From ramakumar1729 at gmail.com Tue Oct 4 05:10:02 2011 From: ramakumar1729 at gmail.com (Rama Kumar Pasumarthi) Date: Tue, 4 Oct 2011 17:40:02 +0530 Subject: [Hotspot] Thermal maps of 2D layers during 3DIC simulation Message-ID: Hi, I am simulating the thermal profile a four-tier 3DIC, which gives me a ttrace file as an output, which has information for each layer. Is there some way for me to use this ttrace file and the floorplan for generating a thermal profile. As I understand it, the grid_thermal_map.pl requires grid temperatures, which are generated in the 2D using the flag -grid_steady_file while running hotspot. How are these grid temperatures to be generated for a single tier in the case of a 3DIC? Regards, Rama -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111004/04cf1561/attachment.html From wh6p at virginia.edu Tue Oct 4 07:45:53 2011 From: wh6p at virginia.edu (Wei Huang) Date: Tue, 4 Oct 2011 09:45:53 -0500 Subject: [Hotspot] Thermal maps of 2D layers during 3DIC simulation In-Reply-To: References: Message-ID: Rama, This thread in the HotSpto mailing list archive might be helpful to you: http://www.cs.virginia.edu/pipermail/hotspot/2011-September/000567.html -Wei On Tue, Oct 4, 2011 at 7:10 AM, Rama Kumar Pasumarthi < ramakumar1729 at gmail.com> wrote: > Hi, > I am simulating the thermal profile a four-tier 3DIC, which gives me a > ttrace file as an output, which has information for each layer. Is there > some way for me to use this ttrace file and the floorplan for generating a > thermal profile. As I understand it, the grid_thermal_map.pl requires grid > temperatures, which are generated in the 2D using the flag -grid_steady_file > while running hotspot. How are these grid temperatures to be generated for a > single tier in the case of a 3DIC? > > Regards, > Rama > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111004/3a714bdc/attachment-0001.html From raidalzobaidi at yahoo.com Fri Oct 21 12:14:49 2011 From: raidalzobaidi at yahoo.com (Raaed) Date: Fri, 21 Oct 2011 12:14:49 -0700 (PDT) Subject: [Hotspot] error: singular matrix in lupdcmp In-Reply-To: References: Message-ID: <1319224489.68189.YahooMailNeo@web110012.mail.gq1.yahoo.com> Greetings, ? I am new to HotSpot so please accept my apology if I am asking a silly question! ? I'm having this error message "error: singular matrix in lupdcmp" when I am trying to run the HotSpot on my processor floorplan!? The processor build using 65nm technology so the dimensions of the blocks are very small (chip area is 21mmx12mm and the smallest block measures 0.75mmx0.5mm). I realized if I multiply the dimensions by a factor of 100 then this message disappear and the program show me some results?! any suggestions? ? Many thanks in advance. ? Ra'ed ? phd research student Newcastle University -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111021/c1a4890f/attachment.html From ks4kk at virginia.edu Fri Oct 21 13:34:56 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 21 Oct 2011 13:34:56 -0700 Subject: [Hotspot] error: singular matrix in lupdcmp In-Reply-To: <1319224489.68189.YahooMailNeo@web110012.mail.gq1.yahoo.com> References: <1319224489.68189.YahooMailNeo@web110012.mail.gq1.yahoo.com> Message-ID: Hi Ra'ed, The singular matrix error typically means that the floorplan has some unaccounted empty spaces / overlaps. Could you please check to make sure there are no gaps/overlaps in your floorplan? Thanks, -karthik On Fri, Oct 21, 2011 at 12:14 PM, Raaed wrote: > Greetings, > > I am new to HotSpot so please accept my apology if I am asking a silly > question! > > I'm having this error message "error: singular matrix in lupdcmp" when I > am trying to run the HotSpot on my processor floorplan!? The processor > build using 65nm technology so the dimensions of the blocks are very small > (chip area is 21mmx12mm and the smallest block measures 0.75mmx0.5mm). I > realized if I multiply the dimensions by a factor of 100 then this message > disappear and the program show me some results?! any suggestions? > > Many thanks in advance. > > Ra'ed > > phd research student > Newcastle University > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111021/ae4aff5c/attachment.html From raidalzobaidi at yahoo.com Wed Oct 26 13:26:18 2011 From: raidalzobaidi at yahoo.com (Raaed) Date: Wed, 26 Oct 2011 13:26:18 -0700 (PDT) Subject: [Hotspot] error: singular matrix in lupdcmp In-Reply-To: References: <1319224489.68189.YahooMailNeo@web110012.mail.gq1.yahoo.com> Message-ID: <1319660778.39913.YahooMailNeo@web110016.mail.gq1.yahoo.com> Thanks karthik ... I doubled checked but I didn't find anything wrong with my floorplan file! any way the same file?work OK with the grid model :-) so I am using the grid model at the moment :D ? I've another important question I need help with! I'm trying to integrate