On Thu, 7 Sep 2000, Alex Holden wrote: > However, I have discovered that the 36067 has support for DMAing RGB data > from system RAM and filtering it through the 36060- using it as a hardware > JPEG accelerator. So what I could do, and admittedly this isn't an ideal > solution but it should be faster than doing the compression in software, > is to capture in RGB format to a contigous buffer, draw the captions to > the buffer (a user space program could mmap() the buffer to get to it and > orchestrate the process using ioctl()s), then feed the data back through > the 36067 and 36060 to compress it. Are there any obvious flaws I've > missed? Any suggestions on a suitable API? Sounds much more complex than necessary. Each frame is also going to involve moving a lot of data across the PCI bus, depending on how much you're trying to capture, this might be too much. Adding the information you want as a comment in the jpeg file would be easy. But, if you need it actually be part of the image, you can do that too. You can add additional MCU blocks to the image without fully uncompressing the image.
