From sachitni at usc.edu  Sun Dec  6 18:30:56 2009
From: sachitni at usc.edu (Samir Chitnis)
Date: Sun, 06 Dec 2009 15:30:56 -0800
Subject: [Hotspot] Integration with SimpleScalar
Message-ID: <de4ec8961e1e6.4b1bce30@usc.edu>

Hi,
I saw the sim-template.c file provided with HotSpot code. 
Following are my conclusion, please correct me if I am wrong:
1. I will have to create a similar code and add it to the SimpleScalar code, so as to initiate the HotSpot functions when SimpleScalar simulator is run.
2. In such a case the HotSpot will use the same benchmark being used by SimpleScalar to simulate the temperature for the provided floorplan.
3. The output of the HotSpot will be a file.

However, for my project I need the HotSpot to provide the temperatures to the SimpleScalar code during the simulation. I need to perform some calculations using these temperature feedback in the SimpleScalar.
Can you provide me some information on how this can be done?

Thanks,
Samir.

From bernauer at informatik.uni-tuebingen.de  Mon Dec  7 00:53:22 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Mon, 7 Dec 2009 06:53:22 +0100
Subject: [Hotspot] Integration with SimpleScalar
In-Reply-To: <de4ec8961e1e6.4b1bce30@usc.edu>
References: <de4ec8961e1e6.4b1bce30@usc.edu>
Message-ID: <c92980110912062153g3431f7b6k9e9b9762cd71ca0a@mail.gmail.com>

Hi Samir,

On Mon, Dec 7, 2009 at 00:30, Samir Chitnis <sachitni at usc.edu> wrote:
> Hi,
> I saw the sim-template.c file provided with HotSpot code.
> Following are my conclusion, please correct me if I am wrong:
> 1. I will have to create a similar code and add it to the SimpleScalar code, so as to initiate the HotSpot functions when SimpleScalar simulator is run.
> 2. In such a case the HotSpot will use the same benchmark being used by SimpleScalar to simulate the temperature for the provided floorplan.
> 3. The output of the HotSpot will be a file.

The output will not be a file, but temperature data in an array, see
lines 155-158: after the call to compute_temp, you can find the
resulting temperatures in the temp[] array; see also the comment in
the source file before these lines.

> However, for my project I need the HotSpot to provide the temperatures to the SimpleScalar code during the simulation. I need to perform some calculations using these temperature feedback in the SimpleScalar.
> Can you provide me some information on how this can be done?

After lines 155-158 in sim-template.c, you can use the temperature
data in temp[] to feed it back into your simulator.


Cheers,

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html


From sachitni at usc.edu  Mon Dec  7 03:07:20 2009
From: sachitni at usc.edu (Samir Chitnis)
Date: Mon, 07 Dec 2009 00:07:20 -0800
Subject: [Hotspot] Integration with SimpleScalar
In-Reply-To: <c92980110912062153g3431f7b6k9e9b9762cd71ca0a@mail.gmail.com>
References: <de4ec8961e1e6.4b1bce30@usc.edu>
	<c92980110912062153g3431f7b6k9e9b9762cd71ca0a@mail.gmail.com>
Message-ID: <dde19bf7187bf.4b1c4738@usc.edu>


Hi Andreas,

Thanks a lot for your help.
I have a few more doubts, it would be nice if you can help me with those.

1. Should the code in sim-templace.c be added to one of the files in the SimpleScalar Simulator?

2. After interfacing HotSpot with SimpleScalar, (for now, i have just linked using hotspot-iface.h) HotSpot functions would be called from the simplescalar code. However, the HotSpot would need the floorplan and power trace file as the inputs. What would be command to initiate the simplescalar in this case.
eg. I am using the following command to run simplescalar: ./sim-outorder tests-pisa/bin.little/test-math
How should this be modified such that HotSpot can run too.

Thanks,
Samir.

----- Original Message -----
From: Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
Date: Sunday, December 6, 2009 9:53 pm
Subject: Re: [Hotspot] Integration with SimpleScalar
To: Samir Chitnis <sachitni at usc.edu>
Cc: hotspot at mail.cs.virginia.edu

> Hi Samir,
> 
> On Mon, Dec 7, 2009 at 00:30, Samir Chitnis <sachitni at usc.edu> wrote:
> > Hi,
> > I saw the sim-template.c file provided with HotSpot code.
> > Following are my conclusion, please correct me if I am wrong:
> > 1. I will have to create a similar code and add it to the 
> SimpleScalar code, so as to initiate the HotSpot functions when 
> SimpleScalar simulator is run.
> > 2. In such a case the HotSpot will use the same benchmark being 
> used by SimpleScalar to simulate the temperature for the provided 
> floorplan.> 3. The output of the HotSpot will be a file.
> 
> The output will not be a file, but temperature data in an array, see
> lines 155-158: after the call to compute_temp, you can find the
> resulting temperatures in the temp[] array; see also the comment in
> the source file before these lines.
> 
> > However, for my project I need the HotSpot to provide the 
> temperatures to the SimpleScalar code during the simulation. I need 
> to perform some calculations using these temperature feedback in 
> the SimpleScalar.
> > Can you provide me some information on how this can be done?
> 
> After lines 155-158 in sim-template.c, you can use the temperature
> data in temp[] to feed it back into your simulator.
> 
> 
> Cheers,
> 
> -- 
> Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
> WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
> http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
> 