the Hotspot with the network simulator I'm using (Noxim) but it keep showing a linking problem? so I did a workaround?as follows: After executing number cycles (sampling time) in the network simulator I will store the power trace in a file and call the HOTSPOT as external exe file. Hotspot then will read the power trace and store the steady state temp in another file which in return going to be read by the network simulator and so on.... am I doing the correct thing or this will affect the internal states of the HOTSPOT in the subsequents execution (keeping in mind that I'm using the old temp as initial for the next run). ? Need you advices? ? Thanks. ? Ra'ed ? ________________________________ From: Karthik Sankaranarayanan To: Raaed Cc: "hotspot at mail.cs.virginia.edu" Sent: Friday, October 21, 2011 9:34 PM Subject: Re: [Hotspot] error: singular matrix in lupdcmp Hi Ra'ed, The singular matrix error typically means that the floorplan has some unaccounted empty spaces / overlaps. Could you please check to make sure there are no gaps/overlaps in your floorplan? Thanks, -karthik On Fri, Oct 21, 2011 at 12:14 PM, Raaed wrote: Greetings, >? >I am new to HotSpot so please accept my apology if I am asking a silly question! >? >I'm having this error message "error: singular matrix in lupdcmp" when I am trying to run the HotSpot on my processor floorplan!? The processor build using 65nm technology so the dimensions of the blocks are very small (chip area is 21mmx12mm and the smallest block measures 0.75mmx0.5mm). I realized if I multiply the dimensions by a factor of 100 then this message disappear and the program show me some results?! any suggestions? >? >Many thanks in advance. >? >Ra'ed >? >phd research student >Newcastle University > > >_______________________________________________ >HotSpot mailing list >HotSpot at mail.cs.virginia.edu >http://www.cs.virginia.edu/mailman/listinfo/hotspot > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111026/f53b6357/attachment.html From smillican at wisc.edu Fri Oct 28 09:53:54 2011 From: smillican at wisc.edu (Spencer Millican) Date: Fri, 28 Oct 2011 11:53:54 -0500 Subject: [Hotspot] : TSV modeling Message-ID: Greetings, I'm new to Hotspot, and I don't know if it can implement my problem. I want to model 3D package-on-package chips. Is there any way for a 2D floor plan not to have a uniform thermal resistance? I want to do this since I want to model through silicon vias to allow heat to pass from one stack to another. Thanks -- Spencer Millican Ph.D. Student,?Electrical Engineering University of Wisconsin - Madison From smillican at wisc.edu Fri Oct 28 12:18:07 2011 From: smillican at wisc.edu (Spencer Millican) Date: Fri, 28 Oct 2011 14:18:07 -0500 Subject: [Hotspot] : TSV modeling In-Reply-To: References: Message-ID: Greetings, I'm new to Hotspot, and I don't know if it can implement my problem. I want to model 3D package-on-package chips. ?Is there any way for a 2D floor plan not to have a uniform thermal resistance? ?I want to do this since I want to model through silicon vias to allow heat to pass from one stack to another. Thanks -- Spencer Millican Ph.D. Student,?Electrical Engineering University of Wisconsin - Madison -- Spencer Millican Ph.D. Student,?Electrical Engineering University of Wisconsin - Madison smillican at wisc.edu 715-321-0066 From ks4kk at virginia.edu Fri Oct 28 16:45:55 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 28 Oct 2011 16:45:55 -0700 Subject: [Hotspot] : TSV modeling In-Reply-To: References: Message-ID: Hi Spencer, The default implementation of HotSpot has fixed thermal conductivity per layer. So, to study TSVs, you would have to modify the code to implement geometry-dependent conductivity. If you are using the grid model, it should be relatively straightforward (inserting your code in the temperature_grid.c line: model->layers[i].rz = getr(...)). Hope this is helpful. Thanks, -karthik On Fri, Oct 28, 2011 at 12:18 PM, Spencer Millican wrote: > Greetings, > > I'm new to Hotspot, and I don't know if it can implement my problem. > I want to model 3D package-on-package chips. Is there any way for a > 2D floor plan not to have a uniform thermal resistance? I want to do > this since I want to model through silicon vias to allow heat to pass > from one stack to another. > > Thanks > > > > -- > Spencer Millican > Ph.D. Student, Electrical Engineering > University of Wisconsin - Madison > > > > -- > Spencer Millican > Ph.D. Student, Electrical Engineering > University of Wisconsin - Madison > smillican at wisc.edu > 715-321-0066 > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111028/2f8a2f31/attachment-0001.html From ks4kk at virginia.edu Fri Oct 28 16:53:43 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 28 Oct 2011 16:53:43 -0700 Subject: [Hotspot] error: singular