From link at cse.psu.edu Wed Oct 13 11:48:17 2004 From: link at cse.psu.edu (Greg Link) Date: Wed Mar 22 16:04:13 2006 Subject: [Hotspot] Abstracted view of the Inv B Matrix? Message-ID: <4D1A646B-1D2F-11D9-8D02-000A95AB9D08@cse.psu.edu> Hotspot List - The Hotspot documentation notes that Temp = PowerVector * InvB which is how the steady state temperature is generated. I'm interested in the values contained in the InvB matrix, and wondered if anyone had any simple abstract ways of thinking of the values therein. For example, is it true to note that the values in the diagonal of the InvB matrix are showing the effective thermal resistance between nodes in the system? And as such, the value located at row 0, column 1 is showing the thermal resistance between functional_block 0 and functional_block 1 (assuming at least two functional blocks in the design). Any insight as to other characteristics of this matrix would be appreciated. - Greg Link From ks4kk at cs.virginia.edu Thu Oct 14 01:25:51 2004 From: ks4kk at cs.virginia.edu (Karthik Sankaranarayanan) Date: Wed Mar 22 16:04:13 2006 Subject: [Hotspot] Abstracted view of the Inv B Matrix? In-Reply-To: <4D1A646B-1D2F-11D9-8D02-000A95AB9D08@cse.psu.edu> References: <4D1A646B-1D2F-11D9-8D02-000A95AB9D08@cse.psu.edu> Message-ID: Greg, The T = P * invB equation comes from the "node voltage method" of classic circuit analysis. invB is the inverse of the conductance matrix B. So, loosely speaking, only the conductances found in B have the meanings you mentioned (diagonal values being sums of node conductances etc.). I am not sure if invB has such a straightforward interpretation. A natural question that comes to mind is the usability of "loop" or "mesh" method of circuit analysis where the resistance matrix (as opposed to the conductance matrix) has meanings similar to what you mention (diagonal elements are actually sums of resistances in a loop). However, as functional blocks in HotSpot are treated as current sources, node voltage method (which deals with current sources easier than voltage sources) is a more appropriate method of analysis for us. Converting our equations to mesh equations is definitely possible (using supermeshes etc.) but I don't think that the mesh structure for our floorplan will be so simple that there is a direct interpretation for the resistance matrix. I might be totally wrong here. Wei might be able to give you a better insight into this. Thanks -karthik On Wed, 13 Oct 2004, Greg Link wrote: > Hotspot List - > The Hotspot documentation notes that > > Temp = PowerVector * InvB > > which is how the steady state temperature is generated. I'm interested > in the values contained in the InvB matrix, and wondered if anyone had > any simple abstract ways of thinking of the values therein. For > example, is it true to note that the values in the diagonal of the InvB > matrix are showing the effective thermal resistance between nodes in > the system? And as such, the value located at row 0, column 1 is > showing the thermal resistance between functional_block 0 and > functional_block 1 (assuming at least two functional blocks in the > design). > > Any insight as to other characteristics of this matrix would be > appreciated. > > - Greg Link > > _______________________________________________ > HotSpot mailing list > HotSpot@cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot >