From mahaboobalishah at gmail.com Thu Jan 5 06:39:19 2012 From: mahaboobalishah at gmail.com (Mahaboob Alisha) Date: Thu, 5 Jan 2012 20:09:19 +0530 Subject: [Hotspot] about hospot In-Reply-To: References: Message-ID: Respected Sir, I have done floorplanning the floorplan(EV6.flp) that you had specified in the Hotspot tool. The processor that you have specified in the floorplan is that a multiprocessor. If not so i have a small doubt. If i want to give a floorplan for the multicore processor say it consists of 3 cores, is that i have to specify all the functional blocks that are present in all the 3 cores accordingly in the given size or i have to allocate separate area for each core and then i have to specify the floorplan for corresponding functional blocks of that repective core in that related area specified to that core. THANKS AND REGARD -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120105/e8a0bcb2/attachment.html From ks4kk at virginia.edu Thu Jan 5 12:43:33 2012 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Thu, 5 Jan 2012 12:43:33 -0800 Subject: [Hotspot] about hospot In-Reply-To: References: Message-ID: The supplied ev6.flp is a single-core floorplan. A few architectural blocks have been sub-divided into sub-blocks for better accuracy of the block-model. You can visualize the floorplan after generating a PDF using the following command (on a linux box): ./tofig.pl ev6.flp | fig2dev -L ps | ps2pdf - t.pdf Hope this helps. -karthik On Thu, Jan 5, 2012 at 6:39 AM, Mahaboob Alisha wrote: > Respected Sir, > > I have done floorplanning the floorplan(EV6.flp) that you > had specified in the Hotspot tool. The processor that you have specified in > the floorplan is that a multiprocessor. If not so i have a small doubt. If > i want to give a floorplan for the multicore processor say it consists of 3 > cores, is that i have to specify all the functional blocks that are present > in all the 3 cores accordingly in the given size or i have to > allocate separate area for each core and then i have to specify the > floorplan for corresponding functional blocks of that repective core in > that related area specified to that core. > > THANKS AND REGARD > > > > _______________________________________________ > 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/20120105/9956ac1a/attachment.html From mahaboobalishah at gmail.com Thu Jan 5 20:22:23 2012 From: mahaboobalishah at gmail.com (Mahaboob Alisha) Date: Fri, 6 Jan 2012 09:52:23 +0530 Subject: [Hotspot] about hospot In-Reply-To: References:

Message-ID: Respected sir, Thanks for your patience reply. I have already visualized the floorplan with using that command. Now from your reply i was confirmed that the floorplan that you have given in the tool is a singlecore. I want to do it for multicore so for that i have to specify a floorplan. Can i know how can i specify the floorplan. For example let me consider that i am having two cores. So, there will be some set of functional blocks for both the cores. so, is that i have to take and place all the subblocks related to those two cores in that particular size(16*16) of chip or i have to allocate a particular area for each core and the subblocks related to that core should be placed in that size that has been allocated to that particular core and same rest of the area for another core and its subblocks. On 1/6/12, Karthik Sankaranarayanan wrote: > The supplied ev6.flp is a single-core floorplan. A few architectural blocks > have been sub-divided into sub-blocks for better accuracy of the > block-model. You can visualize the floorplan after generating a PDF using > the following command (on a linux box): > > ./tofig.pl ev6.flp | fig2dev -L ps | ps2pdf - t.pdf > > Hope this helps. > -karthik > > > > On Thu, Jan 5, 2012 at 6:39 AM, Mahaboob Alisha > wrote: > >> Respected Sir, >> >> I have done floorplanning the floorplan(EV6.flp) that you >> had specified in the Hotspot tool. The processor that you have specified >> in >> the floorplan is that a multiprocessor. If not so i have a small doubt. If >> i want to give a floorplan for the multicore processor say it consists of >> 3 >> cores, is that i have to specify all the functional blocks that are >> present >> in all the 3 cores accordingly in the given size or i have to >> allocate separate area for each core and then i have to specify the >> floorplan for corresponding functional blocks of that repective core in >> that related area specified to that core. >> >> THANKS AND REGARD >> >> >> >> _______________________________________________ >> HotSpot mailing list >> HotSpot at mail.cs.virginia.edu >> http://www.cs.virginia.edu/mailman/listinfo/hotspot >> >> > From ks4kk at virginia.edu Fri Jan 6 10:23:46 2012 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 6 Jan 2012 10:23:46 -0800 Subject: [Hotspot] about hospot In-Reply-To: References:

Message-ID: Hi Mahaboob, HotSpot is agnostic to core vs. functional block boundary. So, to specify a multi-core floorplan, you would have to specify each functional block of each core. See the attached file for an example of a 4-core floorplan. Hope this helps, -karthik On Thu, Jan 5, 2012 at 8:22 PM, Mahaboob Alisha wrote: > Respected sir, > > Thanks for your patience reply. I have already > visualized the floorplan with using that command. Now from your reply > i was confirmed that the floorplan that you have given in the tool is > a singlecore. > > I want to do it for multicore so for that i have > to specify a floorplan. Can i know how can i specify the floorplan. > For example let me consider that i am having two cores. So, there will > be some set of functional blocks for both the cores. so, is that i > have to take and place all the subblocks related to those two cores in > that particular size(16*16) of chip or i have to allocate a particular > area for each core and the subblocks related to that core should be > placed in that size that has been allocated to that particular core > and same rest of the area for another core and its subblocks. > > On 1/6/12, Karthik Sankaranarayanan wrote: > > The supplied ev6.flp is a single-core floorplan. A few architectural > blocks > > have been sub-divided into sub-blocks for better accuracy of the > > block-model. You can visualize the floorplan after generating a PDF using > > the following command (on a linux box): > > > > ./tofig.pl ev6.flp | fig2dev -L ps | ps2pdf - t.pdf > > > > Hope this helps. > > -karthik > > > > > > > > On Thu, Jan 5, 2012 at 6:39 AM, Mahaboob Alisha > > wrote: > > > >> Respected Sir, > >> > >> I have done floorplanning the floorplan(EV6.flp) that > you > >> had specified in the Hotspot tool. The processor that you have specified > >> in > >> the floorplan is that a multiprocessor. If not so i have a small doubt. > If > >> i want to give a floorplan for the multicore processor say it consists > of > >> 3 > >> cores, is that i have to specify all the functional blocks that are > >> present > >> in all the 3 cores accordingly in the given size or i have to > >> allocate separate area for each core and then i have to specify the > >> floorplan for corresponding functional blocks of that repective core in > >> that related area specified to that core. > >> > >> THANKS AND REGARD > >> > >> > >> > >> _______________________________________________ > >> 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/20120106/f3a56958/attachment.html -------------- next part -------------- A non-text attachment was scrubbed... Name: 4core.flp Type: application/octet-stream Size: 4079 bytes Desc: not available Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120106/f3a56958/attachment.obj From dipinprakash at gmail.com Mon Jan 9 00:10:34 2012 From: dipinprakash at gmail.com (dipin prakash) Date: Mon, 9 Jan 2012 13:40:34 +0530 Subject: [Hotspot] working with Openoffice In-Reply-To: References: Message-ID: **sir, Do we have Open Office interface for Hotspot? Regards, DIPIN PRAKASH S -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120109/b5276cb5/attachment-0001.html From sxm2912 at rit.edu Tue Jan 10 13:03:17 2012 From: sxm2912 at rit.edu (SUNDARARAMAN MOHANRAM) Date: Tue, 10 Jan 2012 16:03:17 -0500 Subject: [Hotspot] Regarding the 3D IC with microchannels Message-ID: Hello, I have read the How to in Hotspot manual and understood that it can simulate the thermal behaviour of 3D IC stack with heat sink, airflow and fan. I want to know, how the hotspot can be used to study the thermal behaviour with microchannels with liquid cooling. can you also explain briefly on the manually changing the themal conductivity of the layers, where thermal vias are located. -- thanks and regards, Sundararaman Mohanram Graduate Student Rochester Institute of technology Rochester NY, USA 14623 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120110/738b6370/attachment.html From Runjie at virginia.edu Tue Jan 10 14:50:49 2012 From: Runjie at virginia.edu (Runjie Zhang) Date: Tue, 10 Jan 2012 17:50:49 -0500 Subject: [Hotspot] Regarding the 3D IC with microchannels In-Reply-To: References: Message-ID: Hi, Sundararaman Currently, HotSpot does support 3D modeling but we do not support multiple materials within one layer, which could be critical for microchannel cooling modeling. However, you can check a 3D extension from Prof. Coskun in BU. That extension enables modeling heterogeneous layers. You can find it here: http://lava.cs.virginia.edu/HotSpot/links.htm Thanks! On Tue, Jan 10, 2012 at 4:03 PM, SUNDARARAMAN MOHANRAM wrote: > Hello, > > I have read the How to in Hotspot manual and understood that it can > simulate the thermal behaviour of 3D IC stack with heat sink, airflow and > fan. I want to know, how the hotspot can be used to study the thermal > behaviour with microchannels with liquid cooling. can you also explain > briefly on the manually changing the themal conductivity of the layers, > where thermal vias are located. > > -- > thanks and regards, > Sundararaman Mohanram > Graduate Student > Rochester Institute of technology > Rochester > NY, USA 14623 > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > -- Runjie Zhang Computer Engineering University of Virginia -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120110/84429b81/attachment.html From mohamed_elsawaf at yahoo.com Mon Jan 16 00:19:52 2012 From: mohamed_elsawaf at yahoo.com (Mohamed Elsawaf) Date: Mon, 16 Jan 2012 00:19:52 -0800 (PST) Subject: [Hotspot] 3D Fuzzy DTM controller In-Reply-To: Message-ID: <1326701992.74205.YahooMailClassic@web161302.mail.bf1.yahoo.com> Dear all I would like to tanks hotspot team for their effort. We used the hotspot simulator to build a new DTM controller based on 3D Fuzzy theory.We managed to enhance the power management of a multi core CPU taking into consideratin the correlation between all the cores. ?All simulations results are extracted through hotspot, and looks interesting. This research is now published at http://www.intechopen.com/articles/show/title/multi-core-cpu-air-cooling? Note: Now we are working on the HW implementation of this 3D Fuzzy DTM controller. Also the SW implementation i.e. on Android OS is also possible? ? Thank you very Much for your help and support ?YoursMohamed Elsawaf -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120116/c642e0a4/attachment.html From mohamed_elsawaf at yahoo.com Mon Jan 16 00:19:52 2012 From: mohamed_elsawaf at yahoo.com (Mohamed Elsawaf) Date: Mon, 16 Jan 2012 00:19:52 -0800 (PST) Subject: [Hotspot] 3D Fuzzy DTM controller In-Reply-To: Message-ID: <1326701992.74205.YahooMailClassic@web161302.mail.bf1.yahoo.com> Dear all I would like to tanks hotspot team for their effort. We used the hotspot simulator to build a new DTM controller based on 3D Fuzzy theory.We managed to enhance the power management of a multi core CPU taking into consideratin the correlation between all the cores. ?All simulations results are extracted through hotspot, and looks interesting. This research is now published at http://www.intechopen.com/articles/show/title/multi-core-cpu-air-cooling? Note: Now we are working on the HW implementation of this 3D Fuzzy DTM controller. Also the SW implementation i.e. on Android OS is also possible? ? Thank you very Much for your help and support ?YoursMohamed Elsawaf -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120116/c642e0a4/attachment-0001.html From toshshi at gmail.com Fri Jan 20 10:29:31 2012 From: toshshi at gmail.com (Jerry) Date: Fri, 20 Jan 2012 13:29:31 -0500 Subject: [Hotspot] How to model a simple circuit by using Hotspot? Message-ID: <4F19B28B.5000806@gmail.com> Hi, I have to some questions about how to use Hotspot to model thermal information of a chip. From the example in "How to", it seems hotspot only model the thermal for architecture-level chip. (ev6.flp include cache, memory...) My question is if I want to know the thermal information on the SoC, hot to do that? More specificlly, I wrote a simple circuit, that's say a divider, by verilog, and then use CAD tools to get the layout for it. I also got the power for every block of the layout. Since this layout is composed by the units from the tech library I used (umc180nm), I can also generate the floorplan for it. The problem I meet was when I tried to use the same way as the example mentioned to generate the thermal information for this divider, there was some errors indicated the size and the coordinates of the floorplan are not correct. Also, the thermal information looks wired. I was wondering if there are some manual for teaching how to apply Hotspot in Soc and how to configure the configure file. How to define the size and what's the unit? Thanks a lot for the help, Jerry From skadron at cs.virginia.edu Fri Jan 20 11:32:25 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Fri, 20 Jan 2012 14:32:25 -0500 Subject: [Hotspot] How to model a simple circuit by using Hotspot? In-Reply-To: <4F19B28B.5000806@gmail.com> References: <4F19B28B.5000806@gmail.com> Message-ID: <4F19C149.80900@cs.virginia.edu> There's no specific guide I know of for using HotSpot for SoCs, because that shouldn't really be necessary. In principle, HotSpot should be able to model any types of blocks. There are some constraints. The floorplan must be smaller than the heat spreader and heat sink. Blocks in the floorplan cannot overlap, and there cannot be any space in the floorplan that is unaccounted for (so whitespace has to be explictly accounted as such.) You're right that the thermal packaging currently requires a heat spreader and heat sink. You can make these very thin, but you can't eliminate them. Without specific error messages and other details, we can't give you more detailed help. --Kevin On 1/20/2012 1:29 PM, Jerry wrote: > Hi, > > I have to some questions about how to use Hotspot to model thermal > information of a chip. > From the example in "How to", it seems hotspot only model the thermal > for architecture-level chip. > (ev6.flp include cache, memory...) > > My question is if I want to know the thermal information on the SoC, hot > to do that? > > More specificlly, I wrote a simple circuit, that's say a divider, by > verilog, and then use CAD tools to get the layout for it. > I also got the power for every block of the layout. > Since this layout is composed by the units from the tech library I used > (umc180nm), I can also generate the floorplan for it. > > The problem I meet was when I tried to use the same way as the example > mentioned to generate the thermal information for this divider, > there was some errors indicated the size and the coordinates of the > floorplan are not correct. > Also, the thermal information looks wired. > > I was wondering if there are some manual for teaching how to apply > Hotspot in Soc and how to configure the configure file. > How to define the size and what's the unit? > > Thanks a lot for the help, > Jerry > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot From toshshi at gmail.com Fri Jan 20 14:43:37 2012 From: toshshi at gmail.com (Jerry) Date: Fri, 20 Jan 2012 17:43:37 -0500 Subject: [Hotspot] Few questions about hotspot. Message-ID: <4F19EE19.90203@gmail.com> Hi, Thank you for the reply. I still have a few questions about this. 1. From the hotspot.config file, I can find the heatsink side is 0.06 and the spreader side is 0.03. Does that mean the size of the heatsink is 0.06*0.06 and the size of the spreader is 0.03*0.03? 2.From your previous mail, you said the constraint is the floorplan must be smaller than the heatsink and the spreader. The question is if the floorplan does smaller than the heatsink and the spreader, if the thermal distribution of the chip will still be affect by the size of the heatsink and the spreader? (different heatsink and spreader size might cause different thermal distribution result even the chip is always smaller the these two stuff?) 3. When I use Hotspot to my circuit to calculate steady thermal information, the results for every block are all 318K. I dont know what's the reason for this problem. By the way, what's the unit I should use in ptrace file? Thank you so much for the time and the help, Jerry From wh6p at virginia.edu Fri Jan 20 15:20:08 2012 From: wh6p at virginia.edu (Wei Huang) Date: Fri, 20 Jan 2012 17:20:08 -0600 Subject: [Hotspot] Few questions about hotspot. In-Reply-To: <4F19EE19.90203@gmail.com> References: <4F19EE19.90203@gmail.com> Message-ID: Please see below.... On Fri, Jan 20, 2012 at 4:43 PM, Jerry wrote: > Hi, > > Thank you for the reply. > I still have a few questions about this. > 1. From the hotspot.config file, I can find the heatsink side is 0.06 > and the spreader side is 0.03. > Does that mean the size of the heatsink is 0.06*0.06 and the size > of the spreader is 0.03*0.03? > yes. the unit is meters for each side. So heatsink is 60cmx60cm, heatspreader is 30cmc30cm > > 2.From your previous mail, you said the constraint is the floorplan > must be smaller than the heatsink and the spreader. > The question is if the floorplan does smaller than the heatsink and > the spreader, if the thermal distribution of the chip will still be > affect by the size of the heatsink and the spreader? > (different heatsink and spreader size might cause different thermal > distribution result even the chip is always smaller the these two stuff?) > It wouldn't make sense to have heat spreader or heat sink the same size or smaller than the die. Depending on how much power is dissipated from the silicon, larger heat sink and spreader size would help to better cooling the chip. > > 3. When I use Hotspot to my circuit to calculate steady thermal > information, the results for every block are all 318K. > I dont know what's the reason for this problem. > By the way, what's the unit I should use in ptrace file? > It can be the case your total power is too small, or the power distribution is quite uniform across the chip... We need more information about your design. for example, total power, chip area, the block with highest power density (that is, power divided by area) and the associated size of that block. -Wei > > Thank you so much for the time and the help, > > Jerry > _______________________________________________ > 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/20120120/77089f68/attachment-0001.html From toshshi at gmail.com Sat Jan 21 12:21:00 2012 From: toshshi at gmail.com (Jerry) Date: Sat, 21 Jan 2012 15:21:00 -0500 Subject: [Hotspot] Few questions about hotspot. In-Reply-To: <4F1B1876.7060104@gmail.com> References: <4F1B1876.7060104@gmail.com> Message-ID: <4F1B1E2C.1020408@gmail.com> Hi, I think the problem is the power is too small. The total power is around 7.4*E-4 and the chip area is about 130um. Also, the highest power density is 2.88*E-6 and the area of the block with highest power density is 9.98 um^2. Is there any way to model the chip with small power? Thanks for the help, Chang-Chih On 1/20/2012 6:20 PM, Wei Huang wrote: > Please see below.... > > On Fri, Jan 20, 2012 at 4:43 PM, Jerry > wrote: > > Hi, > > Thank you for the reply. > I still have a few questions about this. > 1. From the hotspot.config file, I can find the heatsink side is 0.06 > and the spreader side is 0.03. > Does that mean the size of the heatsink is 0.06*0.06 and the size > of the spreader is 0.03*0.03? > > > yes. the unit is meters for each side. So heatsink is 60cmx60cm, > heatspreader is 30cmc30cm > > > 2.From your previous mail, you said the constraint is the floorplan > must be smaller than the heatsink and the spreader. > The question is if the floorplan does smaller than the heatsink and > the spreader, if the thermal distribution of the chip will still be > affect by the size of the heatsink and the spreader? > (different heatsink and spreader size might cause different > thermal > distribution result even the chip is always smaller the these two > stuff?) > > > It wouldn't make sense to have heat spreader or heat sink the same > size or smaller than the die. > Depending on how much power is dissipated from the silicon, larger > heat sink and spreader size would help to better cooling the chip. > > > 3. When I use Hotspot to my circuit to calculate steady thermal > information, the results for every block are all 318K. > I dont know what's the reason for this problem. > By the way, what's the unit I should use in ptrace file? > > > It can be the case your total power is too small, or the power > distribution is quite uniform across the chip... We need more > information about your design. for example, > total power, chip area, the block with highest power density (that is, > power divided by area) and the associated size of that block. > > -Wei > > > Thank you so much for the time and the help, > > Jerry > _______________________________________________ > 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/20120121/b1d21002/attachment.html From wh6p at virginia.edu Sun Jan 22 15:45:28 2012 From: wh6p at virginia.edu (Wei Huang) Date: Sun, 22 Jan 2012 17:45:28 -0600 Subject: [Hotspot] Few questions about hotspot. In-Reply-To: <4F1B1E2C.1020408@gmail.com> References: <4F1B1876.7060104@gmail.com> <4F1B1E2C.1020408@gmail.com> Message-ID: Hi, are you simulating only one or a few circuit blocks? if that is the case, then the tiny power would make temperature rise tiny, too. You'll have to have an idea of the final whole chip power and chip area, and assign the right power to other parts of the chip as well. In other words, simulating only the circuits your interested in wouldn't yield correct and useful results. -Wei On Sat, Jan 21, 2012 at 2:21 PM, Jerry wrote: > > Hi, > > I think the problem is the power is too small. > The