RP2XXX USB - replace rom.memcpy with @memcpy

[arkadiuszwojcik]

In commit 68b5ad1 rom.memcpy was replaced by @memcpy by @patryk4815 but one more rom.memcpy left.

[patryk4815]

I think this one rom.memcpy need be left, maybe you want add comment?
Or new zig0.14.0 fixed @memcpy?
On zig0.13.0 @memcpy crash there

[arkadiuszwojcik]

Hmm intresting. @patryk4815 can you tell me if USB CDC example from fresh git clone of microzig and zig build works for you on RP2040? Because for me it doesn't but witch that change from this commit it does. Very bizarre.

[patryk4815]

I have only rp2350, soo I will test this today or tomorrow

[Grazfather]

Why do we want to do this? Don't we think that the rom version will be optimized for the device?

[arkadiuszwojcik]

Why do we want to do this? Don't we think that the rom version will be optimized for the device?

In other place it was already replaced from rom.memcpy to @memcpy, so this change is to keep consistency. But, probably better solution would be to utilze comptime and use rom.memcpy for rp2040 and @memcpy for rp2350

[Grazfather]

Oh I did not know that they are not in the rp2350 rom, though I just checked (they are there but not exported)

[KaroLaunonen]

At least I needed this change for RP2040 usb-hid example on zig 0.14.0. It would crash without the change.

[Grazfather]

I tried the CDC example, and it seems to work with this change. Without it I couldn't get the device to enumerate, but when connecting with jtag I don't see anything crashing.

❯ sudo cu -l /dev/tty.usbmodem1101 -s 115200
Connected.
This is very very long text sent from RP Pico by USB CDC to your device: 70
This is very very long text sent from RP Pico by USB CDC to your device: 71
Your message to me was: h
Your message to me was: e
Your message to me was: l
This is very very long text sent from RP Pico by USB CDC to your device: 72
Your message to me was: l
Your message to me was: o
This is very very long text sent from RP Pico by USB CDC to your device: 73
This is very very long text sent from RP Pico by USB CDC to your device: 74
Your message to me was:
~This is very very long text sent from RP Pico by USB CDC to your device: 75
.
Disconnected.

[Grazfather]

Can someone test this on rp2350? If it works, we can make rom.memcpy for rp2350 make a compile error or just @memset.

I don't think we should have rom.<anything> secretly call builtins, I'd rather have these functions in the HAL, and have those functions determine whether they call a builtin or a rom function.

[arkadiuszwojcik]

@Grazfather I am very confused right now. Without this change RP2350 is working and RP2040 is not, but with this change it is other way around: RP2350 stops working and RP2040 starts working :D Just tested on both devices. This is very odd issue.
@patryk4815 any news from your side?

[Grazfather]

If we move the memcpy to some chip.mempy, we can have a dispatch there that does @memcpy on rp2040 and
std.mem.copyForwards on rp2350. This could be done in rom but that would be misleading.

[arkadiuszwojcik]

@Grazfather I changed solution to use std.mem.copyForwards and created issue for this problem, leaving that in code comment. Tested on both platforms RP2040 and RP2350 - it works.