From ramakumar1729 at gmail.com Mon Aug 1 00:34:20 2011 From: ramakumar1729 at gmail.com (Rama Kumar Pasumarthi) Date: Mon, 1 Aug 2011 13:04:20 +0530 Subject: [Hotspot] Sampling interval in Hotspot.config Message-ID: Does the term "-sampling_intvl" correspond to the timestep of ptrace file? Also, is there any lower bound on the timestep to be used, for sake of accuracy? -- -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110801/f1aa52d0/attachment.html From fatemeh_gharedaghi at yahoo.com Mon Aug 1 01:38:52 2011 From: fatemeh_gharedaghi at yahoo.com (fatemeh gharedaghi) Date: Mon, 1 Aug 2011 01:38:52 -0700 (PDT) Subject: [Hotspot] Initial Temperature Message-ID: <1312187932.59678.YahooMailNeo@web161208.mail.bf1.yahoo.com> Hi all, what is the value of initial temperature as usuall?how to determine initial temperature? ?Thanks. FatemehGharedaghi -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110801/d0b781d1/attachment.html From wh6p at virginia.edu Mon Aug 1 07:52:06 2011 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 1 Aug 2011 09:52:06 -0500 Subject: [Hotspot] Sampling interval in Hotspot.config In-Reply-To: References: Message-ID: On Mon, Aug 1, 2011 at 2:34 AM, Rama Kumar Pasumarthi < ramakumar1729 at gmail.com> wrote: > Does the term "-sampling_intvl" correspond to the timestep of ptrace file? > Yes > Also, is there any lower bound on the timestep to be used, for sake of > accuracy? > Using extremely short interval such as nanoseconds is not necessary and in fact would cause inaccurate results. Hundreds of microseconds, milliseconds or longer are recommended since they are more in line with chip/package thermal time constants. -Wei > -- > > > > _______________________________________________ > 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/20110801/0af15a9c/attachment.html From wh6p at virginia.edu Mon Aug 1 07:53:06 2011 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 1 Aug 2011 09:53:06 -0500 Subject: [Hotspot] Initial Temperature In-Reply-To: <1312187932.59678.YahooMailNeo@web161208.mail.bf1.yahoo.com> References: <1312187932.59678.YahooMailNeo@web161208.mail.bf1.yahoo.com> Message-ID: The initial temperature for a transient thermal simulation can be set in the .config file. -Wei On Mon, Aug 1, 2011 at 3:38 AM, fatemeh gharedaghi < fatemeh_gharedaghi at yahoo.com> wrote: > Hi all, > what is the value of initial temperature as usuall?how to determine initial > temperature? > Thanks. > > > FatemehGharedaghi > > _______________________________________________ > 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/20110801/a36d4645/attachment.html From ivan.ukhov at liu.se Tue Aug 2 00:16:43 2011 From: ivan.ukhov at liu.se (Ivan Ukhov) Date: Tue, 2 Aug 2011 09:16:43 +0200 Subject: [Hotspot] HotSpot's precision in 5.01 Message-ID: Dear Sir or Madam, First of all, let me thank you for your great thermal simulator, I am actively using it in my work. I am writing to you to ask about the following. Recently, I bumped into a strange behavior of the tool. I had an estimated power profile of a random task graph with about 30 tasks. The floorplan was also quite simple ? 4 equal cores placed next to each other in a grid. Both files are attached to this email. I was simulating the transient temperature profile of about ~83 second time period with the sampling interval equal to 0.7e-3 seconds, so the number of steps in the power profile was about ~120 thousands. The configuration file of HotSpot was untouched, except the sampling interval and the ambient/initial temperature set to 318.15 K (45 C). So, the resulting curves looked quite realistic, that was good. Than, I tried to cover the same time period, but with a bit larger sampling interval ? 