Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Optimization: differentiate function calls #724

Merged
merged 20 commits into from
May 28, 2023
Merged

Optimization: differentiate function calls #724

merged 20 commits into from
May 28, 2023

Conversation

Robbepop
Copy link
Member

Closes #711.

@Robbepop
rename FuncBody -> CompiledFunc
6beae00
@Robbepop
make CompiledFunc use u32 internally
3c17500 
This is to make it possible to use CompiledFunc in wasmi bytecode.
@Robbepop
implement custom Default impl for CodeMap
d1c0c58
@Robbepop
rename insts -> instrs in CodeMap
7ed724f 
"insts" is ambiguous in Wasm with "instances" therefore we always want to use "instr" to refer to an instruction.
@Robbepop
refactor InstructionsRef for special 0 value case
4e36a2c
@Robbepop
add FuncHeader::new
43f25dc
@Robbepop
cleanup code slightly
17ef48e
@Robbepop
rename parameter & minor refactoring
f32e1dc
@Robbepop
fix internal doc links with FuncBody -> CompiledFunc
27282e2
@Robbepop
improve docs in Engine
e33039d
@Robbepop
rename resolve_inst -> resolve_instr
3ee9112
@Robbepop
add docs to some EngineInner APIs
0e34f9c
@Robbepop
add doc link
a69cfba
@Robbepop
allocate CompiledFunc prior to translation
381585f 
Now instead of allocation CompiledFunc to the `wasmi` Engine as the result of validation and translation of a Wasm function, the CompiledFunc is already allocated and initialized instead.
This allows to access the CompiledFunc index for function calls in other translation units which finally provides the possibility to optimize calls to internal functions.
Also this will make it simpler to run function translations concurrently if we ever decide to want this feature.
@Robbepop
add new CallInternal instruction
a8ca866
@Robbepop
remove unnecessary import
9a952f2
@Robbepop
add docs
160722e 
Maybe we should specialize Call to CallImported and mandate that it is only used for imported functions. This could further optimize the call infrastructure in the engine by relying on this.
@Robbepop
fix internal doc links
5612869
@paritytech-cicd-pr
Copy link

