Simplify boolops

[charles-cooper]

What I did

Removed unnecessary goto and seq_unchecked from bool ops. Also has the added benefit of making the code more efficient. The following code loses 6 bytes worth of bytecode:

@external
def _and(x: bool, y: bool, z: bool) -> bool:
    return x and y and z

@external
def _or(x: bool, y: bool, z: bool) -> bool:
    return x or y or z

The difference in IR for the x or y or z expression is

before:

[mstore,
  0,
  [seq_unchecked,
    [if,
      [calldataload, 4 <x>],
      [1, [goto, _boolop_146:11:0]],
      [if,
        [calldataload, 36 <y>],
        [1, [goto, _boolop_146:11:0]],
        [calldataload, 68 <z>]]],
    [label, _boolop_146:11:0]]],

after:

[mstore,
  0,
  [if,
    [iszero, [calldataload, 4 <x>]],
    [if, [iszero, [calldataload, 36 <y>]], [calldataload, 68 <z>], 1],
    1]],

The difference for the x and y and z expression is similar.

How to verify it

Tests still pass, also compile the above example

Description for the changelog

Simplify code generation for boolean operations

Cute Animal Picture

[codecov-commenter]

Codecov Report

Merging #2467 (ecbf63f) into master (39036fb) will decrease coverage by 0.16%.
The diff coverage is 100.00%.

❗ Current head ecbf63f differs from pull request most recent head 17c1d6e. Consider uploading reports for the commit 17c1d6e to get more accurate results

@@            Coverage Diff             @@
##           master    #2467      +/-   ##
==========================================
- Coverage   84.52%   84.35%   -0.17%     
==========================================
  Files          93       93              
  Lines        9228     9271      +43     
  Branches     2085     2102      +17     
==========================================
+ Hits         7800     7821      +21     
- Misses        939      954      +15     
- Partials      489      496       +7     

Impacted Files	Coverage Δ
vyper/lll/optimizer.py	82.80% <100.00%> (+0.56%)	⬆️
vyper/old_codegen/expr.py	79.10% <100.00%> (-0.17%)	⬇️
vyper/old_codegen/parser_utils.py	79.46% <0.00%> (-3.60%)	⬇️
vyper/exceptions.py	92.59% <0.00%> (-0.07%)	⬇️
vyper/ast/nodes.py	93.58% <0.00%> (ø)
vyper/cli/utils.py	92.10% <0.00%> (ø)
vyper/ast/natspec.py	98.76% <0.00%> (ø)
vyper/evm/opcodes.py	100.00% <0.00%> (ø)
vyper/ast/expansion.py	94.11% <0.00%> (ø)
... and 40 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 39036fb...17c1d6e. Read the comment docs.

[fubuloubu]

Does this still retain short-circuiting properties?

[charles-cooper]

FYI - this introduces an extra jump for the nested operations, each block jumps out to the enclosing block. For the most nested block it will do something like

if ... {
  if ... {
    if ... {
      goto if_exit1
    }
    label if_exit1
    goto if_exit2
  }
  label if_exit2
  goto if_exit3
}
label if_exit3

We can analyze and optimize it out of the asm. Basically if the destination of a jump is another jump, we just jump to the final destination. But maybe I'll include that analysis in this PR before it's merged

[charles-cooper]

Does this still retain short-circuiting properties?

And yes. If we don't have a test for that, we should

[fubuloubu]

Does this still retain short-circuiting properties?

And yes. If we don't have a test for that, we should

I'm pretty sure we do, I just don't know where.

[charles-cooper]

We can analyze and optimize it out of the asm. Basically if the destination of a jump is another jump, we just jump to the final destination. But maybe I'll include that analysis in this PR before it's merged

done in f6155f7