From wowlyy at hotmail.com  Thu Jun  2 15:25:14 2011
From: wowlyy at hotmail.com (Yangyang Li)
Date: Thu, 2 Jun 2011 13:25:14 -0600
Subject: [Hotspot] the output file
Message-ID: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>


Hi,
I have a question about the out put(.grid.steay) file of HOTSPOT. We're not sure how HotSpot output the grid temperature in terms of ordering.
Here is an example we used for .ptrace file:

Unit1    Unit2    Unit3    Unit4    Unit5    Unit6    Unit7    Unit8    Unit9    Unit10    Unit11    Unit12    Unit13    Unit14    Unit15    Unit16
1    12     5    3    1    5    5    1    35    50    29    80    100    50    90    100    

and we found the output .grid.steady file is

0    455.46
1    437.07
2    457.67
3    460.06
4    419.49
5    431.49
6    424.53
7    445.55
8    393.29
9    399.42
10    399.84
11    396.77
12    386.31
13    393.45
14    390.46
15    388.75

It seems that the first temperature corresponds to the Unit16, not Unit 1(located at (0,0) ). Can you please explain the order of the output file?

Thanks,

Yangyang

 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110602/c9cf92b7/attachment.html 

From wowlyy at hotmail.com  Thu Jun  2 15:25:14 2011
From: wowlyy at hotmail.com (Yangyang Li)
Date: Thu, 2 Jun 2011 13:25:14 -0600
Subject: [Hotspot] the output file
Message-ID: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>


Hi,
I have a question about the out put(.grid.steay) file of HOTSPOT. We're not sure how HotSpot output the grid temperature in terms of ordering.
Here is an example we used for .ptrace file:

Unit1    Unit2    Unit3    Unit4    Unit5    Unit6    Unit7    Unit8    Unit9    Unit10    Unit11    Unit12    Unit13    Unit14    Unit15    Unit16
1    12     5    3    1    5    5    1    35    50    29    80    100    50    90    100    

and we found the output .grid.steady file is

0    455.46
1    437.07
2    457.67
3    460.06
4    419.49
5    431.49
6    424.53
7    445.55
8    393.29
9    399.42
10    399.84
11    396.77
12    386.31
13    393.45
14    390.46
15    388.75

It seems that the first temperature corresponds to the Unit16, not Unit 1(located at (0,0) ). Can you please explain the order of the output file?

Thanks,

Yangyang

 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110602/c9cf92b7/attachment-0001.html 

From wh6p at virginia.edu  Thu Jun  2 15:44:19 2011
From: wh6p at virginia.edu (Wei Huang)
Date: Thu, 2 Jun 2011 14:44:19 -0500
Subject: [Hotspot] the output file
In-Reply-To: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>
References: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>
Message-ID: <BANLkTi=JxwEf7qDvx0nGx1AT7c6cpmEwRQ@mail.gmail.com>

Hi,

If I remember correctly, the grid temperature output file is ordered from
top to bottom and left to right. This is mainly for plotting a thermal map
and accidentally is not in the same order as the floorplan file. A quick way
to verify this is to generate a thermal map using one of the perl scripts
coming together within the release (detailed instructions are in the HOW-TO
page)

-Wei

On Thu, Jun 2, 2011 at 2:25 PM, Yangyang Li <wowlyy at hotmail.com> wrote:

>  Hi,
> I have a question about the out put(.grid.steay) file of HOTSPOT. We're not
> sure how HotSpot output the grid temperature in terms of ordering.
> Here is an example we used for .ptrace file:
>
> Unit1    Unit2    Unit3    Unit4    Unit5    Unit6    Unit7    Unit8
> Unit9    Unit10    Unit11    Unit12    Unit13    Unit14    Unit15    Unit16
> 1    12     5    3    1    5    5    1    35    50    29    80    100
> 50    90    100
>
> and we found the output .grid.steady file is
>
> 0    455.46
> 1    437.07
> 2    457.67
> 3    460.06
> 4    419.49
> 5    431.49
> 6    424.53
> 7    445.55
> 8    393.29
> 9    399.42
> 10    399.84
> 11    396.77
> 12    386.31
> 13    393.45
> 14    390.46
> 15    388.75
>
> It seems that the first temperature corresponds to the Unit16, not Unit
> 1(located at (0,0) ). Can you please explain the order of the output file?
>
> Thanks,
>
> Yangyang
>
>
> _______________________________________________
> 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/20110602/5f6fafa0/attachment.html 

From agayathribits at gmail.com  Thu Jun  2 15:52:06 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Fri, 3 Jun 2011 01:22:06 +0530
Subject: [Hotspot] Technology scaling
Message-ID: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>

Hi

I want to know  what technology does Hotspot 5.0 models (by default). Is the
technology parameter configurable? I want to scale across technology and see
how temperature varies !

