bare1.rs

//! bare1.rs
//!
//! Tracing and assembly code
//!
//! What it covers
//! - tracing over semihosting and ITM
//! - assembly calls and inline assembly
//! - more on arithmetics
//! ---

#![no_main]
#![no_std]

extern crate panic_halt;

use cortex_m::{iprintln, Peripherals};
use cortex_m_rt::entry;
use cortex_m_semihosting::hprintln;

#[entry]
fn main() -> ! {
    let mut p = Peripherals::take().unwrap();
    let stim = &mut p.ITM.stim[0];

    let mut x = 27;
    iprintln!(stim, "x = {:?}", x);
    hprintln!("x = {:?}", x).unwrap();
    loop {
        x += 1;
        cortex_m::asm::nop();
        cortex_m::asm::bkpt();
        x -= 1;
    }
}

// 0. Setup
//    For this example we will use the `nightly` compiler
//    to get true inline assembly.
//
//    > rustup override set nightly
//
//    You may need/want to install addititonal components also),
//    to that end look at the install section in the README.md.
//    If you change toolchain, exit and re-start `vscode`.
//
//    1. Build and run the application
//    Look at the `hello.rs` and `itm.rs` examples to setup the tracing.
//
//    When debugging the application it should get stuck in the
//    loop, (press pause/suspend to verify this).
//    what is the output in the ITM console
//
//    ** your answer here **
//
//    What is the output in the semihosting (openocd) console
//    ** your answer here **
//
//    commit your answers (bare1_1)
//
// 2. Inspecting the generated assembly code
//    If in `vcsode` the gdb console in DEBUG CONSOLE
//    What is the output of:
//    (gdb) disassemble
//
//    ** your answer here **
//
//    commit your answers (bare1_2)
//
// 3. Now remove the comment for `cortex_m::asm::nop()`.
//    Rebuild and debug, pause the program.
//    What is the output of:
//    (gdb) disassemble
//
//    ** your answer here **
//
//    commit your answers (bare1_3)
//
// 4. Now remove the comment for `cortex_m::asm::bkpt()`
//    Rebuild and debug, let the program run until it halts.
//    What is the output of:
//    (gdb) disassemble
//
//    ** your answer here **
//
//    commit your answers (bare1_4)
//
// 5. Release mode (optimized builds).
//    Rebuild `bare1.rs` in release (optimized mode).
//    Compare the generated assembly for the loop
//    between the dev (unoptimized) and release (optimized) build.
//
//    ** your answer here **
//
//    commit your answers (bare1_5)
//
//    Tips: The optimized build should have some 8 instructions
//    while the debug (dev) build should have > 20 instructions
//    (both counting the inner loop only). The debug build
//    should have additional code that call panic if the additon
//    wraps (and in such case call panic).
//
//    Discussion:
//    In release (optimized) mode the addition is unchecked,
//    so there is a semantic difference here in between
//    the dev and release modes. This is motivited by:
//    1) efficiency, unchecked is faster
//    2) convenience, it would be inconvenient to explicitly use
//    wrapping arithmetics, and wrapping is what the programmer
//    typically would expect in any case. So the check
//    in dev/debug mode is just there for some extra safety
//    if your intention is NON-wrapping arithmetics.
//
// 6. *Optional
//    You can pass additional flags to the Rust `rustc` compiler.
//
//    `-Z force-overflow-checks=off`
//
//    Under this flag, code is never generated for oveflow checking.
//    You can enable this flag (uncomment the corresponding flag in
//    the `.cargo/config` file.)
//
//    What is now the disassembly of the loop:
//
//    ** your answer here **
//
//    commit your answers (bare1_6)
//
//    Now restore the `.cargo/config` to its original state.
//
// 7. *Optional
//    There is another way to conveniently use wrapping arithmetics
//    without passing flags to the compiler.
//
//    https://doc.rust-lang.org/std/num/struct.Wrapping.html
//
//    Rewrite the code using this approach.
//
//    What is now the disassembly of the code in dev mode?
//
//    ** your answer here **
//
//    What is now the disassembly of the code in release mode?
//
//    ** your answer here **
//
//    commit your answers (bare1_7)
//
//    Final discussion:
//
//    Embedded code typically is performance sensitve, hence
//    it is important to understand how code is generated
//    to achieve efficient implementations.
//
//    Moreover, arithmetics are key to processing of data,
//    so its important that we are in control over the
//    computations. E.g. comupting checksums, hashes, cryptos etc.
//    all require precise control over wrapping vs. overflow behaviour.
//
//    If you write a library depending on wrapping arithmetics
//    do NOT rely on a compiler flag. (The end user might compile
//    it without this flag enabled, and thus get erronous results.)