Experiment with some wasm-opt options

[Robbepop]

We currently are not using wasm-opt's --strip-debug --vacuum --dce options. They might be useful to us to reduce the resulting .wasm files. Experimentation needed.

[cmichi]

In that regard experimenting with wasm-opt optimization flags could also have some benefits. Currently we just pass 3 by default. If you execute cargo +nightly contract build --help you'll see the other possibilities.

In general we should review the other flags from wasm-opt --help as well. Could be that there are some additional ones in there which have a big impact.

[HCastano]

I've been playing around a bit with some of the optimizations options using some (but not
all) of our example contracts.

For reference, the original sizes with ink! on hc-bump-allocator and cargo-contract on master
are in the table below.

Contract	Original	Optimized
erc20	51.8K	30.5K
trait-erc20	52.3K	31.0K
erc1155	65.2K	45.3K
erc721	68.3K	41.1K
dns	44.2K	25.3K
multisig_plain	86.2K	53.6K

1. Setting cargo-contract to --optimization-passes s
   This saves us ~0.2-0.3K.

2. Setting cargo-contract to --optimization-passes z
   Same as (1).

3. Adding --vaccum option to wasm-opt
   This increases the size by ~0.2-0.3K.

4. Adding --strip-debug option to wasm-opt
   Same as (3).

5. Adding --dae-optimizing option to wasm-opt
   Same as (3).

I think right now changing the opt level default from -O3 to at least -Os is the most
sensible thing to do.

[athei]

There are a lot of optimization options. I have no idea which are on by default and which are not. Would be nice to find that out. I guess the source code should reveal this information.

For example, --strip-debug is on by default. To keep the debug symbols -g is needed.

[cmichi]

If nobody speaks up in the next days I'll close this issue. We've already set

-Oz  execute default optimization passes, super-focusing on code size

as a consequence of this issue and it lead to some savings.

We've done some experimenting of individual flags which seemed reasonable, but no further big finding was found. There are a lot more options we could try:

$ wasm-opt --help | grep "\-\-" | wc -l
167

But it seems that the potential of finding another big thing is low, whereas we have a number of ideas in paritytech/substrate#9354 now, which will definitely lead to significant size improvements. So I think it's reasonable to rather focus our effort there.