Can you please let me know where can I find these information in HotSpot
5.0?

Thanks,
Gayathri
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110603/a4f5c95f/attachment.html 

From agayathribits at gmail.com  Thu Jun  2 15:52:06 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Fri, 3 Jun 2011 01:22:06 +0530
Subject: [Hotspot] Technology scaling
Message-ID: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>

Hi

I want to know  what technology does Hotspot 5.0 models (by default). Is the
technology parameter configurable? I want to scale across technology and see
how temperature varies !

Can you please let me know where can I find these information in HotSpot
5.0?

Thanks,
Gayathri
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110603/a4f5c95f/attachment-0001.html 

From agayathribits at gmail.com  Thu Jun  2 15:50:09 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Fri, 3 Jun 2011 01:20:09 +0530
Subject: [Hotspot] Technology Scaling
Message-ID: <BANLkTik5hGq6gJrwFKjOf9qcs9M_S5M5tg@mail.gmail.com>

Hi

I want to what technology does Hotspot 5.0 models (by default). Is the
technology parameter configurable? I want to scale across technology and see
how temperature varies !

Can you please let me know where can I found these in HotSpot 5.0?

Thanks,
Gayathri
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110603/d87d9003/attachment.html 

From agayathribits at gmail.com  Thu Jun  2 15:50:09 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Fri, 3 Jun 2011 01:20:09 +0530
Subject: [Hotspot] Technology Scaling
Message-ID: <BANLkTik5hGq6gJrwFKjOf9qcs9M_S5M5tg@mail.gmail.com>

Hi

I want to what technology does Hotspot 5.0 models (by default). Is the
technology parameter configurable? I want to scale across technology and see
how temperature varies !

Can you please let me know where can I found these in HotSpot 5.0?

Thanks,
Gayathri
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110603/d87d9003/attachment-0001.html 

From skadron at cs.virginia.edu  Thu Jun  2 16:51:02 2011
From: skadron at cs.virginia.edu (Kevin Skadron)
Date: Thu, 02 Jun 2011 16:51:02 -0400
Subject: [Hotspot] Technology scaling
In-Reply-To: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>
References: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>
Message-ID: <4DE7F7B6.5030701@cs.virginia.edu>

Hi Gayathri,

HotSpot is technology independent.  It takes as inputs a floorplan and 
power values for each block in the floorplan, as well as the geometric 
and material properties of the cooling solution (size of heat spreader, 
etc.).  The floorplan and power values need to be provided by some other 
source.

This way HotSpot is portable across a variety of infrastructures.

--Kevin

On 6/2/11 3:52 PM, Gayathri Ananth wrote:
> Hi
>
> I want to know  what technology does Hotspot 5.0 models (by default). Is
> the technology parameter configurable? I want to scale across technology
> and see how temperature varies !
>
> Can you please let me know where can I find these information in HotSpot
> 5.0?
>
> Thanks,
> Gayathri
>
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot

-- 
Kevin Skadron
Dept. of Computer Science, University of Virginia
http://www.cs.virginia.edu/~skadron

From mamsadegh at googoolia.com  Thu Jun  2 17:04:06 2011
From: mamsadegh at googoolia.com (MohammadSadegh Sadri)
Date: Thu, 02 Jun 2011 17:04:06 -0400
Subject: [Hotspot] Technology scaling
In-Reply-To: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>
References: <BANLkTikpXMMP3MwOVrAVYQOqhunuNzn=jw@mail.gmail.com>
Message-ID: <4DE7FAC6.60502@googoolia.com>

On 06/02/2011 03:52 PM, Gayathri Ananth wrote:
> Hi
>
> I want to know  what technology does Hotspot 5.0 models (by default). 
> Is the technology parameter configurable? I want to scale across 
> technology and see how temperature varies !
>
> Can you please let me know where can I find these information in 
> HotSpot 5.0?
>
> Thanks,
> Gayathri
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
Gayathri,
Can't we scale total chip area instead of technology, keeping the 
transistor count the same between tests?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110602/b426b2a8/attachment-0001.html 

