I was thrilled to see the announcement of the "Highest performance blackfin processor" (http://www.analog.com/en/press-release/03_28_12_ADI_Blackfin_Processors_Accelerated_Visio/press.html), even more when the news said "a series of 1-GHz, dual-core, Blackfin® processors", but several lines later, I was saddened reading "These four new processors all have two 500MHz cores"
I had previous experience with the first blackfin, so when I recently had to make a design for a power sensitive application (who also had bad needs in computing power ) I selected the BF561. Last week, I finished to make it work on the BF561 EZ Kit after 3 months of heavy work in algorithms enhancement and assembly coding, so I'm balancing between the urge to lower consumption with the 609 and the additional workload to try (and may not succeed) to make it work in it.
I'm not familiar to code profiling (or else) tools to extract data to sharpen the analysis, but I'm relying heavily on the both cores, running at 600MHz, with 24x24 convolutions and others calculus in real time. I've tested by decreasing the frequency that the design needs to run no lower than 550MHz in order not to loose frames. So my questions are:
- will the BF561 be EOL soon?
- It's a preliminary datasheet. Is the frequency really limited to 500MHz? Does customers really thought that the BF561 was to performant?
Maybe I'm abused by marketing talks, but getting to lower processing node than the previous generation doesn't improve frequency, like going nearer shortens the travel time?
- What is the estimated guess in coding time to move from BF561 to BF609? I see that the DMA controller, the external memory and the PPI has changed, is there a simple way to move from one to another, knowing that I'm using vdk.
- Is there a simple way to test or demonstrate that there is significant improvement in the internal architecture, other than the PVP to make the news title more appealing?