From bernauer at informatik.uni-tuebingen.de  Mon Dec  7 03:34:47 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Mon, 7 Dec 2009 09:34:47 +0100
Subject: [Hotspot] Integration with SimpleScalar
In-Reply-To: <dde19bf7187bf.4b1c4738@usc.edu>
References: <de4ec8961e1e6.4b1bce30@usc.edu>
	<c92980110912062153g3431f7b6k9e9b9762cd71ca0a@mail.gmail.com>
	<dde19bf7187bf.4b1c4738@usc.edu>
Message-ID: <c92980110912070034x3a06247cud501aef5022dbc@mail.gmail.com>

Hi Samir,

On Mon, Dec 7, 2009 at 09:07, Samir Chitnis <sachitni at usc.edu> wrote:
> Thanks a lot for your help.
> I have a few more doubts, it would be nice if you can help me with those.
>
> 1. Should the code in sim-templace.c be added to one of the files in the SimpleScalar Simulator?
>
> 2. After interfacing HotSpot with SimpleScalar, (for now, i have just linked using hotspot-iface.h) HotSpot functions would be called from the simplescalar code. However, the HotSpot would need the floorplan and power trace file as the inputs. What would be command to initiate the simplescalar in this case.
> eg. I am using the following command to run simplescalar: ./sim-outorder tests-pisa/bin.little/test-math
> How should this be modified such that HotSpot can run too.

I don't use SimpleScalar, so I can't help you with that.

However, if you want to integrate hotspot in some other tool, you have
to do everything by yourself, as outlined in sim-template.c. For
example, you read the floorplan file with read_flp, and provide the
power traces in an array to compute_temp, not via a power trace file.
Have a look again at sim-template.c.

Cheers,

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html


From rehan1984 at gmail.com  Mon Dec  7 12:19:47 2009
From: rehan1984 at gmail.com (Rehan Ahmed)
Date: Mon, 7 Dec 2009 11:19:47 -0600
Subject: [Hotspot] heat sink, thermal interface values in transient analysis
In-Reply-To: <fa23dc170912070806u7599f08dw8eba01d0829de607@mail.gmail.com>
References: <fa23dc170912070806u7599f08dw8eba01d0829de607@mail.gmail.com>
Message-ID: <fa23dc170912070919x624091cta2d037a5e49f8f8a@mail.gmail.com>

Hi,
I noticed that HOTSPOT does not give heat sink, thermal interface
temperature......in transient analysis output. Is there any way of finding
those values using hotspot. Is it simply the case of
HOTSPOT evaluating these values but not outputting them or does it assume
the values constant as far as transient analysis is concerned??

Thanks,

Rehan Ahmed
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091207/635b8ca8/attachment.html 

From wh6p at virginia.edu  Mon Dec  7 19:28:09 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Mon, 07 Dec 2009 19:28:09 -0500
Subject: [Hotspot] heat sink,
 thermal interface values in transient analysis
In-Reply-To: <fa23dc170912070919x624091cta2d037a5e49f8f8a@mail.gmail.com>
References: <fa23dc170912070806u7599f08dw8eba01d0829de607@mail.gmail.com>
	<fa23dc170912070919x624091cta2d037a5e49f8f8a@mail.gmail.com>
Message-ID: <web-9664859@koi.mail.virginia.edu>

Hi Rehan,

These temperatures are available. We didn't output it in the transient 
temperature trace just because most users are not directly using them. You 
can find them in the final temp[] array, which contains all temperatures in 
all layers. You just need to find the correct offset of the nodes at the TIM 
or heatsink layers. Silicon nodes usually start from 0 to n-1 (where n is 
the number of blocks in the block model, or the number of grid cells of the 
silicon layer in the grid model), TIM nodes start from n to 2n-1, etc.

-Wei

On Mon, 7 Dec 2009 11:19:47 -0600
  Rehan Ahmed <rehan1984 at gmail.com> wrote:
> Hi,
> I noticed that HOTSPOT does not give heat sink, thermal interface
> temperature......in transient analysis output. Is there any way of finding
> those values using hotspot. Is it simply the case of
> HOTSPOT evaluating these values but not outputting them or does it assume
> the values constant as far as transient analysis is concerned??
> 
> Thanks,
> 
> Rehan Ahmed


From paulmathew111 at gmail.com  Wed Dec  9 16:03:14 2009
From: paulmathew111 at gmail.com (mathew paul)
Date: Wed, 9 Dec 2009 13:03:14 -0800
Subject: [Hotspot] query regarding input power trace
Message-ID: <f055de000912091303q12d80e21if272ebc62b9e99b1@mail.gmail.com>

Hi,

I am trying to use the Hotspot tool to evaluate the temperatures of various
blocks of cache memory and am having some trouble with the input power trace
file. I have the dynamic energy values for various cache configurations from
CACTI and compute the power for each block based on the number of accesses
and averaged over time of these accesses. What should the unit of power be
for the power trace - watts or milli watts ?
Thanks

Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091209/935a8901/attachment.html 

From wh6p at virginia.edu  Wed Dec  9 17:16:42 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Wed, 09 Dec 2009 17:16:42 -0500
Subject: [Hotspot] query regarding input power trace
In-Reply-To: <f055de000912091303q12d80e21if272ebc62b9e99b1@mail.gmail.com>
References: <f055de000912091303q12d80e21if272ebc62b9e99b1@mail.gmail.com>
Message-ID: <web-7754635@neon.mail.virginia.edu>

Hi,

By default, it is in watts.

