Add multi dimensional multi constraint pricer algorithm #4012

MishkaRogachev · 2025-11-14T14:01:49Z

Fixes NIT-4069
(Temporary includes #3995)

Depends on:

core: Add multigas to EndTxHook and safeDecrement go-ethereum#579

Changes:

Implement multi-gas pricing constraints model
Modify AddToGasPool to accept multigas
Implement GasModelToUse to select a model depending on the configuration
Migrating gas constraints from single-dimensional to multi-dimensional model (open question)

github-actions · 2025-11-17T11:46:12Z

❌ 4 Tests Failed:

Tests completed	Failed	Passed	Skipped
2180	4	2176	0

View the top 3 failed tests by shortest run time

TestVersion30

Stack Traces | 5.430s run time

... [CONTENT TRUNCATED: Keeping last 20 lines]
ERROR[12-02|10:31:50.538] a large batch posting backlog exists     recentlyHitL1Bounds=false currentPosition=9 messageCount=103 messagesPerBatch=1 postedMessages=1 unpostedMessages=94  batchBacklogEstimate=94
ERROR[12-02|10:31:50.553] Dangling trie nodes after full cleanup
    precompile_inclusion_test.go:94: goroutine 454113 [running]:
        runtime/debug.Stack()
        	/opt/hostedtoolcache/go/1.25.4/x64/src/runtime/debug/stack.go:26 +0x5e
        github.com/offchainlabs/nitro/util/testhelpers.RequireImpl({0x40e7f90, 0xc09fbbfdc0}, {0x40a5820, 0xc0ac6b7cb0}, {0x0, 0x0, 0x0})
        	/home/runner/work/nitro/nitro/util/testhelpers/testhelpers.go:29 +0x55
        github.com/offchainlabs/nitro/system_tests.Require(0xc09fbbfdc0, {0x40a5820, 0xc0ac6b7cb0}, {0x0, 0x0, 0x0})
        	/home/runner/work/nitro/nitro/system_tests/common_test.go:1817 +0x5d
        github.com/offchainlabs/nitro/system_tests.testPrecompiles(0xc09fbbfdc0, 0x1e, {0xc0d1199db0, 0x6, 0x0?})
        	/home/runner/work/nitro/nitro/system_tests/precompile_inclusion_test.go:94 +0x371
        github.com/offchainlabs/nitro/system_tests.TestVersion30(0xc09fbbfdc0?)
        	/home/runner/work/nitro/nitro/system_tests/precompile_inclusion_test.go:67 +0x798
        testing.tRunner(0xc09fbbfdc0, 0x3d2a800)
        	/opt/hostedtoolcache/go/1.25.4/x64/src/testing/testing.go:1934 +0xea
        created by testing.(*T).Run in goroutine 1
        	/opt/hostedtoolcache/go/1.25.4/x64/src/testing/testing.go:1997 +0x465
        
    precompile_inclusion_test.go:94: �[31;1m [] execution aborted (timeout = 5s) �[0;0m
--- FAIL: TestVersion30 (5.43s)

TestPrimaryToSecondaryFailover

Stack Traces | 5.440s run time

=== RUN   TestPrimaryToSecondaryFailover
=== PAUSE TestPrimaryToSecondaryFailover
=== CONT  TestPrimaryToSecondaryFailover
INFO [12-02|10:23:11.032] arbitrum websocket broadcast server is listening address=[::]:46579
    broadcastclients_test.go:206: Primary broadcaster listening on: [::]:46579
INFO [12-02|10:23:11.032] arbitrum websocket broadcast server is listening address=[::]:35531
    broadcastclients_test.go:207: Secondary broadcaster listening on: [::]:35531
    broadcastclients_test.go:230: Primary URL: ws://127.0.0.1:46579
    broadcastclients_test.go:231: Secondary URL: ws://127.0.0.1:35531
INFO [12-02|10:23:11.033] connecting to arbitrum inbox message broadcaster url=ws://127.0.0.1:46579
    broadcastclients_test.go:278: Phase 1: Sending messages from primary broadcaster
INFO [12-02|10:23:11.034] Feed connected                           feedServerVersion=2 chainId=1234 requestedSeqNum=0
INFO [12-02|10:23:16.244] secondary feed started                   url=ws://127.0.0.1:35531 startingFromSeq=4
INFO [12-02|10:23:16.244] connecting to arbitrum inbox message broadcaster url=ws://127.0.0.1:35531
INFO [12-02|10:23:16.246] Feed connected                           feedServerVersion=2 chainId=1234 requestedSeqNum=4
    broadcastclients_test.go:308: Timed out waiting for message 5/5 from primary
--- FAIL: TestPrimaryToSecondaryFailover (5.44s)

