From amin.ezhdehakosh at gmail.com Sun Dec 5 10:27:18 2010 From: amin.ezhdehakosh at gmail.com (Amin Ezhdehakosh) Date: Sun, 5 Dec 2010 21:57:18 +0330 Subject: [Hotspot] 3-D Thermal Map Message-ID: Hi, I want to extract 3-D FPGA thermal model with HotSpot. I have 4 layers, when each layer has more than 3-4 blocks I can't extract the result. It needs so many runtime. Could you please help me to solve this problem? -- Regards, Amin Ezhdehakosh -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101205/3d56e130/attachment.html From soartang at gmail.com Thu Dec 9 11:59:16 2010 From: soartang at gmail.com (Aoxiang Tang) Date: Thu, 9 Dec 2010 14:59:16 -0500 Subject: [Hotspot] Negative temperature Message-ID: Hi Hotspot, First thank you for letting me in the mail list, I appreciate it a lot. Now I'm working on the temperature of 3D chip which has 2 silicon layers and 2 TIM. I met a problem and I can't fix it, so I am here for help. In the hotspot.config, I changed the -sampling_intvl to a big value, 0.00073 and the dimensions of heat sink and heat spreader (Size of heat sink is still larger than that of heat spreader). In that configuration, every line in the power trace file will run for 0.00073s, right? However I got some negative temperatures for some blocks at some time points, I don't understand how this happens. It's definitely wrong because the ambient temperature is 318.15 k. Could you please give me some advice? Thank you in advance! Best Aoxiang -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101209/5469d6a6/attachment.html From wh6p at virginia.edu Thu Dec 9 13:34:04 2010 From: wh6p at virginia.edu (Wei Huang) Date: Thu, 9 Dec 2010 15:34:04 -0600 Subject: [Hotspot] Negative temperature In-Reply-To: References: Message-ID: What are the initial temperatures used? It also might be the case that HotSpot assumes Kelvin from the results, and subtracts 273.15 to convert to Celcius. -Wei On Thu, Dec 9, 2010 at 1:59 PM, Aoxiang Tang wrote: > Hi Hotspot, > > First thank you for letting me in the mail list, I appreciate it a lot. > > Now I'm working on the temperature of 3D chip which has 2 silicon layers > and 2 TIM. I met a problem and I can't fix it, so I am here for help. > > In the hotspot.config, I changed the -sampling_intvl to a big value, > 0.00073 and the dimensions of heat sink and heat spreader (Size of heat sink > is still larger than that of heat spreader). In that configuration, every > line in the power trace file will run for 0.00073s, right? However I got > some negative temperatures for some blocks at some time points, I don't > understand how this happens. It's definitely wrong because the ambient > temperature is 318.15 k. Could you please give me some advice? > > Thank you in advance! > Best > > Aoxiang > > > _______________________________________________ > 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/20101209/3ab4f223/attachment.html From soartang at gmail.com Thu Dec 9 14:16:26 2010 From: soartang at gmail.com (Aoxiang Tang) Date: Thu, 9 Dec 2010 17:16:26 -0500 Subject: [Hotspot] Negative temperature In-Reply-To: References: Message-ID: Hi Wei, Thank you very much for your reply. Actually first I generated the steady file as the initial file and fed it to HOTSPOT by -init_file. In the steady file, I think the temperature unit is Kelvin. You say HotSpot may assume kelvin from the results, but why do the temperatures of some blocks seem to be fine, around 50 while temperatures of the other blocks are around -13? Thanks Wei Aoxiang On Thu, Dec 9, 2010 at 4:34 PM, Wei Huang wrote: > What are the initial temperatures used? It also might be the case that > HotSpot assumes Kelvin from the results, and subtracts 273.15 to convert to > Celcius. > > -Wei > > On Thu, Dec 9, 2010 at 1:59 PM, Aoxiang Tang wrote: > >> Hi Hotspot, >> >> First thank you for letting me in the mail list, I appreciate it a lot. >> >> Now I'm working on the temperature of 3D chip which has 2 silicon layers >> and 2 