-Wei


On Wed, 9 Dec 2009 13:03:14 -0800
  mathew paul <paulmathew111 at gmail.com> wrote:
> Hi,
> 
> I am trying to use the Hotspot tool to evaluate the temperatures of 
>various
> blocks of cache memory and am having some trouble with the input power 
>trace
> file. I have the dynamic energy values for various cache configurations 
>from
> CACTI and compute the power for each block based on the number of accesses
> and averaged over time of these accesses. What should the unit of power be
> for the power trace - watts or milli watts ?
> Thanks
> 
> Matt


From fazal256 at yahoo.com  Thu Dec 10 07:21:52 2009
From: fazal256 at yahoo.com (fazal hameed)
Date: Thu, 10 Dec 2009 04:21:52 -0800 (PST)
Subject: [Hotspot] Integrating hotspot with power simulator
Message-ID: <974822.70437.qm@web54207.mail.re2.yahoo.com>

Dear All,
         I am using hotspot integrated with a power simulator. The frequency of the processor is passed as an argument to the function sim_init() and I am initializing the sim_init(double frequency) function provided with hot spot like this
config.base_proc_freq = frequency
config.sampling_intvl = (10000 / frequency);


While the sim_main(double power_input[]) function is invoked by the power simulator after 10K clock cycles by passing aggreggate power values calculated over the span of 10K clock cycles. I have removed the while loop of sim_main function and am using the block model of the hotspot.

power[get_blk_indesx(flp,"LSQ")] = power_input[0];
power[get_blk_indesx(flp,"ITLB")] = power_input[1];

Also i have removed the if statement (cur_time - prev_time) >= model->config->sampling_intvl

and power[i] is calculated as power[i] /= 10000;

The total core area is 10.85 mm2. As I am using the future technology nodes the core area is very small. The rest  of the parameters are set as default.

I have run the benchmark for 1.68 billion instructions with the power values and the max temp i got is 340.73 kelvin and the min temp is 333.15 kelvin and even for power trace for 2.56 billion instruction. I got nearly the same max temp of 340.875kelvin and min temp is still 333.15.

Am I doing the right thing or there is some problem in the experiment with the smaller core area.

regards,
Fazal Hameed



      

From ks4kk at virginia.edu  Thu Dec 10 21:38:18 2009
From: ks4kk at virginia.edu (Karthik Sankaranarayanan)
Date: Thu, 10 Dec 2009 18:38:18 -0800
Subject: [Hotspot] Integrating hotspot with power simulator
In-Reply-To: <974822.70437.qm@web54207.mail.re2.yahoo.com>
References: <974822.70437.qm@web54207.mail.re2.yahoo.com>
Message-ID: <eae341010912101838r479229d3w55f7e670d1ec8fda@mail.gmail.com>

Hi Fazal,

You should try running HotSpot through the trace file interface first to
check if you get the temperature numbers you expect (for the power inputs
you are interested in). Then, try the cycle-level simulator below and check
to see if they match. If the problem still persists, then let us know and we
can help you.

Thanks,
-karthik

On Thu, Dec 10, 2009 at 4:21 AM, fazal hameed <fazal256 at yahoo.com> wrote:

> Dear All,
>         I am using hotspot integrated with a power simulator. The frequency
> of the processor is passed as an argument to the function sim_init() and I
> am initializing the sim_init(double frequency) function provided with hot
> spot like this
> config.base_proc_freq = frequency
> config.sampling_intvl = (10000 / frequency);
>
>
> While the sim_main(double power_input[]) function is invoked by the power
> simulator after 10K clock cycles by passing aggreggate power values
> calculated over the span of 10K clock cycles. I have removed the while loop
> of sim_main function and am using the block model of the hotspot.
>
> power[get_blk_indesx(flp,"LSQ")] = power_input[0];
> power[get_blk_indesx(flp,"ITLB")] = power_input[1];
>
> Also i have removed the if statement (cur_time - prev_time) >=
> model->config->sampling_intvl
>
> and power[i] is calculated as power[i] /= 10000;
>
> The total core area is 10.85 mm2. As I am using the future technology nodes
> the core area is very small. The rest  of the parameters are set as default.
>
> I have run the benchmark for 1.68 billion instructions with the power
> values and the max temp i got is 340.73 kelvin and the min temp is 333.15
> kelvin and even for power trace for 2.56 billion instruction. I got nearly
> the same max temp of 340.875kelvin and min temp is still 333.15.
>
> Am I doing the right thing or there is some problem in the experiment with
> the smaller core area.
>
> regards,
> Fazal Hameed
>
>
>
>
> _______________________________________________
> 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/20091210/018d846c/attachment.html 

From wh6p at virginia.edu  Fri Dec 11 11:44:18 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 11 Dec 2009 11:44:18 -0500
Subject: [Hotspot] [Fwd: Hotspot]
In-Reply-To: <4B226679.6020206@cs.virginia.edu>
References: <4B226679.6020206@cs.virginia.edu>
Message-ID: <web-7754415@tetra.mail.virginia.edu>

Hello Farzad,

HotSpot is a stand-alone temperature model, not a power model. In order to 
use HotSpot, a power trace input is assumed and has to be generated 
elsewhere outside HotSpot. Tools such as Wattch can do that. However, to 
help our users to better use Hotspot, we provided a simple template showing 
the interface (sim-template.c,.h) to integrate HotSpot with performance 
simulators such as SimpleScalar and power models such as Wattch. You may 
find these useful.

-Wei

> -------- Original Message --------
> Subject: 	Hotspot
> Date: 	Fri, 11 Dec 2009 18:29:28 +0330
>From: 	farzad samie <fsamie at gmail.com>
> To: 	Gurumurthi at cs.virginia.edu
> 
> 
> 
> Dear Sudhanva,
> I have a question about Hotspot. how did you produce power trace files? I 
>want to have some large ptrace files that should be realistic. can you 
>please help me.
> 
> -- 
>Farzad Samie
> 


From wh6p at virginia.edu  Fri Dec 11 17:26:03 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 11 Dec 2009 17:26:03 -0500
Subject: [Hotspot] link to earlier HotSpot releases
Message-ID: <web-8648695@guppy.mail.virginia.edu>

Dear HotSpot users,

It occurs to us that several users are interested in obtaining earlier 
HotSpot releases but don't know where to download them. We realize it would 
be useful to have them available -- all previous releases can be found at 
the following URL:

http://lava.cs.virginia.edu/HotSpot/grab/

If you have questions while using HotSpot for your research, please send an 
email to the mailing list. We are sure all HotSpot users, including us, 
would be very happy to help.

