[Compiler] Re-use static to sema conversions in vm. by RZhang05 · Pull Request #4287 · onflow/cadence

RZhang05 · 2025-10-14T18:04:00Z

Work towards #4273

Description

Re-use static to sema conversions from the compiler in the vm. Tried various changes in the linker to no success, tried caching static type ID calls, which was an improvement but affects execution, so not included.

              │ ./bbq/vm/test/old-3.txt │   ./bbq/vm/test/optimized-2.txt   │
              │         sec/op          │   sec/op     vs base              │
FTTransfer-14               39.10µ ± 2%   38.43µ ± 2%  -1.73% (p=0.007 n=8)

              │ ./bbq/vm/test/old-3.txt │ ./bbq/vm/test/optimized-2.txt │
              │          B/op           │     B/op      vs base         │
FTTransfer-14              36.93Ki ± 0%   36.93Ki ± 0%  ~ (p=0.660 n=8)

              │ ./bbq/vm/test/old-3.txt │ ./bbq/vm/test/optimized-2.txt │
              │        allocs/op        │ allocs/op   vs base           │
FTTransfer-14                965.0 ± 0%   965.0 ± 0%  ~ (p=1.000 n=8) ¹

Targeted PR against master branch
Linked to Github issue with discussion and accepted design OR link to spec that describes this work
Code follows the standards mentioned here
Updated relevant documentation
Re-reviewed Files changed in the Github PR explorer
Added appropriate labels

github-actions · 2025-10-14T18:09:46Z

Benchstat comparison

Base branch: onflow:master
Base commit: ed55f55

Results

	old.txt	new.txt
	time/op		delta
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ContractFunctionInvocation-4	402µs ± 0%	399µs ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
Emit-4	4.25ms ± 0%	4.21ms ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ExportType/composite_type-4	258ns ± 0%	267ns ± 0%	~	(p=1.000 n=1+1)
ExportType/simple_type-4	77.7ns ± 0%	77.8ns ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ImperativeFib-4	20.8µs ± 0%	20.7µs ± 0%	~	(p=1.000 n=1+1)
InterpretRecursionFib-4	2.19ms ± 0%	2.06ms ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/bbq/vm/test goos:linux goarch:amd64
InterpreterFTTransfer-4	114µs ± 0%	115µs ± 0%	~	(p=1.000 n=1+1)
InterpreterImperativeFib-4	21.6µs ± 0%	21.3µs ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
NewInterpreter/new_interpreter-4	837ns ± 0%	863ns ± 0%	~	(p=1.000 n=1+1)
NewInterpreter/new_sub-interpreter-4	317ns ± 0%	315ns ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
RuntimeFungibleTokenTransferInterpreter-4	592µs ± 0%	620µs ± 0%	~	(p=1.000 n=1+1)
RuntimeResourceDictionaryValues-4	2.65ms ± 0%	2.65ms ± 0%	~	(p=1.000 n=1+1)
RuntimeResourceTracking-4	12.2ms ± 0%	12.0ms ± 0%	~	(p=1.000 n=1+1)
RuntimeScriptNoop-4	14.6µs ± 0%	14.5µs ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ValueIsSubtypeOfSemaType-4	59.5ns ± 0%	60.1ns ± 0%	~	(p=1.000 n=1+1)

	alloc/op		delta
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ContractFunctionInvocation-4	152kB ± 0%	152kB ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
Emit-4	1.46MB ± 0%	1.46MB ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ExportType/composite_type-4	120B ± 0%	120B ± 0%	~	(all equal)
ExportType/simple_type-4	0.00B	0.00B	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ImperativeFib-4	8.30kB ± 0%	8.30kB ± 0%	~	(all equal)
InterpretRecursionFib-4	1.19MB ± 0%	1.19MB ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/bbq/vm/test goos:linux goarch:amd64
InterpreterFTTransfer-4	39.7kB ± 0%	39.6kB ± 0%	~	(p=1.000 n=1+1)
InterpreterImperativeFib-4	8.29kB ± 0%	8.29kB ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
NewInterpreter/new_interpreter-4	976B ± 0%	976B ± 0%	~	(all equal)
NewInterpreter/new_sub-interpreter-4	232B ± 0%	232B ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
RuntimeFungibleTokenTransferInterpreter-4	157kB ± 0%	158kB ± 0%	~	(p=1.000 n=1+1)
RuntimeResourceDictionaryValues-4	1.76MB ± 0%	1.76MB ± 0%	~	(p=1.000 n=1+1)
RuntimeResourceTracking-4	9.27MB ± 0%	9.27MB ± 0%	~	(p=1.000 n=1+1)
RuntimeScriptNoop-4	7.99kB ± 0%	7.99kB ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ValueIsSubtypeOfSemaType-4	32.0B ± 0%	32.0B ± 0%	~	(all equal)

