From wh6p at virginia.edu Mon Apr 2 04:37:38 2012 From: wh6p at virginia.edu (Wei Huang) Date: Mon, 2 Apr 2012 06:37:38 -0500 Subject: [Hotspot] About Hotspot In-Reply-To: References: <4F70F643.20301@cs.virginia.edu> Message-ID: Hi, I think I already replied that the default convection thermal resistance (r_conv) is reduced in later HotSpot versions, that's the reason you see lower temperatures. Please compare your HS2.0 and HS5.0 config files and see if that is the case. -Wei On Sun, Apr 1, 2012 at 9:00 PM, Chenyun Pan wrote: > Hi Wei, > > I remember that I sent you an Email last week and I still have some > problems about the Hotspot. I hope you can help me look at it. > About the configuration files (attached), you can see that I use 4 blocks > with 100W/cm2 each and also *I use block steady state model*. The > temperature is 64 degree which is quite low. And I saw some processor have > 50 to 100 W/cm2 power density can easily reach 100 degree core temperature. > And if I set the temperature limitation is 90 degree, then the power > density can reach 230 W/cm2 according to Hotspot 5. Is that too high? > Should I reduce some parameters to make it lower? > > Thanks, > > Chenyun > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120402/41624d88/attachment.html From nedahpg at gmail.com Fri Apr 6 10:27:06 2012 From: nedahpg at gmail.com (Neda Hassanpour) Date: Fri, 6 Apr 2012 21:57:06 +0430 Subject: [Hotspot] Calling HotSpot in Another Code Message-ID: Hi, I have used HotSpot to calculate a 3D IC temperature. Using HOWTO page, I ran HotSpot in Linux with specific commands and succeeded to get the temperature of components in different layers of one 3D IC with my own defined floorplan. Now I need to call HotSpot as a function with a power array input in another C code in Windows to generate a list of temperature. This temperature list would be an input for some functions in this C code. My main problem is how to call HotSpot in my C code. Reading some related mails in archives, I guess I should use sim_template.c for doing this (Is that true?). I looked at it but still have some questions. 1. I understood that I should change the content of sim_template.c according to my work. After changing sim_template.c content, Should I import sim_template.c to the C code? And would it be enough to give me the ability to call HotSpot in my C code? 2. There are 3 functions in sim_template.c, sim_init(),sim_main() and sim_exit(). I have some problems in modifying these functions. Can you give me some explanation about them please? Can anyone give me a modified sim_template.c according to a specific floorplan file? So I can use it as an example. 3. After changing sim_template.c properly and importing it to the C code, which of these 3 functions(sim_init(),sim_main(), sim_exit()) should I call in code to run HotSpot? 4.Is the output of running sim_template.c a file containing temperatures? I would be grateful if you help me. Thank you so much Neda -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120406/f07925f7/attachment.html From ks4kk at virginia.edu Fri Apr 6 12:06:10 2012 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Fri, 6 Apr 2012 12:06:10 -0700 Subject: [Hotspot] Calling HotSpot in Another Code In-Reply-To: References: Message-ID: Hi Neda, Please refer to Section C-3 of the HotSpot HOWTO here: http://lava.cs.virginia.edu/HotSpot/HotSpot-HOWTO.htm . It covers the answers to many of your questions. Hope it is useful. Thanks, -karthik On Fri, Apr 6, 2012 at 10:27 AM, Neda Hassanpour wrote: > Hi, > > > I have used HotSpot to calculate a 3D IC temperature. Using HOWTO page, I > ran HotSpot in Linux with specific commands and succeeded to get the > temperature of components in different layers of one 3D IC with my own > defined floorplan. > > > Now I need to call HotSpot as a function with a power array input in another > C code in Windows to generate a list of temperature. This temperature list > would be an input for some functions in this C code. > > > My main problem is how to call HotSpot in my C code. Reading some related > mails in archives, I guess I should use sim_template.c for doing this (Is > that true?). I looked at it but still have some questions. > > > 1. I understood that I should change the content of sim_template.c according > to my