TestVersion40

Stack Traces | 5.460s run time

... [CONTENT TRUNCATED: Keeping last 20 lines]
=== CONT  TestVersion40
�[38;5;48;1myay!! we validated block 3 in 59.83ms�[0;0m
    precompile_inclusion_test.go:94: goroutine 454114 [running]:
        runtime/debug.Stack()
        	/opt/hostedtoolcache/go/1.25.4/x64/src/runtime/debug/stack.go:26 +0x5e
        github.com/offchainlabs/nitro/util/testhelpers.RequireImpl({0x40e7f90, 0xc0a19408c0}, {0x40a5820, 0xc0acffe240}, {0x0, 0x0, 0x0})
        	/home/runner/work/nitro/nitro/util/testhelpers/testhelpers.go:29 +0x55
        github.com/offchainlabs/nitro/system_tests.Require(0xc0a19408c0, {0x40a5820, 0xc0acffe240}, {0x0, 0x0, 0x0})
        	/home/runner/work/nitro/nitro/system_tests/common_test.go:1817 +0x5d
        github.com/offchainlabs/nitro/system_tests.testPrecompiles(0xc0a19408c0, 0x28, {0xc0affabdf8, 0x5, 0x39?})
        	/home/runner/work/nitro/nitro/system_tests/precompile_inclusion_test.go:94 +0x371
        github.com/offchainlabs/nitro/system_tests.TestVersion40(0xc0a19408c0?)
        	/home/runner/work/nitro/nitro/system_tests/precompile_inclusion_test.go:71 +0x64b
        testing.tRunner(0xc0a19408c0, 0x3d2a808)
        	/opt/hostedtoolcache/go/1.25.4/x64/src/testing/testing.go:1934 +0xea
        created by testing.(*T).Run in goroutine 1
        	/opt/hostedtoolcache/go/1.25.4/x64/src/testing/testing.go:1997 +0x465
        
    precompile_inclusion_test.go:94: �[31;1m [] execution aborted (timeout = 5s) �[0;0m
--- FAIL: TestVersion40 (5.46s)

📣 Thoughts on this report? Let Codecov know! | Powered by Codecov

arbos/constraints/multi_dimensional.go

arbos/l2pricing/l2pricing.go

arbos/l2pricing/model.go

gligneul

Looks great

eljobe · 2025-12-01T12:44:42Z

arbos/l2pricing/l2pricing.go

 const GethBlockGasLimit = 1 << 50
-const gasConstraintsMaxNum = 20
-const MaxExponentBips = arbmath.Bips(85_000)
+const GasConstraintsMaxNum = 20


How did we decide on 20?

It feels like a "magic number". Let's document the "why"

Also, I feel like we're probably actually going to want more than 20.

These limits are very tight because of retryables. I added TODO(NIT-4152) for increasing/removing these limits in close future

eljobe · 2025-12-01T12:46:07Z

arbos/l2pricing/l2pricing.go

-const gasConstraintsMaxNum = 20
-const MaxExponentBips = arbmath.Bips(85_000)
+const GasConstraintsMaxNum = 20
+const MultiGasConstraintsMaxNum = 15


Same question about "why?"

eljobe · 2025-12-01T12:46:59Z

arbos/l2pricing/l2pricing.go

-const MaxExponentBips = arbmath.Bips(85_000)
+const GasConstraintsMaxNum = 20
+const MultiGasConstraintsMaxNum = 15
+const MaxPricingExponentBips = arbmath.Bips(85_000)


I also don't know what this means or why it's set to 85,000

eljobe · 2025-12-01T12:48:33Z

arbos/l2pricing/l2pricing.go

+		// that the new model (which normally applies weight normalization) matches the
+		// results of the old single-gas model.
+		//
+		// resourceWeights := make(map[uint8]uint64, len(FeeRelevantResourceKinds))


Is this commented out code, or part of the documentation? I don't think I really follow what's going on from the description.

eljobe · 2025-12-01T12:49:26Z

arbos/l2pricing/l2pricing.go

+		// }
+		resourceWeights := map[uint8]uint64{uint8(multigas.ResourceKindComputation): 1}
+
+		var uint32Window uint32


I don't like that the variable name also says the type. It feels very "not go style."

eljobe · 2025-12-01T12:52:13Z

arbos/l2pricing/model.go

+	return nil
+}
+
+// applyGasDelta adds delta to backlog if growBacklog=true, otherwise subtracts delta (saturating at zero).


