From yigitdemir2015 at u.northwestern.edu Mon Feb 7 12:26:08 2011 From: yigitdemir2015 at u.northwestern.edu (Yigit Demir) Date: Mon, 7 Feb 2011 14:26:08 -0600 Subject: [Hotspot] Simulation of separated microprocessors Message-ID: Hi, I am trying to use HotSpot to figure out the thermal properties of two separate multicore microprocessor which are seperated by a small distance with some standing air between them. These processors are assumed to be placed on separate silicon dies, but they will not be too far from each other, so heat dissipated from one will effect the other processor. I am trying to observe this effect as a function of the distance between them. I wanted to ask, if the Hotspot can model the heat conduction through air in such a system? I realized that the floorplanner does not allow to leave and blank spaces between blocks, so I wanted to ask if it is possible to create a floorplan for two separate processors and simulate them on Hotspot. Is it possible to run this simulation with separate heatsinks placed on those processors? Thank you Yigit Demir Research Assistant EECS Dept. Northwestern University -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110207/d88634b2/attachment.html From wh6p at virginia.edu Mon Feb 7 16:10:50 2011 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 7 Feb 2011 18:10:50 -0600 Subject: [Hotspot] Simulation of separated microprocessors In-Reply-To: References: Message-ID: It is the ambient air from one chip affecting the ambient air of the other chip, which in turn affects the the other chip's temperature. Right now, HotSpot doesn't have ways to model ambient air temperature change and interactions between chips. You may want to check out Prof. Bianchini and his students work called Mercury at Rutgers University. They did some modeling work related to this. Combining that work with chip-level details provided by HotSpot would provide a good solution to the problem you are looking at. Actually, combining the two models together itself would be a very useful tool to the community. Hope this helps. -Wei On Mon, Feb 7, 2011 at 2:26 PM, Yigit Demir < yigitdemir2015 at u.northwestern.edu> wrote: > Hi, > > I am trying to use HotSpot to figure out the thermal properties of > two separate multicore microprocessor which are seperated by a small > distance with some standing air between them. These processors are assumed > to be placed on separate silicon dies, but they will not be too far from > each other, so heat dissipated from one will effect the other processor. I > am trying to observe this effect as a function of the distance between them. > I wanted to ask, if the Hotspot can model the heat conduction through air in > such a system? I realized that the floorplanner does not allow to leave and > blank spaces between blocks, so I wanted to ask if it is possible to create > a floorplan for two separate processors and simulate them on Hotspot. Is it > possible to run this simulation with separate heatsinks placed on those > processors? > Thank you > > > Yigit Demir > > Research Assistant > EECS Dept. Northwestern 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/20110207/5be78d1a/attachment.html From wh6p at virginia.edu Fri Feb 11 19:12:39 2011 From: wh6p at virginia.edu (Wei Huang) Date: Fri, 11 Feb 2011 18:12:39 -0600 Subject: [Hotspot] HotSpot In-Reply-To: <4D55694B.40406@cs.virginia.edu> References: <4D55694B.40406@cs.virginia.edu> Message-ID: Hi, In response to the following question about dumping transient grid temperatures in HotSpot, we haven't implemented that function -- This would usually generate a large data file... If block temperatures translated from grid temperatures are OK, one can use dump_temp() to dump them to a file. If grid temperatures are preferred, it is relatively easy to add a function to dump that by yourself by following the format of this function but use transient temperature instead: void dump_steady_temp_grid (grid_model_t *model, char *file) Hope this helps. -Wei > I would like transient thermal maps (i.e. every > 10000 cycles) to be the input to a verilog ams memory model that I am > developing. I see that there is an option to use a grid model. Is > there any option to write the transient grid temperatures? For example, > with a grid size of 64x64, can I output 64x64 temperatures every 10000 > cycles? > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110211/79ade73c/attachment.html From ks4kk at virginia.edu Fri Feb 11 20:15:10 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 11 Feb 2011 17:15:10 -0800 Subject: [Hotspot] HotSpot In-Reply-To: References: <4D55694B.40406@cs.virginia.edu> Message-ID: Hi, To add to Wei's inputs, please note that the grid model can model full-3D (through multiple layers). If you are interested in the top layer alone, there is a function that will let you print the grid temperatures for it. Look for "dump_top_layer_temp_grid" in temperature_grid.c. If you would like temperatures of other layers, you would have to write your own version based upon the function above. Hope this helps. Thanks, -karthik On Fri, Feb 11, 2011 at 4:12 PM, Wei Huang wrote: > Hi, > > In response to the following question about dumping transient grid > temperatures in HotSpot, we haven't implemented that function -- This would > usually generate a large data file... If block temperatures translated from > grid temperatures are OK, one can use dump_temp() to dump them to a file. If > grid temperatures are preferred, it is relatively easy to add a function to > dump that by yourself by following the format of this function but use > transient temperature instead: > > void dump_steady_temp_grid (grid_model_t *model, char *file) > > Hope this helps. > > -Wei > > > >> I would like transient thermal maps (i.e. every >> 10000 cycles) to be the input to a verilog ams memory model that I am >> developing. I see that there is an option to use a grid model. Is >> there any option to write the transient grid temperatures? For example, >> with a grid size of 64x64, can I output 64x64 temperatures every 10000 >> cycles? >> > > > _______________________________________________ > 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/20110211/a0549c61/attachment-0001.html From skadron at cs.virginia.edu Thu Feb 17 23:50:42 2011 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Thu, 17 Feb 2011 23:50:42 -0500 Subject: [Hotspot] Some general getting-started questions re HotSpot In-Reply-To: References: Message-ID: <4D5DFAA2.1000308@cs.virginia.edu> I'm taking the liberty of posting this on the hotspot mailing list, which is a better forum for such inquiries--more people to answer, plus the answers are archived! In answer to your questions below: 1. Hypothetical. 1 and 2: I'm not aware of any config or floorplan files for other architectures that are publicly available (if anyone else on the list cares to share their configs, please respond). You would have to generate your own using public documentation about these architectures. 3. This requires a simulator with power modeling, such as Wattch. 4. You should have one floorplan for the entire die. Be sure the spreader and sink are larger than the die. /K > I am new to Hotspot. i went through HOWTO for details, started > playing with examples. > but i have some questions regarding hotspot, as i am new to this > environment. i need some occasional interaction with you to clear my doubts. > > *question1)* > *in hotspot , hotspot.config file specifies some data for which machine > ? intel, amd, athlon, powerpc or any hypothetical machine ?* > *if i want to use data for intel or amd how can i get the data ? please > provide links or supply necessary materials*. > > *question2)* > *hotspot requires one floor planing file and power trace file and > initial power file. distribution comes with only one ev6.flp and > gcc.ptrace file* > * how to obtain floor planning file for say ,ARM or > Intel or AMD ?* > > *question3)* > * suppose i have some of my set of programs, HOW power trace file can > be generated (which will be input to hotspot ) for any application like, > bzip2, matmult ETC ? * > **_how can i generate / obtain power trace file FOR ANY APPLICATION / > PROGRAM ?_** > > *question4) * > *suppose i am having a machine with , say, M cores. do i require .flp > file for each M core. or single .flp file* > * my objective is to measure temp for each core at once From hkustliqi at gmail.com Fri Feb 18 21:32:50 2011 From: hkustliqi at gmail.com (Qi Li) Date: Fri, 18 Feb 2011 21:32:50 -0500 Subject: [Hotspot] Transient thermal map In-Reply-To: References: Message-ID: Hi, I am a beginner user of HotSpot and I saw this message in the mailing list. I am also facing this problem of getting transient thermal map using grid model. However I don't really get the reply here. As its name suggests, I assume dump_steady_temp_grid dumps the steady state temperature to a file, so if I want to get transient temperature, I should use some function like dump_transient_temp_grid, but it seems that there is not such a function in HotSpot. Could anyone please elaborate more on this point? That would be really helpful for me. Thanks! Best, Qi Li ------------------------------------------------------------------------------------------------------ Hi, In response to the following question about dumping transient grid temperatures in HotSpot, we haven't implemented that function -- This would usually generate a large data file... If block temperatures translated from grid temperatures are OK, one can use dump_temp() to dump them to a file. If grid temperatures are preferred, it is relatively easy to add a function to dump that by yourself by following the format of this function but use transient temperature instead: void dump_steady_temp_grid (grid_model_t *model, char *file) Hope this helps. -Wei >* I would like transient thermal maps (i.e. every *>* 10000 cycles) to be the input to a verilog ams memory model that I am *>* developing. I see that there is an option to use a grid model. Is *>* there any option to write the transient grid temperatures? For example, *>* with a grid size of 64x64, can I output 64x64 temperatures every 10000 *>* cycles? *> -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110218/25d4a0be/attachment.html From aziabari at soe.ucsc.edu Sun Feb 20 20:39:15 2011 From: aziabari at soe.ucsc.edu (Amirkoushyar Ziabari) Date: Sun, 20 Feb 2011 17:39:15 -0800 (PST) Subject: [Hotspot] Modified code to dump transient grid temperature results In-Reply-To: <591999764.158482.1298252241486.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: <1879304914.158491.1298252355132.JavaMail.root@mail-01.cse.ucsc.edu> By the way, all the lines that are added by me are commented in the text (search for //Added by amir). Please look at them as well. Best, Amirkoushyar Ziabari Graduate Research Assistant Electrical Engineering Department Baskin School of Engineering University of California, Santa Cruz Mail Stop: SOE3-Grads Baskin School of Engineering University of California, Santa Cruz 1156 High Street, Santa Cruz CA95064 Telephone: +1 (831) 459-1292 Email: aziabari at soe.ucsc.edu ----- Original Message ----- From: "Amirkoushyar Ziabari" To: "Qi Li" Cc: hotspot at mail.cs.virginia.edu Sent: Sunday, February 20, 2011 5:37:21 PM Subject: Modified code to dump transient grid temperature results Dear all, I changed the code so it dumps the transient grid level results. Please use the attached file instead of original "temperature_grid.c". Cheers, Amirkoushyar Ziabari Graduate Research Assistant Electrical Engineering Department Baskin School of Engineering University of California, Santa Cruz Mail Stop: SOE3-Grads Baskin School of Engineering University of California, Santa Cruz 1156 High Street, Santa Cruz CA95064 Telephone: +1 (831) 459-1292 Email: aziabari at soe.ucsc.edu ----- Original Message ----- From: "Qi Li" To: hotspot at mail.cs.virginia.edu Sent: Friday, February 18, 2011 6:32:50 PM Subject: Re: [Hotspot] Transient thermal map Hi, I am a beginner user of HotSpot and I saw this message in the mailing list. I am also facing this problem of getting transient thermal map using grid model. However I don't really get the reply here. As its name suggests, I assume dump_steady_temp_grid dumps the steady state temperature to a file, so if I want to get transient temperature, I should use some function like dump_transient_temp_grid, but it seems that there is not such a function in HotSpot. Could anyone please elaborate more on this point? That would be really helpful for me. Thanks! Best, Qi Li ------------------------------------------------------------------------------------------------------ Hi, In response to the following question about dumping transient grid temperatures in HotSpot, we haven't implemented that function -- This would usually generate a large data file... If block temperatures translated from grid temperatures are OK, one can use dump_temp() to dump them to a file. If grid temperatures are preferred, it is relatively easy to add a function to dump that by yourself by following the format of this function but use transient temperature instead: void dump_steady_temp_grid (grid_model_t *model, char *file) Hope this helps. -Wei > I would like transient thermal maps (i.e. every > 10000 cycles) to be the input to a verilog ams memory model that I am > developing. I see that there is an option to use a grid model. Is > there any option to write the transient grid temperatures? For example, > with a grid size of 64x64, can I output 64x64 temperatures every 10000 > cycles? > _______________________________________________ HotSpot mailing list HotSpot at mail.cs.virginia.edu http://www.cs.virginia.edu/mailman/listinfo/hotspot -------------- next part -------------- A non-text attachment was scrubbed... Name: temperature_grid.c Type: text/x-csrc Size: 112620 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110220/a4040968/attachment-0001.bin From aziabari at soe.ucsc.edu Sun Feb 20 20:37:21 2011 From: aziabari at soe.ucsc.edu (Amirkoushyar Ziabari) Date: Sun, 20 Feb 2011 17:37:21 -0800 (PST) Subject: [Hotspot] Modified code to dump transient grid temperature results In-Reply-To: Message-ID: <591999764.158482.1298252241486.JavaMail.root@mail-01.cse.ucsc.edu> Dear all, I changed the code so it dumps the transient grid level results. Please use the attached file instead of original "temperature_grid.c". Cheers, Amirkoushyar Ziabari Graduate Research Assistant Electrical Engineering Department Baskin School of Engineering University of California, Santa Cruz Mail Stop: SOE3-Grads Baskin School of Engineering