work. After changing sim_template.c content, Should I import > sim_template.c to the C code? And would it be enough to give me the ability > to call HotSpot in my C code? > > > 2. There are 3 functions in sim_template.c, sim_init(),sim_main() and > sim_exit(). I have some problems in modifying these functions. Can you give > me some explanation about them please? Can anyone give me a modified > sim_template.c according to a specific floorplan file? So I can use it as an > example. > > > > 3. After changing sim_template.c properly and importing it to the C code, > which of these 3 functions(sim_init(),sim_main(), sim_exit()) should I call > in code to run HotSpot? > > > > 4.Is the output of running sim_template.c a file containing temperatures? > > > > I would be grateful if you help me. > > > > Thank you so much > > Neda > > > > > > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > From nedahpg at gmail.com Sat Apr 7 04:49:00 2012 From: nedahpg at gmail.com (Neda Hassanpour) Date: Sat, 7 Apr 2012 16:19:00 +0430 Subject: [Hotspot] Calling HotSpot in Another Code In-Reply-To: References: Message-ID: * * ---------- Forwarded message ---------- From: Neda Hassanpour Date: Sat, Apr 7, 2012 at 16:17 Subject: Re: [Hotspot] Calling HotSpot in Another Code To: Karthik Sankaranarayanan Hi, Thanks for your quick reply. I read Section C-3 of the HotSpot HOWTO. Below are some my understandings after reading this page. Please let me know if I am wrong. 1. Since I don?t want to integrate HotSpot with a power/performance simulator it seems it would not be necessary for me to use sim_template.c and as the manual says it would be enough to include hotspot-iface.h to my code, is it true? 2. As mentioned in HOWTO, ?A user should be able to integrate HotSpot very quickly by just including hotspot-iface.h header file and linking the existing codebase with 'libhotspot.a'?. Can you please tell me what does ?linking the existing codebase with 'libhotspot.a'? mean and how should I do it? 3. I tried to run the sample code given in HOWTO as a shorter code snippet that demonstrates the high-level use of HotSpot in the user's code. Despite of what the manual has said about including hotspot-iface.h header file, in this code flp.h and temperature.h have been included and not hotspot-iface.h. Can you please let me know what the reason is? 4. After running this sample code, I get some linking errors like ?undefined reference to ?hotspot_vector(RC_model_t_st*)? ? as an example. Can you please help me to understand what the problem is? Thanks again Neda On Fri, Apr 6, 2012 at 23:36, Karthik Sankaranarayanan wrote: > Hi Neda, > > Please refer to Section C-3 of the HotSpot HOWTO here: > http://lava.cs.virginia.edu/HotSpot/HotSpot-HOWTO.htm . It covers the > answers to many of your questions. Hope it is useful. > > Thanks, > -karthik > > > On Fri, Apr 6, 2012 at 10:27 AM, Neda Hassanpour > wrote: > > Hi, > > > > > > I have used HotSpot to calculate a 3D IC temperature. Using HOWTO page, I > > ran HotSpot in Linux with specific commands and succeeded to get the > > temperature of components in different layers of one 3D IC with my own > > defined floorplan. > > > > > > Now I need to call HotSpot as a function with a power array input in > another > > C code in Windows to generate a list of temperature. This temperature > list > > would be an input for some functions in this C code. > > > > > > My main problem is how to call HotSpot in my C code. Reading some related > > mails in archives, I guess I should use sim_template.c for doing this (Is > > that true?). I looked at it but still have some questions. > > > > > > 1. I understood that I should change the content of sim_template.c > according > > to my work. After changing sim_template.c content, Should I import > > sim_template.c to the C code? And would it be enough to give me the > ability > > to call HotSpot in my C code? > > > > > > 2. There are 3 functions in sim_template.c, sim_init(),sim_main() and > > sim_exit(). I have some problems in modifying these functions. Can you > give > > me some explanation about them please? Can anyone give me a modified > > sim_template.c according to a specific floorplan file? So I can use it > as an > > example. > > > > > > > > 3. After changing sim_template.c properly and importing it to the C