From mamsadegh at googoolia.com  Thu Jun  2 15:37:19 2011
From: mamsadegh at googoolia.com (MohammadSadegh Sadri)
Date: Thu, 02 Jun 2011 15:37:19 -0400
Subject: [Hotspot] the output file
In-Reply-To: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>
References: <SNT123-W2468498C89655B81A4A05DCC7C0@phx.gbl>
Message-ID: <4DE7E66F.4080209@googoolia.com>

On 06/02/2011 03:25 PM, Yangyang Li wrote:
> Hi,
> I have a question about the out put(.grid.steay) file of HOTSPOT. 
> We're not sure how HotSpot output the grid temperature in terms of 
> ordering.
> Here is an example we used for .ptrace file:
>
> Unit1    Unit2    Unit3    Unit4    Unit5    Unit6    Unit7    
> Unit8    Unit9    Unit10    Unit11    Unit12    Unit13    Unit14    
> Unit15    Unit16
> 1    12     5    3    1    5    5    1    35    50    29    80    
> 100    50    90    100
>
> and we found the output .grid.steady file is
>
> 0    455.46
> 1    437.07
> 2    457.67
> 3    460.06
> 4    419.49
> 5    431.49
> 6    424.53
> 7    445.55
> 8    393.29
> 9    399.42
> 10    399.84
> 11    396.77
> 12    386.31
> 13    393.45
> 14    390.46
> 15    388.75
>
> It seems that the first temperature corresponds to the Unit16, not 
> Unit 1(located at (0,0) ). Can you please explain the order of the 
> output file?
>
> Thanks,
>
> Yangyang
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
Dear Yangyang,
Sure the developers of HotSpot will answer you , just I was curious if 
this is the output of HotSpot in steady state with Grid Mode enabled?
What is the size of your grid? I mean for example 32*32 points? what? or 
this is the output of "block mode" steady file? ( because the output of 
grid steady state is a very large file containing numbers for each grid 
point). However the output of normal steady file is what you have shown 
here.
thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110602/2c3994e0/attachment.html 

From nima.aghaee.ghaleshahi at liu.se  Sat Jun  4 13:08:37 2011
From: nima.aghaee.ghaleshahi at liu.se (Nima Aghaee)
Date: Sat, 4 Jun 2011 19:08:37 +0200
Subject: [Hotspot] Bug report
In-Reply-To: <33AD87C79048414ABE197DC821432F0B6824540CB9@MAILBOX2.ad.liu.se>
References: <33AD87C79048414ABE197DC821432F0B6824540CB5@MAILBOX2.ad.liu.se>
	<33AD87C79048414ABE197DC821432F0B6824540CB9@MAILBOX2.ad.liu.se>
Message-ID: <5791BC2122C44B95839B0C7DF881FD96@carafe.ida.liu.se>

To my knowledge, the following describes two closely related bugs.


Related Files:         temperature_block.c, temperature_grid.c, and RCutil.c
Related Versions:  3.1, 4, and 5
Main concern:        Wrong solution to the composed differential equation
Primary effect:       Wrong step size is accumulated, and the reported temperature does not correspond to the expected time, but a shorter time

In the following I just address temperature_block.c file. A similar argument exists for temperature_grid.c file.

Description:
line 468 in temperature_block.c version 5
------------------------------------------------------------------------------------------------
/* Obtain temp at time (t+time_elapsed). 
* Instead of getting the temperature at t+time_elapsed directly, we do it 
* in multiple steps with the correct step size at each time 
* provided by rk4
*/
for (t = 0, new_h = MIN_STEP; t + new_h <= time_elapsed; t+=h) {
    h = new_h;
    new_h = rk4(model, temp, model->t_vector, model->n_nodes, h, 
    /* the slope function callback is typecast accordingly */
                temp, (slope_fn_ptr) slope_fn_block);
    #if VERBOSE > 1
    i++;
    #endif
}
------------------------------------------------------------------------------------------------
t += h, accumulates the initial step size for the next RK4 call, instead of the actual step size.

line 310 in RCutil.c version 5. High level representation is given here:
------------------------------------------------------------------------------------------------
/* try until accuracy is achieved    */
do {

    COMPUTATION with h
     
      PRECISION EVALUATION

      STEP SIZE ADJUSTMENT gives new_h

    } BREAK IF DESIRED PRECISION IS ACHIEVED

------------------------------------------------------------------------------------------------
when this function returns new_h, it is not the step size which actually used in computation (h).

