Experimental: Optimize internal ValueStack by introducing a ValueStackPtr type

[Robbepop]

No description provided.

[paritytech-cicd-pr]

BENCHMARKS

	NATIVE			WASMTIME
BENCHMARK	MASTER	PR	DIFF	MASTER	PR	DIFF	WASMTIME OVERHEAD
execute/ bare_call_0	1.47ms	1.44ms	🟢 -2.58%	1.26ms	1.09ms	🟢 -13.36%	🟢 -24%
execute/ bare_call_0/typed	1.04ms	1.02ms	🟢 -2.39%	749.14µs	731.98µs	🟢 -2.20%	🟢 -28%
execute/ bare_call_1	1.51ms	1.48ms	🟢 -1.96%	1.43ms	1.22ms	🟢 -13.91%	🟢 -17%
execute/ bare_call_16	2.35ms	2.32ms	🟢 -1.00%	3.53ms	3.14ms	🟢 -10.95%	🟢 36%
execute/ bare_call_16/typed	1.59ms	1.57ms	🟢 -1.53%	1.64ms	1.48ms	🟢 -9.68%	🟢 -6%
execute/ bare_call_1/typed	1.16ms	1.13ms	🟢 -2.51%	1.28ms	978.80µs	🟢 -23.78%	🟢 -14%
execute/ bare_call_4	1.59ms	1.55ms	🟢 -2.37%	1.69ms	1.57ms	🟢 -6.99%	🟢 1%
execute/ bare_call_4/typed	1.15ms	1.11ms	🟢 -4.03%	1.05ms	1.01ms	🟢 -3.81%	🟢 -9%
execute/ br_table	1.19ms	1.18ms	⚪ 0.03%	1.48ms	1.20ms	🟢 -19.17%	🟢 1%
execute/ count_until	710.86µs	649.86µs	🟢 -8.53%	1.51ms	1.21ms	🟢 -19.97%	🟡 86%
execute/ factorial_iterative	315.43µs	326.42µs	🔴 3.39%	732.97µs	506.15µs	🟢 -31.03%	🟡 55%
execute/ factorial_recursive	637.66µs	652.38µs	🔴 2.28%	1.47ms	1.12ms	🟢 -23.32%	🟡 72%
execute/ fib_iterative	1.42ms	1.47ms	🔴 3.32%	3.49ms	2.45ms	🟢 -29.85%	🟡 67%
execute/ fib_recursive	5.90ms	5.93ms	⚪ 0.60%	12.15ms	9.85ms	🟢 -18.94%	🟡 66%
execute/ global_bump	1.04ms	956.97µs	🟢 -7.61%	2.54ms	2.32ms	🟢 -8.73%	🔴 142%
execute/ global_const	710.90µs	685.33µs	🟢 -3.66%	1.93ms	1.47ms	🟢 -23.73%	🔴 114%
execute/ host_calls	28.39µs	27.43µs	🟢 -3.05%	43.22µs	38.65µs	🟢 -10.37%	🟢 41%
execute/ memory_fill	1.31ms	1.20ms	🟢 -8.44%	2.94ms	2.09ms	🟢 -28.98%	🟡 74%
execute/ memory_sum	1.24ms	1.21ms	🟢 -2.33%	2.82ms	2.13ms	🟢 -24.23%	🟡 75%
execute/ memory_vec_add	2.51ms	2.38ms	🟢 -4.95%	6.58ms	4.71ms	🟢 -28.48%	🟡 98%
execute/ recursive_is_even	1.16ms	1.15ms	⚪ -1.61%	2.08ms	2.13ms	🔴 2.25%	🟡 86%
execute/ recursive_ok	156.99µs	145.27µs	🟢 -7.18%	286.08µs	287.88µs	⚪ 0.49%	🟡 98%
execute/ recursive_scan	186.26µs	176.29µs	🟢 -5.88%	359.48µs	360.05µs	⚪ 0.15%	🔴 104%
execute/ recursive_trap	14.84µs	14.32µs	🟢 -3.25%	29.85µs	29.28µs	🟢 -2.03%	🔴 104%
execute/ regex_redux	532.12µs	559.13µs	🔴 5.15%	1.25ms	1.05ms	🟢 -16.43%	🟡 87%
execute/ rev_complement	478.59µs	468.78µs	🟢 -2.17%	1.07ms	847.99µs	🟢 -21.60%	🟡 81%
execute/ tiny_keccak	344.10µs	328.57µs	🟢 -4.66%	897.14µs	643.22µs	🟢 -28.15%	🟡 96%
execute/ trunc_f2i	711.30µs	786.19µs	🔴 10.48%	1.86ms	1.44ms	🟢 -22.92%	🟡 83%
instantiate/ wasm_kernel	54.34µs	53.20µs	🟢 -2.73%	56.78µs	57.45µs	⚪ 0.05%	🟢 8%
translate/ erc1155	234.97µs	234.52µs	⚪ 0.22%	452.07µs	447.43µs	⚪ -1.27%	🟡 91%
translate/ erc20	113.14µs	113.38µs	⚪ 0.18%	218.46µs	215.95µs	⚪ -1.20%	🟡 90%
translate/ erc721	161.06µs	161.70µs	⚪ 0.45%	317.83µs	314.18µs	⚪ -1.02%	🟡 94%
translate/ spidermonkey	0.00ns	0.00ns	⚪ 0.02%	0.00ns	0.00ns	⚪ -0.60%	🟢 0%
translate/ wasm_kernel	4.05ms	4.06ms	⚪ 0.21%	8.29ms	8.25ms	⚪ -0.81%	🔴 103%

Link to pipeline

[codecov-commenter]

Codecov Report

Merging #688 (ad25645) into master (43f525c) will decrease coverage by 0.01%.
The diff coverage is 96.57%.

@@            Coverage Diff             @@
##           master     #688      +/-   ##
==========================================
- Coverage   81.01%   81.01%   -0.01%     
==========================================
  Files         100      100              
  Lines        8489     8510      +21     
==========================================
+ Hits         6877     6894      +17     
- Misses       1612     1616       +4     

Impacted Files	Coverage Δ
crates/wasmi/src/engine/stack/mod.rs	81.81% <ø> (ø)
crates/wasmi/src/engine/stack/values/sp.rs	94.25% <94.25%> (ø)
crates/wasmi/src/engine/executor.rs	95.08% <100.00%> (-0.24%)	⬇️
crates/wasmi/src/engine/stack/values/mod.rs	87.17% <100.00%> (+2.10%)	⬆️
crates/wasmi/src/engine/stack/values/tests.rs	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Robbepop]

I cannot explain the massive 10% performance regression on the execute/trunc_f2i test case, however, on the contrary we should probably still merge this PR due to the massive performance improvements across the board for Wasm builds and moderate performance wins on native for other test cases.