From arunx003 at umn.edu Fri Jul 2 09:29:36 2010 From: arunx003 at umn.edu (arunx003 at umn.edu) Date: 02 Jul 2010 08:29:36 -0500 Subject: [Hotspot] Thermal R&C extraction In-Reply-To: References: Message-ID: Hi, I'd like to know if there is a way to extract thermal R and thermal C values which the HotSpot uses to calculate the power/temperature. Thanks, Abhi From wh6p at virginia.edu Fri Jul 2 10:10:27 2010 From: wh6p at virginia.edu (Wei Huang) Date: Fri, 2 Jul 2010 09:10:27 -0500 Subject: [Hotspot] Thermal R&C extraction In-Reply-To: References:

Message-ID: Hi, currently HotSpot doesn't provide a direct way to output R and C values, but that can be done with relatively easy hack, since the values are already there. populate_r() and populate_c() for block model and grid model, respectively, should be a good starting point to find where the values are calculated. -Wei On Fri, Jul 2, 2010 at 8:29 AM, wrote: > Hi, > > I'd like to know if there is a way to extract thermal R and thermal C > values which the HotSpot uses to calculate the power/temperature. > > Thanks, > Abhi > _______________________________________________ > 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/20100702/30554183/attachment.html From devendra.rai at tik.ee.ethz.ch Mon Jul 5 10:30:15 2010 From: devendra.rai at tik.ee.ethz.ch (Devendra Rai) Date: Mon, 5 Jul 2010 16:30:15 +0200 Subject: [Hotspot] Hotspot: tempertures on different layers In-Reply-To: References: Message-ID: Hello Hotspot, There was a request last week on finding the conductances in various layers. I did some work, and I am attaching the pdf. Can you see whether the values seem reasonable? (two silicon "units" only). Also, is there a way to see temperature traces on various layers of chip (spreader, heatsink etc). I know on the spreadsheet version it is possible, how about on command line (linux)? Also, a general question: from the experiments, I see that the lateral heat-flow from a core to neighboring core is very minimal. Intuitively, it does not seem right, since the sink seems to get very hot, and I would believe that any core on the neighborhood of a hot core should get affected. How do I see this affect in hotspot? Would appreciate your help. Regards, Devendra Rai -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100705/4de3a7b6/attachment-0001.html -------------- next part -------------- A non-text attachment was scrubbed... Name: G_from_HotSpot_MLB.pdf Type: application/pdf Size: 48299 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100705/4de3a7b6/attachment-0001.pdf From devendra.rai at tik.ee.ethz.ch Tue Jul 6 16:09:38 2010 From: devendra.rai at tik.ee.ethz.ch (Devendra Rai) Date: Tue, 6 Jul 2010 22:09:38 +0200 Subject: [Hotspot] Temperatures on different layers Message-ID: Hello HotSpot, Is there a way to see temperature traces (not only steady state) on various layers of chip (spreader, heatsink etc). I know on the spreadsheet version it is possible, how about on command line (linux)? HotSpot spits out steady state temperature of various layers on the command line, but I also need information on how the temperature evolves. Also, a general question: from the experiments, I see that the lateral heat-flow from a core to neighboring core is very minimal. Intuitively, it does not seem right, since the sink seems to get very hot, and I would believe that any core on the neighborhood of a hot core should get affected. How do I see this affect in hotspot? (Or, are the observations right)? Cheers Devendra Rai -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100706/f27c196d/attachment.html From devendra.rai at tik.ee.ethz.ch Tue Jul 6 16:09:38 2010 From: devendra.rai at tik.ee.ethz.ch (Devendra Rai) Date: Tue, 6 Jul 2010 22:09:38 +0200 Subject: [Hotspot] Temperatures on different layers Message-ID: Hello HotSpot, Is there a way to see temperature traces (not only steady state) on various layers of chip (spreader, heatsink etc). I know on the spreadsheet version it is possible, how about on command line (linux)? HotSpot spits out steady state temperature of various layers on the command line, but I also need information on how the temperature evolves. Also, a general question: from the experiments, I see that the lateral heat-flow from a core to neighboring core is very minimal. Intuitively, it does not seem right, since the sink seems to get very hot, and I would believe that any core on the neighborhood of a hot core should get affected. How do I see this affect in hotspot? (Or, are the observations right)? Cheers Devendra Rai -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100706/f27c196d/attachment-0001.html From ks4kk at