• Content count

  • Joined

  • Last visited

About johnteaton

  • Rank

Profile Information

  • Gender
  • Location
    Brush Prairie, Wa
  1. Reuse 2016 will be held on 1 Dec at the computer history museum in Mt View Ca. Is anyone from this group attending or presenting? John Eaton
  2. I have a question about using the new ip-Xact catalog file. You can use this to get the relative path to any other ip_Xact file including other catalog files. How do we handle pointing to another catalog file that is contained in a different RCS repository? To do this you need to know where the user placed that repository in their design environment relative to your repository and the name that they called it. You have no way of knowing either piece of information. If you mandate the name and location for that repository then you are violating the fundamental IP-Xact principle that says the user has full control of all file and directory names. John Eaton
  3. Olof, Standardized names only work if you have a single organization with enough clout to impose their naming conventions onto the entire industry. Those companies are very rare and suffer from a problem with lag time. A lone developer needs a new name NOW and may not be able to guess the final name is that will be chosen. The beauty of IP-Xact is that anyone with a valid URL can create VLNV descriptors that are unique through out the world. We need an IP-Xact file where you can create a new fileType with your name and VLNV. If anyone else uses the same language with a different name then they can extend your VLNV to alias their name. That way when their components reference their names it gets translated into your names. This will probably be similar to how buses can be extended using a new VLNV. John Eaton
  4. I don't see how you could ever hope to put these items in a standard and not have it obsolete by the next day. I am now working with code written in chisel so I use fileType user for now but what happens when chisel is added as a fileType? Do all my files break or do I have to check for either case? John Eaton
  5. Sachin, I don't understand what you mean by sub-blocks at different locations. A sub-block of an ip-xact component is itself another ip-xact component. You download all of your ip into your design environment and every component is stored in its own place. If it is used by multiple other componennts then they all get its files from that one location. The best way that I have found to use environment variables is to run all off your toolflow scripts from the top level of your design environment. If you do that then $HOME provides the absolute pathname to your design environment and everything else is relative to that. John Eaton
  6. Hello all, I have been using ip-xact since the 2009 to drive my digital tool flows and I am disapointed at the slow rate of adoption that I see in this industry. It is time to admit that our marketing plan of "if we build it then they will come" has not been working and come up with a real plan. Our target customer is not the tool smith,ip creator or silicon vendor. Those groups are already on board. Our target has to be the soc designers who create the designs and bring them to production. How do we convince them to change their tried and true methods for something new? First of all we show them that it can be done. Find some teams that have already adopted ip-xact and publish articles about how they did it with all the plus and minuses. Release a reference design environment with open sourced ip ,tools and toolflows that uses ip-xact to build an embedded system on a fpga demo board Write a handbook. Jtag is a very popular standard but nobody designs it to the 1149.1 spec. Everyone buys a copy of Ken Parkers "Boundary Scan Handbook" and designs to that. The standard tells you WHAT. The handbook tells you WHAT,WHY and HOW. With lots of working examples. Simplify the standard. Jtag consists of a family of standards from 1149.1 up to 1149.7. The dot 1 standard is very simple and only defines the pieces that all the other standards require. A lot of engineers can do everything they need using only 1149.1. We need to split off the fundamentals into a 1685.1 and add a new dot for each major feature. Write an elevator speech. You meet a target customer at some conference and you only have a few minutes to convince them to look further into ip-xact. How do you introduce ip-xact without sounding like your trying to sign them up of some multilevel marketing scheme. What will catch their interest. Learn from our elders. Back in the 90's the PCB world came up with a universal data interchange format called EDIF. It was supposed to take any design created in any schematic capture system and transfer it to any other. It never really met its goals. All the vendors adopted it but did so with their own unique "flavor" of the standard. You never really could get everything to transfer between all the different vendors. Explain how ip-xact is not going to suffer the same fate with vendorExtensions. Scare the crap out of them. They have been designing chips succesfully for 20 years using the same techniques so why should they change now? Because those chips have been growing and we are now to the point where an IC design environment is starting to behave like "Big Data". The rules don't change with "Big Data" but they are enforced more often. You can make a lot of mistakes in a small data environment and still muddle through and get a chip. Try that with "Big Data" and you will fail and your failure will be spectacular. "Big Data" will show us which teams have real engineers and which ones have hackers. Our industry has thousands of IP creators designing for thousands of IP users and ip-xact is the only standard we have to govern that exchange. If you do not use it then you will not survive "Big Data". Comments , Thoughts? John Eaton