Iris incredibly slow

[Harriet Bazley]

Is it normal for Iris to be more or less unusably slow on an ARMX6?
I've tried twice to book an event via Eventbrite; the text entry was at
less than 1 character per second (I think the site was running some kind
of JavaScript check on every single keypress), and when I eventually got
down to 'Select Order' after struggling to correct typos resulting from
extreme delayed reaction, the site just timed out on an attempt to pay
by credit card - not the only time this has happened, although last time
it got further into the credit card check process before to all intents
and purposes freezing up....

At this point it has currently been attempting to open the next popup or
whatever the action is supposed to be for the last ten minutes, and I
don't think it is going to get anywhere with it.

[Paul Sprangers]

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?

Well, at times it is unusably slow even on my 4té2, sometimes when entering
characters indeed, but certainly not always. I'll see if I can find a
pattern.

I did discover however, that Iris consumes a lot of processor power (30 -
35%) when being idle. Not that it explains its slowness, but I'd like to
report it somewhere. Does some sort of an Iris mailing list exist?

Kind regards,
Paul

[Theo]

Post by Paul Sprangers

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?

Well, at times it is unusably slow even on my 4té2, sometimes when entering
characters indeed, but certainly not always. I'll see if I can find a
pattern.
I did discover however, that Iris consumes a lot of processor power (30 -
35%) when being idle. Not that it explains its slowness, but I'd like to
report it somewhere. Does some sort of an Iris mailing list exist?

It would be worth trying to correlate which website(s) were open at the
time. In web browsers there's no such thing as 'idle', because Javascript
events are firing all the time doing things like animating web page elements
(eg ads). It may be that particular sites are better or worse for this.

Judging by the Task Manager in browsers on other platforms, some web pages
have a particularly high background load. Blocking ads can help a lot
there, although I don't think there's a way to run an adblocker on Iris?
A network-based one like a Pi-hole may be a good start though.

Perhaps disabling Javascript in Iris would be worth a shot to see if the
idle CPU usage decreases?

Theo

[Harriet Bazley]

On 14 Jul 2023 as I do recall,
Theo wrote:


[snip]

Post by Theo
Judging by the Task Manager in browsers on other platforms, some web pages
have a particularly high background load. Blocking ads can help a lot
there, although I don't think there's a way to run an adblocker on Iris?
A network-based one like a Pi-hole may be a good start though.
Perhaps disabling Javascript in Iris would be worth a shot to see if the
idle CPU usage decreases?

I don't think Eventbrite would work at all with Javascript disabled; I
can't get beyond the first page of the event with Netsurf.

[Theo]

Post by Harriet Bazley
On 14 Jul 2023 as I do recall,
[snip]

Post by Theo
Judging by the Task Manager in browsers on other platforms, some web pages
have a particularly high background load. Blocking ads can help a lot
there, although I don't think there's a way to run an adblocker on Iris?
A network-based one like a Pi-hole may be a good start though.
Perhaps disabling Javascript in Iris would be worth a shot to see if the
idle CPU usage decreases?

I don't think Eventbrite would work at all with Javascript disabled; I
can't get beyond the first page of the event with Netsurf.

It may not work, but would at least tell you if the site's Javascript is the
cause of high CPU utilisation when doing not-very-much.

Another possibility is the timers are off - eg the site has animation
effects when navigating, and it uses Javascript timers to run them. If the
timer granularity was wrong (eg uses centisecond RISC OS timer rather than a
millisecond timer and so everything rounds up to steps of 10 milliseconds)
then the effect could go drastically slower.

Can you get to the problematic page, then disable Javascript and see if it's
still slow? Or does disabling JS mean you need to reload the page / restart
the browser?

(I'm not familiar with Iris, so have no special knowledge here)

Theo

[Harriet Bazley]

On 14 Jul 2023 as I do recall,
Theo wrote:

[snip]

Post by Theo
Can you get to the problematic page, then disable Javascript and see if it's
still slow? Or does disabling JS mean you need to reload the page / restart
the browser?

Not really an option for experimentation, since it's the 'click here
to pay now' page. :-(

(I tried it again on a mobile phone, and it's still slow, but it turns
out the problematic popup was the Visa verification window where they
want you to use a card reader to prove that you have the card in hand -
what rather alarms me is that I have just realised that, being
flustered, I entered my *old* post code where the site insisted on a
'Zip code', rather than the one associated with my current debit card,
and the payment apparently went through! So much for verification....)

There doesn't seem to be any visible way to disable Javascript on Iris,
unless it's the 'Use JIT' under General iconbar choices....


