Overhaul Pages and Bytes types

[Robbepop]

Problems with previous types were found during fuzzy testing.
These types make it simpler to fix linear memory allocation problems in wasmi for 32-bit and 64-bit architectures for edge cases.

Also we get rid of the memory_units dependency which did a lot more things that we actually needed in wasmi.

[codecov-commenter]

Codecov Report

Merging #449 (889b301) into master (753b05d) will decrease coverage by 0.11%.
The diff coverage is 69.73%.

@@            Coverage Diff             @@
##           master     #449      +/-   ##
==========================================
- Coverage   79.44%   79.33%   -0.12%     
==========================================
  Files          71       72       +1     
  Lines        6193     6241      +48     
==========================================
+ Hits         4920     4951      +31     
- Misses       1273     1290      +17     

Impacted Files	Coverage Δ
wasmi_v1/src/memory/byte_buffer.rs	71.42% <63.63%> (-7.52%)	⬇️
core/src/units.rs	66.66% <66.66%> (ø)
wasmi_v1/src/engine/executor.rs	98.26% <71.42%> (-0.15%)	⬇️
wasmi_v1/src/memory/mod.rs	59.34% <77.77%> (+0.87%)	⬆️
wasmi_v1/src/module/utils.rs	86.36% <100.00%> (ø)
wasmi_v1/tests/spec/context.rs	77.98% <100.00%> (ø)

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

[paritytech-cicd-pr]

BENCHMARKS

	NATIVE			WASMTIME
BENCHMARK	MASTER	PR	DIFF	MASTER	PR	DIFF	WASMTIME OVERHEAD
execute/ bare_call_0	1.29ms	1.32ms	🔴 1.96%	1.12ms	1.10ms	🔴 -1.89%	🟢 -17%
execute/ bare_call_0/typed	703.72µs	754.52µs	🔴 7.16%	537.68µs	544.35µs	🔴 1.19%	🟢 -28%
execute/ bare_call_1	1.40ms	1.44ms	🔴 3.33%	1.38ms	1.36ms	🟢 -1.51%	🟢 -6%
execute/ bare_call_16	3.03ms	3.07ms	⚪ 2.39%	5.23ms	5.09ms	🟢 -2.53%	🟡 66%
execute/ bare_call_16/typed	1.30ms	1.29ms	⚪ -0.91%	2.53ms	2.37ms	🟢 -6.35%	🟡 84%
execute/ bare_call_1/typed	805.39µs	874.88µs	🔴 8.71%	868.49µs	868.98µs	⚪ -0.21%	🟢 -1%
execute/ bare_call_4	1.77ms	1.75ms	🔴 -0.78%	2.18ms	2.14ms	🟢 -2.02%	🟢 22%
execute/ bare_call_4/typed	889.30µs	916.20µs	🔴 2.98%	1.01ms	985.45µs	🟢 -2.22%	🟢 8%
execute/ br_table	907.59µs	945.51µs	🔴 4.13%	1.10ms	1.11ms	⚪ 0.99%	🟢 18%
execute/ count_until	886.66µs	901.67µs	🔴 1.73%	2.21ms	2.33ms	🔴 5.64%	🔴 159%
execute/ factorial_iterative	362.63µs	363.73µs	⚪ 0.35%	895.64µs	930.20µs	🔴 3.80%	🔴 156%
execute/ factorial_recursive	747.18µs	716.73µs	🟢 -3.98%	1.50ms	1.54ms	🔴 2.58%	🔴 115%
execute/ fib_iterative	1.79ms	1.79ms	⚪ -0.22%	4.92ms	5.47ms	🔴 11.38%	🔴 206%
execute/ fib_recursive	7.12ms	8.16ms	🔴 14.62%	13.59ms	14.17ms	🔴 4.25%	🟡 74%
execute/ global_bump	1.45ms	1.36ms	🟢 -5.57%	3.50ms	3.83ms	🔴 9.47%	🔴 181%
execute/ host_calls	33.06µs	34.74µs	⚪ 1.34%	45.26µs	45.49µs	⚪ 0.58%	🟢 31%
execute/ memory_fill	1.53ms	1.56ms	🔴 1.33%	4.23ms	4.24ms	⚪ 0.17%	🔴 172%
execute/ memory_sum	1.52ms	1.56ms	🔴 2.86%	4.17ms	4.25ms	🔴 1.93%	🔴 172%
execute/ memory_vec_add	4.76ms	3.06ms	🟢 -36.16%	8.22ms	8.67ms	🔴 5.35%	🔴 183%
execute/ recursive_is_even	1.31ms	1.32ms	⚪ 1.02%	2.55ms	2.59ms	🔴 1.87%	🟡 96%
execute/ recursive_ok	175.60µs	172.96µs	🟢 -1.43%	351.76µs	362.25µs	🔴 2.93%	🔴 109%
execute/ recursive_scan	219.90µs	215.17µs	🟢 -2.03%	454.72µs	468.96µs	🔴 3.31%	🔴 118%
execute/ recursive_trap	17.34µs	17.33µs	⚪ -0.16%	34.82µs	36.40µs	🔴 4.73%	🔴 110%
execute/ regex_redux	652.39µs	675.53µs	🔴 3.47%	1.57ms	1.60ms	🔴 2.30%	🔴 136%
execute/ rev_complement	591.18µs	596.97µs	⚪ 1.01%	1.52ms	1.54ms	⚪ 1.14%	🔴 158%
execute/ tiny_keccak	481.94µs	471.69µs	⚪ -1.83%	1.33ms	1.37ms	🔴 2.98%	🔴 191%
execute/ trunc_f2i	1.07ms	1.07ms	⚪ 0.55%	2.41ms	2.65ms	🔴 10.16%	🔴 147%
instantiate/ wasm_kernel	68.36µs	72.02µs	🔴 5.69%	107.28µs	120.26µs	🔴 12.44%	🟡 67%
translate/ wasm_kernel	4.11ms	4.16ms	⚪ 1.13%	7.83ms	7.89ms	⚪ 0.67%	🟡 90%

Link to pipeline

[yjhmelody]

Sorry, I could not find this value in spec. Could you give me a ref?

[yjhmelody]

Could you also consider to extract memoryEntity type as a new crate?

[Robbepop]

Could you also consider to extract memoryEntity type as a new crate?

i do not see a reason for this. also what does this has to do with this PR?

[yjhmelody]

Could you also consider to extract memoryEntity type as a new crate?

i do not see a reason for this. also what does this has to do with this PR?

I think it may be better to put Pages/Bytes into a wasmi_memory crate.

[Robbepop]

Could you also consider to extract memoryEntity type as a new crate?

i do not see a reason for this. also what does this has to do with this PR?

I think it may be better to put Pages/Bytes into a wasmi_memory crate.

Yeah I got that but can you give me a compelling example as to why?

[yjhmelody]

Yeah I got that but can you give me a compelling example as to why?

No examples, simply because I need this crate too :(

[Robbepop]

Yeah I got that but can you give me a compelling example as to why?

No examples, simply because I need this crate too :(

Cannot you use wasmi_core instead?