Suggested fix:
define h as a pointer in temperature_block.c,
pass the pointer to h to the rk4 function and use it instead of local h,
use h in temperature_block.c to keep track of time (instead of new_h),

** IMPORTANT:
A second bug exists regarding this issue, which I briefly mention here. I hope the above part is clear enough to be understood easily. Then the second bug will be easily understood.

2nd Bug: 
For the remainder (line 482 temperature_block.c version 5), the adaptive step size could not and should not be used, since we want to have the exact temperature at exact time (having the exact temperature at an unknown time is not helpful).

Suggested fix:
Add a new rk4 function identical to the present rk4 but remove the do loop, precision evaluation, and step size adjustment. It will look like the following as opposed to the above pseudo-code.
------------------------------------------------------------------------------------------------
    COMPUTATION with h     
 ------------------------------------------------------------------------------------------------
This new function should be used for remainder, instead of the present rk4.

Diagnostics:
1. Print out the differential equation (dT + CT = inv_A * Power) coefficients in compute_temp_block function (line 449 in temperature_block.c version 5), before starting to solve (line 468)
2. Use Matlab or a more precise commercial tool to compute the solution
3. Compare with HotSpot results
4. Fix the bug as described above
5. The difference between HotSpot results and the other solver will be substantially reduced


Please keep me informed.

Thanks,
Nima Aghaee
PhD student 
ESLAB, Linkoping University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110604/61e56a2f/attachment.html 

From ks4kk at virginia.edu  Mon Jun  6 11:31:00 2011
From: ks4kk at virginia.edu (Karthik Sankaranarayanan)
Date: Mon, 6 Jun 2011 08:31:00 -0700
Subject: [Hotspot] Bug report
In-Reply-To: <5791BC2122C44B95839B0C7DF881FD96@carafe.ida.liu.se>
References: <33AD87C79048414ABE197DC821432F0B6824540CB5@MAILBOX2.ad.liu.se>
	<33AD87C79048414ABE197DC821432F0B6824540CB9@MAILBOX2.ad.liu.se>
	<5791BC2122C44B95839B0C7DF881FD96@carafe.ida.liu.se>
Message-ID: <BANLkTinOmHwBpeY1V_Ao41sq--V2J65PKg@mail.gmail.com>

Hi Nima,

Good catch! Thank you very much for letting us know. This is indeed a bug
that needs a fix. We have incorporated the fix in the lines of your
suggestion and will be updating the HotSpot release soon (with due credits
to you, of course). In practice however, it didn't affect the typical use
cases of HotSpot because it occurs only when the adaptive step sizing moves
from a larger step to a smaller one. Since we start very conservatively at
MIN_STEP and expand slowly, this bug didn't usually get activated and hence
escaped our eyes. Thank you for the find!

Apart from this bug, please note that the precision target for HotSpot is
0.01 C. In case you need higher precision, please set the RK4_PRECISION
macro in RCutil.c to the desired value.

Thanks,
-karthik



On Sat, Jun 4, 2011 at 10:08 AM, Nima Aghaee
<nima.aghaee.ghaleshahi at liu.se>wrote:

>    To my knowledge, the following describes two closely related bugs.
>
>
> Related Files:         temperature_block.c, temperature_grid.c, and
> RCutil.c
> Related Versions:  3.1, 4, and 5
> Main concern:        Wrong solution to the composed differential equation
> Primary effect:       Wrong step size is accumulated, and the reported
> temperature does not correspond to the expected time, but a shorter time
>
> In the following I just address temperature_block.c file. A similar
> argument exists for temperature_grid.c file.
>
> Description:
> line 468 in temperature_block.c version 5
>
> ------------------------------------------------------------------------------------------------
> /* Obtain temp at time (t+time_elapsed).
> * Instead of getting the temperature at t+time_elapsed directly, we do it
> * in multiple steps with the correct step size at each time
> * provided by rk4
> */
> for (t = 0, new_h = MIN_STEP; t + new_h <= time_elapsed; t+=h) {
>     h = new_h;
>     new_h = rk4(model, temp, model->t_vector, model->n_nodes, h,
>     /* the slope function callback is typecast accordingly */
>                 temp, (slope_fn_ptr) slope_fn_block);
>     #if VERBOSE > 1
>     i++;
>     #endif
> }
>
> ------------------------------------------------------------------------------------------------
> t += h, accumulates the initial step size for the next RK4 call, instead of
> the actual step size.
>
> line 310 in RCutil.c version 5. High level representation is given here:
>
> ------------------------------------------------------------------------------------------------
> /* try until accuracy is achieved    */
> do {
>
>     COMPUTATION with h
>
>       PRECISION EVALUATION
>
>       STEP SIZE ADJUSTMENT gives new_h
>
>     } BREAK IF DESIRED PRECISION IS ACHIEVED
>
>
> ------------------------------------------------------------------------------------------------
> when this function returns new_h, it is not the step size which actually
> used in computation (h).
>
> Suggested fix:
> define h as a pointer in temperature_block.c,
> pass the pointer to h to the rk4 function and use it instead of local h,
> use h in temperature_block.c to keep track of time (instead of new_h),
>
> ** IMPORTANT:
> A second bug exists regarding this issue, which I briefly mention here. I
> hope the above part is clear enough to be understood easily. Then the second
> bug will be easily understood.
>
> 2nd Bug:
> For the remainder (line 482 temperature_block.c version 5), the adaptive
> step size could not and should not be used, since we want to have the exact
> temperature at exact time (having the exact temperature at an unknown time
> is not helpful).
>
> Suggested fix:
> Add a new rk4 function identical to the present rk4 but remove the do loop,
> precision evaluation, and step size adjustment. It will look like the
> following as opposed to the above pseudo-code.
>
> ------------------------------------------------------------------------------------------------
>     COMPUTATION with h
>
> ------------------------------------------------------------------------------------------------
> This new function should be used for remainder, instead of the present rk4.
>
> Diagnostics:
> 1. Print out the differential equation (dT + CT = inv_A * Power)
> coefficients in compute_temp_block function (line 449 in temperature_block.c
> version 5), before starting to solve (line 468)
> 2. Use Matlab or a more precise commercial tool to compute the solution
> 3. Compare with HotSpot results
> 4. Fix the bug as described above
> 5. The difference between HotSpot results and the other solver will be
> substantially reduced
>
>
> Please keep me informed.
>
> Thanks,
> Nima Aghaee
> PhD student
> ESLAB, Linkoping University
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110606/b01a0b77/attachment.html 

From ksiop at microlab.ntua.gr  Mon Jun  6 13:24:39 2011
From: ksiop at microlab.ntua.gr (Kostas Siozios)
Date: Mon, 6 Jun 2011 20:24:39 +0300
Subject: [Hotspot] Model 3D MPSoCs
Message-ID: <006401cc246e$9e128c80$da37a580$@microlab.ntua.gr>

Hi all,

 

I am trying to model thermal diffusion for a 3D MPSoC. 

I found in previous posts, as well as the manual, that Hotspot provides
support for 3D architectures. 

However, I am wondering if it is possible to quantify the impact on thermal
stress by placing TSVs at different spatial locations over each layer?

 

Is it possible to evaluate this with Hotspot?

 

Regards,

Kostas

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110606/0aa1a296/attachment.html 

From agayathribits at gmail.com  Mon Jun 13 22:26:49 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Tue, 14 Jun 2011 10:56:49 +0530
Subject: [Hotspot] Hotspot 5.0 : Lateral heat flow Query
In-Reply-To: <BANLkTinyLKg7Q=dnXGPH2N+CZ2Sn_YKCjA@mail.gmail.com>
References: <BANLkTinyLKg7Q=dnXGPH2N+CZ2Sn_YKCjA@mail.gmail.com>
Message-ID: <BANLkTimOCy34UsVJ2R67bHbOaor2B7px-A@mail.gmail.com>

Hi All,

I have performed some experiments for understanding Lateral heat flow
through silicon using Hotspot 5.0. But from the data I obtained, *I dont see
any lateral heat flow through Silicon.*
*
*
*The following are the experiments I performed:*
*
*
1. I have a 20mm by 20mm chip . In the floorplan I have four cores. I
incrementally ( insteps of 1mm)  increase the distance between the two power
hungry cores (50W for a 8mm by 10mm core).  I have used the *default
configuraton file provided in Hotspot 5.o version. * I am using grid model
and final steady temperature varies as follows

Distance                            Temp
(b/n cores)
4mm                           338.49      338.49
3mm                           338.67      338.69
2mm                           338.87      338.90
1mm                           339.08      339.17
0mm                           339.33      339.33

The above results show a very small change in the temperature, but according
to temperature aware floorplanning concept, the swing should be more.