(The site nearly gave me a heart attack by proceeding to announce that I
had successfully purchased *six* tickets, which I thought must be the
effect of my repeated attempts to access it via Iris and would amount to
a bill of many hundreds of pounds - but it turned out that this simply
refers to the multiple free 'events' on the same day!)

[Chris Hughes]

Post by Theo

Post by Paul Sprangers

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?

Well, at times it is unusably slow even on my 4té2, sometimes when entering
characters indeed, but certainly not always. I'll see if I can find a
pattern.
I did discover however, that Iris consumes a lot of processor power (30 -
35%) when being idle. Not that it explains its slowness, but I'd like to
report it somewhere. Does some sort of an Iris mailing list exist?

It would be worth trying to correlate which website(s) were open at the
time. In web browsers there's no such thing as 'idle', because Javascript
events are firing all the time doing things like animating web page elements
(eg ads). It may be that particular sites are better or worse for this.

This a fair comment, but we have to remember RISC OS and thus Iris can
only run on one core of the processor so that can impact performance.

Post by Theo
Judging by the Task Manager in browsers on other platforms, some web pages
have a particularly high background load. Blocking ads can help a lot
there, although I don't think there's a way to run an adblocker on Iris?
A network-based one like a Pi-hole may be a good start though.

No adblocker available in Iris and will be no use anyway soon if Google
get their way, as they are trying to prevent adblockers from working
properly with various changes they are pushing through.

Post by Theo
Perhaps disabling Javascript in Iris would be worth a shot to see if the
idle CPU usage decreases?

You can't turn it off in Iris as things stand and if you did a lot of
stuff might not work nowadays. Rightly or wrongly.

[Paul Sprangers]

Post by Theo
It would be worth trying to correlate which website(s) were open at the
time. In web browsers there's no such thing as 'idle', because
Javascript events are firing all the time doing things like animating
web page elements (eg ads)

Iris claims between 18% and 52% of the processor usage (it's going up and
down all the time), after just being loaded, so when no page has been
displayed yet.

Displaying the Google search page, without any noticeable adds or
animations, claims around 40% processor usage, divided over
WebKitNetworkProcess, WebKitWebProcess and Iris itself. The same is true
for the built in ROD start page. By comparison: NetSurf doesn't claim any
usage displaying those pages.

Not that it really matters. As soon as another program needs the full
power, Iris and its companions obediently shrink to 0%. But since they do
consume power when doing nothing noticeable, and although an RPI is
extremely power efficient, I prefer to run Iris only when needed.

By the way, my version is 1.027. Where can we get the newer version?

Kind regards,
Paul

[Chris Hughes]

Post by Paul Sprangers

Post by Theo
It would be worth trying to correlate which website(s) were open at the
time. In web browsers there's no such thing as 'idle', because
Javascript events are firing all the time doing things like animating
web page elements (eg ads)

[snip]

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came out in
May 2023, the main change I believe is adding Printing features.

Changes in v1.028 are listed as:

Sat May 20 2023

* Iris (v1.028)
Add support for landscape printing.
Show the RISC OS print dialogue instead of the GTK one when
requested from
Javascript.
Fix issue with hard spaces in pathnames.
Translate UTF8 web page titles to Latin1 for use in the Browser
window title bar.
* GDK
Try to allow through non topbit set characters that are
generated via the ALT key.
Specifically ALT-Q which is used on an Icelandic keyboard layout
to produce '@'.

[Chris Newman]

Post by Harriet Bazley

Post by Paul Sprangers

Post by Theo
It would be worth trying to correlate which website(s) were open at
the time. In web browsers there's no such thing as 'idle', because
Javascript events are firing all the time doing things like
animating web page elements (eg ads)

[snip]

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came out
in May 2023, the main change I believe is adding Printing features.

Is this update not availble to all?
I am not one of the early subscribers. I merely bought it from the
!Store. All I can see doing that is the Apr2023 iteration which I already
have.

[Chris Hughes]

Post by Chris Newman

Post by Harriet Bazley

Post by Paul Sprangers

Post by Theo
It would be worth trying to correlate which website(s) were open at
the time. In web browsers there's no such thing as 'idle', because
Javascript events are firing all the time doing things like
animating web page elements (eg ads)

[snip]

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came out
in May 2023, the main change I believe is adding Printing features.

Is this update not availble to all?
I am not one of the early subscribers. I merely bought it from the
!Store. All I can see doing that is the Apr2023 iteration which I already
have.

It might depend on the link you have on your start page. There appears to
be two different links. I am not going to publish them here.

You could always email ROD asking if the v1.028 is available to download
as you obtained your link via PlingStore.