code, > > which of these 3 functions(sim_init(),sim_main(), sim_exit()) should I > call > > in code to run HotSpot? > > > > > > > > 4.Is the output of running sim_template.c a file containing temperatures? > > > > > > > > I would be grateful if you help me. > > > > > > > > Thank you so much > > > > Neda > > > > > > > > > > > > > > > > > > _______________________________________________ > > 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/20120407/d3750ea3/attachment.html From malreddysid at gmail.com Mon Apr 9 08:14:36 2012 From: malreddysid at gmail.com (Siddarth Malreddy) Date: Mon, 9 Apr 2012 11:14:36 -0400 Subject: [Hotspot] Help. Message-ID: I am currently working on a project relating to Thermal-Aware testing. I am pretty new to this field, and I have to use HotSpot in my work. Is there anyone who is working with the same thing that can help me? Can anyone help me understand the basic working of the software(suggest some book, or paper) ?? Thank you. Siddarth Malreddy. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120409/5c9f3233/attachment.html From ks4kk at virginia.edu Wed Apr 11 17:58:47 2012 From: ks4kk at virginia.edu (Karthik Sankaranarayanan) Date: Wed, 11 Apr 2012 17:58:47 -0700 Subject: [Hotspot] Calling HotSpot in Another Code In-Reply-To: References: Message-ID: ---------- Forwarded message ---------- From: Karthik Sankaranarayanan Date: Tue, Apr 10, 2012 at 9:42 AM Subject: Re: [Hotspot] Calling HotSpot in Another Code To: Neda Hassanpour Hi Neda, > 1.? Since I don?t want to integrate HotSpot with a power/performance > simulator it seems it would not be necessary for me to use sim_template.c > and as the manual says it would be enough to include hotspot-iface.h to my > code, is it true? That is right. > 2. As mentioned in HOWTO, ?A user should be able to integrate HotSpot very > quickly by just including hotspot-iface.h header file and linking the > existing codebase with 'libhotspot.a'?. Can you please tell me what does > ?linking the existing codebase with 'libhotspot.a'? mean and how should I do > it? Just the same way any other library would be linked. Say you are trying to compile trial.c (which includes hotspot-iface.h and calls the functions defined in it) and want to link libhotspot with it. Assuming both trial.c and libhotspot.a are in the same directory, here is the gcc command that would perform the compilation+linking: gcc -o trial trial.c libhospot.a > 3. I tried to run the sample code given in HOWTO as a shorter code snippet > that demonstrates the high-level use of HotSpot in the user's code. Despite > of what the manual has said about including hotspot-iface.h header file, in > this code flp.h and temperature.h have been included and not > hotspot-iface.h. ?Can you please let me know what the reason is? The code snippet shows a shorter version of the "tight integration" similar to sim-template.c where you would have to compile all the HotSpot files along with your source files (not just libhotspot.a). >From your description above, it looks like this is not what you want. > 4. After running this sample code, I get some linking errors like ?undefined > reference to ?hotspot_vector(RC_model_t_st*)? ? as an example. Can you > please help me to understand what the problem is? Please see #3 above. Hope this is useful. Thanks, -karthik From nedahpg at gmail.com Mon Apr 16 06:35:53 2012 From: nedahpg at gmail.com (Neda Hassanpour) Date: Mon, 16 Apr 2012 18:05:53 +0430 Subject: [Hotspot] Question about the steady temperatures Message-ID: Hi, I have a problem in working with Hotspot, which I would be grateful if anyone helps me. I set the ambient temperature to 25 C (298.5 K) and the initial temperature (init_temp) to 313.5 K in the config file. Then I ran the code below to get the steady temperatures of different components in my floorplan. *hotspot -c hotspot.config -f myflp.flp -p my.ptrace -o my.ttrace -steady_file my.steady -model_type grid* The result is strange to me. Changing the value of init_temp parameter has no effect on resulted steady state temperatures which are always around the ambient temperature. This change in initial temperature only affects the values in my.ttrace file. I also tried to read initial temperatures from a file by running the below command but the steady temperatures still remain the