Q: why I am not seeing such a swing? Also except for  LCF file* is there any
place where I can explicitly give a YES for LATERAL HEAT FLOW.*
*
*
*I also performed other experiments like just heating only one core and
seeing the effects on remaining three cores. But I did not see any lateral
heat conduction through "SILICON".*
*
*
*Also I moved the core diagonally from the edge of the chip until other
edge. The difference in temperature I obtain is very minimal. *
*
*
*Can you please help me in this regard and point me if I am wrong anywhere.*
*
*
*Thanking You,*
*Gayathri*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110614/924c7893/attachment.html 

From agayathribits at gmail.com  Mon Jun 13 22:26:49 2011
From: agayathribits at gmail.com (Gayathri Ananth)
Date: Tue, 14 Jun 2011 10:56:49 +0530
Subject: [Hotspot] Hotspot 5.0 : Lateral heat flow Query
In-Reply-To: <BANLkTinyLKg7Q=dnXGPH2N+CZ2Sn_YKCjA@mail.gmail.com>
References: <BANLkTinyLKg7Q=dnXGPH2N+CZ2Sn_YKCjA@mail.gmail.com>
Message-ID: <BANLkTimOCy34UsVJ2R67bHbOaor2B7px-A@mail.gmail.com>

Hi All,

I have performed some experiments for understanding Lateral heat flow
through silicon using Hotspot 5.0. But from the data I obtained, *I dont see
any lateral heat flow through Silicon.*
*
*
*The following are the experiments I performed:*
*
*
1. I have a 20mm by 20mm chip . In the floorplan I have four cores. I
incrementally ( insteps of 1mm)  increase the distance between the two power
hungry cores (50W for a 8mm by 10mm core).  I have used the *default
configuraton file provided in Hotspot 5.o version. * I am using grid model
and final steady temperature varies as follows

Distance                            Temp
(b/n cores)
4mm                           338.49      338.49
3mm                           338.67      338.69
2mm                           338.87      338.90
1mm                           339.08      339.17
0mm                           339.33      339.33

The above results show a very small change in the temperature, but according
to temperature aware floorplanning concept, the swing should be more.

Q: why I am not seeing such a swing? Also except for  LCF file* is there any
place where I can explicitly give a YES for LATERAL HEAT FLOW.*
*
*
*I also performed other experiments like just heating only one core and
seeing the effects on remaining three cores. But I did not see any lateral
heat conduction through "SILICON".*
*
*
*Also I moved the core diagonally from the edge of the chip until other
edge. The difference in temperature I obtain is very minimal. *
*
*
*Can you please help me in this regard and point me if I am wrong anywhere.*
*
*
*Thanking You,*
*Gayathri*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110614/924c7893/attachment-0001.html 

From arihant at iith.ac.in  Tue Jun 14 00:00:05 2011
From: arihant at iith.ac.in (Arihant Jain)
Date: Tue, 14 Jun 2011 12:30:05 +0530
Subject: [Hotspot] Integration of Wattch with Hotspot
Message-ID: <BANLkTinRVGVOJ7M=WaP6=ZE3U5Zn5kenXQ@mail.gmail.com>

Hey,

How can we generate a ptrace file  using Wattch which is required for
Hotspots??

-- 

Arihant Jain
Class & Hostel Representative
4th Year B.Tech Student
Electrical Engineering Department
Indian Institute of Technology,Hyderabad
Ph.No:+919441919112
Email ID: arihant at iith.ac.in
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110614/f1384541/attachment.html 

From skadron at cs.virginia.edu  Tue Jun 14 11:49:54 2011
From: skadron at cs.virginia.edu (Kevin Skadron)
Date: Tue, 14 Jun 2011 14:49:54 -0400
Subject: [Hotspot] Integration of Wattch with Hotspot
In-Reply-To: <BANLkTinRVGVOJ7M=WaP6=ZE3U5Zn5kenXQ@mail.gmail.com>
References: <BANLkTinRVGVOJ7M=WaP6=ZE3U5Zn5kenXQ@mail.gmail.com>
Message-ID: <4DF7AD52.9060305@cs.virginia.edu>

For each timestep, and for each block in your floorplan, you'll need a 
power value.  Wattch tracks power dissipation in various structures, so 
it's straightforward to modify Wattch to print these to a file.

