Branch optimization

[Robertorosmaninho]

FIX: #1588

This PR implements a branch continuation optimization. Previously, we were popping values in the <instrs> cell once we found a br instruction without a label after it, and once we found a label if the IDX of the br instruction wasn't 0, we rewrote it until reaching the base case. This was fairly time-consuming and inefficient.

This new optimization addresses the two issues by sacrificing space efficiency. Here, we introduce a new K list cell, labels, which contains the whole state continuation of a block. So, right now, instead of popping instructions, we just replace the whole <instr> cell, and instead of rewriting the br instruction until it reaches the base case, we just use the list index to "jump" to the correct block and pop the others using the range hook.

So, in programs that don't have many changes in control flow, we may have some slowdown. However, for programs that contain such cases, we expect to see some great speedups.

[dwightguth]

This needs evidence that it's faster on a simple program with some huge basic blocks.

[dwightguth]

Please rebase this on the latest master so I can more effectively review it.

[Robertorosmaninho]

Please rebase this on the latest master so I can more effectively review it.

Done!

[Robertorosmaninho]

For some reason that I couldn't spot easily, after my changes, call.wast started to fail in only one test: (assert_return (invoke "as-unary-operand") (f32.const 0x0p+0)) with:

poetry -C pykwasm run -- kwasm run tests/wasm-tests/test/core/call.wast > tests/wasm-tests/test/core/call.wast.llvm-out
/home/robertorosmaninho/rv/k/k-distribution/bin/../lib/kframework/k-util.sh: line 114: 1037198 Bus error               (core dumped) "$@"
[Error] krun: 
/home/robertorosmaninho/.cache/kdist-efbf2fa/wasm-semantics/llvm/interpreter 
/tmp/.krun-2024-08-08-21-47-27-pYn6dr884r/tmp.in.PMv6wHOLHm -1 
/tmp/.krun-2024-08-08-21-47-27-pYn6dr884r/result.kore
Syntax error at /tmp/.krun-2024-08-08-21-47-27-pYn6dr884r/result.kore:1.135162-135168: Expected: [:, {] Actual: <EOF>
[Error] krun: kore-print --definition 
/home/robertorosmaninho/.cache/kdist-efbf2fa/wasm-semantics/llvm --output 
pretty /tmp/.krun-2024-08-08-21-47-27-pYn6dr884r/result.kore --color off
make: *** [Makefile:51: tests/wasm-tests/test/core/call.wast.run-term] Error 255

[dwightguth]

This looks about what I expected. It's unfortunate that this doesn't end up helping performance, but I guess it's good that the existing code wasn't that bad in the first place.