same. *hotspot -c hotspot.config -init_file my.init -f my.flp -p my.ptrace -o my.ttrace -model_type* *grid* As I understood only changing the ambient temperature will change the steady temperatures. Can you please tell me what the reason is? Why doesn?t the initial temperature and changing it have any effect on resulted steady temperatures? Thank you -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120416/263fd713/attachment.html From skadron at cs.virginia.edu Mon Apr 16 09:37:14 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Mon, 16 Apr 2012 12:37:14 -0400 Subject: [Hotspot] Question about the steady temperatures In-Reply-To: References: Message-ID: <4F8C4ABA.2090107@cs.virginia.edu> The initial conditions don't affect the steady state. That is only determined by the point at which heat injected (power) and heat removed (convection, etc, which depends on the ambient). To illustrate this, consider any object with zero power dissipation. No matter what its starting temperature, it will settle to the ambient (eg, an object moved from a hot environment to a cold environment will settle to the room temperature in that cold environment). /K On 4/16/12 9:35 AM, Neda Hassanpour wrote: > Hi, > > > I have a problem in working with Hotspot, which I would be grateful if > anyone helps me. > > I set the ambient temperature to 25 C (298.5 K) and the initial > temperature (init_temp) to 313.5 K in the config file. Then I ran the > code below to get the steady temperatures of different components in my > floorplan. > > /hotspot -c hotspot.config -f myflp.flp -p my.ptrace -o my.ttrace > -steady_file my.steady -model_type grid/ > > The result is strange to me. Changing the value of init_temp parameter > has no effect on resulted steady state temperatures which are always > around the ambient temperature. This change in initial temperature only > affects the values in my.ttrace file. I also tried to read initial > temperatures from a file by running the below command but the steady > temperatures still remain the same. > > > /hotspot -c hotspot.config -init_file my.init -f my.flp -p my.ptrace > -o my.ttrace -model_type/ /grid/ > > As I understood only changing the ambient temperature will change the > steady temperatures. Can you please tell me what the reason is? Why > doesn?t the initial temperature and changing it have any effect on > resulted steady temperatures? > > > Thank you > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot From nedahpg at gmail.com Mon Apr 16 10:18:35 2012 From: nedahpg at gmail.com (Neda Hassanpour) Date: Mon, 16 Apr 2012 21:48:35 +0430 Subject: [Hotspot] Question about the steady temperatures In-Reply-To: <4F8C4ABA.2090107@cs.virginia.edu> References: <4F8C4ABA.2090107@cs.virginia.edu> Message-ID: Thank you all for your replies. I need to use Hotspot as a sensor and see the change of temperature in each floorplan component according to change in its power and initial temperature. From your response it seems that using the steady state temperatures would not be suitable for me. Can you please let me know how should I extract the temperatures from Hotspot which are appropriate for me? Should I use the transient temperature from ttrace file? Or should I change any parameters from config file? Thanks again Neda On Mon, Apr 16, 2012 at 21:07, Kevin Skadron wrote: > The initial conditions don't affect the steady state. That is only > determined by the point at which heat injected (power) and heat removed > (convection, etc, which depends on the ambient). > > To illustrate this, consider any object with zero power dissipation. No > matter what its starting temperature, it will settle to the ambient (eg, an > object moved from a hot environment to a cold environment will settle to > the room temperature in that cold environment). > > /K > > > On 4/16/12 9:35 AM, Neda Hassanpour wrote: > >> Hi, >> >> >> I have a problem in working with Hotspot, which I would be grateful if >> anyone helps me. >> >> I set the ambient temperature to 25 C (298.5 K) and the initial >> temperature (init_temp) to 313.5 K in the config file. Then I ran the >> code below to get the steady temperatures of different components in my >> floorplan. >> >> /hotspot -c hotspot.config -f myflp.flp -p my.ptrace -o my.ttrace >> -steady_file my.steady -model_type grid/ >> >> >> The