|
PureBytes Links
Trading Reference Links
|
> However, the crux of TW's performance
> problems
> do not stem from the data crunching, but rather from extremely ineffecient
> screen drawing.
Au contraire. The BIGGEST problem is the architecture used for studies.
It was never clear to me if that was a requirement of the program or
just sloppy study code. My impression was that it should have been
possible to work around the problem but it was present in all the canned
studies. Whatever, it would be IMPOSSIBLE for any PC to run any complex
studies, using the format of the canned studies, on a large amount of
data, without slowing to a crawl realtime.
Let's say you want to plot an average on 10,000 bars and you want to
update it every tick. Every time a new tick comes in, TW will
recalculate the whole 10,000 bars. That's an incredibly inefficient way
to do things. You know how long it takes to load a chart the first time?
Imagine having to go through that delay everytime a tick comes in. You
can duplicate that slow performance in TS by saying something like:
{set to update every tick}
if lastbaronchart then for i = 10000 downto 0 begin
plot1[i](average(c,10)[1],"");
end;
I was a very early beta tester. I asked Steve about the problem and his
answer was to run the study on fewer bars and set it to update less
frequently. He offered no workaround and wouldn't even acknowledge that
there was a problem. I immediately lost interest and put it back on the
shelf.
--
Dennis
|