From ashahrour at masdar.ac.ae  Sun Sep  2 15:51:58 2012
From: ashahrour at masdar.ac.ae (Anas Shahrour)
Date: Mon, 3 Sep 2012 02:51:58 +0400
Subject: [Hotspot] Model Extraction
Message-ID: <EA7F4F5873831445B25649EEE38D2E0F019E50BBB4A5@MAILCLUSTER.minet.ae>

Dear HotSpot group,

I would like to know if you have an interface to extract the thermal model of the processor (R and C values).

Best regards,
Anas Shahrour

From ks4kk at virginia.edu  Mon Sep  3 21:37:39 2012
From: ks4kk at virginia.edu (Karthik Sankaranarayanan)
Date: Tue, 4 Sep 2012 00:37:39 -0400
Subject: [Hotspot] Model Extraction
In-Reply-To: <EA7F4F5873831445B25649EEE38D2E0F019E50BBB4A5@MAILCLUSTER.minet.ae>
References: <EA7F4F5873831445B25649EEE38D2E0F019E50BBB4A5@MAILCLUSTER.minet.ae>
Message-ID: <CAPE+kwC40j6CbAiq5gw-=Vc__c4KU-uMGbz=TvjM20CEO+FFpA@mail.gmail.com>

Hi Anas,

Please look for the debug_print_* functions in the code, they should give
you the  routines to output R and C matrices.

-karthik

On Sun, Sep 2, 2012 at 6:51 PM, Anas Shahrour <ashahrour at masdar.ac.ae>wrote:

> Dear HotSpot group,
>
> I would like to know if you have an interface to extract the thermal model
> of the processor (R and C values).
>
> Best regards,
> Anas Shahrour
> _______________________________________________
> 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/20120904/49e13157/attachment.html 

From smilingsantanu at gmail.com  Mon Sep 17 22:18:23 2012
From: smilingsantanu at gmail.com (Santanu Sarma)
Date: Mon, 17 Sep 2012 22:18:23 -0700
Subject: [Hotspot] processor floor plans
Message-ID: <CANq=v5c9vydUk7CVySFki_2+dOwdakqRbJT0MjLtzngfZ9VRoQ@mail.gmail.com>

Dear Hotspot Group,

   Can anyone share the processor floorplan for any of the EV4, EV5 or EV8
( other than EV6 as it is already included) ?

Thanks and regards,
Santanu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120917/12828fc1/attachment.html 

From santanus at uci.edu  Wed Sep 19 17:54:15 2012
From: santanus at uci.edu (santanus at uci.edu)
Date: Wed, 19 Sep 2012 17:54:15 -0700
Subject: [Hotspot] floorplan files
Message-ID: <4354066a3d78d505a2eafd59e9256ecb.squirrel@webmail.uci.edu>

Hello Hotspot Group,


   I am looking for the floor plan files of  EV4 , EV5 and EV8 
processors. Can anyone share or suggest me a source.

Thanks and Regards,
Santanu




From yanwei.song at gmail.com  Fri Sep 21 06:01:32 2012
From: yanwei.song at gmail.com (Yanwei Song)
Date: Fri, 21 Sep 2012 09:01:32 -0400
Subject: [Hotspot] Thermal modeling without heat sink
Message-ID: <CAHBCnFROatFJizeV6U2rLjcabSpqBOdApwiEU2KwYdEvBc0fKQ@mail.gmail.com>

Hi Hotspot group,

I'm trying to model a 3D die architecture (64 mm2) in the absence of heat
sink (using conventional DRAM-like cooling package settings).
According to some previous threads in the maillist, I set the the thickness
of heat sink to 0.00001.

Should I also set the heat spreader also that thin? Though I did see this
suggestion from those threads.
I thought heat spreader should be kind of independent from heat sink,
especially for a dram chip, can it have only the heat spreader (thickness =
0.001)?
In this way, is it similar to a thin heat sink? I found big temperature
difference between t=0.001 and t=0.00001 for heat spreader.

Another question is about the convection resistance, temperature is also
quite sensitive to this number when there is no heat sink.
Wei Huang mentioned in one thread about setting it to "at least 10x", so I
set that to 1.5. Is it a reasonable number for a chip without heat sink?
In other words, how should I derive the convection resistance correctly?


Thanks.

Best regards,
Yanwei
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120921/f52d528a/attachment.html 

From smillican at wisc.edu  Fri Sep 21 07:47:51 2012
From: smillican at wisc.edu (Spencer Millican)
Date: Fri, 21 Sep 2012 09:47:51 -0500
Subject: [Hotspot] Thermal modeling without heat sink
In-Reply-To: <CAHBCnFROatFJizeV6U2rLjcabSpqBOdApwiEU2KwYdEvBc0fKQ@mail.gmail.com>
References: <CAHBCnFROatFJizeV6U2rLjcabSpqBOdApwiEU2KwYdEvBc0fKQ@mail.gmail.com>
Message-ID: <CAFT6x5eDGthGYeoSKF9UU8h1VHdfa-4pVhinU=iu9qW_G+jr6A@mail.gmail.com>

I'm interested in this too.  However, when I set the thickness too low I
get invalid temperature values back (-inf or -nan).  So far, the best I can
do is set the side length to just over the chip side length (setting it
0.000001 over works fine, but equal does not), and doing the same for the
height.

-Spencer

On Fri, Sep 21, 2012 at 8:01 AM, Yanwei Song <yanwei.song at gmail.com> wrote:

> Hi Hotspot group,
>
> I'm trying to model a 3D die architecture (64 mm2) in the absence of heat
> sink (using conventional DRAM-like cooling package settings).
> According to some previous threads in the maillist, I set the the
> thickness of heat sink to 0.00001.
>
> Should I also set the heat spreader also that thin? Though I did see this
> suggestion from those threads.
> I thought heat spreader should be kind of independent from heat sink,
> especially for a dram chip, can it have only the heat spreader (thickness =
> 0.001)?
> In this way, is it similar to a thin heat sink? I found big temperature
> difference between t=0.001 and t=0.00001 for heat spreader.
>
> Another question is about the convection resistance, temperature is also
> quite sensitive to this number when there is no heat sink.
> Wei Huang mentioned in one thread about setting it to "at least 10x", so I
> set that to 1.5. Is it a reasonable number for a chip without heat sink?
> In other words, how should I derive the convection resistance correctly?
>
>
> Thanks.
>
> Best regards,
> Yanwei
>
> _______________________________________________
> 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/20120921/3294b177/attachment-0001.html 

From dip224 at gmail.com  Thu Sep 27 22:35:31 2012
From: dip224 at gmail.com (subhadip kundu)
Date: Fri, 28 Sep 2012 11:05:31 +0530
Subject: [Hotspot] Problem in ttrace file
Message-ID: <CAOuN=+72yHbutpCsfAr8T0jc_TYgqvJ+AyWOXQ3WM_=CRtDktQ@mail.gmail.com>

HI,

I am having trouble in generating transient temperatures. I have given a
power trace file containing power at different time instants. I want to
check the change in transient temperatures. But the ttrace file generated
by hotspot gives "nan" (not a number) values for most of the cases.

Please help me solving this problem and if possible give me the details of
how this ttrace file is generated.

Thanks and Regards,
Subhadip Kundu.
IIT Kharagpur.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120928/2d844303/attachment.html 

From wh6p at virginia.edu  Fri Sep 28 03:12:45 2012
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 28 Sep 2012 05:12:45 -0500
Subject: [Hotspot] Problem in ttrace file
In-Reply-To: <CAOuN=+72yHbutpCsfAr8T0jc_TYgqvJ+AyWOXQ3WM_=CRtDktQ@mail.gmail.com>
References: <CAOuN=+72yHbutpCsfAr8T0jc_TYgqvJ+AyWOXQ3WM_=CRtDktQ@mail.gmail.com>
Message-ID: <CANYeXBoq_M+B55td_=NWgQRDhv19azgurX2+d-jApLJQo0BNwA@mail.gmail.com>

Can you generate steady-state temperatures with the same power trace file
and other configs you used?