/K

On 6/14/2011 3:00 AM, Arihant Jain wrote:
> Hey,
>
> How can we generate a ptrace file  using Wattch which is required for
> Hotspots??
>
> --
>
> Arihant Jain
> Class & Hostel Representative
> 4th Year B.Tech Student
> Electrical Engineering Department
> Indian Institute of Technology,Hyderabad
> Ph.No:+919441919112
> Email ID: arihant at iith.ac.in <mailto:arihant at iith.ac.in>
>
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot

From Runjie at virginia.edu  Tue Jun 14 19:04:56 2011
From: Runjie at virginia.edu (Runjie Zhang)
Date: Tue, 14 Jun 2011 19:04:56 -0700
Subject: [Hotspot] Hotspot 5.0 : Lateral heat flow Query
In-Reply-To: <BANLkTimOCy34UsVJ2R67bHbOaor2B7px-A@mail.gmail.com>
References: <BANLkTinyLKg7Q=dnXGPH2N+CZ2Sn_YKCjA@mail.gmail.com>
	<BANLkTimOCy34UsVJ2R67bHbOaor2B7px-A@mail.gmail.com>
Message-ID: <BANLkTin-A6YM0tbjsb2opq4_cvttUis=ZQ@mail.gmail.com>

Hi, Gayathri

  By default, HotSpot grid mode reports block center temperature as the
overall block temperature. In your case, the temperature differences should
be much more obvious on edges than in centers. To change this, you can
change the parameter -grid_map_mode in .config file.

  There are also other possible reasons:
- lateral heat coupling is more prominent for small blocks that are close to
each other. The core size used in your case are relatively big.
- the silicon is thin in the default config (0.1mm), which also limits the
impact of lateral heat transfer.

A further consideration would be what the power map for each core looks
like.  If hot regions are not near the border with other cores, it won't
have much effect.  We have an ICCD paper describing the change in
temperature as a function of distance from the heat source:K.-J. Lee, K.
Skadron, and W. Huang. ?Analytical Model for Sensor Placement on
Microprocessors.? In Proceedings of the IEEE International Conference on
Computer Design 2005.

Thanks
Runjie


On Mon, Jun 13, 2011 at 10:26 PM, Gayathri Ananth
<agayathribits at gmail.com>wrote:

> Hi All,
>
> I have performed some experiments for understanding Lateral heat flow
> through silicon using Hotspot 5.0. But from the data I obtained, *I dont
> see any lateral heat flow through Silicon.*
> *
> *
> *The following are the experiments I performed:*
> *
> *
> 1. I have a 20mm by 20mm chip . In the floorplan I have four cores. I
> incrementally ( insteps of 1mm)  increase the distance between the two power
> hungry cores (50W for a 8mm by 10mm core).  I have used the *default
> configuraton file provided in Hotspot 5.o version. * I am using grid model
> and final steady temperature varies as follows
>
> Distance                            Temp
> (b/n cores)
> 4mm                           338.49      338.49
> 3mm                           338.67      338.69
> 2mm                           338.87      338.90
> 1mm                           339.08      339.17
> 0mm                           339.33      339.33
>
> The above results show a very small change in the temperature, but
> according to temperature aware floorplanning concept, the swing should be
> more.
>
> Q: why I am not seeing such a swing? Also except for  LCF file* is there
> any place where I can explicitly give a YES for LATERAL HEAT FLOW.*
> *
> *
> *I also performed other experiments like just heating only one core and
> seeing the effects on remaining three cores. But I did not see any lateral
> heat conduction through "SILICON".*
> *
> *
> *Also I moved the core diagonally from the edge of the chip until other
> edge. The difference in temperature I obtain is very minimal. *
> *
> *
> *Can you please help me in this regard and point me if I am wrong
> anywhere.*
> *
> *
> *Thanking You,*
> *Gayathri*
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>


-- 
Runjie Zhang
Computer Engineering
University of Virginia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110614/c9abe2dd/attachment.html 

From bhanu at iith.ac.in  Thu Jun 16 23:52:10 2011
From: bhanu at iith.ac.in (Bhanu Prathap T)
Date: Fri, 17 Jun 2011 12:22:10 +0530
Subject: [Hotspot] 3D stacks using HotSpot
Message-ID: <BANLkTik-WW9C29nVxoRtQKj+Gw2PTmNq9g@mail.gmail.com>

Hello everyone,

                         i am trying to get a ttrace file by using the
following command given in HOWTO for 3D stacks


         *  ./ hotspot -c hotspot.config -f <some_random_file> -p
example.ptrace \*
*           -o example.ttrace -model_type grid \*
*           -grid_layer_file example.lcf*

i have given ev6.flp as the random file and several other floorplan files.
All i was getting was a error sayin :

*error: no. of units in floorplan and trace file differ*

but i already made sure the number of units in both files are same. Got
struck here. Will be grateful given a detailed solution.

Thanks in advance.


