Prop 2 processing power in current form?

DrPop

Curious if anyone knows how many MIPS we're looking at for the P2 with all the revisions to this point?
With the P1 = (8x) 20 to 25 MIPS depending on crystal MHz, and an average Arduino Uno (ATmega328P) ~ 16 MIPS, it would be nice to have a rough comparison.
Thanks!

jmg

On a simple opcode-feed comparison, the P2 is half the SysClk division, and targets 2x the SysCLK max, so that makes it 4x faster at eating opcodes....

The AVR opcodes have (mostly) just 8 bit operands, and the AVR has a small number of registers, go above that and you have to load/store.

- but the P2 is not opcode-feed alone, it adds a lot of 32 bit opcodes that will speed mathops, and smart pins and a streamer that boost IO. eg It can capture to 1/SysCLK, so 10x the precision of an AVR.

It would leave AVR and P1 in the dust, for most Embedded Applications.


See the Opcode list in the spreadsheet here
http://forums.parallax.com/discussion/162298/prop2-fpga-files-updated-2-november-2017-version-26

cgracey

At 160MHz, cogs will hit 80MIPS each. With 8 cogs, that totals to 640MIPS.

DrPop

Thank you, it's clearly a sizable upgrade from P1 and anything Arduino (etc.) currently offers.
Paired with higher-end computer boards running ROS, the P2 could have a place in larger robotics applications performing real-time control of hardware. With some strategic marketing, Parallax could gun for a slice of that growing pie.

Edit: Maybe I should explain what I meant by that. The "lower level" educational robotics / electronics / programming market is important and a good way to familiarize the next generation of engineers, programmers and makers with the Propeller.
How about making a concentrated effort to "follow" those students and their projects, supporting them on the next level in college and then R&D either in grad student positions or in the field? With the P2, Parallax will have the horsepower in the chip to do that. Serious AI robotics seems like another market just waiting to be broken into by the P2.

Imagine the demand for the P2 if you could get it used in (for example), MIT or Google, or Tesla automated driving R&D, and if it were chosen to be utilized in the production version going into hundreds of thousands of cars, etc.

Or a DARPA robotics challenge, funded by the DOD and imagine that contract if the unit with a Propeller design inside won?

Yes, I realize most of all that will be done with ARM and heavy lifting GPU chips, but there must be a use for a small, direct access, real-time hardware controlling chip in there somewhere.

Along that line of thought, has anyone here written ROS drivers for any of Parallax's products or used the Propeller chip under ROS in any capacity?

DrPop

This doesn't seem to be much of a popular topic, and not trying to step on any toes, but here's just one example: TurtleBot. The TurtleBot in its various guises has been a mainstay of ROS invention for years. Both the TurtleBot 3 Burger and Waffle use an OpenCR1.0 board, and the Burger complements this with a Raspberry Pi 3. The Waffle adds a now discontinued Intel Joule board.

Parallax could approach their Northern Cal ROS friends with a prototype next-gen design of the TurtleBot 4 utilizing something like an Nvidia TX2 and a P2 combo. The ROS crowd would then spread the word and the "work" of globalizing the use of the P2.