I think I'm closer to the answer. It seems to lie in the register AGPIO_CFG at address 0x1000003C. It switches the GPIOs 14 through 29 between the digital pins and the analog ethernet counterparts. Would this modification to AGPIO be done in the DTSI file, some bootloader/kernel parameter or do I need to make a patch to something somewhere? I'm aware I could use something like devmem to directly set this register from userspace but I assume there is something a bit cleaner?

A good socket is always necessary - no doubt about it. Sometimes it’s also a good idea to buy high-quality power supplies . This is the basis, especially when it comes to CNC machines.

@Chris-Stratton said in C++ raw GPIO?: @Costas-Costas the approach chosen by Kit and now Travis won't work on standard firmware until the /dev/mem issue is fixed. Yes, one of the many reasons that I am holding off upgrading the code to Omega2 The other main constraint (other than time availability) is that the endianess differs on the Omega1 and Omega2 and I need to check out the places where this is significant in the code (I know of at least one place)

@fossette Yeap that was it. I figured out the relay I had on these pins was causing a massive Amps' draw for a few milliseconds. That of course cause Omega to reboot due to "power loss". Thanks anyway!

@Johan-Simons I previously used a similar I2C I/O expander with my Arduino - it worked well. In case you are interested, the expander break out I used was http://www.gravitech.us/i2c16gpex.html which uses a PCA9535 chip which provides similar capability to the MCP23008 but with 16 I/O pins rather than 8. Also, the documentation on PCA9535 says: No glitch on power-up Internal power-on reset It worked well for me.

http://www.aliexpress.com/item/Free-Shipping-10pcs-2mm-40-Pin-Male-Single-Row-Pin-Header-Strip/32327259863.html?spm=2114.30010308.3.1.OSreCh&ws_ab_test=searchweb201556_3,searchweb201644_1_10001_10002_10005_301_10006_10003_10004_62,searchweb201560_5,searchweb1451318400_6150,searchweb1451318411_6448