> On Jan 25, 2019, at 11:11 PM, Calvin Slater <[email protected]> 
> wrote:
> 
> Looks like I am a lucky dog as well. My x15 also has the industrial temp 
> rated AM5729. This one was from the first public production run back in late 
> 2017. I did some searching around and found the following:

All Rev C boards used AM5729. The industrial part is a surprise to me. 

> 
> 1) The AM5729 is a non-catalog part, which is why there is no real public 
> datasheet.

Correct. Having 4 EVEs is the only difference from AM5728. I have hopes the 
AM5729 will be public some time this year. 

> 
> 2) This chip has four EVEs but only one is enabled for the purpose of 
> hacking/experimenting.

All 4 are enabled. 

> 
> Is the above correct?
> 
>> On Wednesday, 16 January 2019 05:30:11 UTC-8, Mark A. Yoder wrote:
>> My x15 has an AM5729 on it and it has 2 DSPs and 4 EVEs[1] (Embedded Vision 
>> Engines).  I don't know much about the EVEs, but I read somewhere that each 
>> EVE can do 16 multiply accumulates per clock cycle.  The tidl gives you some 
>> control over which processors (DSP or EVE) works on what part of the problem.
>> 
>> I'm often seeing 15 to 30 frames per second wile recognizing objects.
>> 
>> --Mark
>> 
>> [1] http://processors.wiki.ti.com/index.php/EVE
>> 
>>> On Thursday, January 10, 2019 at 4:40:06 PM UTC-5, Calvin Slater wrote:
>>> That's fantastic!
>>> 
>>> I was just wondering about this a couple weeks ago. 
>>> 
>>> I heard the AM5728 had TIDL support this whole time and uses the DSPs right?
>>> 
>>> 
>>> 
>>>> On Wednesday, 9 January 2019 18:08:28 UTC-8, Mark A. Yoder wrote:
>>>> It was recently pointed out to me that the BeagleBoard-X15 has hardware 
>>>> that supports Deep Learning and TI has already created several examples of 
>>>> how to use it.
>>>> 
>>>> I've created a wiki page that gives a quick guide for installing and 
>>>> running the examples.
>>>> 
>>>> https://elinux.org/EBC_Exercise_39_Setting_Up_tidl_on_X15
>>>> 
>>>> All the examples are pretrained and the X15 is just running the inference 
>>>> engine.  It's been trained to recognize 1000 objects from a live video 
>>>> stream.
>>>> Using a simple webcam, I've shown it several objects (tennis ball, 
>>>> baseball, coffee mug, beer bottle, etc.) and it has recognized them all.
>>>> 
>>>> I'm impressed.
>>>> 
>>>> Has anyone else played with this? What do you think?
>>>> 
>>>> --Mark
>>>> 
>>>> 
>>>> 
> 
> -- 
> For more options, visit http://beagleboard.org/discuss
> --- 
> You received this message because you are subscribed to the Google Groups 
> "BeagleBoard" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected].
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/beagleboard/b1fdf594-4be4-460f-881d-7b489bf1aaba%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

-- 
For more options, visit http://beagleboard.org/discuss
--- 
You received this message because you are subscribed to the Google Groups 
"BeagleBoard" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/beagleboard/D06174DA-783B-43F4-8607-E533D8529C61%40gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to