Bhanu Prathap.T
B.Tech Final year
IIT Hyderabad.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20110617/a7707d43/attachment.html 

From wh6p at virginia.edu  Fri Jun 17 07:54:34 2011
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 17 Jun 2011 09:54:34 -0500
Subject: [Hotspot] 3D stacks using HotSpot
In-Reply-To: <BANLkTik-WW9C29nVxoRtQKj+Gw2PTmNq9g@mail.gmail.com>
References: <BANLkTik-WW9C29nVxoRtQKj+Gw2PTmNq9g@mail.gmail.com>
Message-ID: <BANLkTi=F9Gq4b5petzBxosXb9sOK7J8PHA@mail.gmail.com>

Hi Bhanu,

The power trace file should have all the units in all the layers, So please
make sure that is the case.
Another possibility is the newline character at the end of the file,
depending on whether you edit the file in DOS or Linux and different
editors. One simple way to fix this is to copy and paste the end part in the
example file into your own file.

-Wei

On Fri, Jun 17, 2011 at 1:52 AM, Bhanu Prathap T <bhanu at iith.ac.in> wrote:

> Hello everyone,
>
>                          i am trying to get a ttrace file by using the
> following command given in HOWTO for 3D stacks
>
>
>          *  ./ hotspot -c hotspot.config -f <some_random_file> -p
> example.ptrace \*
> *           -o example.ttrace -model_type grid \*
> *           -grid_layer_file example.lcf*
>
> i have given ev6.flp as the random file and several other floorplan files.
> All i was getting was a error sayin :
>
> *error: no. of units in floorplan and trace file differ*
>
> but i already made sure the number of units in both files are same. Got
> struck here. Will be grateful given a detailed solution.
>
> Thanks in advance.
>
>
> Bhanu Prathap.T
> B.Tech Final year
> IIT Hyderabad.
>
>
>
>
> _______________________________________________
> 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/20110617/fffda1ab/attachment.html 

From dangerisgo at gmail.com  Fri Jun 24 06:27:46 2011
From: dangerisgo at gmail.com (Devon Ostrowski)
Date: Fri, 24 Jun 2011 09:27:46 -0400
Subject: [Hotspot] HS integration with Simplescalar
Message-ID: <4E0490D2.5060902@gmail.com>

Hi,

I am trying to integrate HotSpot 5.0 into the SimpleScalar processor 
model and I am running into some difficulty.  After using the code 
provided in sim-template.c into sim-outorder.c (in the appropriate 
locations), I find that the steady state output file obtained when the 
simulation completes contains the functional units with a '-nan' next to 
each of them.  I have tried both the block and the grid model utilizing 
the default settings (and subsequently, the init temp).  Is there 
anything that I am missing or doing incorrectly?  Also, the power trace 
file is not explained well in the HOWTO.  Is this file necessary for 
anything, or just for obtaining the steady state without integrating 
into a model?  Are there other ways of obtaining a power trace file 
without integration?  Thank you very much for your time.

-Devon O.

From wh6p at virginia.edu  Thu Jun 23 20:04:25 2011
From: wh6p at virginia.edu (Wei Huang)
Date: Thu, 23 Jun 2011 22:04:25 -0500
Subject: [Hotspot] HS integration with Simplescalar
In-Reply-To: <4E0490D2.5060902@gmail.com>
References: <4E0490D2.5060902@gmail.com>
Message-ID: <BANLkTim_f80Cxihyc6j7wLG-WaBh6amTJA@mail.gmail.com>

power trace file is necessary to get a meaningful temperature output from
HotSpot. One method most researchers do is to use Wattch to get a
architecture-level power trace to HotSpot.

-Wei

On Fri, Jun 24, 2011 at 8:27 AM, Devon Ostrowski <dangerisgo at gmail.com>wrote:

> Hi,
>
> I am trying to integrate HotSpot 5.0 into the SimpleScalar processor
> model and I am running into some difficulty.  After using the code
> provided in sim-template.c into sim-outorder.c (in the appropriate
> locations), I find that the steady state output file obtained when the
> simulation completes contains the functional units with a '-nan' next to
> each of them.  I have tried both the block and the grid model utilizing
> the default settings (and subsequently, the init temp).  Is there
> anything that I am missing or doing incorrectly?  Also, the power trace
> file is not explained well in the HOWTO.  Is this file necessary for
> anything, or just for obtaining the steady state without integrating
> into a model?  Are there other ways of obtaining a power trace file
> without integration?  Thank you very much for your time.
>
> -Devon O.
> _______________________________________________
> 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/20110623/0cc381dd/attachment.html