Regards,

The HotSpot team

From wh6p at virginia.edu  Fri Dec 11 17:59:35 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 11 Dec 2009 17:59:35 -0500
Subject: [Hotspot] about on-chip metal layer's impact on silicon temperature
Message-ID: <web-8648848@guppy.mail.virginia.edu>

Dear HotSpot users,

One research question that we have been asked a number of times is -- what 
is the impact of on-chip metal layers on silicon temperature distribution? 
We don't yet have a definitive answer to it. So far, we have done some 
preliminary HotSpot simulations by adding metal layers on top of silicon 
bulk. Due to the tiny cross-sectional area of the metal layer, and therefore 
the large lateral thermal resistance, we found that metal interconnects play 
a negligible role in laterally spreading heat between different silicon 
blocks. However, this is only our preliminary result, and we haven't 
considered factors such as metal self-heating, busses, clock tree, etc. We 
are open to exchanging of ideas on this research problem, and would be glad 
to hear different voices.

(This was also added as a new FAQ entry to the HotSpot website)

Thanks,

-Wei

From David.Ratchkov at lsi.com  Fri Dec 11 18:10:30 2009
From: David.Ratchkov at lsi.com (Ratchkov, David)
Date: Fri, 11 Dec 2009 16:10:30 -0700
Subject: [Hotspot] about on-chip metal layer's impact on silicon
 temperature
In-Reply-To: <web-8648848@guppy.mail.virginia.edu>
References: <web-8648848@guppy.mail.virginia.edu>
Message-ID: <C5277CB418429641BC1498607A9F4805A6779AF4@cosmail01.lsi.com>

Hi Wei,

I'm curious, does this also apply to wide and dense power and/or shielding type of structures?

Regards,
-dave

> -----Original Message-----
> From: hotspot-bounces at cs.virginia.edu [mailto:hotspot-
> bounces at cs.virginia.edu] On Behalf Of Wei Huang
> Sent: Friday, December 11, 2009 3:00 PM
> To: hotspot at cs.virginia.edu
> Subject: [Hotspot] about on-chip metal layer's impact on silicon
> temperature
> 
> Dear HotSpot users,
> 
> One research question that we have been asked a number of times is --
> what
> is the impact of on-chip metal layers on silicon temperature
> distribution?
> We don't yet have a definitive answer to it. So far, we have done some
> preliminary HotSpot simulations by adding metal layers on top of
> silicon
> bulk. Due to the tiny cross-sectional area of the metal layer, and
> therefore
> the large lateral thermal resistance, we found that metal interconnects
> play
> a negligible role in laterally spreading heat between different silicon
> blocks. However, this is only our preliminary result, and we haven't
> considered factors such as metal self-heating, busses, clock tree, etc.
> We
> are open to exchanging of ideas on this research problem, and would be
> glad
> to hear different voices.
> 
> (This was also added as a new FAQ entry to the HotSpot website)
> 
> Thanks,
> 
> -Wei
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot


From wh6p at virginia.edu  Fri Dec 11 19:36:46 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 11 Dec 2009 19:36:46 -0500
Subject: [Hotspot] about on-chip metal layer's impact on silicon
 temperature
In-Reply-To: <C5277CB418429641BC1498607A9F4805A6779AF4@cosmail01.lsi.com>
References: <web-8648848@guppy.mail.virginia.edu>
	<C5277CB418429641BC1498607A9F4805A6779AF4@cosmail01.lsi.com>
Message-ID: <web-8649102@guppy.mail.virginia.edu>

Hi Dave,

If it is only passive metal layers and layers of oxide with poor thermal 
conductivity, my simulations show that it doesn't help much with heat 
spreading even if the layers are dense with metal. I am not sure how this 
would change if we consider self-heating. Comments are welcome...

Regards,

-Wei

On Fri, 11 Dec 2009 16:10:30 -0700
  "Ratchkov, David" <David.Ratchkov at lsi.com> wrote:
> Hi Wei,
> 
> I'm curious, does this also apply to wide and dense power and/or shielding 
>type of structures?
> 
> Regards,
> -dave
> 
>> -----Original Message-----
>> From: hotspot-bounces at cs.virginia.edu [mailto:hotspot-
>> bounces at cs.virginia.edu] On Behalf Of Wei Huang
>> Sent: Friday, December 11, 2009 3:00 PM
>> To: hotspot at cs.virginia.edu
>> Subject: [Hotspot] about on-chip metal layer's impact on silicon
>> temperature
>> 
>> Dear HotSpot users,
>> 
>> One research question that we have been asked a number of times is --
>> what
>> is the impact of on-chip metal layers on silicon temperature
>> distribution?
>> We don't yet have a definitive answer to it. So far, we have done some
>> preliminary HotSpot simulations by adding metal layers on top of
>> silicon
>> bulk. Due to the tiny cross-sectional area of the metal layer, and
>> therefore
>> the large lateral thermal resistance, we found that metal interconnects
>> play
>> a negligible role in laterally spreading heat between different silicon
>> blocks. However, this is only our preliminary result, and we haven't
>> considered factors such as metal self-heating, busses, clock tree, etc.
>> We
>> are open to exchanging of ideas on this research problem, and would be
>> glad
>> to hear different voices.
>> 
>> (This was also added as a new FAQ entry to the HotSpot website)
>> 
>> Thanks,
>> 
>> -Wei
>> _______________________________________________
>> HotSpot mailing list
>> HotSpot at mail.cs.virginia.edu
>> http://www.cs.virginia.edu/mailman/listinfo/hotspot


