Has anyone done any text scrolling on the matrix with this code base?

[mggates39]

I am about to start adding scrolling text to the Stock Ticker and I am looking for guidance. I am not as familiar with the GFX library as I should be. I have done some scrolling with the Micro Python libraries but I do not see their equivalents here. Any tips and pointers would be appreciated.

Thanks,
Marshall

[heidepiek]

Mr. gates, maybe this will help you.

On youtube you have a channel called greatprojects.

This gentleman has quite some experience with hub75 panels. Maybe you can learn something from his codes.

He also has a github site: https://github.com/3tawi

Here is a movie with scrolling text on a hub75, with some in the background. But maybe you can filter it out. https://www.youtube.com/watch?v=4pQMjzJ9uJ8

Regards Harrie

[davepl]

Check out the MoveX and MoveY functions. If you just need to move part of the screen, we could add default args that specify which rows, for example.

[mggates39]

Thank you. I will start there. The fun part, I think, will be putting in new text at the edge as space is made available. I will think about it this week. Now if I could just see what I am doing.

[robertlipe]

We could add code that rendered the letter from the character cells one column at a time and added that to either edge before shifting. Most of that it outsourced down to SmartMattrix, so go grump around in the Layer and Font code in .pio/libdeps/mesmerizer/SmartMatrix/src or the original. I'll admit I don't know a lot about that code specifically, but I've professionally done the kinds of things I know it's doing. Depending on the memory requirement (a 64x64x24bpp pixel buffer * double buffer isn't cheap) maybe we could afford to overdraw the entire pixel buffer and just render a WIDTHxHEIGHT offset to the actual part that gets blitted out to the DMACs. Since these wouldn't be animating at 60Hz, maybe that's an option. Kind of like the old 6845 fast scroll trick where scrolling up and down was "just" telling the video controller to start drawing the buffer from a different offset instead of zero. Maybe this is already in SmartMatrix somewhere - it's pretty fully featured. For HUB75, it's usually not the actulal draw that's the killer, but the need to keep refreshing the cells as they aren't latched and have no 'memory' - you have to keep refreshing the pixels like you were feeding an electron gun from a UART. :-/ Also, the functions that Dave cited are interesting candidates for the Async Memory Copy API, especially if you're on S2, C3 (probably not, since those are both single core) or S3. The ESP devices have a ridiculous number of actual memory busses and just the arbitration of getting a read/write cycle (use [q]words, not bytes) on and off the bus can dwarf the cost of the transaction itiself, so it's worth programming the DMACs to do it if there's a lot to move since they can reduce the # of arbitration requests. For a different project (not this one) I was able to help a friend move the ESP32-S3 with *octal* SPI RAM (very much *Not A Mesmerizer...*) from about 9MB/sec to about 320MB/sec for bursts <https://docs.espressif.com/projects/esp-idf/en/latest/esp32s3/api-reference/system/async_memcpy.html> of each. Now we had assistance from hardware engineering and Espressif themselves to help coach us on every bus cycle, alignment, arbitration request, and opcode. The point isn't to brag (I couldn't do it again without access to all those things and code that wasn't mine to keep) but to say that 40x performance increases on these things really ARE possible if you're willing to help stack the deck in every possible way to make it go fast. Even the choice of which AHB GDMA channel to use required coaching. Wrapping the async_memcpy_isr_cb_t in a lambda for callbacks to actually notify the other thread that the data had arrive was just Chef's Kiss. In their case, it saved them from having to think about moving to a custom IOP....and it was on hardware that the Mesmerizer isn't. But even a "fast" scroll should be between those numbers, right? I'll admit that I wasn't there for every step of the process and I'm pretty sure that the numbers we achieved aren't possible with the older ESP32 part parts. I'm just saying that if clock cycles DO matter, there are some incredible options out there and that for HUB75 development, the hard part probably isn't actually the scrolling; it's keeping the pixels refreshed. Even the "old" esp32's have some pretty awesome DMA options waiting to be used. Not everything needs to be VAX or ATMega-compatible. This class of parts has lots of features to "help". Memcpy-coprocessors are (essentially) things you can use these days. Handing a user-space scgth list of (mumble, mumble) std::spans that are locked into (please don't page out these addresses while I'm transferring to/from them...) core to a DMAC and popping the "wakeup" when you receive the completion interrupt is pretty much the system programmer's dream. JH-7110 has a couple of coprocessors on board that follow this same mantra. Enjoy! On Tue, Dec 12, 2023 at 12:06 PM Marshall G Gates ***@***.***> wrote:

…

Thank you. I will start there. The fun part, I think, will be putting in new text at the edge as space is made available. I will think about it this week. On Tue, Dec 12, 2023 at 1:02 PM David W Plummer ***@***.***> wrote: > Check out the MoveX and MoveY functions. If you just need to move part of > the screen, we could add default args that specify which rows, for example. > > — > Reply to this email directly, view it on GitHub > < #566 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AA6PTXAHCF2ID7QO5QC2AE3YJCLZVAVCNFSM6AAAAABAPCLSSWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TQMZTHAZDS> > . > You are receiving this because you authored the thread.Message ID: > <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/566/comments/7833829 > @github.com> > -- *Marshall G Gates* *Android Developer* — Reply to this email directly, view it on GitHub <#566 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD36NHUTRYGHGN4WU2T3YJCMLVAVCNFSM6AAAAABAPCLSSWVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TQMZTHA3TM> . You are receiving this because you are subscribed to this thread.Message ID: <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/566/comments/7833876 @github.com>

[mggates39]

I have two more days of work, then I am off for the rest of the year. I know what I am going to be doing between family visits. I really need to get my matrix board wired up to my development board so I can test some of this.