`EIP-7823` and `EIP-7883` implementation --- Osaka `MODEXP` changes #2477

OlivierBBB · 2025-11-12T17:58:23Z

Note

Implements fork-aware MODEXP per Osaka (EIP‑7823), introducing London/Osaka MODEXP metadata, pricing and XBS OOB calls, updating MMU and counters to fork-specific sizes/costs, wiring fork into ZkCounter via RPC, and adjusting tests.

zktracer (core):
- Make ZkCounter fork-aware; derive OpCodes/Trace from fork and use fork in MODEXP logic and line counts.
- Split MODEXP into forked implementations: LondonModexpSubsection/OsakaModexpSubsection, with London/OsakaModexpMetadata.
- Add forked MODEXP ops: London/OsakaBlakeModexpDataOperation, London/OsakaModexpPricingOobCall, London/OsakaModexpXbsOobCall.
- Update exponent-log and return sizing to use fork-specific multipliers and max input sizes.
MMU:
- Pass fork into Mmu/MmuInstructions; adjust ModexpData/ModexpZero micro-rows and component sizes per fork.
Precompile processing:
- Route PRC_MODEXP to London/Osaka subsections; count rows via ModexpMetadata.getNumberOfRowsForModexp(); make MODEXP success depend on forked bounds and pricing.
- Use fork-specific MODEXP rows and Blake rows via unified BlakeModexpData API.
RPC / Counter:
- GenerateLineCountsV2 and ConflatedCountTracesV2: construct ZkCounter with resolved fork.
Tests:
- Update unit/integration tests to reflect fork-aware MODEXP limits, pricing, gas, and helper utilities.

^{Written by Cursor Bugbot for commit 2fedec6. This will update automatically on new commits. Configure here.}

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java

...ktracer/module/hub/fragment/imc/oob/precompiles/modexp/xbsOobCall/OsakaModexpXbsOobCall.java

...sensys/linea/zktracer/module/hub/section/call/precompileSubsection/PrecompileSubsection.java

...tization/src/main/java/net/consensys/linea/zktracer/module/hub/fragment/imc/mmu/MmuCall.java

...net/consensys/linea/zktracer/module/hub/precompiles/modexpMetadata/LondonModexpMetadata.java

...n/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/LondonBlakeModexpOperation.java

With Lorenzo and François

arithmetization/src/main/java/net/consensys/linea/zktracer/ZkCounter.java

One failing test (MODEXP limits)

the old tests were overwriting mbs in memory !

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java

...module/hub/fragment/imc/oob/precompiles/modexp/pricingOobCall/OsakaModexpPricingOobCall.java

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java

...ion/src/main/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/BlakeComponents.java

...ion/src/main/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/BlakeModexpData.java

...ain/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/BlakeModexpDataOperation.java

...c/main/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/LondonBlakeModexpData.java

...tization/src/main/java/net/consensys/linea/zktracer/module/hub/fragment/imc/mmu/MmuCall.java

arithmetization/src/main/java/net/consensys/linea/zktracer/module/hub/Hub.java

...test/java/net/consensys/linea/zktracer/module/limits/precompileLimits/ModexpLimitsTests.java

arithmetization/src/test/resources/junit-platform.properties

testing/src/main/java/net/consensys/linea/reporting/TracerTestBase.java

plan.md

cursor · 2025-11-19T17:36:12Z

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java

+  }
+
+  protected boolean allXbsesAreInBounds() {
+    return true;


Bug: Missing maxNumberOfLines override in LondonModexpSubsection

The LondonModexpSubsection class doesn't override the maxNumberOfLines() method that was present in the original ModexpSubsection class. This causes it to inherit the default implementation from PrecompileSubsection which returns 0, instead of the correct value NB_ROWS_HUB_PRC_MODEXP (13). This affects line counting for MODEXP precompile operations and could lead to incorrect trace sizing calculations.

…tyThreeSixtyFourthsPrecompileTests

cursor · 2025-11-19T19:53:07Z

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java

+  }
+
+  protected boolean allXbsesAreInBounds() {
+    return true;


Bug: Missing maxNumberOfLines override in LondonModexpSubsection

The LondonModexpSubsection class was refactored from ModexpSubsection but the maxNumberOfLines() method override was removed. The diff shows the old ModexpSubsection had this method returning NB_ROWS_HUB_PRC_MODEXP (value 13), but the new LondonModexpSubsection doesn't override it, causing it to inherit the default return value of 0 from PrecompileSubsection. This breaks the precompile subsection initialization in the constructor which calls new ArrayList<>(maxNumberOfLines()), potentially causing incorrect memory allocation and fragment list sizing for MODEXP operations.