From stleger at mail.ustc.edu.cn  Mon Dec 14 02:19:36 2009
From: stleger at mail.ustc.edu.cn (stleger at mail.ustc.edu.cn)
Date: Mon, 14 Dec 2009 15:19:36 +0800 (CST)
Subject: [Hotspot] about the ptrace file
Message-ID: <28264796.900241260775176467.JavaMail.coremail@mailweb>

Hi,
I am trying to use the data from wattch to yield a ptrace file, but I find that not every item in HotSpot's gcc.ptrace can be founded in wattch's output statistic file. How can I construct such file? How should I set the every component's power value of ptrace file according to the power statistic of wattch?


Thanks very much if anyone can help.

Best Regards,

St.Leger




From fazal256 at yahoo.com  Fri Dec 18 11:23:13 2009
From: fazal256 at yahoo.com (fazal hameed)
Date: Fri, 18 Dec 2009 08:23:13 -0800 (PST)
Subject: [Hotspot] About thermal warm up
Message-ID: <45023.99663.qm@web54203.mail.re2.yahoo.com>

Dear All,
         I want to invoke hotspot simulator(version 5.0) from a power simulator with the sim-template.c file and want to simulate 100 million instructions (About .1 sec simulation time). 

1. Do I need a thermal warm up for the hotspot or the default configuration will work well with the power traces ?. 

2. In case warm up is necessary, What should be the minimum time for warm up(in second) ?

3. If I want to initialize the steady temp file as as init temp file, how long simulation time(in second) should I run in order to generate a steady temp file ?

regards,
Fazal Hameed



      

From wh6p at virginia.edu  Fri Dec 18 13:34:52 2009
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 18 Dec 2009 13:34:52 -0500
Subject: [Hotspot] About thermal warm up
In-Reply-To: <45023.99663.qm@web54203.mail.re2.yahoo.com>
References: <45023.99663.qm@web54203.mail.re2.yahoo.com>
Message-ID: <web-7812672@neon.mail.virginia.edu>

Hi Fazal,

Steady-state thermal simulation is a one-time simulation, independent of 
time. You don't need to actually simulate seconds or minutes to reach steady 
state. As for thermal and architectural warm-up, there is no golden rule. I 
would suggest use the setting in our ISCA'03 paper as a starting point and 
adjust according your own simulated architecture and thermal configs.

Hope this helps.

-Wei

On Fri, 18 Dec 2009 08:23:13 -0800 (PST)
  fazal hameed <fazal256 at yahoo.com> wrote:
> Dear All,
>         I want to invoke hotspot simulator(version 5.0) from a power 
>simulator with the sim-template.c file and want to simulate 100 million 
>instructions (About .1 sec simulation time). 
> 
> 1. Do I need a thermal warm up for the hotspot or the default 
>configuration will work well with the power traces ?. 
> 
> 2. In case warm up is necessary, What should be the minimum time for warm 
>up(in second) ?
> 
> 3. If I want to initialize the steady temp file as as init temp file, how 
>long simulation time(in second) should I run in order to generate a steady 
>temp file ?
> 
> regards,
>Fazal Hameed
> 
> 
> 
>      
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot


From bernauer at informatik.uni-tuebingen.de  Fri Dec 18 13:42:19 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Fri, 18 Dec 2009 19:42:19 +0100
Subject: [Hotspot] About thermal warm up
In-Reply-To: <45023.99663.qm@web54203.mail.re2.yahoo.com>
References: <45023.99663.qm@web54203.mail.re2.yahoo.com>
Message-ID: <c92980110912181042k28deaa33w4763433bbd8bce93@mail.gmail.com>

On Fri, Dec 18, 2009 at 17:23, fazal hameed <fazal256 at yahoo.com> wrote:

> Dear All,
>         I want to invoke hotspot simulator(version 5.0) from a power
> simulator with the sim-template.c file and want to simulate 100 million
> instructions (About .1 sec simulation time).
>
> 1. Do I need a thermal warm up for the hotspot or the default configuration
> will work well with the power traces ?.
>
> 2. In case warm up is necessary, What should be the minimum time for warm
> up(in second) ?
>
> 3. If I want to initialize the steady temp file as as init temp file, how
> long simulation time(in second) should I run in order to generate a steady
> temp file ?
>

The default configuration usually won't work well for your particular power
traces.
If you omit the '-o' parameter (that is, you don't write transient
temperatures), the steady file will be generated rather instantly.
The steady file will be a good starting point, so you probably don't need
any or only little warmup.

Hope this helps.

Cheers,

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091218/a22b0679/attachment.html 

From lsubrama at andrew.cmu.edu  Sat Dec 19 13:11:10 2009
From: lsubrama at andrew.cmu.edu (Lavanya Subramanian)
Date: Sat, 19 Dec 2009 13:11:10 -0500
Subject: [Hotspot] Temperature variation on a chip
Message-ID: <68c803b44147ef8c78c851e9edc6aff5.squirrel@webmail.andrew.cmu.edu>

Hi,

I replicate the same block several times in a floorplan and have the same
kind of power profiles for all of these replicas.

I observe that the steady state temperatures of the blocks at the centre
is higher than the ones at the corners, which I expect to see. However,
the blocks at the right corner have substantially lower steady state
temperatures.

