-01 18:05 ` Eli Zaretskii
2020-11-22 11:19 ` [PATCH v2 00/16] Speeding up DEFVAR_PER_BUFFER Kévin Le Gouguec
2020-11-22 16:12 ` Eli Zaretskii
2020-11-22 16:28 ` Spencer Baugh
2020-11-22 17:13 ` Eli Zaretskii
2020-11-29 17:41 ` Spencer Baugh
2020-11-30 18:32 ` Eli Zaretskii
2020-11-30 20:11 ` Spencer Baugh
2020-11-30 22:10 ` Stefan Monnier [this message]
2020-11-30 22:26 ` Andrea Corallo via Emacs development discussions.
2020-12-01 15:15 ` Eli Zaretskii
2020-12-01 15:56 ` Stefan Monnier
2020-12-01 15:10 ` Eli Zaretskii
2020-12-01 15:20 ` Spencer Baugh
2020-12-01 16:01 ` Stefan Monnier
2020-12-01 16:16 ` Eli Zaretskii
2020-12-01 18:10 ` Eli Zaretskii
2020-11-22 20:11 ` Stefan Monnier
2020-11-19 15:38 ` [PATCH 01/10] Take buffer field name in DEFVAR_PER_BUFFER Spencer Baugh
2020-11-19 15:38 ` [PATCH 02/10] Add bset_save_length and use it Spencer Baugh
2020-11-19 15:38 ` [PATCH 03/10] Use bset_last_selected_window everywhere Spencer Baugh
2020-11-19 15:38 ` [PATCH 04/10] Use bset_enable_multibyte_characters everywhere Spencer Baugh
2020-11-19 15:38 ` [PATCH 05/10] Add BVAR_DEFAULT for access to buffer defaults Spencer Baugh
2020-11-19 15:38 ` [PATCH 06/10] Disallow using BVAR as an lvalue Spencer Baugh
2020-11-19 15:38 ` [PATCH 07/10] Reorder buffer.h for upcoming rework of BVAR Spencer Baugh
2020-11-19 15:38 ` [PATCH 08/10] Make cache_long_scans buffer-local when setting it Spencer Baugh
2020-11-19 15:38 ` [PATCH 09/10] Access buffer_defaults in BVAR if there's no local binding Spencer Baugh
2020-11-19 18:08 ` Stefan Monnier
2020-11-19 18:21 ` Eli Zaretskii
2020-11-19 15:38 ` [PATCH 10/10] Don't iterate over all buffers in set_default_internal Spencer Baugh
2020-11-19 18:21 ` Stefan Monnier
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
List information: https://www.gnu.org/software/emacs/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=jwvh7p6v7m5.fsf-monnier+emacs@gnu.org \
--to=monnier@iro.umontreal.ca \
--cc=arnold@tdrhq.com \
--cc=dgutov@yandex.ru \
--cc=eliz@gnu.org \
--cc=emacs-devel@gnu.org \
--cc=rms@gnu.org \
--cc=sbaugh@catern.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.
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/emacs.git
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).