• Make the logic more clean and clear. - check recursion_limit by >=

BENCHMARKS

	NATIVE			WASMTIME
BENCHMARK	MASTER	PR	DIFF	MASTER	PR	DIFF	WASMTIME OVERHEAD
execute/ bare_call_0	1.27ms	1.27ms	:white_circle: 0.43%	1.12ms	1.09ms	:green_circle: -2.98%	:green_circle: -14%
execute/ bare_call_0/typed	698.58µs	705.30µs	:white_circle: 0.90%	547.87µs	551.10µs	:white_circle: 0.60%	:green_circle: -22%
execute/ bare_call_1	1.40ms	1.40ms	:green_circle: -0.82%	1.37ms	1.35ms	:green_circle: -1.58%	:green_circle: -4%
execute/ bare_call_16	3.13ms	2.90ms	:green_circle: -7.07%	4.99ms	5.00ms	:green_circle: 0.32%	:yellow_circle: 72%
execute/ bare_call_16/typed	1.32ms	1.29ms	:green_circle: -2.45%	2.34ms	2.28ms	:green_circle: -2.44%	:yellow_circle: 77%
execute/ bare_call_1/typed	800.90µs	799.92µs	:white_circle: -0.08%	874.15µs	865.19µs	:white_circle: -0.96%	:green_circle: 8%
execute/ bare_call_4	1.76ms	1.78ms	:red_circle: 1.39%	2.14ms	2.16ms	:green_circle: 1.12%	:green_circle: 21%
execute/ bare_call_4/typed	891.43µs	886.81µs	:white_circle: -0.47%	1.00ms	984.30µs	:green_circle: -1.56%	:green_circle: 11%
execute/ br_table	983.74µs	908.23µs	:green_circle: -7.64%	1.14ms	1.10ms	:green_circle: -4.01%	:green_circle: 21%
execute/ count_until	903.65µs	903.43µs	:white_circle: 0.05%	2.40ms	2.21ms	:green_circle: -7.87%	:red_circle: 144%
execute/ factorial_iterative	365.06µs	376.29µs	:red_circle: 2.97%	946.61µs	901.84µs	:green_circle: -4.67%	:red_circle: 140%
execute/ factorial_recursive	730.68µs	711.02µs	:green_circle: -2.72%	1.58ms	1.45ms	:green_circle: -8.78%	:red_circle: 103%
execute/ fib_iterative	1.79ms	1.79ms	:white_circle: 0.03%	5.00ms	4.90ms	:green_circle: -1.89%	:red_circle: 174%
execute/ fib_recursive	7.05ms	6.75ms	:green_circle: -4.33%	14.29ms	13.68ms	:green_circle: -4.27%	:red_circle: 103%
execute/ global_bump	1.36ms	1.38ms	:white_circle: 1.05%	3.65ms	3.35ms	:green_circle: -8.26%	:red_circle: 143%
execute/ global_const	999.89µs	999.49µs	:white_circle: -0.02%	2.83ms	2.52ms	:green_circle: -11.15%	:red_circle: 152%
execute/ host_calls	32.63µs	32.39µs	:white_circle: -0.77%	44.69µs	44.82µs	:white_circle: 0.48%	:green_circle: 38%
execute/ memory_fill	1.55ms	1.55ms	:white_circle: 0.09%	4.20ms	4.07ms	:green_circle: -3.00%	:red_circle: 162%
execute/ memory_sum	1.55ms	1.55ms	:white_circle: 0.10%	4.18ms	4.05ms	:green_circle: -3.10%	:red_circle: 161%
execute/ memory_vec_add	3.11ms	3.05ms	:green_circle: -1.86%	8.75ms	8.02ms	:green_circle: -8.39%	:red_circle: 163%
execute/ recursive_is_even	1.29ms	1.29ms	:white_circle: 0.15%	2.63ms	2.50ms	:green_circle: -5.15%	:yellow_circle: 93%
execute/ recursive_ok	170.56µs	167.63µs	:green_circle: -1.72%	377.40µs	359.20µs	:green_circle: -4.95%	:red_circle: 114%
execute/ recursive_scan	213.76µs	209.21µs	:green_circle: -2.13%	466.29µs	451.00µs	:green_circle: -3.26%	:red_circle: 116%
execute/ recursive_trap	16.93µs	16.26µs	:green_circle: -3.98%	36.09µs	34.52µs	:green_circle: -4.36%	:red_circle: 112%
execute/ regex_redux	669.97µs	662.32µs	:green_circle: -1.24%	1.60ms	1.45ms	:green_circle: -9.38%	:red_circle: 118%
execute/ rev_complement	590.19µs	595.15µs	:white_circle: 0.94%	1.56ms	1.43ms	:green_circle: -8.18%	:red_circle: 141%
execute/ tiny_keccak	450.54µs	482.95µs	:red_circle: 7.28%	1.39ms	1.29ms	:green_circle: -7.12%	:red_circle: 167%
execute/ trunc_f2i	1.08ms	1.01ms	:green_circle: -6.67%	2.66ms	2.44ms	:green_circle: -8.11%	:red_circle: 142%
instantiate/ wasm_kernel	67.57µs	70.66µs	:white_circle: 4.00%	91.92µs	94.48µs	:red_circle: 2.72%	:green_circle: 34%
translate/ wasm_kernel	4.19ms	4.22ms	:white_circle: 0.74%	7.74ms	7.88ms	:red_circle: 1.80%	:yellow_circle: 87%

Link to pipeline

I checked out your PR locally after seeing some regressions according to our benchmarking CI and unfortunately I can confirm locally that some benchmarks are slightly regressed with those changes. I see that they are some kind of a clean up but I am hesistant merging a change if it introduces regressions that I do not understand.

@Robbepop I do not see regressions in my local

Codecov Report

Merging #508 (ec90097) into master (0d087e2) will decrease coverage by 0.00%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #508      +/-   ##
==========================================
- Coverage   79.73%   79.73%   -0.01%     
==========================================
  Files          75       75              
  Lines        6292     6291       -1     
==========================================
- Hits         5017     5016       -1     
  Misses       1275     1275              

Impacted Files	Coverage Δ
crates/wasmi/src/engine/mod.rs	83.52% <100.00%> (ø)
crates/wasmi/src/engine/stack/frames.rs	92.00% <100.00%> (-0.60%)	:arrow_down:
crates/wasmi/src/engine/stack/mod.rs	79.71% <100.00%> (+0.29%)	:arrow_up:

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

:( execute/memory_vec_add meet a lot regressions. Why?

:( execute/memory_vec_add meet a lot regressions. Why?

I don't know. Could be an artifact. Unfortunately wasmi these days is extremely sensible to performance regressions.

Hi, seems now it at least not make regressions. @Robbepop

Why is red when improved? Such as 🔴 -3.70% 🔴 -0.35%

Strange, wasm_kernel is not stable, I see a big improvement last time.

@Robbepop It now seems is a good improvement that could be merged now.

@Robbepop It now seems is a good improvement that could be merged now.

The problem with those benchmarks is that they are currently extremely flaky/noisy for this PR in particular. I don't know why that is but we cannot really trust those benchmarks fully with respect to this PR. I will take a look today on it and decide if we want to merge it. Generally I think the code of the PR is an improvement, however, I can remember that I programmed the structure so far in this way due to performance regressions back then. However, code has changed a lot since then and FuncFrame also became much smaller and therefore moving a FuncFrame around is no longer that bad.