Is this because of some parameters in the config file? I played around
with r_convec and s_spreader a bit. But, this does not seem to help much.
Could you please help me out with this?

Thanks,
Lavanya.


From fsamie at gmail.com  Sun Dec 20 00:38:36 2009
From: fsamie at gmail.com (farzad samie)
Date: Sun, 20 Dec 2009 09:08:36 +0330
Subject: [Hotspot] linking hotspot with Wattch or simplescalar
Message-ID: <ffd054460912192138g20ee7ab2m9585b3c7ad14537a@mail.gmail.com>

Dear all,
when I was reading FAQs, I find that many of users want to link Hotspot with
some performance simulators like Wattch and SimpleScalar and I know that
there is an interface file (sim-template.c) to do it. But when I was
investigating, I didn't find a precise and complete solution to do it.
specifically there are some question that I didn't find their reply:
1) whether we should include first lines of sim-template.c in WATTCH (e.g.
sim-outorder.c) and add other lines in sim-main, sim-init, sim-exit methods?
2)Is there anyone that do it completely to help us?
 many thanks for this kindly mailing list and quick replies.

Farzad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091220/e55e0dbe/attachment.html 

From bernauer at informatik.uni-tuebingen.de  Sun Dec 20 03:37:53 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Sun, 20 Dec 2009 09:37:53 +0100
Subject: [Hotspot] Temperature variation on a chip
In-Reply-To: <68c803b44147ef8c78c851e9edc6aff5.squirrel@webmail.andrew.cmu.edu>
References: <68c803b44147ef8c78c851e9edc6aff5.squirrel@webmail.andrew.cmu.edu>
Message-ID: <c92980110912200037k6b0d76e4p6230fd19633078b6@mail.gmail.com>

On Sat, Dec 19, 2009 at 19:11, Lavanya Subramanian
<lsubrama at andrew.cmu.edu>wrote:

> Hi,
>
> I replicate the same block several times in a floorplan and have the same
> kind of power profiles for all of these replicas.
>
> I observe that the steady state temperatures of the blocks at the centre
> is higher than the ones at the corners, which I expect to see. However,
> the blocks at the right corner have substantially lower steady state
> temperatures.
>

I could imagine that if your chip is not exactly of the same size as the
heat spreader, one corner of your chip will have a slightly larger cooling
area. Can you check this?

Cheers,

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091220/d5c7f558/attachment.html 

From lsubrama at andrew.cmu.edu  Sun Dec 20 23:34:14 2009
From: lsubrama at andrew.cmu.edu (Lavanya Subramanian)
Date: Sun, 20 Dec 2009 23:34:14 -0500
Subject: [Hotspot] Temperature variation on a chip
Message-ID: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>

Hi,

I made the chip the same size as the spreader and the same kind of spatial
steady state temperature difference exists.

Thanks,
Lavanya.

On Sun, December 20, 2009 3:37 am, Andreas Bernauer wrote:
> On Sat, Dec 19, 2009 at 19:11, Lavanya Subramanian
> <lsubrama at andrew.cmu.edu>wrote:
>
>
>> Hi,
>>
>>
>> I replicate the same block several times in a floorplan and have the
>> same kind of power profiles for all of these replicas.
>>
>> I observe that the steady state temperatures of the blocks at the
>> centre is higher than the ones at the corners, which I expect to see.
>> However,
>> the blocks at the right corner have substantially lower steady state
>> temperatures.
>>
>
> I could imagine that if your chip is not exactly of the same size as the
> heat spreader, one corner of your chip will have a slightly larger cooling
>  area. Can you check this?
>
> Cheers,
>
>
> --
> Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
> WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
> http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
>
>



From bernauer at informatik.uni-tuebingen.de  Mon Dec 21 03:08:06 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Mon, 21 Dec 2009 09:08:06 +0100
Subject: [Hotspot] Temperature variation on a chip
In-Reply-To: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>
References: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>
Message-ID: <c92980110912210008s18b8a22bq4fe5f169a22ad671@mail.gmail.com>

On Mon, Dec 21, 2009 at 05:34, Lavanya Subramanian
<lsubrama at andrew.cmu.edu>wrote:

> I made the chip the same size as the spreader and the same kind of spatial
> steady state temperature difference exists.
>

Hi Lavanya,

I repeated the simulation according to your description, but I don't see the
problem you are describing.

Attached you find the files I used for the simulation. The command to
generate the steady file was
hotspot -c hotspot.config -f flp36 -p power36 -steady_file steady36

The floorplan describes 6x6 blocks named Unit0..Unit35.

In the steady file, the unit temperatures are symmetrical, as expected.

I can't see what is going wrong at your side.

Cheers,

Andreas.

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/0ddd2083/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hotspot.config
Type: application/octet-stream
Size: 4517 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/0ddd2083/attachment.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: flp36
Type: application/octet-stream
Size: 1537 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/0ddd2083/attachment-0001.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: power36
Type: application/octet-stream
Size: 350 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/0ddd2083/attachment-0002.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: steady36
Type: application/octet-stream
Size: 2733 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/0ddd2083/attachment-0003.obj 

From bernauer at informatik.uni-tuebingen.de  Mon Dec 21 03:38:54 2009
From: bernauer at informatik.uni-tuebingen.de (Andreas Bernauer)
Date: Mon, 21 Dec 2009 09:38:54 +0100
Subject: [Hotspot] Temperature variation on a chip
In-Reply-To: <c92980110912210008s18b8a22bq4fe5f169a22ad671@mail.gmail.com>
References: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>
	<c92980110912210008s18b8a22bq4fe5f169a22ad671@mail.gmail.com>
Message-ID: <c92980110912210038h4251f274sa894e6636843fd17@mail.gmail.com>

