Slow-down in typing/mouse?

[Ron Kaplan]

I have the impression that there is slight but noticeable delay in the echo from typing and the effect of mouseclicking is a little slower in the current master. Has anyone else had that impression?

M2 Mac running Tahoe 26.1.

[Paolo Amoroso]

Yes, I've noticed a similar slow-down on my Linux Mint 22.1 Cinnamon PC. My benchmark is clicking the NOTECARDS button in an apps sysout session: now it may take up to 3-4 clicks for the action to take effect, whereas a few weeks ago it was instantaneous.

On Mon, Nov 24, 2025 at 2:44 AM Ron Kaplan <ron.k...@post.harvard.edu> wrote:

I have the impression that there is slight but noticeable delay in the echo from typing and the effect of mouseclicking is a little slower in the current master.  Has anyone else had that impression?

M2 Mac running Tahoe 26.1.

--
You received this message because you are subscribed to the Google Groups "Medley Interlisp core" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lispcore+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/lispcore/ECDFCE07-29F9-44D5-B479-1E6084546FCC%40post.harvard.edu.

--

Paolo Amoroso | Blog | Mastodon

[Paolo Amoroso]

When I change BACKGROUND-YIELD to also call CAUSE-INTERRUPT responsiveness improves for both the keyboard and mouse. However, the system GCs continuously: the mouse pointer switches to the GC bitmap and flickers continuously. The pointer never flickered like that prior to the BACKGROUND-YIELD changes.

This happens on both master and rmk143--Flush-font-caches.

[Nick Briggs]

Did you compile it after you changed it?

To view this discussion visit https://groups.google.com/d/msgid/lispcore/CAGi1hzuCMxbCvH%3DqaSt2Yx%3DorCmmLUzGfFb88ySeMoueLqiZCQ%40mail.gmail.com.

[Paolo Amoroso]

No, I forgot. After compiling the change I get the responsiveness improvement without continuous GC activity.

[Nick Briggs]

Try starting it without the cause-interrupt but with -timer 5000 as an option (I know run-medley passes that through to lde, not sure if all the startup scripts do).  I think the default timer interrupts are infrequent enough that we see the input delays, and the SIGIO for input available isn’t reliable and/or responsive enough to keep it feeling snappy 

— Nick

Sent from my iPhone

On Nov 25, 2025, at 08:28, Paolo Amoroso <paolo....@gmail.com> wrote:



[Paolo Amoroso]

When starting Medley with medley -a -e -n -v + -- -timer 5000 I get the responsiveness improvement too.

[Matt Heffron]

I just pulled the latest master branch and built full loadup.

Before, with older saved .virtualmem files, I didn’t see the slow-down issues.

With this new full.sysout, now I do.

I tried adding the  -- -timer 5000 and I do see some improvement, but still not as responsive as previously.

 

Matt

To view this discussion visit https://groups.google.com/d/msgid/lispcore/CAGi1hzuPrdDE%2BRhEeHqueZjJQy55jKJ3qp3wPV7bbrzdsKOWqA%40mail.gmail.com.

[Nick Briggs]

Try some smaller timer values, see what the largest value is that still makes it feel responsive enough, and let me know.   The default for the timer value is 10000.

-- Nick

[Matt Heffron]

I tried 2000, 1000, 500, and 100.

They were better than 5000 (but less than proportionate improvement).

None were as good as opening an older .virtualmem file (without any -timer).

The lag was still noticeable.

 

Matt

[Paolo Amoroso]

I played a bit with timer values and agree with Matt's conclusions. In general it's hard to notice differences under 5000.

[Nick Briggs]

I've been looking at the code and the timing - the keyboard (X event queue) gets checked every time that interrupts are checked - which is triggered by the (virtual time) timer firing, by a SIGIO signal, and when cause-interrupt [that is, (SUBRCALL CAUSE-INTERRUPT)] is executed.  cause-interrupt is called about 100 times more frequently off the backgroundfns (via BACKGROUND-YIELD) than anything else.

The X code has never been (in my knowledge of recent times) enabled the SIGIO interrupt for the display connection - I'm going to see if that works on macOS and whether it makes a difference to the responsiveness.  Turning the -timer option to larger values makes the slowdown more visible to me.

-- Nick

To view this discussion visit https://groups.google.com/d/msgid/lispcore/f525b04e-5cfd-4988-b6f0-1bb9136fbb9en%40googlegroups.com.

[Matt Heffron]

Also, just for complete information, I’m running on Windows WSL Ubuntu 24.04.3 using X, not VNC.

 

Matt

To view this discussion visit https://groups.google.com/d/msgid/lispcore/FFB98F55-1E51-422F-ABCA-8DFCED971AF0%40gmail.com.

[Nick Briggs]

Enabling SIGIO on the X11 connection makes it responsive even if the timer value is set to 1000000.  SIGIO on the X11 connection works on macOS, I'm guessing it will work on most of the other systems we tend to use.  Not sure about more obscure(?) systems like Haiku.  I will have to look at what happens with SDL, which has a different I/O architecture.

-- Nick