small wasm files example is actually huge!

[IcanDivideBy0]

While the documentation claims the add_bg.wasm would be 710 byte, it's more like 700KB

Either using the online demo or building my own with --release flag produces a huge binary

Current demo file: https://rustwasm.github.io/wasm-bindgen/exbuild/add/23a8ffdd74ff21744bf0.module.wasm

Putting it into wasm2wat show that a huge zero filled data string is embeded inside the wasm file

rustc 1.34.0-nightly (f47ec2ad5 2019-02-14)

[alexcrichton]

Thanks for the report! The extra size here is dwarf debugging information which is typically stripped by default in release builds but we has a misconfiguration that was fixed recently in #1255 and deployed pretty recently as well. The examples haven't fully rebuilt and redeployed yet but they'll be taken care of soon!

[IcanDivideBy0]

Updated to 0.2.37, everything looks good!
As always, thanks for your reactivity :)

[jlb6740]

Just ran into this ... @IcanDivideBy0 what did you update to correct this? @alexcrichton looks like the fix #1255 is merged right? I am using 0.2.38 version wasm-bindgen but the issue is still there:

Compiling proc-macro2 v0.4.27
Compiling unicode-xid v0.1.0
Compiling wasm-bindgen-shared v0.2.38
Compiling cfg-if v0.1.7
Compiling lazy_static v1.3.0
Compiling wasm-bindgen v0.2.38
Compiling log v0.4.6
Compiling quote v0.6.11
Compiling syn v0.15.27
Compiling wasm-bindgen-backend v0.2.38
Compiling wasm-bindgen-macro-support v0.2.38
Compiling wasm-bindgen-macro v0.2.38

[IcanDivideBy0]

@jlb6740 Simply updated wasm-bindgen to 0.2.37
did you compiled with the --release flag ?
also running the binaryen's utility wasm-opt can help a lot

[alexcrichton]

@jlb6740 in addition to @IcanDivideBy0's suggestions, can you gist an example to reproduce what you're seeing? That might help us debug!

[jlb6740]

Hi .. Yes thanks @IcanDivideBy0 for the suggestion. I use the release flag and the file size is much smaller than the debug, but still larger than expected (180K instead of 989K). I have an example function where wasm-opt seemed to strip out code that I didn't consider dead .. I need to look at the options for this command more carefully but I agree this is a good suggestion. For now I have been doing wasm2wat and then wat2wasm instead of using bindgen or wasm-opt. @alexcrichton I put the code I am compiling here: https://gist.github.com/jlb6740/e31cf87f0906663995e654f3dea0caf5 . I have not been working with rust for long so I wouldn't be surprised if it is something trivial I am overlooking. Thanks.

[alexcrichton]

@jlb6740 you may want to be sure to consult https://rustwasm.github.io/docs/book/reference/code-size.html when shrinking wasm, but if I check out that gist and run wasm-pack build it produces a 862 byte binary which I think is probably what you're looking for?

WebAssembly is a fundamentally low-level format that requires a lot of tools to work in concert right now so if you're just using rustc without wasm-bindgen and/or wasm-pack you'll need to reimplement the functionality that they bundle in with your workflow as well.

[IcanDivideBy0]

As @alexcrichton pointed it out, after running the wasm-bindgen cli, you .wasm file shrink down to 862 bytes!
And wasm-opt make it 228 bytes

[jlb6740]

Hi ... 862 Bytes is a lot smaller than what I see with bindgen (180K). I don't know then what is different with my environment ... seems pretty straight forward but I would like to know. I am compiling with cargo build --release.

What you get from bindgen though is still discouraging. Nnot to get side tracked, but I am getting much smaller sizes ... 170 bytes by using wasm2wat:

cd wasm_binaries; cp fib.wasm fib.wasm.orig; wasm2wat fib.wasm.orig > fib.wasm; wat2wasm fib.wasm; cd ../;

Even if ignore all the \00 , the code is not as barebones as when I exclude bindgen. Code without bindgen:

