AddressableLED ColorOrder #2329
Conversation
|
Three years, do we want to do anything with this? What was this blocked on? (not that I expect anyone to remember...) |
|
It needs a C++ version, but the problem is that you can't do a C++ version of this in the same way, since the LEDData class in C++ is a bit-for-bit wrapper around the C structure. We could maybe make a templated LEDData class (or simply separate LEDData classes) to do the different color order variants? |
|
What's the use case for this, anyway? |
Most COTS led strips are not RGB color order, and many differ from the typical BGR too. I have a few GRB strips |
calcmogul
added
component: wpilibj
wpilibj/src/main/java/edu/wpi/first/wpilibj/AddressableLEDBuffer.java
Outdated
Show resolved
Hide resolved
| default -> { | ||
| // Unreachable | ||
| } |
There was a problem hiding this comment.
According to https://docs.oracle.com/en/java/javase/17/language/switch-expressions.html this shouldn't be required for enums?
There was a problem hiding this comment.
Well, for some reason, a bunch of places in WPILib have default cases for enums anyway.
There was a problem hiding this comment.
This isn't a switch expression, it's still a statement.
There was a problem hiding this comment.
The empty default is functionally equivalent to not handling it
|
Implementing the c++ version via templates is now significantly more difficult due to the LED pattern library. |
|
I think the right answer for this may be to do the color channel swizzling when the buffer is copied to the FPGA, rather than on every write, using NEON SIMD to make it fast using code similar to https://github.com/ermig1979/Simd/blob/master/src/Simd/SimdNeonBgrToRgb.cpp (effectively the same code as discussed here: https://developer.arm.com/documentation/102159/0400/Load-and-store---example-RGB-conversion). We'd want to align and pad out the internal buffers to avoid boundary conditions. |
|
In the interest of preventing duplicated work- I've got the simd solution mostly implemented, just need to take some time and finish up all the little details. |
|
Superseded by #7102. |
No description provided.