I need some help in understanding qlikview's internal working machanism. Am I right to say that one qlikview document will only run as a single process which takes one core of the CPU? In order to take advantage of a multi-core processor, I need multiple qlikview document to be run at the same time?
The calculations for a single document may be automatically distributed across several cores at the same time. For example, different charts on the same sheet may be calculated in parallel by distributing each chart calculation to a separate processor. Even a single expression like "sum(Sales)" may be distributed across multiple cores to calculate different chunks of the table and then aggregating the results.
A single document will typically take advantage of multiple cores with no special effort on your part.
My experience is that QV is good about automatically using parallell cores. The details are usually important in only very large or troublesome applications.
(This is an opportunity for me to get on my soapbox about avoiding using macros. Macros tend to inhibit parallel processing).
Thanks for the prompt reply. Sorry for not being clear, but I am trying to load a large amount of data from csv or qvd file. Basically, the process is something like below.
Load master file--->inner join(masterfile) with fileA---> store combined table into fileA--->inner join(masterfile) with fileB--->store combined table into fileB--->...
I am wondering if I should split the process into different documents like
document 1: Load master file--->inner join(masterfile) with fileA--->store combined table into fileA
document 2: Load master file--->inner join(masterfile) with fileB--->store combined table into fileB
...
And use a scheduler to make them run at the same time. The master file is the bulk here with around 1 million records.
I misunderstood. You are asking about parallel processing in script execution (reloading), not document rendering. There is some automatic parallelization in loading, but I think it's mostly about calculation operations, not parallel load statements (although this has been discussed on the Ideas Forum).
My best advice is to test both methods. My guess is that two qvws/tasks will complete sooner, but it's worth testing.
It is worth monitoring the CPU usage across all cores whilst monitoring the reload progress window. A well designed load script should be using multiple cores, but if you see the load process just using one core for a significant time then this could indicate an area of the script that could be optimised, perhaps by replacing joins with applymap statements.