result is strange to me. Changing the value of init_temp parameter >> has no effect on resulted steady state temperatures which are always >> around the ambient temperature. This change in initial temperature only >> affects the values in my.ttrace file. I also tried to read initial >> temperatures from a file by running the below command but the steady >> temperatures still remain the same. >> >> >> /hotspot -c hotspot.config -init_file my.init -f my.flp -p my.ptrace >> -o my.ttrace -model_type/ /grid/ >> >> >> As I understood only changing the ambient temperature will change the >> steady temperatures. Can you please tell me what the reason is? Why >> doesn?t the initial temperature and changing it have any effect on >> resulted steady temperatures? >> >> >> Thank you >> >> >> >> ______________________________**_________________ >> 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/20120416/904ff546/attachment.html From skadron at cs.virginia.edu Mon Apr 16 10:36:55 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Mon, 16 Apr 2012 13:36:55 -0400 Subject: [Hotspot] Question about the steady temperatures In-Reply-To: References: <4F8C4ABA.2090107@cs.virginia.edu> Message-ID: <4F8C58B7.4050006@cs.virginia.edu> If there is anything dynamic (ie, not static/steady state) then indeed, you need to use the transient capabilities. Dynamic changes in power can be specified via the power trace, and the corresponding dynamic behavior observed in the temperature trace. Initial conditions are set in the config file. Alternatively, if the behavior you're trying to model is driven by some kind of simulation, dynamic changes in power, and the corresponding dynamic changes in temperature, can be modeled within a simulator by making the appropriate calls each time step from your simulator into the HotSpot library. Our simulator example shows how you can modify your simulator to do this. /K On 4/16/2012 1:18 PM, Neda Hassanpour wrote: > Thank you all for your replies. > > > I need to use Hotspot as a sensor and see the change of temperature in > each floorplan component according to change in its power and initial > temperature. From your response it seems that using the steady state > temperatures would not be suitable for me. Can you please let me know > how should I extract the temperatures from Hotspot which are appropriate > for me? Should I use the transient temperature from ttrace file? Or > should I change any parameters from config file? > > > Thanks again > > Neda > > > On Mon, Apr 16, 2012 at 21:07, Kevin Skadron > wrote: > > The initial conditions don't affect the steady state. That is only > determined by the point at which heat injected (power) and heat > removed (convection, etc, which depends on the ambient). > > To illustrate this, consider any object with zero power dissipation. > No matter what its starting temperature, it will settle to the > ambient (eg, an object moved from a hot environment to a cold > environment will settle to the room temperature in that cold > environment). > > /K > > > On 4/16/12 9:35 AM, Neda Hassanpour wrote: > > Hi, > > > I have a problem in working with Hotspot, which I would be > grateful if > anyone helps me. > > I set the ambient temperature to 25 C (298.5 K) and the initial > temperature (init_temp) to 313.5 K in the config file. Then I > ran the > code below to get the steady temperatures of different > components in my > floorplan. > > /hotspot -c hotspot.config -f myflp.flp -p my.ptrace -o my.ttrace > -steady_file my.steady -model_type grid/ > > > The result is strange to me. Changing the value of init_temp > parameter > has no effect on resulted steady state temperatures which are always > around the ambient temperature. This change in initial > temperature only > affects the values in my.ttrace file. I also tried to read initial > temperatures from a file by running the below command but the steady > temperatures still remain the same. > > > /hotspot -c hotspot.config -init_file my.init -f my.flp -p > my.ptrace > -o my.ttrace -model_type/ /grid/ > > > As I understood only changing the ambient temperature will > change the > steady temperatures. Can you please tell me what the reason is? Why > doesn?t the initial temperature and changing it have any effect on > resulted steady temperatures? > > > Thank