TIM. I met a problem and I can't fix it, so I am here for help. >> >> In the hotspot.config, I changed the -sampling_intvl to a big value, >> 0.00073 and the dimensions of heat sink and heat spreader (Size of heat sink >> is still larger than that of heat spreader). In that configuration, every >> line in the power trace file will run for 0.00073s, right? However I got >> some negative temperatures for some blocks at some time points, I don't >> understand how this happens. It's definitely wrong because the ambient >> temperature is 318.15 k. Could you please give me some advice? >> >> Thank you in advance! >> Best >> >> Aoxiang >> >> >> _______________________________________________ >> 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/20101209/8c8d8165/attachment-0001.html From ks7h at virginia.edu Thu Dec 9 20:10:24 2010 From: ks7h at virginia.edu (Kevin Skadron) Date: Thu, 09 Dec 2010 23:10:24 -0500 Subject: [Hotspot] Negative temperature In-Reply-To: References: Message-ID: <4D01A830.4010802@virginia.edu> Aoxiang, Please check re K vs C. Other than that, there are too many possibilities to speculate. We need the specific inputs. Thanks, --Kevin On 12/9/2010 5:16 PM, Aoxiang Tang wrote: > Hi Wei, > > Thank you very much for your reply. Actually first I generated the > steady file as the initial file and fed it to HOTSPOT by -init_file. In > the steady file, I think the temperature unit is Kelvin. > > You say HotSpot may assume kelvin from the results, but why do the > temperatures of some blocks seem to be fine, around 50 while > temperatures of the other blocks are around -13? > > Thanks Wei > > Aoxiang > > On Thu, Dec 9, 2010 at 4:34 PM, Wei Huang > wrote: > > What are the initial temperatures used? It also might be the case > that HotSpot assumes Kelvin from the results, and subtracts 273.15 > to convert to Celcius. > > -Wei > > On Thu, Dec 9, 2010 at 1:59 PM, Aoxiang Tang > wrote: > > Hi Hotspot, > > First thank you for letting me in the mail list, I appreciate it > a lot. > > Now I'm working on the temperature of 3D chip which has 2 > silicon layers and 2 TIM. I met a problem and I can't fix it, > so I am here for help. > > In the hotspot.config, I changed the -sampling_intvl to a big > value, 0.00073 and the dimensions of heat sink and heat spreader > (Size of heat sink is still larger than that of heat spreader). > In that configuration, every line in the power trace file will > run for 0.00073s, right? However I got some > negative temperatures for some blocks at some time points, I > don't understand how this happens. It's definitely wrong because > the ambient temperature is 318.15 k. Could you please give me > some advice? > > Thank you in advance! > Best > > Aoxiang > > > _______________________________________________ > 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 From wh6p at virginia.edu Fri Dec 10 08:34:14 2010 From: wh6p at virginia.edu (Wei Huang) Date: Fri, 10 Dec 2010 10:34:14 -0600 Subject: [Hotspot] Negative temperature In-Reply-To: References:

Message-ID: You can't remove the spreader and heatsink in HotSpot. But you can "fake" that by using thin and small spreader and heatsink in the config file, and change r_convec (heatsink to air thermal resistance) to a much higher value (at least 10x, I would suggest), to account for the fact the heat convention surface area is much smaller now. To get a more accurate estimation, you should calculate the convection resistance based on your chip size and air flow speed. -Wei On Thu, Dec 9, 2010 at 10:58 PM, Aoxiang Tang wrote: > Hi Wei, > > According to you answer, I no longer have negative temperatures. Thanks a > lot. > > Here I have another question. Is it possible that I take away the heat > spreader or heat sink? I want to get the temperatures under the worst > situation. > Really thank you for your patience! > > Aoxiang > > > On Thu, Dec 9, 2010 at 6:20 PM, Aoxiang Tang wrote: > >> Hi wei, >> I get it. My layer is square and actually