Support for GPIO16/D0 #222
Conversation
There was a problem hiding this comment.
Howdy. I’d be happy to merge this functionality, but would request changes. I’ll summarize below if you’d like to do yourself, otherwise I’ll merge as-is and then follow up with my own edits. In either case, the commit history would acknowledge you as the author. If I don’t hear back by next time I’m making a pass through PRs, will do the latter.
Rather than compensating for the branches by subtracting cycles, I’d suggest simply having a distinct loop for the (pin == 16) case — it could all go in the test starting around line 66. Yes this would increase code size by a couple dozen bytes, but I feel it’s worth it to avoid anything cycle-counting-related, as cycles can be thrown off by optimizer settings, clock speed, etc.
(The irony of that last part isn’t lost on me, since like 90% of the code is cycle-counting nastiness, but any opportunity to make that better instead of worse is welcome, like when the SAMD51 code got cleaned up.)
Thanks!
|
Hi, feel free to modify or change anything, no need to mention my name or anything. Maybe this is interesting for you too, kinda clean rewrite of my library while I was working on FastLED on the issues with framework-arduinoespressif8266 3.0.0 https://github.com/sascha432/NeoPixelEspEx |
Adds support for GPIO16/D0 for ESP8266 by using the RTC register address instead the GPIO register address.
The changes can be tested with any example code.
I tested it with an 8 pixel strip (WS2812 800kHz, CPU 80MHz) and measured timings with an oscilloscope. The timings match the timings without modifications and are within the datasheets specifications.
Ports tested 12/D7 and 16/D0.
ESP32 code isn't affected.