you > > > > _________________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/__mailman/listinfo/hotspot > > > From zaeem.baig at mavs.uta.edu Mon Apr 23 17:58:14 2012 From: zaeem.baig at mavs.uta.edu (Baig, Zaeem) Date: Tue, 24 Apr 2012 00:58:14 +0000 Subject: [Hotspot] Hotspot (Thermal Resistance) In-Reply-To: References: Message-ID: Hi, I want to get the temperature distribution on a planar chip with multiple heat sources. Now my question is that, is it possible that instead of specifying the interface material specs, heat sink specs, spreader specs, e.t.c seperately, I could use a single value for the thermal resistance which I have calculated experimentally and encompasses all the thermal resistances? For example, I specify the chip specs and input the floor plan and use thermal resistance= 2K/W or 5000 W/m2K (Equivalent thermal resistance), and generate the temperature profile. Is this possible in hotspot? Thank you for your prompt response. Regards, Zaeem -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120424/1f650a23/attachment.html From zaeem.baig at mavs.uta.edu Mon Apr 23 18:04:47 2012 From: zaeem.baig at mavs.uta.edu (Baig, Zaeem) Date: Tue, 24 Apr 2012 01:04:47 +0000 Subject: [Hotspot] Hotspot (Thermal Resistance) In-Reply-To: References: Message-ID: Hi, I want to get the temperature distribution on a planar chip with multiple heat sources. Now my question is that, is it possible that instead of specifying the interface material specs, heat sink specs, spreader specs, e.t.c seperately, I could use a single value for the thermal resistance which I have calculated experimentally and encompasses all the thermal resistances? For example, I specify the chip specs and input the floor plan and use thermal resistance= 2K/W or 5000 W/m2K (Equivalent thermal resistance), and generate the temperature profile. Is this possible in hotspot? Thank you for your prompt response. Regards, Zaeem -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120424/b9cf0a18/attachment.html From skadron at cs.virginia.edu Mon Apr 23 18:25:21 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Mon, 23 Apr 2012 21:25:21 -0400 Subject: [Hotspot] Hotspot (Thermal Resistance) In-Reply-To: References: Message-ID: <4F960101.4070104@cs.virginia.edu> Hi Zaeem, I'm afraid HotSpot can't quite work that way. We can't get the temperatures at the internal nodes without knowing the vertical and lateral resistances for each layer. As long as you know the dimensions and materials, HotSpot calculates all the internal properties. /K On 4/23/2012 8:58 PM, Baig, Zaeem wrote: > > Hi, > > I want to get the temperature distribution on a planar chip with > multiple heat sources. Now my question is that, is it possible that > instead of specifying the interface material specs, heat sink specs, > spreader specs, e.t.c seperately, I could use a single value for the > thermal resistance which I have calculated experimentally and > encompasses all the thermal resistances? > > For example, I specify the chip specs and input the floor plan and use > thermal resistance= 2K/W or 5000 W/m2K (Equivalent thermal resistance), > and generate the temperature profile. Is this possible in hotspot? > > Thank you for your prompt response. > > Regards, > Zaeem > > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot From dipinprakash at gmail.com Thu Apr 26 02:43:00 2012 From: dipinprakash at gmail.com (dipin prakash) Date: Thu, 26 Apr 2012 02:43:00 -0700 Subject: [Hotspot] Error : line too long Message-ID: Sir, How can i remove the Error: *Line too long* when i try to create the grid.steady file for some of the ISCAS benchmark circuits. When i try to create the steady file alone the error is *Hotspot: util.c:165 dmatrix : Assetion m[0] != ((void*)0) failed. * Regards, DIPIN PRAKASH S -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120426/23db7dc8/attachment-0001.html From solidrepellent at gmail.com Fri Apr 27 05:40:30 2012 From: solidrepellent at gmail.com (Solid Repel) Date: Fri, 27 Apr 2012 15:40:30 +0300 Subject: [Hotspot] Boundary Conditions: Need a re-check In-Reply-To: References: Message-ID: Hi Wei and hotspot, HotSpot assumes no heat transfer path on the edges of silicon. That is why > it reports higher corner temperatures. If there a parameter in the config file which can be used to