[cc] Compose ActionRecorder outputs into a single C file for Emscripten

[archibate]

Related issue = #

[Click here for the format server]

---

Usage e.g.:

TI_ACTION_RECORD=mpm88.yml TI_ARCH=cc python examples/mpm88.py
ti cc_compose mpm88.yml mpm88.c
emcc mpm88.c -o mpm88.js

Then write a mpm88.html that includes taichi.js and mpm88.js and the following inline scripts:

let taichi = new Taichi(Module);
let N = 8192;
let RES = 512;

taichi.ready(function() {

    let canvas = document.getElementById('canvas');
    let gui = new TaichiGUI(canvas, RES);

    let matrix_to_ext_arr = taichi.get('matrix_to_ext_arr_c20_0');
    let substep = taichi.get('substep_c4_0');
    let init = taichi.get('init_c6_0');

    init();

    gui.animation(function() {
        for (let i = 0; i < 5; i++) {  // fps extreme when 50
            substep();
        }
        let extr = taichi.set_ext_arr(0, [N, 2]);  // index, array shape
        matrix_to_ext_arr();

        gui.circles(extr);  // updated, since extr is a ref
    });
});

See https://github.com/taichi-dev/taichi.js for client, please run python server.py to set up a simple http server, it won't work in file:///.
Btw, I forget to grant myself with admin permission while transfer... LOL, @yuanming-hu Would you mind restore me with admin and make the taichi.js repo to be public so that people could share it?

[codecov]

Codecov Report

Merging #1629 into master will decrease coverage by 0.70%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #1629      +/-   ##
==========================================
- Coverage   45.00%   44.29%   -0.71%     
==========================================
  Files          40       41       +1     
  Lines        5739     5858     +119     
  Branches     1002     1017      +15     
==========================================
+ Hits         2583     2595      +12     
- Misses       3002     3116     +114     
+ Partials      154      147       -7     

Impacted Files	Coverage Δ
python/taichi/cc_compose.py	0.00% <0.00%> (ø)
python/taichi/core/util.py	0.29% <0.00%> (-0.01%)	⬇️
python/taichi/main.py	23.73% <0.00%> (-0.40%)	⬇️
python/taichi/testing.py	76.92% <ø> (ø)
python/taichi/lang/__init__.py	54.75% <0.00%> (-0.46%)	⬇️
python/taichi/misc/gui.py	0.00% <0.00%> (ø)
python/taichi/lang/kernel.py	68.71% <0.00%> (ø)
python/taichi/lang/impl.py	66.18% <0.00%> (+0.09%)	⬆️
python/taichi/lang/matrix.py	67.47% <0.00%> (+0.15%)	⬆️
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 81437d6...5ec8dc6. Read the comment docs.

[yuanming-hu]

Great! Just some minor issues that need a little clarification.

[yuanming-hu]

There are still people using information like set_kernel_arg_int64. Can we keep these for now?

[archibate]

Well, it doesn't even include dtype... They are just spams in our compile-only no-launch action parser, both in CC and [数据删除] in my pratice...

[yuanming-hu]

But there are still people depending on this. I don't think we should remove them immediately. Maybe a better solution is to mark them as a different type of message.

[archibate]

I thought the feature is purposely undocumented and used by only you and me and the [数据删除]? Also did you confirmed this doesn't harm performance? IMO we'd remove all the run-time action output iapr, since only compile-time kernel information are required for reconstructing a Taichi program outside, e.g., this PR, did that easily.

[yuanming-hu]

Thanks, much better now. I still suggest keeping the kernel argument information.

[yuanming-hu]

But there are still people depending on this. I don't think we should remove them immediately. Maybe a better solution is to mark them as a different type of message.

[yuanming-hu]

LGTM. Thank you so much!