on a buffer performance test on Windows GNU EMACS 28, 27, 26 and 24

on a buffer performance test on Windows GNU EMACS 28, 27, 26 and 24

[Wayne Harris via Emacs development discussions.]

On Windows, I said M-x run-python, then said

  for i in range(100000): print(i)

on both GNU EMACS 28, 27, 26 and GNU EMACS 24.  It turns out GNU EMACS
24 is the slowest.  There doesn't seem to be a difference between 28 and
27, but 26 is clearly slower too.

I timed the speed of the buffer to scroll up.  I used my own phone's
stop watch.  I started out the slow one first, which was EMACS 24, only
after it was running I started the clock, then I started GNU EMACS 28's
code.  The result was GNU EMACS 28 finishes in less than 8.00 seconds.
GNU EMACS 24 finishes after 24.44 seconds.  I didn't time 27 and 26, but
I couldn't tell any difference between 28 and 27.  And 26 was slower
than 27 and 28.

What is the reason for the difference?  Thanks!

GNU Emacs 24.3.1 (i386-mingw-nt6.2.9200) of 2013-03-17 on MARVIN
GNU Emacs 26.3 (build 1, x86_64-w64-mingw32) of 2019-08-29
GNU Emacs 27.1 (build 1, x86_64-w64-mingw32) of 2020-08-21
GNU Emacs 28.0.50 (build 1, i686-w64-mingw32) of 2020-07-05

Re: on a buffer performance test on Windows GNU EMACS 28, 27, 26 and 24

[Eli Zaretskii]

> Date: Tue, 25 Aug 2020 00:35:50 -0300
> From: Wayne Harris via "Emacs development discussions." <emacs-devel@gnu.org>
> 
> On Windows, I said M-x run-python, then said
> 
>   for i in range(100000): print(i)
> 
> on both GNU EMACS 28, 27, 26 and GNU EMACS 24.  It turns out GNU EMACS
> 24 is the slowest.  There doesn't seem to be a difference between 28 and
> 27, but 26 is clearly slower too.
> 
> I timed the speed of the buffer to scroll up.  I used my own phone's
> stop watch.  I started out the slow one first, which was EMACS 24, only
> after it was running I started the clock, then I started GNU EMACS 28's
> code.  The result was GNU EMACS 28 finishes in less than 8.00 seconds.
> GNU EMACS 24 finishes after 24.44 seconds.  I didn't time 27 and 26, but
> I couldn't tell any difference between 28 and 27.  And 26 was slower
> than 27 and 28.
> 
> What is the reason for the difference?  Thanks!

We changed the value of w32-pipe-read-delay to zero in Emacs 27,
that sounds like the likely reason.

Re: on a buffer performance test on Windows GNU EMACS 28, 27, 26 and 24

[Wayne Harris via Emacs development discussions.]

Eli Zaretskii <eliz@gnu.org> writes:

>> Date: Tue, 25 Aug 2020 00:35:50 -0300
>> From: Wayne Harris via "Emacs development discussions." <emacs-devel@gnu.org>
>> 
>> On Windows, I said M-x run-python, then said
>> 
>>   for i in range(100000): print(i)
>> 
>> on both GNU EMACS 28, 27, 26 and GNU EMACS 24.  It turns out GNU EMACS
>> 24 is the slowest.  There doesn't seem to be a difference between 28 and
>> 27, but 26 is clearly slower too.
>> 
>> I timed the speed of the buffer to scroll up.  I used my own phone's
>> stop watch.  I started out the slow one first, which was EMACS 24, only
>> after it was running I started the clock, then I started GNU EMACS 28's
>> code.  The result was GNU EMACS 28 finishes in less than 8.00 seconds.
>> GNU EMACS 24 finishes after 24.44 seconds.  I didn't time 27 and 26, but
>> I couldn't tell any difference between 28 and 27.  And 26 was slower
>> than 27 and 28.
>> 
>> What is the reason for the difference?  Thanks!
>
> We changed the value of w32-pipe-read-delay to zero in Emacs 27,
> that sounds like the likely reason.

Lol.  I've been using ``slow'' GNU Emacs 24 for years without ever
thinking I could set a variable and have it speed up quite a bit.

As you know, w32-pipe-read-delay is 50 in GNU Emacs 24 and setting it to
zero makes it as fast as GNU Emacs 27.  I don't have to upgrade after
all.  But now I did.  It's very good to know that I can go back if
somehow I need to.  Thank you for the information.