change this default behavior. What is the motivation of the hotspot developers to model it like that in the first place. I always felt that there is air or other gas (and not vaccum) in between the silicon and the edge of heat spreader/sink. Thank you very much for your time. Naveen. On Tue, Apr 5, 2011 at 1:11 PM, Devendra Rai wrote: > Hello HotSpot > > I am using the default EV6/alpha floorplan which ships with hotspot. I > found some multi-processor platforms with the ASU's magma project, and > I picked up an 8x8 EV6 floorplan. > > I am running power numbers (again, magma project, bzip2 load), and I > find a strange issue. The cores on the corner show the highest > temperature while the cores towards inside show a lower temperature. > This is contrary to the perception which says that for the same power > dissipation, cores on the corner will be the coolest. > > Attached are the matlab fig files for your review. Can you please let > me know if there is any explanation for this? The hotspot "engine" is > un-modified. > > Best, > > > Devendra Rai > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120427/7597a2f0/attachment.html From skadron at cs.virginia.edu Fri Apr 27 06:04:38 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Fri, 27 Apr 2012 09:04:38 -0400 Subject: [Hotspot] Boundary Conditions: Need a re-check In-Reply-To: References: Message-ID: <4F9A9966.5080403@cs.virginia.edu> There's no direct support for this, but you could certainly add it by changing the model to have an appropriate set of thermal resistances and capacitances off the edges of each layer. We omitted this because the heat transfer in this direction will be negligible, because the heat transfer is proportional to the surface area and the air flow would be poor. You can do a back of the envelope analysis to verify this. /K On 4/27/12 8:40 AM, Solid Repel wrote: > Hi Wei and hotspot, > > HotSpot assumes no heat transfer path on the edges of silicon. That > is why it reports higher corner temperatures. > > > If there a parameter in the config file which can be used to change this > default behavior. What is the motivation of the hotspot developers to > model it like that in the first place. > > I always felt that there is air or other gas (and not vaccum) in between > the silicon and the edge of heat spreader/sink. > > Thank you very much for your time. > > Naveen. > > > > > On Tue, Apr 5, 2011 at 1:11 PM, Devendra Rai > > wrote: > > Hello HotSpot > > I am using the default EV6/alpha floorplan which ships with hotspot. I > found some multi-processor platforms with the ASU's magma project, and > I picked up an 8x8 EV6 floorplan. > > I am running power numbers (again, magma project, bzip2 load), and I > find a strange issue. The cores on the corner show the highest > temperature while the cores towards inside show a lower temperature. > This is contrary to the perception which says that for the same power > dissipation, cores on the corner will be the coolest. > > Attached are the matlab fig files for your review. Can you please let > me know if there is any explanation for this? The hotspot "engine" is > un-modified. > > Best, > > > Devendra Rai > > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot From devendra.rai at tik.ee.ethz.ch Fri Apr 27 06:30:11 2012 From: devendra.rai at tik.ee.ethz.ch (Devendra Rai) Date: Fri, 27 Apr 2012 15:30:11 +0200 Subject: [Hotspot] Boundary Conditions: Need a re-check In-Reply-To: References: Message-ID: Ha... I asked exactly the same question a year ago.. On Fri, Apr 27, 2012 at 14:40, Solid Repel wrote: > Hi Wei and hotspot, > > HotSpot assumes no heat transfer path on the edges of silicon. That is why >> it reports higher corner temperatures. > > > If there a parameter in the config file which can be used to change this > default behavior. What is the motivation of the hotspot developers to model > it like that in the first place. > > I always felt that there is air or other gas (and not vaccum) in between > the silicon and the edge of heat spreader/sink. > > Thank you very much for your time. > > Naveen. > > > > > > On Tue, Apr 5, 2011 at 1:11 PM, Devendra Rai wrote: > >> Hello HotSpot >> >> I am using the default EV6/alpha floorplan which ships with hotspot. I >> found some multi-processor platforms with the ASU's magma project, and >> I picked up an 8x8 EV6 floorplan. >> >> I am running power numbers (again, magma project, bzip2 load), and I >> find a strange issue. The cores on the corner show the highest >> temperature while the cores towards inside show a lower temperature. >> This is contrary to the perception which says that for the same power >> dissipation, cores on the corner will be the coolest. >> >> Attached are the matlab fig files for your review. Can you please let >> me know if there is any explanation for this? The hotspot "engine" is >> un-modified. >> >> Best, >> >> >> Devendra Rai >> > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120427/d89cd6c1/attachment.html From wh6p at virginia.edu Fri Apr 27 07:40:21 2012 From: wh6p at virginia.edu (Wei Huang) Date: Fri, 27 Apr 2012 09:40:21 -0500 Subject: [Hotspot] Boundary Conditions: Need a re-check In-Reply-To: References: Message-ID: I agree with Kevin's explanation -- that's the reason we didn't model this. Most of the heat will be conducted vertically and spread out via the underneath heat spreader. Another fact that people usually neglect or are not aware of is that there is some blank or low-power area along the edge of most processor dies. HotSpot do have a parameter in the config file to add those areas to make a more realistic construction of the model. So in reality, the impact of heat conduction via the edge of the chip is indeed negligible. -Wei On Fri, Apr 27, 2012 at 8:30 AM, Devendra Rai wrote: > Ha... I asked exactly the same question a year ago.. > > > > On Fri, Apr 27, 2012 at 14:40, Solid Repel wrote: > >> Hi Wei and hotspot, >> >> HotSpot assumes no heat transfer path on the edges of silicon. That is >>> why it reports higher corner temperatures. >> >> >> If there a parameter in the config file which can be used to change this >> default behavior. What is the motivation of the hotspot developers to model >> it like that in the first place. >> >> I always felt that there is air or other gas (and not vaccum) in between >> the silicon and the edge of heat spreader/sink. >> >> Thank you very much for your time. >> >> Naveen. >> >> >> >> >> >> On Tue, Apr 5, 2011 at 1:11 PM, Devendra Rai > > wrote: >> >>> Hello HotSpot >>> >>> I am using the default EV6/alpha floorplan which ships with hotspot. I >>> found some multi-processor platforms with the ASU's magma project, and >>> I picked up an 8x8 EV6 floorplan. >>> >>> I am running power numbers (again, magma project, bzip2 load), and I >>> find a strange issue. The cores on the corner show the highest >>> temperature while the cores towards inside show a lower temperature. >>> This is contrary to the perception which says that for the same power >>> dissipation, cores on the corner will be the coolest. >>> >>> Attached are the matlab fig files for your review. Can you please let >>> me know if there is any explanation for this? The hotspot "engine" is >>> un-modified. >>> >>> Best, >>> >>> >>> Devendra Rai >>> >> >> >> _______________________________________________ >> HotSpot mailing list >> HotSpot at mail.cs.virginia.edu >> http://www.cs.virginia.edu/mailman/listinfo/hotspot >> >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120427/e3449a28/attachment.html From skadron at cs.virginia.edu Sun Apr 29 07:10:18 2012 From: skadron at cs.virginia.edu (Kevin Skadron) Date: Sun, 29 Apr 2012 10:10:18 -0400 Subject: [Hotspot] Error : line too long In-Reply-To: References: Message-ID: <4F9D4BCA.7020704@cs.virginia.edu> It looks like your model has too many units; we only provisioned for 64 KB per line. In fact, the assertion also suggests that the input is so large that your system can't allocate enough memory. You could easily modify HotSpot to support larger models, but it looks like you also need to deal with the memory capacity problem. /K On 4/26/2012 5:43 AM, dipin prakash wrote: > > Sir, > > How can i remove the Error: *Line too long* when i try to create the > grid.steady file for some of the ISCAS benchmark circuits. > When i try to create the steady file alone the error is *Hotspot: > util.c:165 dmatrix : Assetion m[0] != ((void*)0) failed. > * > > Regards, > > DIPIN PRAKASH S > > > > > > > _______________________________________________ > HotSpot mailing list > HotSpot at mail.cs.virginia.edu > http://www.cs.virginia.edu/mailman/listinfo/hotspot