fix my sloppy oversight of asUint8 missing unchecked suffix in asBool…

[nonergodic]

I made sloppy mistakes in my proposal here by using asUint8 instead of asUint8Unchecked.

Additionally, as it turns out, the inline assembly is not safe because solc is actually no cleaning up the upper 31 bytes of val (as I expected it would), so there's no way to avoid the cost of the two ISZERO opcodes after all. I missed that because it happened to insert cleanup instructions in my Playground example, which I believed to be equivalent but then it wasn't...

[kcsongor]

if val is not cleaned up then val != 0 (assuming it only operates on the last byte, otherwise it would be incorrect) involves a mask in addition to the ISZERO check right? we might as well do the cleaning mask and keep the assembly assignment to at least save the ISZERO check

[kcsongor]

also, if solc doesn't assume that the variables are cleaned up anyway, then what's wrong with returning a bool that's not cleaned up?
(maybe there's an assumption that functions return cleaned words? I need to look this up)

[nonergodic]

if val is not cleaned up then val != 0 (assuming it only operates on the last byte, otherwise it would be incorrect) involves a mask in addition to the ISZERO check right? we might as well do the cleaning mask and keep the assembly assignment to at least save the ISZERO check

That's a good point, I hadn't considered that the cleaning will also require masking. My first way to fix it was

uint valAsUint = uint(val);
bool ret;
assembly ("memory-safe") {
  ret := valAsUint
}

But I figured not only was it less readable (and more error prone) but also 6 gas (1 PUSH, 1 AND) just like 2 ISZERO (and 3 bytes in deployed opcode (PUSH 0xFF AND) vs just 2).

But given what you're saying, perhaps this is the preferred impl after all?

[nonergodic]

Ok, pushed the above implementation (tested it to make sure it works as intended, but didn't look at the assembly)