it's not fully covered by the >> blocks and there's still some blank space left. So maybe that's the case. >> Thank you! >> >> It's like this. The layer has 33 cores and I place them in 6*6 so the last >> line has 3 blank blocks. What do you suggest I put in the white space? >> Thanks a lot for your quick reply. It helps a lot >> >> Aoxiang >> >> On Thu, Dec 9, 2010 at 6:04 PM, Wei Huang wrote: >> >>> By the way, one requirement for HotSpot to calculate correctly is that >>> the floorplan has to be exactly filled. That is, no overlapping blocks and >>> no white spaces. Is that the case for your floorplans? >>> >>> -Wei >>> >>> >>> On Thu, Dec 9, 2010 at 4:42 PM, Wei Huang wrote: >>> >>>> PS: also the sequence of commands you used to reach your results. >>>> thanks, >>>> >>>> -Wei >>>> >>>> >>>> On Thu, Dec 9, 2010 at 4:41 PM, Wei Huang wrote: >>>> >>>>> Can you send the necessary files (floorplans, config, power trace, >>>>> layer config, steady temperature, etc)? I need to replicate your results >>>>> first. >>>>> Are you using HS5.0? >>>>> >>>>> Regards, >>>>> >>>>> -Wei >>>>> >>>>> >>>>> On Thu, Dec 9, 2010 at 4:16 PM, Aoxiang Tang wrote: >>>>> >>>>>> Hi Wei, >>>>>> >>>>>> Thank you very much for your reply. Actually first I generated the >>>>>> steady file as the initial file and fed it to HOTSPOT by -init_file. In the >>>>>> steady file, I think the temperature unit is Kelvin. >>>>>> >>>>>> You say HotSpot may assume kelvin from the results, but why do the >>>>>> temperatures of some blocks seem to be fine, around 50 while temperatures of >>>>>> the other blocks are around -13? >>>>>> >>>>>> Thanks Wei >>>>>> >>>>>> Aoxiang >>>>>> >>>>>> >>>>>> On Thu, Dec 9, 2010 at 4:34 PM, Wei Huang wrote: >>>>>> >>>>>>> What are the initial temperatures used? It also might be the case >>>>>>> that HotSpot assumes Kelvin from the results, and subtracts 273.15 to >>>>>>> convert to Celcius. >>>>>>> >>>>>>> -Wei >>>>>>> >>>>>>> On Thu, Dec 9, 2010 at 1:59 PM, Aoxiang Tang wrote: >>>>>>> >>>>>>>> Hi Hotspot, >>>>>>>> >>>>>>>> First thank you for letting me in the mail list, I appreciate it a >>>>>>>> lot. >>>>>>>> >>>>>>>> Now I'm working on the temperature of 3D chip which has 2 silicon >>>>>>>> layers and 2 TIM. I met a problem and I can't fix it, so I am here for >>>>>>>> help. >>>>>>>> >>>>>>>> In the hotspot.config, I changed the -sampling_intvl to a big value, >>>>>>>> 0.00073 and the dimensions of heat sink and heat spreader (Size of heat sink >>>>>>>> is still larger than that of heat spreader). In that configuration, every >>>>>>>> line in the power trace file will run for 0.00073s, right? However I got >>>>>>>> some negative temperatures for some blocks at some time points, I don't >>>>>>>> understand how this happens. It's definitely wrong because the ambient >>>>>>>> temperature is 318.15 k. Could you please give me some advice? >>>>>>>> >>>>>>>> Thank you in advance! >>>>>>>> Best >>>>>>>> >>>>>>>> Aoxiang >>>>>>>> >>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> 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/20101210/9cd4f289/attachment.html From aziabari at soe.ucsc.edu Fri Dec 10 17:35:33 2010 From: aziabari at soe.ucsc.edu (Amirkoushyar Ziabari) Date: Fri, 10 Dec 2010 17:35:33 -0800 (PST) Subject: [Hotspot] HotSpot time complexity In-Reply-To: <374082142.1213361292029731365.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: <1595189140.1213981292031333082.JavaMail.root@mail-01.cse.ucsc.edu> Dear all, would you please let me know what is the order of computation time complexity of algorithm used in HotSpot? 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 From solidrepellent at gmail.com Sat Dec 11 17:19:13 2010 From: solidrepellent at gmail.com (Solid Repel) Date: Sun, 12 Dec 2010 03:19:13 +0200 Subject: [Hotspot] Excel file and visual thermal map Message-ID: Dear Hotspot, I have a couple of beginners questions. 1) Can the provided excel sheet (hotspotUI-5.0.xls) be used for 3D chip simulation? or it is only for 2D ? 