	allocs/op		delta
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ContractFunctionInvocation-4	2.46k ± 0%	2.46k ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
Emit-4	39.0k ± 0%	39.0k ± 0%	~	(p=1.000 n=1+1)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
ExportType/composite_type-4	3.00 ± 0%	3.00 ± 0%	~	(all equal)
ExportType/simple_type-4	0.00	0.00	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ImperativeFib-4	176 ± 0%	176 ± 0%	~	(all equal)
InterpretRecursionFib-4	17.7k ± 0%	17.7k ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/bbq/vm/test goos:linux goarch:amd64
InterpreterFTTransfer-4	833 ± 0%	833 ± 0%	~	(all equal)
InterpreterImperativeFib-4	175 ± 0%	175 ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
NewInterpreter/new_interpreter-4	15.0 ± 0%	15.0 ± 0%	~	(all equal)
NewInterpreter/new_sub-interpreter-4	4.00 ± 0%	4.00 ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/runtime goos:linux goarch:amd64
RuntimeFungibleTokenTransferInterpreter-4	2.98k ± 0%	2.98k ± 0%	~	(all equal)
RuntimeResourceDictionaryValues-4	36.7k ± 0%	36.7k ± 0%	~	(all equal)
RuntimeResourceTracking-4	159k ± 0%	159k ± 0%	~	(all equal)
RuntimeScriptNoop-4	113 ± 0%	113 ± 0%	~	(all equal)
pkg:github.com/onflow/cadence/interpreter goos:linux goarch:amd64
ValueIsSubtypeOfSemaType-4	1.00 ± 0%	1.00 ± 0%	~	(all equal)

SupunS

It's surprising how/why caching doesn't improve the performance, given that first iteration of caching in #4054 gave us some improvement.

Given that we already cache it after the very first conversion, maybe that initial conversion overhead is trivial?
Could you please add some more stat: for example, the number of cache hits/misses before and after the improvement?

Additionally, do we also need to cache types coming from the imports / for the imported programs?

SupunS · 2025-10-17T21:26:46Z

bbq/program.go

+	Variables     []Variable[E]
+	Types         []T
+	Globals       []Global
+	SemaTypeCache map[sema.TypeID]sema.Type


bbq.Program is currently serializable (and deserializable), even though we don't make use of it at the moment. But, we would lose this capability if we include sema.Type information in bbq.Program.

Could we instead pass this mapping externally? i.e: maybe as part of the vm.Config?

github-actions · 2025-11-28T17:42:12Z

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

RZhang05 · 2025-12-01T19:22:39Z

Given that we already cache it after the very first conversion, maybe that initial conversion overhead is trivial? Could you please add some more stat: for example, the number of cache hits/misses before and after the improvement?

For FTTransfer, cache hits/miss before:
701/115

vs cache hits/miss after:
815/1

I would guess the initial conversion overhead is insignificant as you said. Benchmarking FTTransfer reveals no real improvement.

Additionally, do we also need to cache types coming from the imports / for the imported programs?

Ideally yes, since for FTTransfer the same compiler config is used for all contracts, all the imported types should be cached in the stats above.

RZhang05 added 2 commits October 10, 2025 16:44

Reuse sema types from compiler, remove sema type in member access check.

85c0635

Add comments.

5d2ff5c

RZhang05 changed the title ~~Reuse sema types from compiler, remove sema type in member access check.~~ [Compiler] Re-use static to sema conversions in vm. Oct 14, 2025

Update type conversion memory metering test.

a11e16b

RZhang05 self-assigned this Oct 14, 2025

RZhang05 added 2 commits October 14, 2025 14:56

Update runtime test.

2962daa

Add type cache of linked programs.

751e32e

RZhang05 added the Improvement label Oct 14, 2025

Fix linker changes.

d6167e7

RZhang05 force-pushed the raymond/optimize-static-to-sema branch from 84d40f6 to d6167e7 Compare October 14, 2025 19:23

RZhang05 added 2 commits October 14, 2025 17:04

Revert check member access changes.

273842a

Remove linker changes to reuse compiler types.

7bc4611

RZhang05 force-pushed the raymond/optimize-static-to-sema branch from 22c2edb to 7bc4611 Compare October 15, 2025 19:30

Fix runtime tests.

b7fad6b

RZhang05 force-pushed the raymond/optimize-static-to-sema branch from 3c317a6 to b7fad6b Compare October 15, 2025 19:37

RZhang05 marked this pull request as ready for review October 15, 2025 20:22

RZhang05 requested review from SupunS and turbolent as code owners October 15, 2025 20:22

SupunS reviewed Oct 17, 2025

View reviewed changes

Merge branch 'master' into raymond/optimize-static-to-sema

5079a79

RZhang05 added 2 commits December 1, 2025 13:13

Merge branch 'master' into raymond/optimize-static-to-sema

06e8e19

Move type cache externally.

cd93e3d

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Compiler] Re-use static to sema conversions in vm.#4287

[Compiler] Re-use static to sema conversions in vm.#4287
RZhang05 wants to merge 12 commits intomasterfrom
raymond/optimize-static-to-sema

RZhang05 commented Oct 14, 2025 •

edited by SupunS

Loading

Uh oh!

github-actions bot commented Oct 14, 2025 •

edited

Loading

Uh oh!

SupunS left a comment •

edited

Loading

Uh oh!

SupunS Oct 17, 2025

Uh oh!

github-actions bot commented Nov 28, 2025 •

edited

Loading

Uh oh!

RZhang05 commented Dec 1, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

RZhang05 commented Oct 14, 2025 • edited by SupunS Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Uh oh!

github-actions bot commented Oct 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchstat comparison

Uh oh!

SupunS left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

SupunS Oct 17, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Nov 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Dependency Review

Scanned Files

Uh oh!

RZhang05 commented Dec 1, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

RZhang05 commented Oct 14, 2025 •

edited by SupunS

Loading

github-actions bot commented Oct 14, 2025 •

edited

Loading

SupunS left a comment •

edited

Loading

github-actions bot commented Nov 28, 2025 •

edited

Loading