virginia.edu Tue Jul 6 19:02:32 2010 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Tue, 6 Jul 2010 16:02:32 -0700 Subject: [Hotspot] Temperatures on different layers In-Reply-To: References: Message-ID: Hi, While there is not a command line option to output transient internal node temperatures, adding such functionality is very simple. Currently, the "-steady_temp" option outputs the steady state temperatures including the internal node temperatures. It does so by using the "dump_temp" function call and passing the steady state temperature vector argument. Passing the transient temperature vector instead will result in the functionality you want. Regarding the lateral heat flow in silicon, what is the thickness of the die you are modeling? Since the thickness is usually much smaller than the lateral dimensions, lateral heat flow is usually lower than vertical. Low-conductivity of interface material can exacerbate this issue. These are just general observations, the actual behavior depends on the geometry/material properties of the package/cooling solution you are modeling. Hope this helps. Thanks, -karthik On Tue, Jul 6, 2010 at 1:09 PM, Devendra Rai wrote: > Hello HotSpot, > > Is there a way to see temperature traces (not only steady state) on various > layers of chip (spreader, heatsink etc). I know on the spreadsheet version > it is possible, how about on command line (linux)? HotSpot spits out steady > state temperature of various layers on the command line, but I also need > information on how the temperature evolves. > > Also, a general question: from the experiments, I see that the lateral > heat-flow from a core to neighboring core is very minimal. Intuitively, it > does not seem right, since the sink seems to get very hot, and I would > believe that any core on the neighborhood of a hot core should get affected. > How do I see this affect in hotspot? (Or, are the observations right)? > > > Cheers > Devendra Rai > > > _______________________________________________ > 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/20100706/11b2e5dd/attachment.html From eggerxu at gmail.com Tue Jul 6 19:51:40 2010 From: eggerxu at gmail.com (Zichuan XU) Date: Wed, 7 Jul 2010 07:51:40 +0800 Subject: [Hotspot] Integration with scheduling algorithm Message-ID: Dear HotSpot, Can HotSpot be integrated with the scheduling algorithm. That's, the power traces are generated dynamically by scheduling algorithm, and then they are input to HotSpot to get the temperature. Is it possible? If possible, do you got any guidelines? Thanks very much. Sincerely yours, Zichuan Xu --------------------- Zichuan Xu Software College, Dalian University of Technology,China MSN:eggerxu at hotmail.com QQ:286567346 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100707/3015d9e9/attachment.html From skadron at cs.virginia.edu Tue Jul 6 20:26:33 2010 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Tue, 06 Jul 2010 20:26:33 -0400 Subject: [Hotspot] Integration with scheduling algorithm In-Reply-To: References: Message-ID: <4C33C9B9.2070406@cs.virginia.edu> Yes, HotSpot should be straightforward to integrate. The system would work pretty much as you describe it: at each scheduling step, you would compute the average power dissipation in each block over the last time step and pass that to HotSpot, which would then return the new temperatures for use by the scheduler. Note that the power values passed to HotSpot have to match the floorplan. /K On 7/6/2010 7:51 PM, Zichuan XU wrote: > Dear HotSpot, > > Can HotSpot be integrated with the scheduling algorithm. > That's, the power traces are generated dynamically by scheduling > algorithm, and then they are input to HotSpot to get the temperature. > Is it possible? If possible, do you got any guidelines? > Thanks very much. > > > Sincerely yours, > Zichuan Xu > --------------------- > Zichuan Xu > Software College, Dalian University of Technology,China > MSN:eggerxu at hotmail.com > QQ:286567346 > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot -- Kevin Skadron Assoc. Professor of Computer Science, University of Virginia Please note that all statements are mine alone and do not represent the views of the institution From bernauer at informatik.uni-tuebingen.de Wed Jul 7 01:58:09 2010 From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer) Date: Wed, 7 Jul 2010 07:58:09 +0200 (CEST) Subject: [Hotspot] Integration with scheduling algorithm In-Reply-To: Message-ID: <7818921.4621278482284667.JavaMail.bernauer@lava> ----- "Zichuan XU" wrote: > Dear HotSpot, > > Can HotSpot be integrated with the scheduling algorithm. > That's, the power traces are generated dynamically