1e-3 seconds (the number of steps dropped to about ~80 thousands), and the curves became mad. I have attached these two temperature profiles plotted in MatLab for you to see. I have checked the previous version of HotSpot, 5.0, it works fine. I looked at the changes since 5.0 and saw that you had modified the remainder part, line 483 in temperature_block.c (the same should be with the grid model). I think now HotSpot can produce considerable errors in the end of each time interval when power/temperature varies to rapidly, since now we do not have any adjustable steps and do not control the final error, line 487 rkf(? TRUE). If you try to change TRUE to FALSE here, everything will come back to normal. Of course, in this case we do not cover the whole reminder. That it is not very good either, here I totally agree with my colleague, Nima Aghaee, who also contacted you recently, but I do not think it is reasonable to try to cover it with that price. Therefore, I suggest to bring back the algorithm with adjustable steps for the reminder, but with an extra condition (instead of "int disable") that ensures that we come to the end of the interval. For example, we can loop more in rk4, or when we escape the loop, we can do yet another rk4_core if we really need to. May be my input data is too unrealistic to happen in the real life, anyway, it is you to decide, but please let me know if it makes any sense. Thank you. Sincerely yours, Ivan Ukhov ESLAB, IDA, Linkoping University, Sweden [cid:fe5240b9-e80d-4b9f-be9b-4aa65782305f at ad.liu.se] [cid:6fe7a999-57c1-4999-bde2-efb7d36cad7c at ad.liu.se] -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110802/802c1e07/attachment-0001.html -------------- next part -------------- A non-text attachment was scrubbed... Name: temperature_good.jpg Type: image/jpg Size: 86560 bytes Desc: temperature_good.jpg Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110802/802c1e07/attachment-0002.jpg -------------- next part -------------- A non-text attachment was scrubbed... Name: temperature_bad.jpg Type: image/jpg Size: 97874 bytes Desc: temperature_bad.jpg Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110802/802c1e07/attachment-0003.jpg -------------- next part -------------- A non-text attachment was scrubbed... Name: hotspot.zip Type: application/zip Size: 134641 bytes Desc: hotspot.zip Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110802/802c1e07/attachment-0001.zip From yishuihan.zc at gmail.com Tue Aug 9 23:58:57 2011 From: yishuihan.zc at gmail.com (Chao Zhang) Date: Wed, 10 Aug 2011 14:58:57 +0800 Subject: [Hotspot] image for 3D temperature Message-ID: Hi all~ I have a question. As is said in "HOWTO", there is a way to show temperature of stacked layers by hotspot. I tried the example and it gives the grid temperature out, but no floor plan could be used to make a svg file! How to make a svg file for this stacked IC? have you ever met this before? Thanks! -- Chao Zhang Electronic Engineering of EECS Department Yuanpei Program of Peking University Tel?+86 151-200-78471 Email?yishuihan.zc at gmail.com -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110810/38f41aac/attachment.html From wh6p at virginia.edu Wed Aug 10 00:40:46 2011 From: wh6p at virginia.edu (Wei Huang) Date: Wed, 10 Aug 2011 02:40:46 -0500 Subject: [Hotspot] image for 3D temperature In-Reply-To: References: Message-ID: Hi Chao, Glad you are using HotSpot. HotSpot can only plot 2D temperature map for each layer (if you provide the grid temperatures and the floorplan file). If you are having problems generating 2D temperature plot using the examples, please post detailed command line you used and the error message for us to take a look. -Wei On Wed, Aug 10, 2011 at 1:58 AM, Chao Zhang wrote: > Hi all~ > I have a question. > As is said in "HOWTO", there is a way to show temperature of stacked layers > by hotspot. > I tried the example and it gives the grid temperature out, but no floor > plan could be used to make a svg file! > How to make a svg file for this stacked IC? > have you ever met this before? > > Thanks! > > -- > Chao Zhang > Electronic Engineering of EECS Department > Yuanpei Program of Peking University > Tel?