[Frederick]

[snip]

Post by Chris Hughes

Post by Chris Newman

Post by Harriet Bazley
[snip]

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came
out in May 2023, the main change I believe is adding Printing
features.

Is this update not availble to all?
I am not one of the early subscribers. I merely bought it from the
!Store. All I can see doing that is the Apr2023 iteration which I
already have.

It might depend on the link you have on your start page. There appears
to be two different links. I am not going to publish them here.
You could always email ROD asking if the v1.028 is available to download
as you obtained your link via PlingStore.

Just to add, the link given as a perk at the SW Show also doesn't show
v1.028.

[Chris Newman]

Post by Chris Hughes

Post by Chris Hughes

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came
out in May 2023, the main change I believe is adding Printing
features.

Is this update not availble to all? I am not one of the early
subscribers. I merely bought it from the !Store. All I can see doing
that is the Apr2023 iteration which I already have.

It might depend on the link you have on your start page. There appears
to be two different links. I am not going to publish them here.
You could always email ROD asking if the v1.028 is available to
download as you obtained your link via PlingStore.

Ta. Interestingly, the vers 1.027 I have on my FAST machine stiffs the
machine when I quit Iris about 3 times out of 5. I've tried closing all
other running things before I shut Iris but seems to make no difference.
I can't fathom why this is happening. Nobody else seems to have reported
it happening to them.

[Chris Hughes]

Post by Chris Newman

Post by Chris Hughes

Post by Chris Hughes

Post by Paul Sprangers
By the way, my version is 1.027. Where can we get the newer version?

Go to the download link you will have on your start page, it came
out in May 2023, the main change I believe is adding Printing
features.

Is this update not availble to all? I am not one of the early
subscribers. I merely bought it from the !Store. All I can see doing
that is the Apr2023 iteration which I already have.

It might depend on the link you have on your start page. There appears
to be two different links. I am not going to publish them here.
You could always email ROD asking if the v1.028 is available to
download as you obtained your link via PlingStore.

Ta. Interestingly, the vers 1.027 I have on my FAST machine stiffs the
machine when I quit Iris about 3 times out of 5. I've tried closing all
other running things before I shut Iris but seems to make no difference.
I can't fathom why this is happening. Nobody else seems to have reported
it happening to them.

I have seen this a couple of times in the past, usually where a page has
failed to load for some reason.

Found it was a stuck Webkit entry in Application tasks, that was still
trying to load the page in the background.

[Chris Newman]

<snip>

Post by Chris Hughes

Post by Chris Newman
Ta. Interestingly, the vers 1.027 I have on my FAST machine stiffs
the machine when I quit Iris about 3 times out of 5. I've tried
closing all other running things before I shut Iris but seems to make
no difference. I can't fathom why this is happening. Nobody else
seems to have reported it happening to them.

I have seen this a couple of times in the past, usually where a page
has failed to load for some reason.
Found it was a stuck Webkit entry in Application tasks, that was still
trying to load the page in the background.

Ta. Is there a cure?

[Steve Fryatt]

On 15 Jul, Paul Sprangers wrote in message

By comparison: NetSurf doesn't claim any usage displaying those pages.

Unless the page has animated GIFs on it, NetSurf has very little to do in
the background once a page has loaded. As others have pointed out, that
isn't necessarily true once Javascript is included in the mix.

Not that it really matters. As soon as another program needs the full
power, Iris and its companions obediently shrink to 0%.

That's the nature of RISC OS's cooperative multitasking, which has pros and
cons.

But since they do consume power when doing nothing noticeable, and
although an RPI is extremely power efficient, I prefer to run Iris only
when needed.

That may not be relevant. What would matter is whether they prevent the
system from throttling back the clock speed when idle, which something like
CPUClock will tell you.

[Theo]

Post by Paul Sprangers
Iris claims between 18% and 52% of the processor usage (it's going up and
down all the time), after just being loaded, so when no page has been
displayed yet.
Displaying the Google search page, without any noticeable adds or
animations, claims around 40% processor usage, divided over
WebKitNetworkProcess, WebKitWebProcess and Iris itself. The same is true
for the built in ROD start page. By comparison: NetSurf doesn't claim any
usage displaying those pages.

Google is full of Javascript: it is an ad network after all, so it's doing
plenty of tracking. Not just that, but things like suggestions as you type
will be in operation.

I just tried it and simply loading the front page and pressing Reject on the
cookies banner loaded 1.7MB of Javascript. To display the front page on my
x86 laptop running Linux, it took Firefox 600ms of CPU time running the
Javascript.