2) If I have the temperature trace file of the 3D simulation (with at least 3 layers of silicon), then how can I obtain a visually appealing thermal map with all the gradients and stuff for each layer of silicon? Thanks and Regards, Kamesh -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101212/1f8203e7/attachment.html From nicolas.peltier at doceapower.com Mon Dec 13 00:18:34 2010 From: nicolas.peltier at doceapower.com (Nicolas Peltier) Date: Mon, 13 Dec 2010 09:18:34 +0100 Subject: [Hotspot] Grid mode Message-ID: <4D05D6DA.1010304@doceapower.com> Hi HotSpot, If my understanding is correct, in grid mode, the silicon block is meshed with specified constraints (by default 64*64) in the xy plane. Then, the size of mesh elements in the x and y directions will be : (silicon size / 64). 1- What will happen if the first block (at x=0, y=0) of the floorplan has a size of : (silicon size / 192) : one third of the mesh element size ? 1- a/ Will this block be taken into account ? 1- b/ Where power of this block will be applied : only on the first mesh point at x=0, y=0 or distributed on other mesh nodes ? 1- c/ If power is distributed on several nodes, is it a weighted distribution and with which weights ? 2- Is the silicon (die) the only block that will be meshed or spreader, heatsink, bumps and board will also be meshed ? Thanks you in advance for your help. Nicolas -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 2803 bytes Desc: S/MIME Cryptographic Signature Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101213/921b44db/attachment-0001.bin From bernauer at informatik.uni-tuebingen.de Mon Dec 13 01:12:02 2010 From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer) Date: Mon, 13 Dec 2010 10:12:02 +0100 (CET) Subject: [Hotspot] HotSpot time complexity In-Reply-To: <26686306.2361292231384287.JavaMail.bernauer@lava> Message-ID: <28455726.2381292231518995.JavaMail.bernauer@lava> ----- "Amirkoushyar Ziabari" wrote: > Dear all, > > would you please let me know what is the order of computation time > complexity of algorithm used in HotSpot? > Last time I checked, HotSpot used a 4th order Adaptive Runge-Kutta for integration. http://en.wikipedia.org/wiki/Runge%E2%80%93Kutta_methods#Adaptive_Runge.E2.80.93Kutta_methods Implementation is at RCutil.c, line 297ff. in HotSpot 4.0. May I ask why you would like to know the computation time complexity? 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 ksiop at microlab.ntua.gr Mon Dec 13 02:21:20 2010 From: ksiop at microlab.ntua.gr (Kostas Siozios) Date: Mon, 13 Dec 2010 12:21:20 +0200 Subject: [Hotspot] hotfloorplan tool Message-ID: <004701cb9aaf$7c155c00$74401400$@microlab.ntua.gr> Hello, I am trying to create a Leon3 floorplan with hotfloorplan tool. I 'm giving a valid description file, as well as the power consumption of each block with no syntax errors (tabs etc..) The aspect ratio for each block ranges between 1-3, and rotation is enabled. The final floorplan contains a number of unknown blocks (named, _0, _1, _2, ...). My questions are: 1) Does hotfloorplan creates these boxes as black boxes. Do I have to add them to power trace file for hotspot use? Where can I find info for these? 2) Which is the meaning of wire density of connections on description file? Does it mean logical connections (netlist info) or physical connections (after P&R)? I suppose the first one, but could you provide me an info source? Is there any limitation on the wire density value? Regards, kostas -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101213/ceadfb6d/attachment.html From wh6p at virginia.edu Mon Dec 13 08:18:05 2010 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 13 Dec 2010 10:18:05 -0600 Subject: [Hotspot] Excel file and visual thermal map In-Reply-To: References: Message-ID: On Sat, Dec 11, 2010 at 7:19 PM, Solid Repel wrote: > Dear Hotspot, > > I have a couple of beginners questions. > > 1) Can the provided excel sheet (hotspotUI-5.0.xls) be used for 3D chip > simulation? or it is only for 2D ? > no, it can't right now. But using the command line is quite straightforward, and can apply 3D thermal simulation. Please refer to HOW-TO and FAQs on the HotSpot website. > > 2) If I have the temperature trace file of the 3D simulation (with at least > 3 layers of silicon), then how can I obtain a visually appealing thermal map > with all the gradients and stuff for each layer of silicon? > In the grid model (using command line, not Excel), you can have temperatures for each grid cell, for each silicon layer respectively. Then you can use the perl script in the release to plot temperature map for each silicon layer. More details and can be found in HOW-TO and FAQs. -Wei > > Thanks and Regards, > Kamesh > > _______________________________________________ > 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/20101213/ad9bb1dd/attachment.html From wh6p at virginia.edu Mon Dec 13 08:27:23 2010 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 13 Dec 2010 10:27:23 -0600 Subject: [Hotspot] Grid mode In-Reply-To: <4D05D6DA.1010304@doceapower.com> References: <4D05D6DA.1010304@doceapower.com> Message-ID: On Mon, Dec 13, 2010 at 2:18 AM, Nicolas Peltier < nicolas.peltier at doceapower.com> wrote: > Hi HotSpot, > > If my understanding is correct, in grid mode, the silicon block is meshed > with specified constraints (by default 64*64) in the xy plane. > Then, the size of mesh elements in the x and y directions will be : > (silicon size / 64). > correct. > > 1- What will happen if the first block (at x=0, y=0) of the floorplan has a > size of : (silicon size / 192) : one third of the mesh element size ? > that block will be accounted by one cell (if it falls entirely inside a grid cell), or divided into subblocks striding multiple cells. Basically, HotSpot uses grid cell to divide a block > 1- a/ Will this block be taken into account ? > Yes. > 1- b/ Where power of this block will be applied : only on the first mesh > point at x=0, y=0 or distributed on other mesh nodes ? > The power will be divided and added to the power of the grid cells the block belongs to. > 1- c/ If power is distributed on several nodes, is it a weighted > distribution and with which weights ? > HotSpot uses grid cell to divide blocks, and sums up power of all the partial blocks that fall inside a grid cell. It then uses power per grid cell to derive the cell temperature (you can choose average, center or max cell temperature), and remaps the grid temperature (again, area weighted) back to block temperature. > > 2- Is the silicon (die) the only block that will be meshed or spreader, > heatsink, bumps and board will also be meshed ? > silicon and the center area of spreader/TIM/heatsink that is covered by the silicon are all meshed. > > Thanks you in advance for your help. > > Hope this helps. -Wei > Nicolas > > > _______________________________________________ > 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/20101213/3c4a9837/attachment.html From wh6p at virginia.edu Mon Dec 13 08:36:32 2010 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 13 Dec 2010 10:36:32 -0600 Subject: [Hotspot] hotfloorplan tool In-Reply-To: <004701cb9aaf$7c155c00$74401400$@microlab.ntua.gr> References: <004701cb9aaf$7c155c00$74401400$@microlab.ntua.gr> Message-ID: On Mon, Dec 13, 2010 at 4:21 AM, Kostas Siozios wrote: > Hello, > > > > I am trying to create a Leon3 floorplan with hotfloorplan tool. > > I 'm giving a valid description file, as well as the power consumption of each block with no syntax errors (tabs etc..) > > The aspect ratio for each block ranges between 1-3, and rotation is enabled. > > > > The final floorplan contains a number of unknown blocks (named, _0, _1, _2, ...). > > > > My questions are: > > > > I am not the best person to answer these questions (Karthik please correct me if I am making incorrect comments...), but I will try... > 1) Does hotfloorplan creates these