+86 151-200-78471 > Email?yishuihan.zc at gmail.com > > > _______________________________________________ > 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/20110810/5b88d1e7/attachment.html From ks4kk at virginia.edu Wed Aug 10 14:34:58 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Wed, 10 Aug 2011 14:34:58 -0700 Subject: [Hotspot] HotSpot's precision in 5.01 In-Reply-To: References: Message-ID: Hi Ivan, Thank you for taking the time to bring this error to our notice. My apologies for the oversight during the previous bugfix. We have rectified the problem and a new release is on its way! Sincerely, -karthik On Tue, Aug 2, 2011 at 12:16 AM, Ivan Ukhov wrote: > Dear Sir or Madam, > > > First of all, let me thank you for your great thermal simulator, I am > actively using it in my work. > > I am writing to you to ask about the following. Recently, I bumped into a > strange behavior of the tool. I had an estimated power profile of a random > task graph with about 30 tasks. The floorplan was also quite simple ? 4 > equal cores placed next to each other in a grid. Both files are attached to > this email. I was simulating the transient temperature profile of about ~83 > second time period with the sampling interval equal to 0.7e-3 seconds, so > the number of steps in the power profile was about ~120 thousands. The > configuration file of HotSpot was untouched, except the sampling interval > and the ambient/initial temperature set to 318.15 K (45 C). So, the > resulting curves looked quite realistic, that was good. Than, I tried to > cover the same time period, but with a bit larger sampling interval ? 1e-3 > seconds (the number of steps dropped to about ~80 thousands), and the curves > became mad. I have attached these two temperature profiles plotted in MatLab > for you to see. > > I have checked the previous version of HotSpot, 5.0, it works fine. I > looked at the changes since 5.0 and saw that you had modified the remainder > part, line 483 in temperature_block.c (the same should be with the grid > model). I think now HotSpot can produce considerable errors in the end of > each time interval when power/temperature varies to rapidly, since now we do > not have any adjustable steps and do not control the final error, line 487 > rkf(? TRUE). If you try to change TRUE to FALSE here, everything will come > back to normal. Of course, in this case we do not cover the whole reminder. > That it is not very good either, here I totally agree with my colleague, > Nima Aghaee, who also contacted you recently, but I do not think it is > reasonable to try to cover it with that price. > > Therefore, I suggest to bring back the algorithm with adjustable steps for > the reminder, but with an extra condition (instead of "int disable") that > ensures that we come to the end of the interval. For example, we can loop > more in rk4, or when we escape the loop, we can do yet another rk4_core if > we really need to. > > May be my input data is too unrealistic to happen in the real life, anyway, > it is you to decide, but please let me know if it makes any sense. Thank > you. > > > Sincerely yours, > Ivan Ukhov > > > ESLAB, IDA, > Linkoping University, > Sweden > > > > > _______________________________________________ > 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/20110810/1848ea62/attachment-0001.html -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: image/jpg Size: 86560 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110810/1848ea62/attachment-0002.jpg -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: image/jpg Size: 97874 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110810/1848ea62/attachment-0003.jpg From apangshuextc at gmail.com Fri Aug 12 02:42:20 2011 From: apangshuextc at gmail.com (Apangshu Das) Date: Fri, 12 Aug 2011 15:12:20 +0530 Subject: [Hotspot] Query about the input files of hotspot Message-ID: Thank you sir for giving me chance to be a part of HotSpot.I have some query, those are listed below: 1. How can i get the hotspot result interpretation. Is there any tutorial or manual to use this tool other than the readme file. 2. I have installed the hotspot & run one example file but how can i generate floorplan file(.flp) & power trace file(.ptrace). which tools are to be preferred to produce these files. -- Thanks & Regards Apangshu Das M.Tech(M.S) Student National