I don't see a growBacklog parameter. Stale comment?

eljobe · 2025-12-01T12:56:17Z

arbos/l2pricing/model.go

+			// compared to the legacy single-gas model. In the compatibility tests we
+			// sometimes build a constraint where only ResourceKindComputation has
+			// weight=1 (all other weights are 0); in that case sumWeights=1 and this
+			// reduces to the legacy exponent. The commented alternative below shows the


I'm not sure it's useful to have the comment about the unnormalized divisor.

eljobe · 2025-12-01T12:59:16Z

precompiles/ArbRetryableTx.go

 	// This ensures that the gas pool has enough gas to run the retryable attempt.
-	return retryTxHash, c.State.L2PricingState().AddToGasPool(arbmath.SaturatingCast[int64](gasToDonate))
+	// TODO(NIT-4120): clarify the gas dimension for gasToDonate
+	return retryTxHash, c.State.L2PricingState().UpdateBacklog(l2pricing.Shrink, gasToDonate, multigas.ComputationGas(gasToDonate))


Why not "l2pricing.ShrinkBacklog" and "l2pricing.GrowBacklog" Are we going to support other operations?

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from a3cb346 to 5dc32f7 Compare November 17, 2025 10:00

MishkaRogachev mentioned this pull request Nov 17, 2025

core: Add multigas to EndTxHook and safeDecrement OffchainLabs/go-ethereum#579

Draft

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from 5dc32f7 to a7acb86 Compare November 17, 2025 11:21

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch 2 times, most recently from 454b599 to 28a7a5f Compare November 18, 2025 13:44

MishkaRogachev marked this pull request as ready for review November 19, 2025 15:04

MishkaRogachev requested review from eljobe and gligneul November 19, 2025 15:04

MishkaRogachev assigned gligneul Nov 19, 2025

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from fa8781a to 4f2c34e Compare November 21, 2025 12:04

MishkaRogachev assigned MishkaRogachev and unassigned gligneul Nov 21, 2025

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from 8ec77cd to 0611789 Compare November 21, 2025 14:03

MishkaRogachev assigned gligneul and unassigned gligneul Nov 21, 2025

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch 2 times, most recently from b14c464 to 46e002a Compare November 24, 2025 10:55

MishkaRogachev assigned gligneul and unassigned MishkaRogachev Nov 24, 2025

gligneul requested changes Nov 24, 2025

View reviewed changes

gligneul assigned MishkaRogachev and unassigned gligneul Nov 24, 2025

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch 5 times, most recently from fd0e681 to 752c710 Compare November 26, 2025 15:43

MishkaRogachev requested a review from gligneul November 26, 2025 15:58

MishkaRogachev assigned gligneul Nov 26, 2025

MishkaRogachev requested a review from gligneul November 28, 2025 15:34

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch 2 times, most recently from 09d57c1 to 45af934 Compare December 1, 2025 09:46

gligneul previously approved these changes Dec 1, 2025

View reviewed changes

gligneul assigned eljobe and unassigned gligneul Dec 1, 2025

MishkaRogachev dismissed gligneul’s stale review via 92ac6ef December 1, 2025 12:05

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from 45af934 to 92ac6ef Compare December 1, 2025 12:05

eljobe requested changes Dec 1, 2025

View reviewed changes

eljobe assigned MishkaRogachev and unassigned eljobe Dec 1, 2025

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch 3 times, most recently from 5ba72e8 to f68c58c Compare December 1, 2025 17:48

MishkaRogachev assigned eljobe and unassigned MishkaRogachev Dec 1, 2025

MishkaRogachev added 12 commits December 2, 2025 11:11

Replace ShouldUseGasConstraints with GasModelToUse

8ad8e59

Correct AddToGasPool signature to support multi-gas

a621ee1

Implement multi-gas multi-constraints pricing formula

d3f683c

Minor pricer improvments

d9cc2ce

Refactor and improve multi-gas price calculations

e727f57

Add log to addToGasPoolWithMultiGasConstraints

f4a5d6c

Fix flaky TestSetAndGetMultiGasPricingConstraints

bcca891

Post-merge mixes for redeem

2a5c47f

Separate AddToGasPool/SubstractGasFromPool in l2 pricing

5d3aeeb

Adjust backlog update cost and better namings

1efa7a3

Merge constrains package to l2pricing & review fixes

a978edf

Review fixes: renamings and updated comments

7b28238

MishkaRogachev force-pushed the add-multi-dimensional-multi-constraint-pricer-algorithm branch from f68c58c to 7b28238 Compare December 2, 2025 10:12

MishkaRogachev marked this pull request as draft December 2, 2025 12:10

Add multi dimensional multi constraint pricer algorithm #4012

Are you sure you want to change the base?

Add multi dimensional multi constraint pricer algorithm #4012

Uh oh!

Conversation

MishkaRogachev commented Nov 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Nov 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ 4 Tests Failed:

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gligneul left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

MishkaRogachev commented Nov 14, 2025 •

edited

Loading

github-actions bot commented Nov 17, 2025 •

edited

Loading