On Mon, Dec 21, 2009 at 09:08, Andreas Bernauer <
bernauer at informatik.uni-tuebingen.de> wrote:

> I repeated the simulation according to your description, but I don't see
> the problem you are describing.
>
> Attached you find the files I used for the simulation. The command to
> generate the steady file was
> hotspot -c hotspot.config -f flp36 -p power36 -steady_file steady36
>
> The floorplan describes 6x6 blocks named Unit0..Unit35.
>

line 93ff in temperature_block.c says:

    double w_chip = get_total_width (flp);    /* x-axis    */
    double l_chip = get_total_height (flp);    /* y-axis    */

    /* sanity check on floorplan sizes    */
    if (w_chip > s_sink || l_chip > s_sink ||
        w_chip > s_spreader || l_chip > s_spreader) {
        print_flp(flp);
        print_flp_fig(flp);
        fatal("inordinate floorplan size!\n");
    }


The sanity check sometimes fails, although everything is OK.
This happened to me with the previously mentioned 6x6 block example.
I had to slightly enlarge the heat spreader side length to 0.0300000000001
to bypass the sanity check.

If the head spreader side is left as 0.03, the sanity check fails on
conditions like 0.030000000000000002 >  0.029999999999999999     (w_chip >
s_spreader).
This is because of floating point rounding errors in get_total_width (or
get_total_height).

A common fix is to add a small epsilon in the comparison of floating points,
like this

--- temperature_block.c.orig    2009-12-21 09:31:08.000000000 +0100
+++ temperature_block.c    2009-12-21 09:29:43.000000000 +0100
@@ -94,8 +94,8 @@
     double l_chip = get_total_height (flp);    /* y-axis    */

     /* sanity check on floorplan sizes    */
-    if (w_chip > s_sink || l_chip > s_sink ||
-        w_chip > s_spreader || l_chip > s_spreader) {
+    if (w_chip > s_sink + __DBL_EPSILON__ || l_chip > s_sink +
__DBL_EPSILON__ ||
+        w_chip > s_spreader + __DBL_EPSILON__  || l_chip > s_spreader +
__DBL_EPSILON__ ) {
         print_flp(flp);
         print_flp_fig(flp);
         fatal("inordinate floorplan size!\n");

where __DBL_EPSILON__ is predefined (in the gcc, at least) to be the
smallest unit representable in a double (2.2204460492503131e-16 on my
system).

-- 
Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20091221/ac3134d8/attachment.html 

From lsubrama at andrew.cmu.edu  Mon Dec 28 10:37:13 2009
From: lsubrama at andrew.cmu.edu (Lavanya Subramanian)
Date: Mon, 28 Dec 2009 10:37:13 -0500
Subject: [Hotspot] Temperature variation on a chip
In-Reply-To: <c92980110912210038h4251f274sa894e6636843fd17@mail.gmail.com>
References: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>
	<c92980110912210008s18b8a22bq4fe5f169a22ad671@mail.gmail.com>
	<c92980110912210038h4251f274sa894e6636843fd17@mail.gmail.com>
Message-ID: <7d4f3103c9aaeaaf2606e3a87e5b4b23.squirrel@webmail.andrew.cmu.edu>

Hi Andreas,

Making the spreader exactly the same size at the die helps.
However, I figured out that the one different thing in my floorplan is
that I leave some parts of the floorplan empty, to account for
interconnect.
If I remove these, I don't see the huge temperature variations I was
seeing earlier.
Is this behavior expected?

Thanks a lot for your detailed responses.

Lavanya.

On Mon, December 21, 2009 3:38 am, Andreas Bernauer wrote:
> On Mon, Dec 21, 2009 at 09:08, Andreas Bernauer <
> bernauer at informatik.uni-tuebingen.de> wrote:
>
>> I repeated the simulation according to your description, but I don't
>> see the problem you are describing.
>>
>> Attached you find the files I used for the simulation. The command to
>> generate the steady file was hotspot -c hotspot.config -f flp36 -p
>> power36 -steady_file steady36
>>
>> The floorplan describes 6x6 blocks named Unit0..Unit35.
>>
>>
>
> line 93ff in temperature_block.c says:
>
> double w_chip = get_total_width (flp);    /* x-axis    */ double l_chip =
> get_total_height (flp);    /* y-axis    */
>
> /* sanity check on floorplan sizes    */
> if (w_chip > s_sink || l_chip > s_sink || w_chip > s_spreader || l_chip >
> s_spreader) { print_flp(flp); print_flp_fig(flp); fatal("inordinate
> floorplan size!\n"); }
>
>
>
> The sanity check sometimes fails, although everything is OK.
> This happened to me with the previously mentioned 6x6 block example.
> I had to slightly enlarge the heat spreader side length to 0.0300000000001
>  to bypass the sanity check.
>
> If the head spreader side is left as 0.03, the sanity check fails on
> conditions like 0.030000000000000002 >  0.029999999999999999     (w_chip >
>  s_spreader). This is because of floating point rounding errors in
> get_total_width (or get_total_height).
>
> A common fix is to add a small epsilon in the comparison of floating
> points, like this
>
> --- temperature_block.c.orig    2009-12-21 09:31:08.000000000 +0100
> +++ temperature_block.c    2009-12-21 09:29:43.000000000 +0100
> @@ -94,8 +94,8 @@
> double l_chip = get_total_height (flp);    /* y-axis    */
>
> /* sanity check on floorplan sizes    */
> -    if (w_chip > s_sink || l_chip > s_sink ||
> -        w_chip > s_spreader || l_chip > s_spreader) {
> +    if (w_chip > s_sink + __DBL_EPSILON__ || l_chip > s_sink +
> __DBL_EPSILON__ ||
> +        w_chip > s_spreader + __DBL_EPSILON__  || l_chip > s_spreader +
> __DBL_EPSILON__ ) {
> print_flp(flp); print_flp_fig(flp); fatal("inordinate floorplan size!\n");
>
> where __DBL_EPSILON__ is predefined (in the gcc, at least) to be the
> smallest unit representable in a double (2.2204460492503131e-16 on my
> system).
>
> --
> Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
> WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
> http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>



