eclic: move eclic-mode-hack.S into Rust#55
Conversation
| ".weak TIMER0_UP", | ||
| ".weak TIMER0_TRG_CMT", | ||
| ".weak TIMER0_CHANNEL", | ||
| ".weak TIMER1", |
There was a problem hiding this comment.
Raises a warning when compiling longan-nano/examples/interrupt
warning: TIMER1 changed binding to STB_WEAK
|
note: instantiated into assembly here
--> <inline asm>:33:1
|
33 | .weak TIMER1
The timer interrupt fires, but only once at the beginning of execution.
Not sure how to solve this...
Using global_asm raises a TIMER1 changed binding to STB_GLOBAL error, and the build fails.
There was a problem hiding this comment.
This part and vectors should be global_asm because declaring it as a function may add a few instructions in the beginning and all the handler addresses will become shifted in this case. In any case, this is not code, so should not be callable.
There was a problem hiding this comment.
This part and vectors should be global_asm because declaring it as a function may add a few instructions in the beginning and all the handler addresses will become shifted in this case.
I completely agree, but when I tried to make it global_asm it raises an error in the longan-nano example code:
error: TIMER1 changed binding to STB_GLOBAL
That error occurs both with the interrupt macro from gd32vf103-pac, and the current code from the master branch.
Do you think it would help to implement a hand-written proc-macro, similar to what's in cortex-m and avr-device?
There was a problem hiding this comment.
Something is strange here. I got different results for debug and release builds:
% nm target/riscv32imac-unknown-none-elf/debug/examples/interrupt | grep " TIMER1"
08000568 T TIMER1
% nm target/riscv32imac-unknown-none-elf/release/examples/interrupt | grep " TIMER1"
080001c2 W TIMER1
It's ok when STB_GLOBAL is used (like in debug here), but STB_WEAK looks wrong. Maybe each interrupt handler needs a default implementation like was done in stm32f1, for example. It provides device.x with a lot of lines of the following format PROVIDE(TIM4 = DefaultHandler); and then an array is created with all the handlers: https://docs.rs/stm32f1/latest/src/stm32f1/stm32f103/mod.rs.html#82-177
There was a problem hiding this comment.
Moved everything into a global_asm block, and use a device.x file in the longan-nano example.
Using the device.x file to define the default handlers clears the STB_WEAK/GLOBAL errors, but the blinky still stays static (constantly on after 1 second). Not sure what is causing the error.
Edit: Maybe the lack of ilp32 calling convention/ABI? IDK, but I'm putting this work on hold for now. There doesn't seem to be much benefit (it's still assembly, just using the Rust/cargo build system). Ultimately, I think waiting for something like rust-lang/rust#111891 to land would probably be better.
There was a problem hiding this comment.
I got to the same point as you: the interrupt fires, but execution doesn't resume afterwards. What do you mean about the ABI? Why would it influence the resumption? I thought hardware was responsible for jumping to th right piece of code.
There was a problem hiding this comment.
I found an oddity in _irq_handler, where msubm was replaced with pushmsubm compared to the working version:
// Load x5 value into PUSHMSUBM system status register
"csrw 0x7eb, x5",
but changing it doesn't make interrupts finish.
There was a problem hiding this comment.
I managed to get the interrupt to finish by using vectored mode. The interrupt won't trigger again and I'm kind of out of ideas.
There was a problem hiding this comment.
It's not triggering again because RTC alarm counter wraps after several days. Duh.
See #71
Updates cargo dependencies, and removes the deprecated `bare_metal::Nr` trait. The `bare_metal::Nr` is no longer present in the latest version, and all usage in this HAL are easily replaced by directly converting the `Interrupt` enum to a base integer type.
Adds `.option +zicsr` to `eclic-mode-hack.S` to fix a failure to assemble with recent versions of `riscv64-unknown-elf-gcc` (e.g. 12.2.0).
Updates the precompiled `bin/gd32vf103xx-hal.a` part of the library.
rmsyn
force-pushed
the
eclic-asm
branch
2 times, most recently
from
d210ecf to
51a9fbd
Compare
Uses inline assembly to move the code from `eclic-mode-hack.S` into Rust. The generated assembly is different. This commit should be reverted if regressions are noticed on real hardware.
3 participants
Uses inline assembly to move the code from
eclic-mode-hack.Sinto Rust.The generated assembly is different. This commit should be reverted if regressions are noticed on real hardware.