University of California, Santa Cruz 1156 High Street, Santa Cruz CA95064 Telephone: +1 (831) 459-1292 Email: aziabari at soe.ucsc.edu ----- Original Message ----- From: "Qi Li" To: hotspot at mail.cs.virginia.edu Sent: Friday, February 18, 2011 6:32:50 PM Subject: Re: [Hotspot] Transient thermal map Hi, I am a beginner user of HotSpot and I saw this message in the mailing list. I am also facing this problem of getting transient thermal map using grid model. However I don't really get the reply here. As its name suggests, I assume dump_steady_temp_grid dumps the steady state temperature to a file, so if I want to get transient temperature, I should use some function like dump_transient_temp_grid, but it seems that there is not such a function in HotSpot. Could anyone please elaborate more on this point? That would be really helpful for me. Thanks! Best, Qi Li ------------------------------------------------------------------------------------------------------ Hi, In response to the following question about dumping transient grid temperatures in HotSpot, we haven't implemented that function -- This would usually generate a large data file... If block temperatures translated from grid temperatures are OK, one can use dump_temp() to dump them to a file. If grid temperatures are preferred, it is relatively easy to add a function to dump that by yourself by following the format of this function but use transient temperature instead: void dump_steady_temp_grid (grid_model_t *model, char *file) Hope this helps. -Wei > I would like transient thermal maps (i.e. every > 10000 cycles) to be the input to a verilog ams memory model that I am > developing. I see that there is an option to use a grid model. Is > there any option to write the transient grid temperatures? For example, > with a grid size of 64x64, can I output 64x64 temperatures every 10000 > cycles? > _______________________________________________ HotSpot mailing list HotSpot at mail.cs.virginia.edu http://www.cs.virginia.edu/mailman/listinfo/hotspot -------------- next part -------------- A non-text attachment was scrubbed... Name: temperature_grid.c Type: text/x-csrc Size: 112620 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110220/7f6dcb34/attachment-0001.bin From ks4kk at virginia.edu Tue Feb 22 00:22:10 2011 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Mon, 21 Feb 2011 21:22:10 -0800 Subject: [Hotspot] Fwd: Modified code to dump transient grid temperature results In-Reply-To: <1879304914.158491.1298252355132.JavaMail.root@mail-01.cse.ucsc.edu> References: <591999764.158482.1298252241486.JavaMail.root@mail-01.cse.ucsc.edu> <1879304914.158491.1298252355132.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: Thank you for your contribution Amirkoushyar! Qi Li - Below is an earlier response that points to a relevant function that might help you: "dump_top_layer_temp_grid". Apologies for posting it in the wrong forum before! Thanks, -karthik ---------- Forwarded message ---------- From: Karthik Sankaranarayanan Date: Fri, Feb 11, 2011 at 5:15 PM Subject: Re: [Hotspot] HotSpot To: hotspot at mail.cs.virginia.edu Cc: Wei Huang Hi, To add to Wei's inputs, please note that the grid model can model full-3D (through multiple layers). If you are interested in the top layer alone, there is a function that will let you print the grid temperatures for it. Look for "dump_top_layer_temp_grid" in temperature_grid.c. If you would like temperatures of other layers, you would have to write your own version based upon the function above. Hope this helps. Thanks, -karthik On Fri, Feb 11, 2011 at 4:12 PM, Wei Huang wrote: > Hi, > > In response to the following question about dumping transient grid > temperatures in HotSpot, we haven't implemented that function -- This would > usually generate a large data file... If block temperatures translated from > grid temperatures are OK, one can use dump_temp() to dump them to a file. If > grid temperatures are preferred, it is relatively easy to add a function to > dump that by yourself by following the format of this function but use > transient temperature instead: > > void dump_steady_temp_grid (grid_model_t *model, char *file) > > Hope this helps. > > -Wei > > > >> I would like transient thermal maps (i.e. every >> 10000 cycles) to be the input to a verilog ams memory model that I am >> developing. I see that there is an option to use a grid model. Is >> there any option to write the transient grid temperatures? For example, >> with a grid size of 64x64, can I output 64x64 temperatures every 10000 >> cycles? >> > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > ---------- Forwarded message ---------- From: Amirkoushyar Ziabari Date: Sun, Feb 20, 2011 at 5:39 PM Subject: Re: [Hotspot] Modified code to dump transient grid temperature results To: Qi Li Cc: hotspot at mail.cs.virginia.edu By the way, all the lines that