Adding "optimization-passes" option

[trace-andreason]

Addressing issue #169

[ascjones]

I'd like to see this restricted to the valid values, perhaps by introducing an enum type. Also if you look at wasm-opt --help there are a couple of non integer values possible:

 -O0                                          execute no optimization passes 
   -O1                                          execute -O1 optimization passes
                                                (quick&useful opts, useful for 
                                                iteration builds) 
   -O2                                          execute -O2 optimization passes
                                                (most opts, generally gets most
                                                perf) 
   -O3                                          execute -O3 optimization passes
                                                (spends potentially a lot of 
                                                time optimizing) 
   -O4                                          execute -O4 optimization passes
                                                (also flatten the IR, which can
                                                take a lot more time and memory,
                                                but is useful on more nested / 
                                                complex / less-optimized input)
   -Os                                          execute default optimization 
                                                passes, focusing on code size 
   -Oz                                          execute default optimization 
                                                passes, super-focusing on code 
                                                size 

[trace-andreason]

@ascjones ahhh, I had assumed it was limited to integers as it was typed that way in the do_optimization function. I'll give the enum a try.

[trace-andreason]

oh, if we wanted to go the enum route, it might create some problems with the binaryen-as-dependency optionsfor installation which seems to only take a u32 for optimization passes: https://docs.rs/binaryen/0.12.0/src/binaryen/lib.rs.html#21-28

[ascjones]

You can always convert from the enum to the u32 at the call site. Would be worth finding out if 5 -> s and 6 -> z. At the very least we should validate the input value so it can't just be any u32.

[trace-andreason]

did a little more digging on the binaryen crate:

/// Codegen configuration.
#[derive(Default)]
pub struct CodegenConfig {
    /// 0, 1, 2 correspond to -O0, -Os, -Oz
    pub shrink_level: u32,
    /// 0, 1, 2 correspond to -O0, -O1, -O2, etc.
    pub optimization_level: u32,
    /// If set, the names section is emitted.
    pub debug_info: bool,
}

I made the shrink options default to 3 optimization passes in the case of binaryen as a dependency

[ascjones]

Awesome, I'd like @cmichi to take a look at this too since he's more familiar with the wasm-opt integration

[cmichi]

Right now this is the cli help output:

$ cargo run -- contract build --help
…
        --optimization-passes <optimization-passes>    
             [default: 3]

What I would like to see is either

• a note with possible params and a sentence noting that they are simply forwarded to wasm-opt or
• a list of all possible values with each ones description (like the list which @ascjones posted above).

[trace-andreason]

@cmichi added. Here is what the new cli help output looks like:

        --optimization-passes <optimization-passes>
            number of optimization passes, passed as an argument to wasm-opt
            
            - `0`: execute no optimization passes
            
            - `1`: execute 1 optimization pass (quick & useful opts, useful for iteration builds)
            
            - `2`, execute 2 optimization passes (most opts, generally gets most perf)
            
            - `3`, execute 3 optimization passes (spends potentially a lot of time optimizing)
            
            - `4`, execute 4 optimization passes (also flatten the IR, which can take a lot more time and memory but is
            useful on more nested / complex / less-optimized input)
            
            - `s`, execute default optimization passes, focusing on code size
            
            - `z`, execute default optimization passes, super-focusing on code size
            
            - [default: 3]

[cmichi]

Thanks @trace-andreason !