boxes as black boxes. Do I have to add them to power trace file for hotspot use? Where can I find info for > > these? > > These are the blank spaces, and are added to the floorplan file. You'll need to add them in the power trace file with zero power, otherwise HotSpot will complain. > > > 2) Which is the meaning of wire density of connections on description file? Does it mean logical connections (netlist info) or physical > > connections (after P&R)? I suppose the first one, but could you provide me an info source? Is there any limitation on the wire density value? > > I think you are basically right. It is essentially a chicken-and-egg problem: you need a floorplanner to reach a reasonable floorplan first. After that you can have more accurate estimation of communication delay between blocks, which should further help HotFloorplan to reach a more accurate floorplan (if needed). One purpose of developing HotFloorplan is to help users reach a reasonably good and thermally-optimized floorplan with limited information about the physical design. After that, more detailed P&R CAD tools can be used to fine tune the floorplan. Hope this helps. -Wei > > > Regards, > > kostas > > > > > > _______________________________________________ > 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/20101213/900c7589/attachment-0001.html From aziabari at soe.ucsc.edu Tue Dec 14 11:35:52 2010 From: aziabari at soe.ucsc.edu (Amirkoushyar Ziabari) Date: Tue, 14 Dec 2010 11:35:52 -0800 (PST) Subject: [Hotspot] HotSpot time complexity In-Reply-To: <713557221.1313571292355152602.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: <607673139.1313841292355352385.JavaMail.root@mail-01.cse.ucsc.edu> Thanks for the email. I am comparing our method (the Power Blurring method) with HotSpot and some other Architectural level methods and this is why I wanted to know about time complexity of HotSpot. Thanks, 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: "Andreas Bernauer" To: "Amirkoushyar Ziabari" Cc: hotspot at cs.virginia.edu Sent: Monday, December 13, 2010 1:12:02 AM GMT -08:00 US/Canada Pacific Subject: Re: [Hotspot] HotSpot time complexity ----- "Amirkoushyar Ziabari" wrote: > Dear all, > > would you please let me know what is the order of computation time > complexity of algorithm used in HotSpot? > Last time I checked, HotSpot used a 4th order Adaptive Runge-Kutta for integration. http://en.wikipedia.org/wiki/Runge%E2%80%93Kutta_methods#Adaptive_Runge.E2.80.93Kutta_methods Implementation is at RCutil.c, line 297ff. in HotSpot 4.0. May I ask why you would like to know the computation time complexity? 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 aziabari at soe.ucsc.edu Wed Dec 15 17:12:15 2010 From: aziabari at soe.ucsc.edu (Amirkoushyar Ziabari) Date: Wed, 15 Dec 2010 17:12:15 -0800 (PST) Subject: [Hotspot] Transient simulation grid level response Message-ID: <1413489310.45111292461935700.JavaMail.root@mail-01.cse.ucsc.edu> Dear all, It seems that for grid-level transient simulation, HotSpot only reports the block level response rather than temperature value of each grid. Would you please let me know how we can extract the grid temperature response of transient simulation? 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 From wh6p at virginia.edu Thu Dec 16 07:55:15 2010 From: wh6p at virginia.edu (Wei Huang) Date: Thu, 16 Dec 2010 09:55:15 -0600 Subject: [Hotspot] Transient simulation grid level response In-Reply-To: <1413489310.45111292461935700.JavaMail.root@mail-01.cse.ucsc.edu> References: <1413489310.45111292461935700.