-Wei

On Fri, Sep 28, 2012 at 12:35 AM, subhadip kundu <dip224 at gmail.com> wrote:

> HI,
>
> I am having trouble in generating transient temperatures. I have given a
> power trace file containing power at different time instants. I want to
> check the change in transient temperatures. But the ttrace file generated
> by hotspot gives "nan" (not a number) values for most of the cases.
>
> Please help me solving this problem and if possible give me the details of
> how this ttrace file is generated.
>
> Thanks and Regards,
> Subhadip Kundu.
> IIT Kharagpur.
>
> _______________________________________________
> 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/20120928/35c019b1/attachment.html 

From wh6p at virginia.edu  Fri Sep 28 03:36:17 2012
From: wh6p at virginia.edu (Wei Huang)
Date: Fri, 28 Sep 2012 05:36:17 -0500
Subject: [Hotspot] Problem in ttrace file
In-Reply-To: <CAOuN=+7Cm+E20tCEf=yVDTfQv=bqp2KTuCGZBn58rm90W8sXsw@mail.gmail.com>
References: <CAOuN=+72yHbutpCsfAr8T0jc_TYgqvJ+AyWOXQ3WM_=CRtDktQ@mail.gmail.com>
	<CANYeXBoq_M+B55td_=NWgQRDhv19azgurX2+d-jApLJQo0BNwA@mail.gmail.com>
	<CAOuN=+6dnGE7-sNv6eQjz_=aH+gh2ZbnsVWVrtiHjyzyJY0rsg@mail.gmail.com>
	<CAOuN=+7Cm+E20tCEf=yVDTfQv=bqp2KTuCGZBn58rm90W8sXsw@mail.gmail.com>
Message-ID: <CANYeXBqvGPVN2ebSpYPLAchMKbxrj7YdwBz8Vmzr7PzGnpXVUA@mail.gmail.com>

Please send your command line and config file, I can take a look.

-Wei

On Fri, Sep 28, 2012 at 5:21 AM, subhadip kundu <dip224 at gmail.com> wrote:

> I can send the files if u want ...
>
>
> On Fri, Sep 28, 2012 at 3:50 PM, subhadip kundu <dip224 at gmail.com> wrote:
>
>> Yes, I can generate Steady state Temperature...
>>
>>
>> On Fri, Sep 28, 2012 at 3:42 PM, Wei Huang <wh6p at virginia.edu> wrote:
>>
>>> Can you generate steady-state temperatures with the same power trace
>>> file and other configs you used?
>>>
>>> -Wei
>>>
>>> On Fri, Sep 28, 2012 at 12:35 AM, subhadip kundu <dip224 at gmail.com>wrote:
>>>
>>>> HI,
>>>>
>>>> I am having trouble in generating transient temperatures. I have given
>>>> a power trace file containing power at different time instants. I want to
>>>> check the change in transient temperatures. But the ttrace file generated
>>>> by hotspot gives "nan" (not a number) values for most of the cases.
>>>>
>>>> Please help me solving this problem and if possible give me the details
>>>> of how this ttrace file is generated.
>>>>
>>>> Thanks and Regards,
>>>> Subhadip Kundu.
>>>> IIT Kharagpur.
>>>>
>>>> _______________________________________________
>>>> HotSpot mailing list
>>>> HotSpot at mail.cs.virginia.edu
>>>> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>>>>
>>>>
>>>
>>
>>
>> --
>> Subhadip Kundu.
>> IIT Kharagpur.
>>
>
>
>
> --
> Subhadip Kundu.
> IIT Kharagpur.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120928/e816680d/attachment.html 

From devendra.rai at tik.ee.ethz.ch  Fri Sep 28 07:09:44 2012
From: devendra.rai at tik.ee.ethz.ch (Devendra Rai)
Date: Fri, 28 Sep 2012 16:09:44 +0200
Subject: [Hotspot] Most common supported architectures
Message-ID: <CAM3NbWHFqp0X5m7FP2n5jX2qR1+30unfCaZzGRD81qXcGD+wAw@mail.gmail.com>

Hello,

I am curious, as to how most people use hotspot these days.

