As we bid farewell to 2020, here are 5 highly-rated InTime features added during a very eventful year.
- Auto Pilot - Automated recipes selection
- Project-specific AI database
- Support for Microchip FPGAs and Libero
- Training Data Filter - precision control of Machine Learning data
- New Analysis Charts - SLR Crossings, Fanout & Logic Levels Charts
1. Auto Pilot - Automated Recipes Selection
Let's say you want to run the "Hot Start" optimization recipe, followed by three rounds of "Default" and finally "Placement Exploration", but only if the WNS is better than -300ps. Before "Auto Pilot", you would have had to write a custom Tcl script or manually work the GUI at the end of each recipe.
The new Auto Pilot recipe gives you hands-free optimization and learning. InTime ships with ready-made Auto Pilot templates to get you started. For more details, see our earlier post on this topic.
2. Project-Specific Database
InTime was originally designed and structured around a user workspace, meaning that optimization results and meta-data are generated and saved within the user's folders, for example under your home folder. In some cases, this made data-sharing and collaboration more difficult as users had to export and import data from different users' folders to share optimization insights.
To overcome this restriction, we created a new database mode which centers around a project instead. If the project-specific database is enabled, the InTime design database will be saved within each individual project's folder instead of being associated with a single user or a single repository.
3. Welcome, Libero!
We are delighted to add support for the Libero software from Microchip in InTime and optimize Microchip FPGA designs.
New Tool Support - Libero, starting from version 12.4
New recipe: “Region Exploration” Examines module placements and finds more efficient regions for them. For example, a design may be squeezed into a flatter shape in order to be closer to critical resources.
4. Training Data Filter
As InTime uses machine learning methods to predict strategies, the type of data fed into the InTime determines the quality of its prediction.
Previously, InTime has been selecting data for its users in the background. With this new feature, users can understand what data is included in the prediction and what is excluded.
For example, there are different versions of the designs saved in the same database. And for a new run, InTime only needs to take into account the latest jobs. In this case, you can specify in the Local Job ID filter, e.g. ">5", which only use/show Job Id that are more than 5.
5. New Charts & Analysis
Who doesn't love more charts? In the latest 3.1.2, we have 3 new types of charts and analysis.
The newest addition is the Multi-die Analysis. This chart gives us an idea of the percentage of the paths crossing the die versus not crossing the die. Ideally, the lesser the crossings, the better is the timing performance.
There are 3 tables below this chart, that displays the following:
- Number of Crossings Versus Non-crossing Paths
- The worst slack of the SLR-crossing paths in each SLR
- Which instances has the most paths that are crossing SLR.
Using the information above can assist the designer in minimizing unnecessary crossings, floorplanning of the instances.
Logic Levels and Fanout Charts
InTime classifies the fanout and logic levels into 4 levels of criticality.
A larger “class” number indicates that the particular result has higher levels of fanout or logic levels. This helps to identify bottlenecks in the design. For example, for Fanout, InTime classifies them as such for the chart.
|Class 1||Fanout is less than 125|
|Class 2||Fanout is greater than 125 and less than 250|
|Class 3||Fanout is greater than 250 and less than 500|
|Class 4||Fanout greater than 500|
That's it. These are the top 5 features for 2020.
If you are interested to know more or try out InTime, please proceed to https://cloud.plunify.com/ to get a free evaluation copy.
Leave a Reply