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: For steady-state grid temperatures, you can use the option -grid_steady_file (detailed info can be found in the HOW-TO page of HotSpot website). For transient grid temperatures, you may do a simple hack of the code by following the way the steady grid temperatures are dumped into a file. Hope this helps. -Wei On Wed, Dec 15, 2010 at 7:12 PM, Amirkoushyar Ziabari wrote: > Dear all, > > It seems that for grid-level transient simulation, HotSpot only reports the > block level response rather than temperature value of each grid. Would you > please let me know how we can extract the grid temperature response of > transient simulation? > > 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 > _______________________________________________ > 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/20101216/d8a200b8/attachment.html From ks4kk at virginia.edu Thu Dec 16 11:44:10 2010 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Thu, 16 Dec 2010 11:44:10 -0800 Subject: [Hotspot] HotSpot time complexity In-Reply-To: <1595189140.1213981292031333082.JavaMail.root@mail-01.cse.ucsc.edu> References: <374082142.1213361292029731365.JavaMail.root@mail-01.cse.ucsc.edu> <1595189140.1213981292031333082.JavaMail.root@mail-01.cse.ucsc.edu> Message-ID: Hi Amirkoushyar, There are three types of solvers in HotSpot: 1) Steady state solver of the block-based model: this is a direct-matrix solver based upon LUP decomposition. For an input matrix size of nxn (i.e., n-nodes), this is O(n^3) using the CLR-algorithm. On the other hand, if you have installed one of the Math acceleration engines and have enabled the MATHACCEL #define in HotSpot, it could be faster but I do not know the complexity of those engines. 2) Steady state solver of the grid model: this is an iterative solver using a multigrid approach with gauss-siedel as its fundamental step. This algorithm is linear in the grid size. If grid size N= n x n x k (n= no. of rows and cols; k is the constant no. of layers), then the order is O(n^2). Note that orthogonal to the no. of steps, the rate of convergence within a single step (which is dependent on the largest eigenvalue of the matrix that describes the difference equations) is very quick as well. The numerical error basically goes down geometrically with the ratio being related to this largest eigenvalue. 3) Transient solver for both the models: this is a fourth order Runge-Kutta based numerical integrator with adaptive step sizing. There are two parts to its computational complexity: a) The number of operations in a single time step: This is linear in the input size (i.e., for the block model with n blocks, it is O(n^2) and for the grid model with an nxn grid, it is O(n^2)) b) The total number of steps taken for the total interval of integration: Due to adaptive step-sizing, this depends on the initial temperature and the power input. For power inputs approximated by a sequence of pulses of differing widths and heights, the thermal responses are usually exponentials. So, one can say that the no. of steps taken is roughly O(log(t)) where t is the total interval of integration Hope this clarifies your question. For more information, please take a look at: http://www.nrbook.com/a/bookcpdf.php: Chapters 16.0-2, 19.5-6 https://engineering.purdue.edu/ME608/webpage/main.pdf Thanks, -karthik On Fri, Dec 10, 2010 at 5:35 PM, Amirkoushyar Ziabari wrote: > > Dear all, > > would you please let me know what is the order of computation time > complexity of algorithm used in HotSpot? > > 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 > _______________________________________________ > 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/20101216/1c468414/attachment.html From amin.ezhdehakosh at gmail.com Sun Dec 26 03:50:03 2010 From: amin.ezhdehakosh at gmail.com (Amin Ezhdehakosh) Date: Sun, 26 Dec 2010 15:20:03 +0330 Subject: [Hotspot] Remove Heat Sink and Heat Spreader Message-ID: Hi, How can I remove Heatsink and HeatSpreader in HotSpot? -- Regards, Amin Ezhdehakosh -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20101226/fd16703e/attachment.html From raghavan.ramadoss at gmail.com Thu Dec 30 06:57:06 2010 From: raghavan.ramadoss at gmail.com (RagZ) Date: Thu, 30 Dec 2010 06:57:06 -0800 Subject: [Hotspot] (no subject) Message-ID: http://discount2011.co.cc/o5r2f8