diff options
| author | Paul Eggert | 2025-12-19 09:09:16 -0800 |
|---|---|---|
| committer | Paul Eggert | 2025-12-19 09:10:31 -0800 |
| commit | 4d11449c96c80e2ea402c88fe637958b8c0cf940 (patch) | |
| tree | bb82b10e60078128623a9f33e4dbdbfb0bea932d /admin/notes/java | |
| parent | a6c2ae3a3fb89cd0eedd1b6b4e81fb90ce910f96 (diff) | |
| download | emacs-4d11449c96c80e2ea402c88fe637958b8c0cf940.tar.gz emacs-4d11449c96c80e2ea402c88fe637958b8c0cf940.zip | |
Revert to simpler (and we hope faster) TAGGEDP
Suggested by Mattias Engdegård (bug#80021#17).
* src/lisp.h (lisp_h_FIXNUMP, lisp_h_TAGGEDP):
Go back to the simpler (X&7) == TAG approach for checking object tags.
This reverses my commit ccdb08ef4ed8f96e79aa06cf5e806c9c487d58ad
“Improve performance of CONSP, FIXNUMP, etc.”
dated 2018-08-25 13:39:18 -0700,
though it keeps the TAGGEDP function the older commit introduced.
Although the older commit improved performance on its circa 2010
platform, when I ran today’s ‘make -C lisp compile-always’
benchmark on Ubuntu 25.10 which uses gcc (Ubuntu 15.2.0-4ubuntu4)
on an circa-2021 Intel Xeon W-1350, this patch makes the
‘make -C lisp compile-always’ benchmark 3.1% faster. Although the
patch unfortunately also makes the Emacs text segment 0.6% larger,
in this case speed and simplicity beat text size in importance.
Diffstat (limited to 'admin/notes/java')
0 files changed, 0 insertions, 0 deletions