Institute of Technology,Agartala Barjala, Tripura-799055 From rehan1984 at gmail.com Sat Aug 13 23:29:02 2011 From: rehan1984 at gmail.com (Rehan Ahmed) Date: Sun, 14 Aug 2011 01:29:02 -0500 Subject: [Hotspot] Analytic solution Message-ID: Hi, I was attempting to get the analytic solution to the system of differential which are numerically solved by hotspot to come up with temperature estimates. To do this, I dumped the C metrix from hotspot and used its eigen values/vectors to come up with my analytic solution. However, the analytic solution matches up with the the result given by hotspot only if I time scale the analytic solution by 1/0.758 ; or multiply the C matrix by 0.758. I am not sure I understand the significance of this constant. Is this one of the fitting constants? Thanks for your help. Regards, Rehan -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110814/3c5aeffe/attachment.html From wh6p at virginia.edu Sun Aug 14 12:16:34 2011 From: wh6p at virginia.edu (Wei Huang) Date: Sun, 14 Aug 2011 14:16:34 -0500 Subject: [Hotspot] Analytic solution In-Reply-To: References: Message-ID: Hi Rehan, The thermal R-C model used in HotSpot is a so-called "lumped" circuit model. A "lumped" model is slightly different from the physical entity which has a "distributive" nature . In HotSpot, we try to use a scaling factor to account for the thermal time constant difference between the lumped model and the real distributed nature of heat transfer. What you observed I believe is the result of that factor. Right now, we derive that factor thru a combination of theoretical analysis and detailed thermal simulations from tools like ANSYS. We may further improve that factor in the future if necessary. Hope this answers your question. -Wei On Sun, Aug 14, 2011 at 1:29 AM, Rehan Ahmed wrote: > Hi, > I was attempting to get the analytic solution to the system of differential > which are numerically solved by hotspot to come up with temperature > estimates. To do this, I dumped the C metrix from hotspot and used its eigen > values/vectors to come up with my analytic solution. However, the analytic > solution matches up with the the result given by hotspot only if I time > scale the analytic solution by 1/0.758 ; or multiply the C matrix by 0.758. > I am not sure I understand the significance of this constant. Is this one of > the fitting constants? Thanks for your help. > > Regards, > Rehan > > _______________________________________________ > 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/20110814/75e3ac5a/attachment.html From skadron at cs.virginia.edu Mon Aug 15 18:40:11 2011 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Mon, 15 Aug 2011 21:40:11 -0400 Subject: [Hotspot] Query about the input files of hotspot In-Reply-To: References: Message-ID: <4E49CA7B.5020007@cs.virginia.edu> I'm afraid the only documentation we have are the various materials on the website. There are many ways to generate floorplans and power traces. For floorplans, you might look at the HotFloorplan tool that is available on the website. For power traces, you need a power model such as Wattch, McPAT, etc. /K On 8/12/2011 5:42 AM, Apangshu Das wrote: > Thank you sir for giving me chance to be a part of HotSpot.I have some > query, those are listed below: > 1. How can i get the hotspot result interpretation. Is there any > tutorial or manual to use this tool other than the readme file. > 2. I have installed the hotspot& run one example file but how can i > generate floorplan file(.flp)& power trace file(.ptrace). which tools > are to be preferred to produce these files. > From bhanu at iith.ac.in Thu Aug 18 00:52:38 2011 From: bhanu at iith.ac.in (Bhanu Prathap T) Date: Thu, 18 Aug 2011 13:22:38 +0530 Subject: [Hotspot] empty spaces Message-ID: Hello everyone, I had to reproduce a floorplan used in certain paper. So, i took the image of it and manually calculated the coordinates of each module using image viewer software and scaled the coordinates appropriately. The floorplan had empty spaces. I carried on with the floorplan (with empty spaces) and simulated the