paritytech-cicd-pr commented May 28, 2023 •
edited
Loading

BENCHMARKS

NATIVEWASMTIME
BENCHMARKMASTERPRDIFFMASTERPRDIFFWASMTIME OVERHEAD
execute/
bare_call_0		 1.53ms 1.54ms 🟢 0.83% 1.06ms 1.05ms 🟢 -1.20% 🟢 -32%
execute/
bare_call_0/typed		 1.18ms 1.15ms 🟢 -2.23% 693.11µs 680.56µs 🟢 -1.77% 🟢 -41%
execute/
bare_call_1		 1.58ms 1.58ms 🟢 -0.15% 1.27ms 1.19ms 🟢 -6.63% 🟢 -25%
execute/
bare_call_16		 2.57ms 2.52ms 🟢 -1.81% 3.49ms 3.27ms 🟢 -5.94% 🟢 30%
execute/
bare_call_16/typed		 1.57ms 1.59ms ⚪ 0.99% 1.91ms 1.82ms 🟢 -4.88% 🟢 15%
execute/
bare_call_1/typed		 1.25ms 1.22ms 🟢 -2.43% 1.02ms 1.01ms ⚪ -0.96% 🟢 -18%
execute/
bare_call_4		 1.78ms 1.75ms ⚪ -1.33% 1.63ms 1.60ms 🟢 -2.48% 🟢 -9%
execute/
bare_call_4/typed		 1.21ms 1.21ms ⚪ -0.12% 1.00ms 966.41µs 🟢 -2.98% 🟢 -20%
execute/
br_table		 1.33ms 1.40ms 🔴 5.28% 1.26ms 1.21ms 🟢 -3.67% 🟢 -14%
execute/
count_until		 578.09µs 577.36µs ⚪ 0.02% 1.16ms 1.16ms ⚪ -0.34% 🟡 100%
execute/
factorial_iterative		 319.97µs 318.87µs ⚪ -0.57% 554.71µs 539.69µs 🟢 -2.76% 🟡 69%
execute/
factorial_recursive		 595.65µs 506.33µs 🟢 -15.21% 842.94µs 675.28µs 🟢 -19.71% 🟢 33%
execute/
fibonacci_iter		 1.40ms 1.36ms 🟢 -3.21% 3.18ms 2.94ms 🟢 -7.50% 🔴 117%
execute/
fibonacci_rec		 4.94ms 4.20ms 🟢 -15.09% 7.62ms 6.40ms 🟢 -15.96% 🟡 53%
execute/
fibonacci_tail		 1.08ms 899.00µs 🟢 -16.62% 1.90ms 1.51ms 🟢 -20.58% 🟡 68%
execute/
global_bump		 723.35µs 738.35µs 🔴 2.09% 1.41ms 1.54ms 🔴 9.47% 🔴 109%
execute/
global_const		 683.54µs 659.48µs 🟢 -3.46% 1.31ms 1.31ms ⚪ 0.28% 🟡 99%
execute/
host_calls		 36.36µs 36.35µs ⚪ 0.41% 40.00µs 41.44µs 🔴 3.53% 🟢 14%
execute/
memory_fill		 1.14ms 1.15ms ⚪ 0.28% 2.26ms 2.21ms 🟢 -2.28% 🟡 93%
execute/
memory_sum		 1.13ms 1.13ms ⚪ -0.17% 2.47ms 2.43ms 🟢 -1.40% 🔴 114%
execute/
memory_vec_add		 2.24ms 2.24ms ⚪ 0.07% 4.77ms 4.36ms 🟢 -8.45% 🟡 95%
execute/
recursive_is_even		 1.02ms 695.66µs 🟢 -32.24% 1.61ms 1.03ms 🟢 -35.86% 🟢 48%
execute/
recursive_ok		 122.69µs 99.39µs 🟢 -19.19% 196.37µs 149.07µs 🟢 -24.10% 🟢 50%
execute/
recursive_scan		 161.19µs 135.24µs 🟢 -15.68% 242.42µs 194.45µs 🟢 -19.65% 🟢 44%
execute/
recursive_trap		 12.44µs 9.60µs 🟢 -23.48% 20.70µs 14.71µs 🟢 -28.92% 🟡 53%
execute/
regex_redux		 491.74µs 458.48µs 🟢 -6.62% 889.69µs 846.71µs 🟢 -4.81% 🟡 85%
execute/
rev_complement		 442.12µs 435.15µs ⚪ -1.34% 870.68µs 839.67µs 🟢 -3.52% 🟡 93%
execute/
tiny_keccak		 320.50µs 320.77µs ⚪ 0.15% 661.45µs 658.96µs ⚪ -0.29% 🔴 105%
execute/
trunc_f2i		 731.41µs 736.33µs ⚪ 1.19% 1.67ms 1.73ms 🔴 3.55% 🔴 135%
instantiate/
wasm_kernel		 55.21µs 52.79µs 🟢 -3.23% 55.45µs 58.72µs 🔴 5.32% 🟢 11%
translate/
erc1155		 197.45µs 198.79µs ⚪ 0.51% 417.94µs 442.34µs 🔴 5.71% 🔴 123%
translate/
erc20		 97.58µs 97.16µs ⚪ -0.48% 200.51µs 210.01µs 🔴 4.80% 🔴 116%
translate/
erc721		 138.92µs 137.99µs ⚪ -0.74% 292.31µs 305.87µs 🔴 4.60% 🔴 122%
translate/
spidermonkey		 62.18ms 62.19ms ⚪ 0.12% 0.00ns 0.00ns 🔴 5.77% 🟢 -100%
translate/
wasm_kernel		 3.53ms 4.12ms 🔴 16.28% 7.51ms 7.93ms 🔴 5.48% 🟡 93%

Link to pipeline

@codecov-commenter
Copy link

codecov-commenter commented May 28, 2023 •
edited
Loading

Codecov Report

Merging #724 (ef6b5e6) into master (74c173c) will increase coverage by 0.03%.
The diff coverage is 85.71%.

@@            Coverage Diff             @@
##           master     #724      +/-   ##
==========================================
+ Coverage   79.03%   79.06%   +0.03%     
==========================================
  Files         102      102              
  Lines        8684     8757      +73     
==========================================
+ Hits         6863     6924      +61     
- Misses       1821     1833      +12
Impacted Files Coverage Δ
crates/wasmi/src/engine/bytecode/mod.rs 94.11% <ø> (ø)
crates/wasmi/src/engine/func_builder/translator.rs 86.31% <61.53%> (-0.35%) ⬇️
crates/wasmi/src/engine/executor.rs 78.30% <72.72%> (-0.20%) ⬇️
crates/wasmi/src/module/parser.rs 72.10% <80.00%> (-0.39%) ⬇️
crates/wasmi/src/engine/code_map.rs 90.47% <93.18%> (+7.97%) ⬆️
...ates/wasmi/src/engine/func_builder/inst_builder.rs 80.64% <100.00%> (ø)
crates/wasmi/src/engine/func_builder/mod.rs 91.66% <100.00%> (ø)
crates/wasmi/src/engine/mod.rs 77.72% <100.00%> (+0.43%) ⬆️
crates/wasmi/src/engine/tests.rs 99.36% <100.00%> (ø)
crates/wasmi/src/func/mod.rs 77.16% <100.00%> (ø)
... and 3 more

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

@Robbepop Robbepop merged commit b68995e into master May 28, 2023
@Robbepop Robbepop deleted the rf-differentiate-calls branch May 28, 2023 13:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Optimization: Differentiate between imported and internal calls
3 participants
@Robbepop @paritytech-cicd-pr @codecov-commenter