Hey there 🤗, Welcome to my GSoC Journey

This is a series of blogs that I am writing to share my amazing GSoC 22 (Google Summer of Code 2022) Journey with you all.

Extending WASM and WAT Backends

In this phase, I focused on extending the WASM and WAT Backends. Specifically, I added support for

  • i64 constants/literals
  • unary minus for i32 and i64
  • floating points f32 and f64
    • constants/literals
    • unary minus
    • basic arithmetic operations

I also refactored the WASM and WAT Backends to hopefully improve them.

A challenge I faced during this phase was during the decoding the i64 constants in the WASM->WAT backend. I was following Wikipedia/LEB128 for the encoding and decoding of u32 and i32 constants. I followed the same for i64 constants for encoding as well as decoding. Even after following the same/similar algorithm as given in Wikipedia/LEB128, the decoding was still not working correctly. After much research, I found this which helped me find out the bug.

The bug was in the following code:

byte = next byte in input;
result |= (byte & 0x7f) << shift;
shift += 7;

We see that in the middle line in the above code, we have (byte & 0x7f) << shift. Since byte is of 8 bits and 0x7f is of 32 bits, the result will be of 32 bits and this is fine when shift < 32. Therefore, this code works for i32 and u32, but did not work for i64 as we might/would lose bits once the shift which might be >= 32 is applied .

The following code fixes the above issue:

byte = next byte in input;
uint64_t slice = (byte & 0x7f);
result |= slice << shift;
shift += 7;

This phase was mostly/approximately from 13-06-2022 to 17-06-2022.

The MRs during this phase are as follows:

That’s all for this blog. Thank you for your time. We continue this series in the next blog.