Hotspot requires power numbers (generally from Wattch), which requires
architectural simulator (Simplescalar). Since simplescalar was originally
based on Alpha/Pisa architectures (DEC Alpha?) , so the entire chain works
for Alpa/PISA architectures.

Now, that Simplescalar supports more diverse architectures, the bottleneck
seems to be detailed micro-architectural information which Hotspot
requires. How is this done, for say, modern processors? Going by recent
activity on the mailing list, I am assuming that this information is hard
to come by.

Also, can someone tell me which technology node does Wattch currently
simulate?

The point is, is it fair to assume that Simplescalar-Wattch-Hotspot
toolchain works best with Alpha/PISA architectures, for which all details
are known?

Would be great to hear from anyone.

Devendra Rai
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120928/540fb700/attachment.html 

From Runjie at virginia.edu  Fri Sep 28 10:42:42 2012
From: Runjie at virginia.edu (Runjie Zhang)
Date: Fri, 28 Sep 2012 13:42:42 -0400
Subject: [Hotspot] Thermal modeling without heat sink
In-Reply-To: <CAFT6x5eDGthGYeoSKF9UU8h1VHdfa-4pVhinU=iu9qW_G+jr6A@mail.gmail.com>
References: <CAHBCnFROatFJizeV6U2rLjcabSpqBOdApwiEU2KwYdEvBc0fKQ@mail.gmail.com>
	<CAFT6x5eDGthGYeoSKF9UU8h1VHdfa-4pVhinU=iu9qW_G+jr6A@mail.gmail.com>
Message-ID: <CAHxZ6YAefFX26BqLHy4YZxg40uBJkJKabp7vQnF39K=2L8s1-g@mail.gmail.com>

Hi, Yanwei and Spencer

  Using small thickness for heatsink and heatspreader are OK for such
studies (as long as HotSpot doesn't complain).

  For convection resistance, we do not have a specific number yet, but it
should be available online.

Runjie

  For  convection resistance, I am not sure what value should people use
because different heatsink has different resistance value. Does an
arbitrary large value work?

On Fri, Sep 21, 2012 at 10:47 AM, Spencer Millican <smillican at wisc.edu>wrote:

> I'm interested in this too.  However, when I set the thickness too low I
> get invalid temperature values back (-inf or -nan).  So far, the best I can
> do is set the side length to just over the chip side length (setting it
> 0.000001 over works fine, but equal does not), and doing the same for the
> height.
>
> -Spencer
>
> On Fri, Sep 21, 2012 at 8:01 AM, Yanwei Song <yanwei.song at gmail.com>wrote:
>
>> Hi Hotspot group,
>>
>> I'm trying to model a 3D die architecture (64 mm2) in the absence of heat
>> sink (using conventional DRAM-like cooling package settings).
>> According to some previous threads in the maillist, I set the the
>> thickness of heat sink to 0.00001.
>>
>> Should I also set the heat spreader also that thin? Though I did see this
>> suggestion from those threads.
>> I thought heat spreader should be kind of independent from heat sink,
>> especially for a dram chip, can it have only the heat spreader (thickness =
>> 0.001)?
>> In this way, is it similar to a thin heat sink? I found big temperature
>> difference between t=0.001 and t=0.00001 for heat spreader.
>>
>> Another question is about the convection resistance, temperature is also
>> quite sensitive to this number when there is no heat sink.
>> Wei Huang mentioned in one thread about setting it to "at least 10x", so
>> I set that to 1.5. Is it a reasonable number for a chip without heat sink?
>> In other words, how should I derive the convection resistance correctly?
>>
>>
>> Thanks.
>>
>> Best regards,
>> Yanwei
>>
>> _______________________________________________
>> 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
>
>


-- 
Runjie Zhang
Computer Engineering
University of Virginia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120928/71b88240/attachment-0001.html