I have a csv with a million rows to be processed though a workspace. How to reduce the run time.

Have you tried turning off featurecaching? (Run--> Enable Feature Caching)

It would really depend on what you're doing to the data. CSV reading would be lightning quick, even with a million rows. Please add some more details about the data transformation and writing.

To be more specific, it is getting more time at one of the attribute manager where I am using +23rd Feature value of date attribute

I think I have tried that, but it has not given significant time saving​

I have ticked​ "Enable Adjacent Feature Attributes" with "23" number of subsequent features

With adjacent features enabled, the AttributeManager becomes a blocking transformer and bulk mode is lost. FYI this is a known issue - FMEENGINE-63455

I want to perform an operation where I want to add a new attribute which will have value of next 23rd attribute. Is there any quicker way to do this?

Sounds like you have relationships between rows. I would then try to write the data in to a relational database. cause it would reduce any blocking. Then read the relational data instead.

Before writing to the relational database you should try to create a clear design on how to make the data clean and valid. Maybe showing ChatGPT a piece of your datastructure would help.