(module
(type (;0;) (func))
(type (;1;) (func (param i32) (result i32)))
(func (;0;) (type 0))
(func (;1;) (type 1) (param i32) (result i32)
(local i32)
i32.const 1
set_local 1
block ;; label = @1
get_local 0
i32.const -1
i32.add
tee_local 0
i32.const 2
i32.lt_u
br_if 0 (;@1;)
i32.const 1
set_local 1
loop ;; label = @2
get_local 0
call 1
get_local 1
i32.add
set_local 1
get_local 0
i32.const -2
i32.add
tee_local 0
i32.const 1
i32.gt_u
br_if 0 (;@2;)
end
end
get_local 1)
(table (;0;) 1 1 anyfunc)
(memory (;0;) 16)
(global (;0;) (mut i32) (i32.const 1048576))
(global (;1;) i32 (i32.const 1048576))
(global (;2;) i32 (i32.const 1048576))
(export "memory" (memory 0))
(export "__heap_base" (global 1))
(export "__data_end" (global 2))
(export "_call" (func 1)))

Code with bindgen includes a lot of references to bindgen.. including a lot of calls some placeholder import stuff which I am not sure the purpose.

@alexcrichton @IcanDivideBy0 Any comments on this later size discrepancy? What is Bindgen adding for you that is not included in my wat printout above?

[IcanDivideBy0]

cargo build --release
wasm-bindgen target/wasm32-unknown-unknown/release/fib.wasm --out-dir .
wasm-opt -Os  fib_bg.wasm -o fib_bg.wasm
stat -c %s fib_bg.wasm 
228

[IcanDivideBy0]

and here is the text format i get

(module
  (type $t0 (func (param i32) (result i32)))
  (func $f0 (type $t0) (param $p0 i32) (result i32)
    (local $l1 i32)
    (local.set $l1
      (i32.const 1))
    (if $I0
      (i32.ge_u
        (local.tee $p0
          (i32.add
            (local.get $p0)
            (i32.const -1)))
        (i32.const 2))
      (then
        (loop $L1
          (local.set $l1
            (i32.add
              (call $f0
                (local.get $p0))
              (local.get $l1)))
          (br_if $L1
            (i32.gt_u
              (local.tee $p0
                (i32.add
                  (local.get $p0)
                  (i32.const -2)))
              (i32.const 1))))))
    (local.get $l1))
  (func $_call (export "_call") (type $t0) (param $p0 i32) (result i32)
    (call $f0
      (local.get $p0)))
  (memory $memory (export "memory") 17))

[IcanDivideBy0]

Using --remove-producers-section flag, i went down to 104 bytes.
So I guess that's where the difference with the WAT size comes from.

wasm-bindgen \
  target/wasm32-unknown-unknown/release/fib.wasm \
  --remove-producers-section \
  --out-dir .

[jlb6740]

@IcanDivideBy0 Ok, cool. That actually looks like it should be fewer bytes smaller in binary than the file I am producing. I am not using wasm-opt though for the reason I mentioned. What is the size before wasm_opt .. just using bindgen? If I use wasm opt I can get the binary down further to 142 bytes instead of 170 bytes and my wat printout is also a little smaller. I definitely will try to use this after understanding the one example it appeared to break for me.

W.r.t your binary size, I suppose the WASM binary you have has some metadata that is not output here. My WASM binary size for what I posted is only 170 bytes. The size of the WASM file is critical for me because I am taking the files, dumping the file with hexdump and then embedding this as a data array in a tiny footprint interpreter I am running. I am curious now between bindgen vs wasm-opt w.r.t. the wat file you show and which is having the greater impact on wat size. I am also curious if I need to keep running wasm2wat and wat2wasm to get a small binary or if either bindgen or wasm-opt can get rid of whatever data is making your binary larger than it needs to be for my use case.

[jlb6740]

@IcanDivideBy0 lol .. cool. We had the same question. I will give this a try as soon as I can get past my first size problem and get down to 228bytes
.

[JeanMertz]

I'm seeing something similar, with latest versions. On top of seeing zero-filled data, I also see some assertion failure data creeping in the binary:

use wasm_bindgen::prelude::*;

#[wasm_bindgen(start)]
pub fn run() -> Result<(), JsValue> {
    Ok(())
}

$ cargo build --release
$ wasm-bindgen --version
wasm-bindgen 0.2.47
$ wasm-bindgen --remove-producers-section --remove-name-section --out-dir . target/wasm32-unknown-unknown/release/test.wasm

(module
  (type (;0;) (func))
  (func (;0;) (type 0))
  (memory (;0;) 17)
  (export "memory" (memory 0))
  (export "run" (func 0))
  (export "__wbindgen_start" (func 0))
  (data (;0;) (i32.const 1048576) "\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00")
  (data (;1;) (i32.const 1049152) "\00\00\00\00assertion failed: `(left == right)`\0a  left: ``,\0a right: ``\00\00D\02\10\00-\00\00\00q\02\10\00\0c\00\00\00}\02\10\00\01\00\00\00/Users/jean/.cargo/registry/src/git.luolix.top-1ecc6299db9ec823/wasm-bindgen-0.2.47/src/lib.rs\00\00\98\02\10\00Z\00\00\003\04\00\00\09\00\00\00\03\00\00\00\04\00\00\00\04\00\00\00\04\00\00\00\05\00\00\00\06\00\00\00\07\00\00\00\00\00\00\00\01\00\00\00\08\00\00\00called `Option::unwrap()` on a `None` valuesrc/libcore/option.rs,\03\10\00+\00\00\00W\03\10\00\15\00\00\00[\01\00\00\15\00\00\00\09\00\00\00\10\00\00\00\04\00\00\00\0a\00\00\00\0b\00\00\00\0c\00\00\00\0c\00\00\00\04\00\00\00\0d\00\00\00src/liballoc/raw_vec.rscapacity overflow\bf\03\10\00\11\00\00\00\a8\03\10\00\17\00\00\00\ec\02\00\00\05\00\00\00\10\00\00\00\00\00\00\00\01\00\00\00\11\00\00\00index out of bounds: the len is  but the index is \00\00\f8\03\10\00 \00\00\00\18\04\10\00\12\00\00\00called `Option::unwrap()` on a `None` valuesrc/libcore/option.rs<\04\10\00+\00\00\00g\04\10\00\15\00\00\00[\01\00\00\15\00\00\00src/libcore/slice/mod.rsslice index starts at  but ends at \00\ac\04\10\00\16\00\00\00\c2\04\10\00\0d\00\00\00\94\04\10\00\18\00\00\00\01\0a\00\00\05\00\00\000x00010203040506070809101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899\00\00\00\00\00\00src/libcore/fmt/mod.rs\00\00\c0\05\10\00\16\00\00\00H\04\00\00(\00\00\00\c0\05\10\00\16\00\00\00T\04\00\00\11\00\00\00"))

[package]
name = "test"
version = "0.1.0"
edition = "2018"

[lib]
crate-type = ["cdylib"]

[dependencies]
wasm-bindgen = "0.2"

$ cargo tree
test v0.1.0 (/Users/jean/code/test)
└── wasm-bindgen v0.2.47
    └── wasm-bindgen-macro v0.2.47
        ├── quote v0.6.12
        │   └── proc-macro2 v0.4.30
        │       └── unicode-xid v0.1.0
        └── wasm-bindgen-macro-support v0.2.47
            ├── proc-macro2 v0.4.30 (*)
            ├── quote v0.6.12 (*)
            ├── syn v0.15.38
            │   ├── proc-macro2 v0.4.30 (*)
            │   ├── quote v0.6.12 (*)
            │   └── unicode-xid v0.1.0 (*)
            ├── wasm-bindgen-backend v0.2.47
            │   ├── bumpalo v2.4.3
            │   ├── lazy_static v1.3.0
            │   ├── log v0.4.6
            │   │   └── cfg-if v0.1.9
            │   ├── proc-macro2 v0.4.30 (*)
            │   ├── quote v0.6.12 (*)
            │   ├── syn v0.15.38 (*)
            │   └── wasm-bindgen-shared v0.2.47
            └── wasm-bindgen-shared v0.2.47 (*)

running wasm-opt removes the zero-filled data object but the second one is still there.

[Pauan]

@JeanMertz Do you get the same result when using the rust-webpack-template?

(Make sure to use npm run build for maximum optimizations)

[alexcrichton]

@JeanMertz the issue about panic/assertion strings should be fixed by d9e53ac, it was a recent regression but hasn't been published to crates.io yet