APPROVE COMMIT NOTE: This patch has been committed. This will also improve performance for large VM buffers, and I wouldn’t be surprised if it gave the largest improvement of the changes I’ve made. It does this by reducing the calls to Lstream_read(), and by matching the buffer size to the coding buffer size, so there’s less (though not no) need for adjustment of the character_tell() result due to the the two getting out of sync. # HG changeset patch # User Aidan Kehoe <kehoea(a)parhasard.net&gt; # Date 1389903987 0 # Node ID 319e18d08654e2b6b6a20bc496349fcc959ed69f # Parent ccaa851ae712e870ceb65d19226426d40d8de48c Increase READ_BUF_SIZE, used in #'insert-file-contents-internal, to #x10000 src/ChangeLog addition: 2014-01-16 Aidan Kehoe <kehoea(a)parhasard.net&gt; * fileio.c (READ_BUF_SIZE): Stack sizes > 2**16 elicited bugs back in August 1996, but there's nothing to indicate this is still the case. GNU uses a block size of #x10000, and that size works better with the coding system buffering, improving performance there; move to this value. diff -r ccaa851ae712 -r 319e18d08654 src/ChangeLog --- a/src/ChangeLog Thu Jan 16 19:56:06 2014 +0000 +++ b/src/ChangeLog Thu Jan 16 20:26:27 2014 +0000 ＠＠ -1,3 +1,11 ＠＠ +2014-01-16 Aidan Kehoe <kehoea(a)parhasard.net&gt; + + * fileio.c (READ_BUF_SIZE): + Stack sizes > 2**16 elicited bugs back in August 1996, but there's + nothing to indicate this is still the case. GNU uses a block size + of #x10000, and that size works better with the coding system + buffering, improving performance there; move to this value. + 2014-01-16 Aidan Kehoe <kehoea(a)parhasard.net&gt; * file-coding.c: diff -r ccaa851ae712 -r 319e18d08654 src/fileio.c --- a/src/fileio.c Thu Jan 16 19:56:06 2014 +0000 +++ b/src/fileio.c Thu Jan 16 20:26:27 2014 +0000 ＠＠ -2870,9 +2870,7 ＠＠ }  -/* Stack sizes > 2**16 is a good way to elicit compiler bugs */ -/* #define READ_BUF_SIZE (2 << 16) */ -#define READ_BUF_SIZE (1 << 15) +#define READ_BUF_SIZE (2 << 16) DEFUN ("insert-file-contents-internal", Finsert_file_contents_internal, 1, 7, 0, /* -- ‘Liston operated so fast that he once accidentally amputated an assistant’s fingers along with a patient’s leg, […] The patient and the assistant both died of sepsis, and a spectator reportedly died of shock, resulting in the only known procedure with a 300% mortality.’ (Atul Gawande, NEJM, 2012) _______________________________________________ XEmacs-Patches mailing list XEmacs-Patches(a)xemacs.org http://lists.xemacs.org/mailman/listinfo/xemacs-patches