ttrace. I have obtained a wrong output. Is it because the floorplan had empty spaces ? If so , what is the effective way of removing those ? (by empty spaces i mean, free slots which are not even named ) Or , is there any mistake in the way i generated the floorplan? Please comment ! Bhanu Prathap.T IV th B.Tech , Electrical Engineering, IIT Hyderabad -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110818/494bfda2/attachment.html From wh6p at virginia.edu Thu Aug 18 07:34:38 2011 From: wh6p at virginia.edu (Wei Huang) Date: Thu, 18 Aug 2011 09:34:38 -0500 Subject: [Hotspot] empty spaces In-Reply-To: References: Message-ID: Hi Bhanu, You need to define the empty space in your floorplan as well, as assign zero power to it in your power trace. Hope this helps. -Wei On Thu, Aug 18, 2011 at 2:52 AM, Bhanu Prathap T wrote: > Hello everyone, > I had to reproduce a floorplan used in certain > paper. So, i took the image of it and > manually calculated the coordinates of each module using image viewer > software and scaled the > coordinates appropriately. The floorplan had empty spaces. I carried on > with the floorplan (with empty spaces) > and simulated the ttrace. I have obtained a wrong output. > > Is it because the floorplan had empty spaces ? > If so , what is the effective way of removing those ? (by empty spaces i > mean, free slots which are not even named ) > Or , is there any mistake in the way i generated the floorplan? > > Please comment ! > > > Bhanu Prathap.T > IV th B.Tech , > Electrical Engineering, > IIT Hyderabad > > _______________________________________________ > 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/20110818/41b6f60a/attachment.html From bhanu at iith.ac.in Wed Aug 24 06:06:35 2011 From: bhanu at iith.ac.in (Bhanu Prathap T) Date: Wed, 24 Aug 2011 18:36:35 +0530 Subject: [Hotspot] 3D ttrace Message-ID: Hi everyone, I have been using hotspot to get temperature profile of 3D IC's. As mentioned in HOWTO i have run the following commands. *./hotspot -c hotspot.config -f orisc_1X4.flp -p orisc_4X4_125Mhz.p \* * -o orisc4X4.ttrace -model_type grid \* * -grid_layer_file orisc.lcf* * * The ttrace that was generated has following problems.. 1. The maximum was around 60 degrees and minimum around 56 degree Celsius. Isn't this range very small ? 2. I guess the initial temperature is 60 degrees. ( all blocks with 0 input power gave the same 60 degrees temperature out.) Some modules have 56 degrees C of temperature. How is this possible , i mean the temperature lowering ? 3. Can you comment on the ttrace that i generated. Is there any mistake that i did or a step that i missed in generating it ? 4. HOWTO gives .init file generation in grid model for 2D IC's . Why not for 3D ? If there are such commands , what are they ? I am attaching the necessary input files apart from the hotspot setup. Hoping that atleast some of you will respond to this mail as soon as possible. Thanking You, Bhanu Prathap.T EE08B008 , IIT Hyderabad -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110824/ebd6ea8f/attachment-0001.html -------------- next part -------------- A non-text attachment was scrubbed... Name: inputs.zip Type: application/zip Size: 4854 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110824/ebd6ea8f/attachment-0001.zip From skadron at cs.virginia.edu Wed Aug 24 08:20:09 2011 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Wed, 24 Aug 2011 11:20:09 -0400 Subject: [Hotspot] hotspot for mac In-Reply-To: <1313743946.55234.YahooMailNeo@web121507.mail.ne1.yahoo.com> References: <1313743946.55234.YahooMailNeo@web121507.mail.ne1.yahoo.com> Message-ID: <4E5516A9.5080402@cs.virginia.edu> We have actually never worked with Macs for HotSpot. It should be similar to compiling on any unix/linux system, although you might have to change the compiler's search path for include libraries. If anyone on this list has compiled for Mac and has any pointers, please chime in! --Kevin On 8/19/11 4:52 AM, Sohaib Majzoub wrote: > Hello, > Please can you direct me into how can I compile Hotspot on a mac OS? > thank you for the