branch master updated: * tp/TODO: notes on possible optimization after looking at callgrind output for HTML output with C.

[Patrice Dumas]

This is an automated email from the git hooks/post-receive script.

pertusus pushed a commit to branch master
in repository texinfo.

The following commit(s) were added to refs/heads/master by this push:
     new 695f870ff2 * tp/TODO: notes on possible optimization after looking at 
callgrind output for HTML output with C.
695f870ff2 is described below

commit 695f870ff29b776d417a1d47c4de318e86469359
Author: Patrice Dumas <pertusus@free.fr>
AuthorDate: Mon Apr 29 15:40:40 2024 +0200

    * tp/TODO: notes on possible optimization after looking at callgrind
    output for HTML output with C.
---
 ChangeLog |  5 +++++
 tp/TODO   | 17 +++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/ChangeLog b/ChangeLog
index 2f94061022..6d9a119130 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,8 @@
+2024-04-29  Patrice Dumas  <pertusus@free.fr>
+
+       * tp/TODO: notes on possible optimization after looking at callgrind
+       output for HTML output with C.
+
 2024-04-29  Patrice Dumas  <pertusus@free.fr>
 
        * tp/Texinfo/Config.pm (GNUT_get_stage_handlers),
diff --git a/tp/TODO b/tp/TODO
index dc8e8081dd..e2622069c8 100644
--- a/tp/TODO
+++ b/tp/TODO
@@ -101,9 +101,23 @@ converter could check if the punctuation was present 
simply by checking
 $current->{'extra'}->{'following_punctuation'} or similar.
 
 
+Using callgrind to find the time used by functions
+
 C code could be checked to see if using an hash map implementation,
 by compiling C code as C++ and using the standard C++ library hash map
 could be interesting (Patrice 2023-10-14).
+Could be interesting for find_string:
+ unique_target -> find_string
+ (and, though much less used output_files_open_out -> find_string)
+
+Other possibility for optimisation, do something similar as in Perl with
+shared index information when parser is called from gdt/convert_substring,
+through simple_parser, to avoid the init_index_command call (in particular
+the call to add_index_command).
+
+Another possibility for optimization would be to call Perl less, as it
+still uses about 40% of the time (for html) even though it should mainly be
+for code called once (and not for Perl functions called from C).
 
 
 hyphenation: should only appear in toplevel.
@@ -167,6 +181,9 @@ to specify a locale and use it with strxfrm_l to sort, but 
it is only
 for testing and should not be kept in the long term, the plan is to replace by
 C code that sets Variable Weighting to Non-ignorable and before that keep
 calling Perl.
+Related glibc enhancement request:
+ request for Non-Ignorable Variable Weighting Unicode collation
+ https://sourceware.org/bugzilla/show_bug.cgi?id=31658
 
 
 HTML5 validation tidy errors that do not need fixing