Le 05/12/2021 à 05:32, sovhist a écrit :
My last mail is stil waiting moderation due to big atachments, so I'm
resending mail with lesser attachments.
Nothing changed with TNR DVMS and DVSM fonts.
OK. Good to know.
I'm sending 3 svg from Hotspot. I can see more in svg,
than in screen photo of Hotspot, but not all viewers can open them. I
can do screenshots if you
need them.
No this is very good.
I opended Lyx, hugenote document, scrolled one huge inset up and down, cliced
to other tab and scrolled usual document, after returned to hugenote and
scrolled a bit, if I remember properly.
Full are all graph, flame_1 is TextMetrics selected from first hugenote
scrolling and flame_2 is from last scrolling in hugenote, selected
lyx::frontend.
So, here is what I see (and do not see):
In flame_2, a lot of time is spent to draw the misspelling marks.
Indeed, for that it is necessary to find the position for beginning and
end of each misspelt word. What happens if you turn off on-screen spell
checking?
In flame_1, I do not see much, except maybe the fact that a lot of time
is spent acessing the cache (the part with QCache and QHash) of already
computing rows. This can be good (already is already computed) or bad
(the cache access are very slow), depending on the importance of this
part of code in the whole picture.
In full.svg, I do not see the time spent painting the screen (what is in
flame_2), so I cannot compare it to the time spend computing the
position of each element (basically the redoParagraph stuff). It would
be good to understand what is too slow.
Finally, could you check whether LyX is the only thing using cpu time?
What about the X server, for example?
JMarc
--
lyx-users mailing list
lyx-users@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-users