help, > -Sohaib From ivan.ukhov at liu.se Wed Aug 24 08:48:44 2011 From: ivan.ukhov at liu.se (Ivan Ukhov) Date: Wed, 24 Aug 2011 17:48:44 +0200 Subject: [Hotspot] hotspot for mac In-Reply-To: <4E5516A9.5080402@cs.virginia.edu> References: <1313743946.55234.YahooMailNeo@web121507.mail.ne1.yahoo.com> <4E5516A9.5080402@cs.virginia.edu> Message-ID: Hello, I am working with Mac OS. I have successfully compiled HotSpot without any extra adjustments in both Snow Leopard and Lion. The only thing that you need to do is to install the developer tools, Xcode, which you can download from App Store for free. Best wishes, Ivan On Aug 24, 2011, at 5:20 PM, Kevin Skadron wrote: > We have actually never worked with Macs for HotSpot. It should be > similar to compiling on any unix/linux system, although you might have > to change the compiler's search path for include libraries. > > If anyone on this list has compiled for Mac and has any pointers, please > chime in! > > --Kevin > > On 8/19/11 4:52 AM, Sohaib Majzoub wrote: >> Hello, >> Please can you direct me into how can I compile Hotspot on a mac OS? >> thank you for the help, >> -Sohaib > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot From wh6p at virginia.edu Wed Aug 24 11:37:59 2011 From: wh6p at virginia.edu (Wei Huang) Date: Wed, 24 Aug 2011 13:37:59 -0500 Subject: [Hotspot] 3D ttrace In-Reply-To: References: Message-ID: Hello Bhanu, It is possible that the transient temperatures can get lower than the initial temperatures. This happens when the amount of power you burn is not enough to heat up the chip to 60C. A steady-state simulation should be able to confirm that. In your case, if what you'd want is to simulate the warm-up of a chip, then it would make sense to set the initial temperatures to the ambient temperature. -Wei On Wed, Aug 24, 2011 at 8:06 AM, Bhanu Prathap T wrote: > Hi everyone, > I have been using hotspot to get temperature profile of > 3D IC's. > As mentioned in HOWTO i have run the following commands. > > *./hotspot -c hotspot.config -f orisc_1X4.flp -p orisc_4X4_125Mhz.p \* > * -o orisc4X4.ttrace -model_type grid \* > * -grid_layer_file orisc.lcf* > * > * > The ttrace that was generated has following problems.. > 1. The maximum was around 60 degrees and minimum around 56 degree Celsius. > > Isn't this range very small ? > > 2. I guess the initial temperature is 60 degrees. ( all blocks with 0 > input power gave the same 60 degrees temperature out.) > Some modules have 56 degrees C of temperature. How is this possible , i > mean the temperature lowering ? > > 3. Can you comment on the ttrace that i generated. Is there any mistake > that i did or a step that i missed in generating it ? > > 4. HOWTO gives .init file generation in grid model for 2D IC's . Why not > for 3D ? If there are such commands , what are they ? > > I am attaching the necessary input files apart from the hotspot setup. > Hoping that atleast some of you will respond to this mail as soon as > possible. > > Thanking You, > > Bhanu Prathap.T > EE08B008 , IIT Hyderabad > > _______________________________________________ > 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/20110824/dda9a24d/attachment.html From bhanu at iith.ac.in Thu Aug 25 01:28:05 2011 From: bhanu at iith.ac.in (Bhanu Prathap T) Date: Thu, 25 Aug 2011 13:58:05 +0530 Subject: [Hotspot] difference in ttraces Message-ID: hello everyone , I have developed ttrace for a 3D IC in two cases 1) @ 125 Mhz with all cores working 2) @ 125 Mhz with alternate cores working but to my surprise , i found smaller range of temperatures in the first case than the second one. Initial temperature was set to 318.15 K (same as ambient temp ) 1) 41 degree C to 45 odd. 2) 45 degree C to 222 odd. But these results seem to be wrong. how is this possible ? Please comment Bhanu Prathap.T EE08B008, IIT Hyderabad -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110825/787dee0d/attachment-0001.html -------------- next part -------------- A non-text attachment was scrubbed... Name: orisc_4X4_125Mhz.ttrace Type: application/octet-stream Size: 4800 