by scheduling > algorithm, and then they are input to HotSpot to get the temperature. > Is it possible? If possible, do you got any guidelines? For "guideline", have a look at sim-template.c in the HotSpot directory. Cheers, -- Andreas Bernauer WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940 http://www.ti.uni-tuebingen.de/?id=Andreas_Bernauer From eggerxu at gmail.com Wed Jul 7 10:41:12 2010 From: eggerxu at gmail.com (Zichuan XU) Date: Wed, 7 Jul 2010 22:41:12 +0800 Subject: [Hotspot] Comparison with other chip temperature model Message-ID: Dear HotSpot, Is there any results of the comparison results with other chip temperature models, like ATMI. Thanks very much. Sincerely yours, Zichuan Xu ------- Zichuan Xu Software College, Dalian University of Technology,China MSN:eggerxu at hotmail.com QQ:286567346 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100707/75b1c1a3/attachment.html From wh6p at virginia.edu Wed Jul 7 11:06:55 2010 From: wh6p at virginia.edu (Wei Huang) Date: Wed, 7 Jul 2010 10:06:55 -0500 Subject: [Hotspot] Comparison with other chip temperature model In-Reply-To: References: Message-ID: We have done several validation efforts for HotSpot. Some results are published in our DAC'04, WDDD'07 and ISPASS'09 papers. Paper links are available at Prof. Skadron's homepage. -Wei On Wed, Jul 7, 2010 at 9:41 AM, Zichuan XU wrote: > Dear HotSpot, > > Is there any results of the comparison results with other chip temperature > models, like ATMI. > Thanks very much. > > Sincerely yours, > Zichuan Xu > > > > ------- > > Zichuan Xu > Software College, Dalian University of Technology,China > MSN:eggerxu at hotmail.com > QQ:286567346 > > _______________________________________________ > 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/20100707/c5a03335/attachment-0001.html From bernauer at informatik.uni-tuebingen.de Wed Jul 7 11:33:09 2010 From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer) Date: Wed, 7 Jul 2010 17:33:09 +0200 (CEST) Subject: [Hotspot] Comparison with other chip temperature model In-Reply-To: Message-ID: <8857147.4981278516784908.JavaMail.bernauer@lava> ----- "Zichuan XU" wrote: > Dear HotSpot, > > Is there any results of the comparison results with other chip > temperature > models, like ATMI. Have you checked http://lava.cs.virginia.edu/HotSpot/documentation.htm ? There, some papers are referenced. Cheers, -- Andreas Bernauer WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940 http://www.ti.uni-tuebingen.de/?id=Andreas_Bernauer From kop7646 at gmail.com Thu Jul 15 23:19:46 2010 From: kop7646 at gmail.com (Yi Xiang) Date: Fri, 16 Jul 2010 11:19:46 +0800 Subject: [Hotspot] HotSpot Digest, Vol 37, Issue 3 In-Reply-To: References: Message-ID: Hello HotSpot, I want to visualize 3D-stacked thermal map of Hotspot. As far as I know, SVG has no 3D support. Do you have suggestions on what kind of tools can be used to plot 3D thermal map. Thanks! Sincerely yours Yi From wh6p at virginia.edu Mon Jul 19 01:35:23 2010 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 19 Jul 2010 00:35:23 -0500 Subject: [Hotspot] HotSpot Digest, Vol 37, Issue 3 In-Reply-To: References:

Message-ID: You can plot a 2D temperature map for each layer (silicons, TIM, spreader, heatsink, etc). "Full" 3D plot is not supported in HotSpot, as each layer is discrete and an interpolation between layers is not trivial. If you want to visualize 2D thermal map in 3D (like a contour map), IBM has a free tool DataExplorer which is quite powerful and can do much more than 3D visualization. -Wei On Thu, Jul 15, 2010 at 10:19 PM, Yi Xiang wrote: > Hello HotSpot, > > I want to visualize 3D-stacked thermal map of Hotspot. As far as I > know, SVG has no 3D support. Do you have suggestions on what kind of > tools can be used to plot 3D thermal map. > > Thanks! > > Sincerely yours > Yi > _______________________________________________ > 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/20100719/35ecb92a/attachment.html From wowlyy at hotmail.com Wed Jul 28 15:33:28 2010 From: wowlyy at hotmail.com (LiYangyang) Date: Thu, 29 Jul 2010 03:33:28 +0800 Subject: [Hotspot] the effect of vias Message-ID: Hello Hotspot, I'm trying to use Hotspot to estimate the effect of adding vias to the chip temperature, is there any suggestion on how to do this? Thank you. Yangyang Li -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100729/75a7fb3c/attachment.html From wh6p at virginia.edu Wed Jul 28 18:09:17 2010 From: wh6p at virginia.edu (Wei Huang) Date: Wed, 28 Jul 2010 17:09:17 -0500 Subject: [Hotspot] the effect of vias In-Reply-To: References: Message-ID: Hello Yangyang, Right now, each layer