are added by me are commented in the text (search for //Added by amir). Please look at them as well. Best, Amirkoushyar Ziabari Graduate Research Assistant Electrical Engineering Department Baskin School of Engineering University of California, Santa Cruz Mail Stop: SOE3-Grads Baskin School of Engineering University of California, Santa Cruz 1156 High Street, Santa Cruz CA95064 Telephone: +1 (831) 459-1292 Email: aziabari at soe.ucsc.edu ----- Original Message ----- From: "Amirkoushyar Ziabari" To: "Qi Li" Cc: hotspot at mail.cs.virginia.edu Sent: Sunday, February 20, 2011 5:37:21 PM Subject: Modified code to dump transient grid temperature results Dear all, I changed the code so it dumps the transient grid level results. Please use the attached file instead of original "temperature_grid.c". Cheers, Amirkoushyar Ziabari Graduate Research Assistant Electrical Engineering Department Baskin School of Engineering University of California, Santa Cruz Mail Stop: SOE3-Grads Baskin School of Engineering University of California, Santa Cruz 1156 High Street, Santa Cruz CA95064 Telephone: +1 (831) 459-1292 Email: aziabari at soe.ucsc.edu ----- Original Message ----- From: "Qi Li" To: hotspot at mail.cs.virginia.edu Sent: Friday, February 18, 2011 6:32:50 PM Subject: Re: [Hotspot] Transient thermal map Hi, I am a beginner user of HotSpot and I saw this message in the mailing list. I am also facing this problem of getting transient thermal map using grid model. However I don't really get the reply here. As its name suggests, I assume dump_steady_temp_grid dumps the steady state temperature to a file, so if I want to get transient temperature, I should use some function like dump_transient_temp_grid, but it seems that there is not such a function in HotSpot. Could anyone please elaborate more on this point? That would be really helpful for me. Thanks! Best, Qi Li ------------------------------------------------------------------------------------------------------ Hi, In response to the following question about dumping transient grid temperatures in HotSpot, we haven't implemented that function -- This would usually generate a large data file... If block temperatures translated from grid temperatures are OK, one can use dump_temp() to dump them to a file. If grid temperatures are preferred, it is relatively easy to add a function to dump that by yourself by following the format of this function but use transient temperature instead: void dump_steady_temp_grid (grid_model_t *model, char *file) Hope this helps. -Wei > I would like transient thermal maps (i.e. every > 10000 cycles) to be the input to a verilog ams memory model that I am > developing. I see that there is an option to use a grid model. Is > there any option to write the transient grid temperatures? For example, > with a grid size of 64x64, can I output 64x64 temperatures every 10000 > cycles? > _______________________________________________ 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 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110221/d32e9922/attachment.html From baikeina2 at gmail.com Tue Feb 22 21:34:19 2011 From: baikeina2 at gmail.com (haha haha) Date: Wed, 23 Feb 2011 10:34:19 +0800 Subject: [Hotspot] Leakage model Message-ID: Hi all, I'm a Ph.D student in Zhejiang University. In the Hotspot there is a leakage power model, Pl = Area* alpha *exp(beta*(T-Tbase)). The values of parameter alpha and beta are for 180nm. Does anyone have the values for 45nm ? Or if anyone has another leakage power model, could you please provide the values of the parameters ? We don't have a SPICE simulator yet. we just wanna perform some high level estimations about the temperature dependent leakage power. Thanks very much Best Xiaohang -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110223/12636799/attachment.html From wh6p at virginia.edu Tue Feb 22 22:51:10 2011 From: wh6p at virginia.edu (Wei Huang) Date: Tue, 22 Feb 2011 21:51:10 -0600 Subject: [Hotspot] Leakage model In-Reply-To: References: Message-ID: Hi, We don't have up to date numbers for the mentioned leakage model. But you can try to use ITRS projection to estimate alpha and beta for 45nm. -Wei On Tue, Feb 22, 2011 at 8:34 PM, haha haha wrote: > Hi all, > I'm a Ph.D student in Zhejiang University. > In the Hotspot there is a leakage power model, Pl = Area* alpha > *exp(beta*(T-Tbase)). The values of parameter alpha and beta are for 180nm. > Does anyone have the values for 45nm ? Or if anyone has another leakage > power model, could you please provide the values of the parameters ? > We don't have a SPICE simulator yet. we just wanna perform some high level > estimations about the temperature dependent leakage power. > Thanks very much > Best > Xiaohang > > _______________________________________________ > 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/20110222/7e242b07/attachment.html