Comments on: Describe is not the same as Design

By: Observations from Uppsala » Modeling Endianness

Observations from Uppsala » Modeling Endianness — Sun, 26 Dec 2010 19:10:52 +0000

[...] of the hardware at the software level. In a way, I would say it is more geared towards the “design” use of modeling, rather than “describe“. Leave a comment | Trackback No comments [...]

By: Jakob

Jakob — Thu, 25 Mar 2010 10:17:41 +0000

@Phrank Yes, that is how I understand it too. There is nothing to describe WHAT a register does, only WHERE it is (and some of its simpler properties).

By: Phrank

Phrank — Thu, 25 Mar 2010 05:08:14 +0000

As I understand it, IP-XACT is just a way to describe connections between highlevel blocks. Seems more like a netlist language.

By: Jakob

Jakob — Tue, 16 Feb 2010 07:04:05 +0000

Everett Lumpkin : I suspect that the Carbon folks may disagree with the sentence “There is no automagic getting to a fast abstract model from a design-oriented description.”

I don't think so. They are perfectly clear that their models are in the 10KHz range, at best, while a fast simulation is at 100MHz to 1GHz... faster than RTL certainly, but not a fast virtual platform by any stretch.

While not really a language should we consider the meta-data of IP-XACT also as a “description language”?

It lacks semantics for the registers, so let's say it is half a description language.

By: Everett Lumpkin

Everett Lumpkin — Tue, 16 Feb 2010 04:03:53 +0000

I like your comparison of a “design language” vs. “description language”, and one of the faults of systemC for target software simulation is the (miss) use of the abstraction level I’ve heard referred to as systemC-ilog. It is just too common to create systemC models at too low of a level of abstraction for target SW simulation. Maybe Simics DML is a solution, but currently proprietary?

I suspect that the Carbon folks may disagree with the sentence “There is no automagic getting to a fast abstract model from a design-oriented description.”

While not really a language should we consider the meta-data of IP-XACT also as a “description language”?