matrix in lupdcmp In-Reply-To: <1319660778.39913.YahooMailNeo@web110016.mail.gq1.yahoo.com> References: <1319224489.68189.YahooMailNeo@web110012.mail.gq1.yahoo.com> <1319660778.39913.YahooMailNeo@web110016.mail.gq1.yahoo.com> Message-ID: Hi Ra'ed, That is surprising indeed! Please do send us your floorplan file and we will take a look. As for your second question, the approach you use might not be fully accurate since the steady state temperature file only stores temperatures at a block granularity. By storing and resuming from it, you lose the spatial gradients within a block. Whether this loss of accuracy is significant or not is your judgment call. The alternative would be to write the entire grid model vector (grid cell temperatures of all layers) to a file and resume from it. Thanks, -karthik On Wed, Oct 26, 2011 at 1:26 PM, Raaed wrote: > Thanks karthik ... I doubled checked but I didn't find anything wrong with > my floorplan file! any way the same file work OK with the grid model :-) > so I am using the grid model at the moment :D > > I've another important question I need help with! I'm trying to integrate > the Hotspot with the network simulator I'm using (Noxim) but it keep showing > a linking problem? so I did a workaround as follows: After executing > number cycles (sampling time) in the network simulator I will store the > power trace in a file and call the HOTSPOT as external exe file. Hotspot > then will read the power trace and store the steady state temp in another > file which in return going to be read by the network simulator and so on.... > am I doing the correct thing or this will affect the internal states of the > HOTSPOT in the subsequents execution (keeping in mind that I'm using the > old temp as initial for the next run). > > Need you advices? > > Thanks. > > Ra'ed > > > *From:* Karthik Sankaranarayanan > *To:* Raaed > *Cc:* "hotspot at mail.cs.virginia.edu" > *Sent:* Friday, October 21, 2011 9:34 PM > *Subject:* Re: [Hotspot] error: singular matrix in lupdcmp > > Hi Ra'ed, > > The singular matrix error typically means that the floorplan has some > unaccounted empty spaces / overlaps. Could you please check to make sure > there are no gaps/overlaps in your floorplan? > > Thanks, > -karthik > > > On Fri, Oct 21, 2011 at 12:14 PM, Raaed wrote: > > Greetings, > > I am new to HotSpot so please accept my apology if I am asking a silly > question! > > I'm having this error message "error: singular matrix in lupdcmp" when I > am trying to run the HotSpot on my processor floorplan!? The processor > build using 65nm technology so the dimensions of the blocks are very small > (chip area is 21mmx12mm and the smallest block measures 0.75mmx0.5mm). I > realized if I multiply the dimensions by a factor of 100 then this message > disappear and the program show me some results?! any suggestions? > > Many thanks in advance. > > Ra'ed > > phd research student > Newcastle University > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20111028/63e9c800/attachment.html From skadron at cs.virginia.edu Sun Oct 30 08:04:24 2011 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Sun, 30 Oct 2011 12:04:24 -0400 Subject: [Hotspot] : TSV modeling In-Reply-To: References: Message-ID: <4EAD7588.6090708@cs.virginia.edu> Ayse Coskun has done some extensions to HotSpot to better support 3D organizations, including what is discussed below. She has sent us the code but we haven't yet had the chance to incorporate this into a new release. We have linked to a couple of 3D tools on the HotSpot website. See http://lava.cs.virginia.edu/HotSpot/links.htm and stay tuned for future releases... --Kevin On 10/28/2011 7:45 PM, Karthik Sankaranarayanan wrote: > Hi Spencer, > > The default implementation of HotSpot has fixed thermal conductivity per > layer. So, to study TSVs, you would have to modify the code to implement > geometry-dependent conductivity. If you are using the grid model, it > should be relatively straightforward (inserting your code in the > temperature_grid.c line: model->layers[i].rz = getr(...)). > > Hope this is helpful. > Thanks, > -karthik > > On Fri, Oct 28, 2011 at 12:18 PM, Spencer Millican > wrote: > > Greetings, > > I'm new to Hotspot, and I don't know if it can implement my problem. > I want to model 3D package-on-package chips. Is there any way for a > 2D floor plan not to have a uniform thermal resistance? I want to do > this since I want to model through silicon vias to allow heat to pass > from one stack to another. > > Thanks > > > > -- > Spencer Millican > Ph.D. Student, Electrical Engineering > University of Wisconsin - Madison > > > > -- > Spencer Millican > Ph.D. Student, Electrical Engineering > University of Wisconsin - Madison > smillican at wisc.edu > 715-321-0066 > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot