Peek command seems to alter memory

[econtrerasd]

Version:
BASIC816 v0.2.0-alpha+274

Description:
When you peek at a memory address (at least in the graphics area memory &hb0000) the memory position at the location+4 that you peek is zeroed.
For a graphical demo of this behavior the provided code shows sprite 1 on screen and then proceeds to poke 1s to create a solid sprite. Just after this the code peeks at all its values.
You'll see the sprite appear and dissapear magically!

Code:

5 CLS
10 GRAPHICS 47
20 spb%=&hb00000: REM sprite block constant
30 SPRITE 1,0,spb%
40 SPRITEAT 1,100,100
50 SPRITESHOW 1,1
60 FOR c%=0 to 1024:POKE spb%+c%,1:NEXT
70 FOR c%=0 to 1024:a%=PEEK(spb%+c%):NEXT

[pweingar]

Interesting. It's not doing it on my machine. What version of the C256 are you using? I doubt the issue is PEEK itself, but it may have to do with how the CPU accesses video memory, which has to go through Vicky. I may have to modify how PEEK works when the address is in video memory.

[econtrerasd]

That makes sense, I have a Foenix U
Kernel Date March 13th, 2021 v.0.3.0-alpha+1371

I'm sharing a video so you can see what happens when I execute the code

Err.34.mp4

[econtrerasd]

Update:
I tried poking & peeking on a different memory address (&h100000) and can confirm that peek does not modify memory (you get 1s printed).
If you change the address to &hb00000 in the two loops you get all zeroes

Code:

5 CLS
10 FOR c%=0 to 100:POKE &h100000+c%,1:NEXT
20 FOR c%=0 to 100:a%=PEEK(&h100000+c%):PRINT a%;:NEXT

[pweingar]

Thanks. I've not been able to see that on the U+, but I am probably using a slightly different version of the FPGA. I do see that only zeros comes back. I have looked into how to use the Vicky FIFO to read data, and I think I have a fix. I'll need to apply it to all the PEEK variants.