[parsec-users] Canneal Program

Jason F snorlaxgb at gmail.com
Sat Sep 14 12:34:42 EDT 2013


Hello,
The weird issue here is ( I should of specified) is I am using an ideal
machine and I am also only executing ROI. So I have unlimited memory.

The way I measure ILP is I get total instructions executed / Total Cycle
Time.

Where unit of time is basically mapping the dependencies among instructions
(to form a tree). And the length of the tree is basically cycle time.
I am pretty baffled.

Also thank you so much for your response, it is greatly appreciated!!



On Sat, Sep 14, 2013 at 1:56 AM, George Kurian
<georgekurian.086 at gmail.com>wrote:

> I know that canneal has a huge sequential initialization portion where it
> sets up the netlist. Did you measure the ILP in that too? To not measure
> the initialization region, compile using gcc-hooks and measure the ILP in
> just the ROI (region-of-interest).
>
>
> On Fri, Sep 13, 2013 at 10:04 PM, Jason F <snorlaxgb at gmail.com> wrote:
>
>> I am using a proprietary simulator that measures the amount of ILP in a
>> given benchmark. I noticed when I go from small to large instruction set
>> for Canneal, the ILP is significantly smaller for large.
>>
>> Anyone know why that is?
>> I assumed a large set provides more parallelism.
>>
>>
>> _______________________________________________
>> parsec-users mailing list
>> parsec-users at lists.cs.princeton.edu
>> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>>
>>
>
>
> --
> George Kurian
> 5th year Graduate Student,
> Computer Science and Artificial Intelligence Laboratory (CSAIL),
> Massachusetts Institute of Technology.
>
> _______________________________________________
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20130914/917bedf2/attachment.html>


More information about the parsec-users mailing list