specified in HotSpot has to be "homogenous", that is, with the same material. Since vias and surrounding dielectrics are not the same materials, there is no direct way of knowing the detailed temperature as per via(s) in HotSpot. However, if your intention is to find the average effect, such as "if I add x% more vias uniformly, how much temperature drop I would get", you can always use "equivalent" thermal conductivity and specific heat etc. Of course this is less accurate compared to modeling individual vias, but it should provide a good estimation. We are considering possibilities of supporting multiple materials in one layer in future releases. Hope this helps. -Wei 2010/7/28 LiYangyang > Hello Hotspot, > I'm trying to use Hotspot to estimate the effect of adding vias to the > chip temperature, is there any suggestion on how to do this? > Thank you. > > Yangyang Li > > > > _______________________________________________ > 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/20100728/9cf8ac0d/attachment.html From mrs8n at cms.mail.virginia.edu Wed Jul 28 18:55:23 2010 From: mrs8n at cms.mail.virginia.edu (Mircea R. Stan) Date: Wed, 28 Jul 2010 18:55:23 -0400 Subject: [Hotspot] HotSpot users survey In-Reply-To: <4C509CEF.90303@cs.virginia.edu> References: <4C4F2FBA.50307@cs.virginia.edu> <9484B911-8294-43BC-9F34-600C3F8A4672@virginia.edu> <4DD48AD1-E325-44AA-9C51-45096143A0F8@virginia.edu> <4C509CEF.90303@cs.virginia.edu> Message-ID: Dear Hotspot user, We are at an inflection point where we would like to decide on the future directions for Hotspot development and would like to engage the Hotspot user base in the decision process. Here are some potential directions for future Hotspot development that we envision: a. enhancements of the core Hotspot internals (faster solvers, parallel algorithms, etc.) b. enhancements of the Hotspot interfaces (GUI, integration with other tools, including EDA/CAD, spreadsheet interface, etc.) c. "horizontal" extension to other than "pure thermal" (variations, reliability, power delivery, etc.) d. "vertical" extension to other than microarchitecture ("lower" circuit level, and/or "higher" system level) e. "turn-key" simulation environments (3D, micro channels, SOC/embedded, exotic package solutions, etc.) f. closer integration with empirical/experimental infrastructure (test chips, IR camera, etc.) We hope you can spend a couple of minutes and respond to us (please don't respond to the entire list, just to Kevin and I) in order to help us prioritize the above list, and provide suggestions/feedback on the (un)desirability of any of the above for the Hotspot infrastructure. If you feel all of the above are important please say so, if you can provide an ordered list even better, if you think some of them don't make sense please let us know, we really appreciate your time and effort. Looking forward to receiving your feedback! Thanks a lot! Cheers, Kevin and Mircea From aytackubilay at gmail.com Thu Jul 29 13:34:46 2010 From: aytackubilay at gmail.com (=?ISO-8859-1?Q?Ayta=E7_Kubilay?=) Date: Thu, 29 Jul 2010 19:34:46 +0200 Subject: [Hotspot] transient power trace file using Wattch? Message-ID: I have installed Wattch and found some benchmark files in order to run a simulation to see how the power fluctuates for different components of the processor. My aim was to get a .ptrace file for Hotspot through time (or instructions). However, the output statistics of Wattch only give average and total power of various blocks. Is it possible to create a .ptrace file with more than one line of power values by using Wattch? Thanks in advance, Aytac -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20100729/d1f32db4/attachment.html From skadron at cs.virginia.edu Thu Jul 29 14:29:00 2010 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Thu, 29 Jul 2010 14:29:00 -0400 Subject: [Hotspot] transient power trace file using Wattch? In-Reply-To: References: Message-ID: <4C51C86C.1020403@cs.virginia.edu> Yes, but it requires minor modifications to the Wattch code, to print out the power in each block after each timestep. /K On 7/29/2010 1:34 PM, Ayta? Kubilay wrote: > I have installed Wattch and found some benchmark files in order to run a > simulation to see how the power fluctuates for different components of > the processor. My aim was to get a .ptrace file for Hotspot through time > (or instructions). However, the output statistics of Wattch only give > average and total power of various blocks. Is it possible to create a > .ptrace file with more than one line of power values by using Wattch? > > Thanks in advance, > > Aytac > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot