vfLabs

vfLabs is a way for us to provide you with select features in advance of their full release. We do this when we’re trying out something new and we want to confirm whether it’s useful and whether it has been implemented in a way that’s natural for you.

You access vfLabs by clicking the vfLabs button in the vfAnalyst GUI. There you can see what features are in vfLabs and selectively enable the features you want to try.

Because the purpose of vfLabs is to gauge the benefit of certain features, we encourage feedback, offering up by email a quick survey of your impressions (typically less than six questions) some time after you enable and use a vfLabs feature. Based on the feedback we receive, we may release the feature into full production, we may rework it, or we may abandon it outright.

The features in vfLabs are of production quality; vfLabs is not a beta software area. We’re just making sure that vfLabs features are as effective as possible before committing them permanently to vfAnalyst.

Currently in vfLabs:

vfStream library

This C library provides streaming communication capabilities for synchronizing data between parallel tasks. vfStream supports both FIFO and Windowed FIFO streams (as depicted with a green line in the dependency pane).

CPU Skyline view

This feature adds a skyline view that shows the CPU usage for each invocation. Use the CPU Skyline to track how many CPUs you are using and to balance the load.

Execution Schedule view

This feature adds an Execution Schedule view for partitioned loops in the 1D and 2D profiles. This view displays an ASAP (or eager) execution schedule of the tasks in the partitioned loop. This shows you how the partitioned loop achieves its speedup and what dependencies may limit any further speedup.