PIO push direction...

[JackCode0011]

Hello everyone

I'm really lost and just can't see the solution here anymore.
I'm not sure, if using SHIFT LEFT or SHIFT RIGHT is the right way for my problem.

I'm reading Pins 0,1,2 and 3 with in_(pins, 4).
Let's assume, the pins have the following logic:
Pin 0 - 1
Pin 1 - 1
Pin 2 - 0
Pin 3 - 0

I want to read everything into a bufferarray and than later on convert it to binary.
When converting to binary I want as output:
1100 1100 1100 1100
1st in_ 2nd in_ 3rd in_ 4th in_

Also when pushing ISR into FIFO and FIFO to Bytearray, new push should be at the end of the bytearray:
11001100110011001100110011001100.... 11001100110011001100110011001100....
1st FIFO to Bytearray 2nd FIFO to Bytearray

Now, in my understanding I have to shift left. Because as I understand it:

ISR before first in_
00000000000000000000000000000000
in Pin 0: 1
00000000000000000000000000000001
in Pin 1: 1
00000000000000000000000000000011
in Pin 2: 0
00000000000000000000000000000110
in Pin 3: 0
00000000000000000000000000001100

and so one until autopush with 32. So everything moves to the left.

So, am I right with SHIFT LEFT or should it be SHIFT RIGHT?

I'm asking, because I'm reading an LCD-Signal who outputs 4 bit pixel at once:
So every clock is 4 bits (Pin 0,1,2,3) and the 4 bits are 4 pixel from left to right.
Now to check the picture, when using framebuf.MONO_HLSB or HMSB in both cases the bits are not in the right places.
So with SHIFT LEFT and HLSB I get everything wrong.
With SHIFT LEFT and HMSB I get the right order for bytes, but the bits are still reversed (so 00110011).

The only way of getting the right picture is using SHIFT RIGHT with PIO and then HMSB. But as I said, in my head it doesn't make sense to me.

Greetings

[JackCode0011]

To clarify my thoughts:

With shift left I would expect out of the RX FIFO:
1 2 3 4 5 6 7 8 9 10 ... 32 => so one byte in the bytearray: 11001100, 204 or HEX CC
and with shift right
32 .. 10 9 8 7 6 5 4 3 2 1 => so one byte in the bytearray: 00110011, 41, HEX 33

But maybe I see, that there could be another problem:
When RX FIFO has 4 words (32 bits) and DMA puts it into the bytearray:
Is it bytearray: BYTE BYTE BTYE BYTE ...
or
WORD WORD WORD WORD ... (WORD = 4 BYTES, so whole RX FIFO)

[robert-hh]

The DMA puts the 32 bit word into the memory, hight order byte first with bswap=True, or the opposite with bswap=False. The DMA does not know what a bytearray is, it just knows a target address.

[robert-hh]

You can as well tell the PIO to push every byte with push_thresh=8 and set the DMA word size to byte.

[JackCode0011]

Moin Robert
Thank you very much for this tipp! Mein Retter des Tages. It's the first time I wrote code with DMA from scratch and I was not aware of that. I was checking the documentation beforehand but obviously missed bswap or just forgot it again. So pack_ctrl with bswap helped me, solve the problem. I was not aware, that changing the out order out of DMA was possible with bswap. So now everything is working as it should. And yes, I also tried push_tresh= 8, but with ISR always pushing 32 bit to RX FIFO anyway, I would later need one more step to get rid of 24 bits. So push_tresh= 32 is the "easier" way, as I need 480 pushes with 32 bits anyway (15'360 bits per cycle, 120 cycles per second). With push_tresh= 8 I would need to deal with getting rid of 24 unwanted bits 230'400 times per second (15'360 bits / 8 bit = 1'920 pushes; 1920 * 120 cycles =230'400 pushes per second). And getting the byte and bit-order right from the beginning is also easier, because afterwards I can just take every byte and bit in the order it already gets out of DMA into the bytearray and I don't have to change to order before creating the image respectively flushing the data out of USB to the screen. With the speed of micropython I have made the experience, that it helps getting things done as quickly as possible, by letting PIO and DMA do as much work as possible beforehand, so the output is already as useable as possible so the CPU has as little to do as possible.
Even with just creating a framebuffer and nothing else with micropython, I have already to skip cycles and just create every third frame (so screen output is 40 Hz and not 120 Hz).
Thank you very much!