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110825/787dee0d/attachment-0002.obj -------------- next part -------------- A non-text attachment was scrubbed... Name: orisc_4X4_125Mhz_alternate.ttrace Type: application/octet-stream Size: 4840 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110825/787dee0d/attachment-0003.obj From TAYW0034 at e.ntu.edu.sg Fri Aug 26 21:23:54 2011 From: TAYW0034 at e.ntu.edu.sg (#TAY WEI CHOON#) Date: Sat, 27 Aug 2011 04:23:54 +0000 Subject: [Hotspot] Questions Message-ID: <9FC1D3EE5F47314887481926A2836C9106EB8C@SINPRD0104MB138.apcprd01.prod.exchangelabs.com> Hi, I would like to enquire on 1) What type of boundary does the program use? Do you have any equation for it? I'm currently simulating thermal modelling using ADI-FDTD method. It has a thermal boundary condition of kappa*del(T)/del(n) + h*T = H*T_ambient. 2) How to reduce the program into a single layer (just chip layer)? I would like to remove away all the heatsink, thermal interface layer and so on. 3) How is the source(gcc.ptrace) added into the program while doing transient simulation. Is it line by line, or just the first line of the gcc.ptrace? Do you divide the power evening by the area and thickness before adding into the program? Hope to hear from you soon. Thank you. Shawn Tay -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110827/c2d3c5ed/attachment.html From anandmay22 at gmail.com Sat Aug 27 10:22:21 2011 From: anandmay22 at gmail.com (Anand J stephen) Date: Sat, 27 Aug 2011 22:52:21 +0530 Subject: [Hotspot] Need help regarding to temperature aware design. In-Reply-To: References: Message-ID: Hi to all I am started to working in thermal aware design. Previously i did fixed outline floorplaning. I just want to extent as temp design. I just want to use MCNC benchmark. I just want to know Wther Temp design is possoble with mcnc. What kind of issues i have to concentrate. Regards S. Anand -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110827/6ae1260e/attachment.html From wh6p at virginia.edu Sun Aug 28 04:21:21 2011 From: wh6p at virginia.edu (Wei Huang) Date: Sun, 28 Aug 2011 06:21:21 -0500 Subject: [Hotspot] Questions In-Reply-To: <9FC1D3EE5F47314887481926A2836C9106EB8C@SINPRD0104MB138.apcprd01.prod.exchangelabs.com> References: <9FC1D3EE5F47314887481926A2836C9106EB8C@SINPRD0104MB138.apcprd01.prod.exchangelabs.com> Message-ID: On Fri, Aug 26, 2011 at 11:23 PM, #TAY WEI CHOON# wrote: > Hi, > > > > > I would like to enquire on > > 1) What type of boundary does the program use? Do you have any equation for > it? > I'm currently simulating thermal modelling using ADI-FDTD method. It has a > thermal boundary condition of kappa*del(T)/del(n) + h*T = H*T_ambient. > It uses a similar convection boundary. A common heat transfer coefficient (h) is applied to all the heat sink surface area, and the heat sink if further divided into multiple regions, each has its own convection resistance (calculated from h and region area) and conduction resistance to the rest of the package/chip. > > 2) How to reduce the program into a single layer (just chip layer)? I would > like to remove away all the heatsink, thermal interface layer and so on. > The existence of the heatsink/spreader/TIM layers are hard-coded, so they can't be fully removed. However, you can always specify a very thin layer (10s of micron for sink/spreader and a few microns for TIM) for these layers. That should make the impact of these layers on thermal simulation negligible. > > 3) How is the source(gcc.ptrace) added into the program while doing > transient simulation. Is it line by line, or just the first line of the > gcc.ptrace? Do you divide the power evening by the area and thickness before > adding into the program? > It is line by line. each line represents the power consumption during that sampling interval. Sampling interval can be set in the .config file. Hope this helps. -Wei > > Hope to hear from you soon. > > Thank you. > > Shawn Tay > > _______________________________________________ > 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/20110828/b132e091/attachment.html