Login Page - Create Account

Support Board


Date/Time: Thu, 16 May 2024 05:00:19 +0000



Post From: Scale beyond one CPU

[2018-04-27 12:22:12]
QnReally - Posts: 181
Hello,

During the startup and a few other times SC is able to utilize multiple CPUs to speed things up. However, once it starts up and reaches a steady state where it continues to receive and process tick data for its charts and other things, SC is only able to run on a single CPU (as far as I can tell -- at least, if I add a few more studies beyond its one CPU capability, SC can really get bogged down). Here are some suggestions for the future consideration if and when you get time to look into these:
1) Look for some tasks that can be done in parallel, even when you have inter-chart dependencies and inter-study dependencies that force you to become sequential
2) possibly, delegate the screen painting job to either some additional CPUs or a GPU, when feasible

Nowadays, every computer has multiple CPU cores by default and most computers have dedicated GPU hardware. It would be great to further optimize SC to take advantage of these unused resources.

I understand this can be a major undertaking but this will make SC much more scalable. The current way of distributing the load among multiple instances of SC just to get around this adds an overhead and burdensome on the user to keep track of all the dependencies. This makes the setup fairly brittle and it makes the user really think through when they need to make a change. This is especially true for those of us who write their own custom studies.

Thanks.