From ks4kk at virginia.edu  Mon Dec 28 19:28:13 2009
From: ks4kk at virginia.edu (Karthik Sankaranarayanan)
Date: Mon, 28 Dec 2009 16:28:13 -0800
Subject: [Hotspot] Temperature variation on a chip
In-Reply-To: <7d4f3103c9aaeaaf2606e3a87e5b4b23.squirrel@webmail.andrew.cmu.edu>
References: <6a36e3f4be56515502cc13a0c649b1ef.squirrel@webmail.andrew.cmu.edu>
	<c92980110912210008s18b8a22bq4fe5f169a22ad671@mail.gmail.com>
	<c92980110912210038h4251f274sa894e6636843fd17@mail.gmail.com>
	<7d4f3103c9aaeaaf2606e3a87e5b4b23.squirrel@webmail.andrew.cmu.edu>
Message-ID: <eae341010912281628h6e4db76ay7551fee4e444e1c@mail.gmail.com>

Hi Lavanya,

HotSpot assumes that the floorplan is fully specified (i.e., does not have
any empty spaces). If you want certain blocks to be dissipating zero power,
please do specify them in the floorplan file and assign zero power to them.
If you are using the grid-based model, you could also uncomment the
"test_bgmap" function call in temperature_grid.c to do a sanity check on
your floorplan.

Hope this helps.

Thanks,
-karthik

On Mon, Dec 28, 2009 at 7:37 AM, Lavanya Subramanian <
lsubrama at andrew.cmu.edu> wrote:

> Hi Andreas,
>
> Making the spreader exactly the same size at the die helps.
> However, I figured out that the one different thing in my floorplan is
> that I leave some parts of the floorplan empty, to account for
> interconnect.
> If I remove these, I don't see the huge temperature variations I was
> seeing earlier.
> Is this behavior expected?
>
> Thanks a lot for your detailed responses.
>
> Lavanya.
>
> On Mon, December 21, 2009 3:38 am, Andreas Bernauer wrote:
> > On Mon, Dec 21, 2009 at 09:08, Andreas Bernauer <
> > bernauer at informatik.uni-tuebingen.de> wrote:
> >
> >> I repeated the simulation according to your description, but I don't
> >> see the problem you are describing.
> >>
> >> Attached you find the files I used for the simulation. The command to
> >> generate the steady file was hotspot -c hotspot.config -f flp36 -p
> >> power36 -steady_file steady36
> >>
> >> The floorplan describes 6x6 blocks named Unit0..Unit35.
> >>
> >>
> >
> > line 93ff in temperature_block.c says:
> >
> > double w_chip = get_total_width (flp);    /* x-axis    */ double l_chip =
> > get_total_height (flp);    /* y-axis    */
> >
> > /* sanity check on floorplan sizes    */
> > if (w_chip > s_sink || l_chip > s_sink || w_chip > s_spreader || l_chip >
> > s_spreader) { print_flp(flp); print_flp_fig(flp); fatal("inordinate
> > floorplan size!\n"); }
> >
> >
> >
> > The sanity check sometimes fails, although everything is OK.
> > This happened to me with the previously mentioned 6x6 block example.
> > I had to slightly enlarge the heat spreader side length to
> 0.0300000000001
> >  to bypass the sanity check.
> >
> > If the head spreader side is left as 0.03, the sanity check fails on
> > conditions like 0.030000000000000002 >  0.029999999999999999     (w_chip
> >
> >  s_spreader). This is because of floating point rounding errors in
> > get_total_width (or get_total_height).
> >
> > A common fix is to add a small epsilon in the comparison of floating
> > points, like this
> >
> > --- temperature_block.c.orig    2009-12-21 09:31:08.000000000 +0100
> > +++ temperature_block.c    2009-12-21 09:29:43.000000000 +0100
> > @@ -94,8 +94,8 @@
> > double l_chip = get_total_height (flp);    /* y-axis    */
> >
> > /* sanity check on floorplan sizes    */
> > -    if (w_chip > s_sink || l_chip > s_sink ||
> > -        w_chip > s_spreader || l_chip > s_spreader) {
> > +    if (w_chip > s_sink + __DBL_EPSILON__ || l_chip > s_sink +
> > __DBL_EPSILON__ ||
> > +        w_chip > s_spreader + __DBL_EPSILON__  || l_chip > s_spreader +
> > __DBL_EPSILON__ ) {
> > print_flp(flp); print_flp_fig(flp); fatal("inordinate floorplan
> size!\n");
> >
> > where __DBL_EPSILON__ is predefined (in the gcc, at least) to be the
> > smallest unit representable in a double (2.2204460492503131e-16 on my
> > system).
> >
> > --
> > Andreas Bernauer <bernauer at informatik.uni-tuebingen.de>
> > WSI/TI, Sand 13, B202, 72076 T?bingen, +49 70 71 29 75 940
> > http://www.ti.uni-tuebingen.de/Andreas_Bernauer.227.0.html
> > _______________________________________________
> > 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/20091228/a36c2476/attachment.html