A more basic site to try would be https://example.com/ which is static
HTML+CSS with no JS.

Theo

[Chris Hughes]

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?

Well, at times it is unusably slow even on my 4te2, sometimes when entering
characters indeed, but certainly not always. I'll see if I can find a
pattern.

The issue of slow text input is already known to the developer and ROD,
they have done some tweaks to try an improve it.

I did discover however, that Iris consumes a lot of processor power (30 -
35%) when being idle. Not that it explains its slowness, but I'd like to
report it somewhere. Does some sort of an Iris mailing list exist?

Have you looked at how much memory is being used by Iris and associated
bits. Just looked at the tasks window and it using 100+K plus Shared
Libraries over 116K and that while it is just sitting on the icon bar.

It will be your biggest application on your RISC OS machine.

You do not say which version of Iris you are using I am on v1.028

The ARMX6 is considered the minimum spac RISC OS computer for use with
Iris, faster machine like the Pi4 (4te2) running at 2 Ghz are better, but
does depend a lot on the site and how much junk is being downloaded in the
background from any given websites, the more tabs or Iris windows you have
open the more speed can be impact.

Remember we are only using one core of the processor.

[Harriet Bazley]

On 14 Jul 2023 as I do recall,
Chris Hughes wrote:

[snip]

Post by Chris Hughes
Have you looked at how much memory is being used by Iris and associated
bits. Just looked at the tasks window and it using 100+K plus Shared
Libraries over 116K and that while it is just sitting on the icon bar.
It will be your biggest application on your RISC OS machine.

The biggest by miles, yes. I was running it from RAMdisc in an attempt
to speed it up; it is using 142Mb of temporary disc space plus 6304K
for Iris plus 8960K 'WebKitNetworkProcess" plus three dynamic areas of
2Mb each, plus 116580K of Shared Libraries... plus 7072K for 'Cairo',
which I assume is probably something to do with it as well!

Post by Chris Hughes
You do not say which version of Iris you are using I am on v1.028

It reports itself as v1.027 of April 2023 - I wasn't aware there was a
more recent release.

Post by Chris Hughes
The ARMX6 is considered the minimum spac RISC OS computer for use with
Iris, faster machine like the Pi4 (4te2) running at 2 Ghz are better, but
does depend a lot on the site and how much junk is being downloaded in the
background from any given websites, the more tabs or Iris windows you have
open the more speed can be impact.

This was from a fresh reboot with only the Iris window open on screen.
It did seem a little faster under those circumstances than my prior
attempt with lots of other applications loaded, but that may have been
because I'd just been through all the forms once already and wasn't
wasting a lot of time scrolling around trying to work out to type where!

[J Peachey]

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?
I've tried twice to book an event via Eventbrite; the text entry was at
less than 1 character per second (I think the site was running some kind
of JavaScript check on every single keypress), and when I eventually got
down to 'Select Order' after struggling to correct typos resulting from
extreme delayed reaction, the site just timed out on an attempt to pay
by credit card - not the only time this has happened, although last time
it got further into the credit card check process before to all intents
and purposes freezing up....
At this point it has currently been attempting to open the next popup or
whatever the action is supposed to be for the last ten minutes, and I
don't think it is going to get anywhere with it.

I've just had paly with the eventbright site, and buying a ticket, it
didn't seem any slower than any other site. The site does seem to have a
mass of chunky graphics which may be slowing things down. Doesn't appear
to be lots of animations etc.
This test done with a Titanium and Iris 1.027

john

[Harriet Bazley]

On 14 Jul 2023 as I do recall,

Post by J Peachey

Post by Harriet Bazley
Is it normal for Iris to be more or less unusably slow on an ARMX6?
I've tried twice to book an event via Eventbrite; the text entry was at
less than 1 character per second (I think the site was running some kind
of JavaScript check on every single keypress), and when I eventually got
down to 'Select Order' after struggling to correct typos resulting from
extreme delayed reaction, the site just timed out on an attempt to pay
by credit card - not the only time this has happened, although last time
it got further into the credit card check process before to all intents
and purposes freezing up....

[snip]

Post by J Peachey
I've just had paly with the eventbright site, and buying a ticket, it
didn't seem any slower than any other site. The site does seem to have a
mass of chunky graphics which may be slowing things down. Doesn't appear
to be lots of animations etc.

I didn't notice any animations at all - and I remember I did have the
same issue when trying to buy visitors' parking permits from my local
council's website (which is otherwise very accessible). Iris simply
timed out again at the final stage of the purchase process; whatever
Visa is doing is clearly very computationally complex.

I suppose it might have worked if I had been using the Paypal or Google
Pay options!