total power is around 7.4*E-4 and the chip area is about 130um. > Also, the highest power density is 2.88*E-6 and the area of the block with > highest power density is 9.98 um^2. > Is there any way to model the chip with small power? > > Thanks for the help, > Chang-Chih > > > On 1/20/2012 6:20 PM, Wei Huang wrote: > > Please see below.... > > On Fri, Jan 20, 2012 at 4:43 PM, Jerry wrote: > >> Hi, >> >> Thank you for the reply. >> I still have a few questions about this. >> 1. From the hotspot.config file, I can find the heatsink side is 0.06 >> and the spreader side is 0.03. >> Does that mean the size of the heatsink is 0.06*0.06 and the size >> of the spreader is 0.03*0.03? >> > > yes. the unit is meters for each side. So heatsink is 60cmx60cm, > heatspreader is 30cmc30cm > >> >> 2.From your previous mail, you said the constraint is the floorplan >> must be smaller than the heatsink and the spreader. >> The question is if the floorplan does smaller than the heatsink and >> the spreader, if the thermal distribution of the chip will still be >> affect by the size of the heatsink and the spreader? >> (different heatsink and spreader size might cause different thermal >> distribution result even the chip is always smaller the these two stuff?) >> > > It wouldn't make sense to have heat spreader or heat sink the same size > or smaller than the die. > Depending on how much power is dissipated from the silicon, larger heat > sink and spreader size would help to better cooling the chip. > >> >> 3. When I use Hotspot to my circuit to calculate steady thermal >> information, the results for every block are all 318K. >> I dont know what's the reason for this problem. >> By the way, what's the unit I should use in ptrace file? >> > > It can be the case your total power is too small, or the power > distribution is quite uniform across the chip... We need more information > about your design. for example, > total power, chip area, the block with highest power density (that is, > power divided by area) and the associated size of that block. > > -Wei > >> >> Thank you so much for the time and the help, >> >> Jerry >> _______________________________________________ >> 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/20120122/76133fca/attachment.html From ramakumar1729 at gmail.com Fri Jan 27 22:08:12 2012 From: ramakumar1729 at gmail.com (Rama Kumar Pasumarthi) Date: Sat, 28 Jan 2012 11:38:12 +0530 Subject: [Hotspot] Ambient and initial temperature in hotspot.config Message-ID: Hi, Hotspot,config file uses 45 C as ambient temperature and 60 C as initial temperature. Is there any motivation for these specific values?Also, why is initial temperature higher than ambient temperature? Do these reflect realistic values for chips during functioning? Thanks and regards, Rama Kumar -- Rama Kumar Pasumarthi -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120128/628dfa64/attachment.html From wh6p at virginia.edu Sat Jan 28 11:21:53 2012 From: wh6p at virginia.edu (Wei Huang) Date: Sat, 28 Jan 2012 13:21:53 -0600 Subject: [Hotspot] Ambient and initial temperature in hotspot.config In-Reply-To: References: Message-ID: Hi, If you are simulating a chip and package that is heated up from ambient temperature, the initial temperature should be the same as ambient temperature, especially if you are not interested in temperature-related computer architecture research. On the other hand, part of the reason that the default initial temperatures are higher than ambient temperature is because the thermal "warm up" phase of simulation can be quite long for architectural simulations. A higher initial temperature is one way to speed up that phase. A more accurate way is to do a steady-state (i.e. static) thermal simulation first, and use the steady-state temperatures as initial temperature, and then do the usual architectural "warm-up" simulation to reach a more consistent architectural state. Hope this helps. -Wei On Sat, Jan 28, 2012 at 12:08 AM, Rama Kumar Pasumarthi < ramakumar1729 at gmail.com> wrote: > Hi, > Hotspot,config file uses 45 C as ambient temperature and 60 C as initial > temperature. Is there any motivation for these specific values?Also, why is > initial temperature higher than ambient temperature? Do these reflect > realistic values for chips during functioning? > > Thanks and regards, > Rama Kumar > -- > Rama Kumar Pasumarthi > > > _______________________________________________ > 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/20120128/6fa0e9a2/attachment.html