cursor · 2025-11-20T10:28:42Z

.../java/net/consensys/linea/zktracer/module/hub/precompiles/modexpMetadata/ModexpMetadata.java


+  public short getNumberOfRowsForModexp() {
+    return (short) (4 * (getMaxInputSize() / LLARGE));
+  }


Bug: Mismatch between row count formula and actual indices

The getNumberOfRowsForModexp() method calculates rows as 4 * (getMaxInputSize() / LLARGE), assuming all four MODEXP phases have equal-sized index arrays. However, the actual row count depends on INDEX_MAX_MODEXP_BASE, INDEX_MAX_MODEXP_EXPONENT, INDEX_MAX_MODEXP_MODULUS, and INDEX_MAX_MODEXP_RESULT constants from fork-specific Trace classes. If these constants aren't uniformly derived from getMaxInputSize() / LLARGE, this formula produces incorrect line counts, leading to miscounting in blakemodexp.updateTally() calls in ZkCounter.tracePrecompileCall().

amkCha · 2025-11-20T10:41:53Z

...ion/src/main/java/net/consensys/linea/plugins/rpc/batchlinecount/ConflatedCountTracesV2.java


    return tracerSharedConfiguration.isLimitless()
-        ? new ZkCounter(l1L2BridgeSharedConfiguration)
+        ? new ZkCounter(l1L2BridgeSharedConfiguration, fork)


This will require a PR in the monorepo as well to change the ZkCounter interface, everywhere ZkCounter is called https://github.com/Consensys/linea-monorepo/blob/5aed83a5f904a1b19906dac4aa0d45656074dfd3/besu-plugins/linea-sequencer/sequencer/src/main/java/net/consensys/linea/rpc/methods/LineaEstimateGas.java#L437 - might also be the case for Besu-shomei plugin

Signed-off-by: F Bojarski <[email protected]>

letypequividelespoubelles

LGTM

…2477) * feat: constraints update * constraints update * first commit * various ... * stuff ... * clean up With Lorenzo and François * fix: 99% up to par + @EqualsAndHash(callSuper = true) One failing test (MODEXP limits) * fix: ModexpTests the old tests were overwriting mbs in memory ! * feat: fix ZkCounter event detection * spotless * spotless ... again ? * feat: Osaka MODEXP pricing OOB call * stuff * spotless * setting TOT in MODEXP's MMU instructions (DATA and ZERO) * small stuff * bug: use OsakaBlakeModexpOperation for Osaka rather than LondonBla * feat: most tests are passing * feat: constraints update * merge arith-dev * constraints update * feat: byteMultiplier of leadLog in MODEXP made fork dependent leadLog is essentially defined as log_2(exponent_leading_word) + multiplier * (ebs - 32) with multiplier = 8 pre-Osaka, and 16 post-Osaka * feat: ugly test fixes * feat: fork dependent MODEXP_PRICING byte multiplier m * (ebs - 32) * feat: fork dependent xbsNonzero for MODEXP_XBS * ras: constraints commit update * feat: add fork field to ZkCounter class * spotless * fix: use fork appropriate gasLimit in LowGasStipendPrecompileCallTests * fix: disable parallelism (enabled by mistake) * wip: fix of huge gas consumption in 'preCallProgramGasMap' * added meaningful input for call to POINT_EVALUATION precompile in SixtyThreeSixtyFourthsPrecompileTests * fix: initialize KZG * constraints update * ras: TracerTestBase now on OSAKA * review comments * update constraints Signed-off-by: F Bojarski <[email protected]> --------- Signed-off-by: F Bojarski <[email protected]> Co-authored-by: Lorenzo Gentile <[email protected]> Co-authored-by: F Bojarski <[email protected]> Co-authored-by: François Bojarski <[email protected]>

OlivierBBB added 3 commits November 7, 2025 15:49

feat: constraints update

089f81c

constraints update

bd40fca

first commit

4718cd1

OlivierBBB self-assigned this Nov 12, 2025

OlivierBBB added hardfork Osaka modexp labels Nov 12, 2025

cursor bot reviewed Nov 12, 2025

View reviewed changes

letypequividelespoubelles reviewed Nov 13, 2025

View reviewed changes

...tization/src/main/java/net/consensys/linea/zktracer/module/hub/fragment/imc/mmu/MmuCall.java Show resolved Hide resolved

letypequividelespoubelles reviewed Nov 13, 2025

View reviewed changes

...net/consensys/linea/zktracer/module/hub/precompiles/modexpMetadata/LondonModexpMetadata.java Show resolved Hide resolved

various ...

0bd9eef

cursor bot reviewed Nov 13, 2025

View reviewed changes

...n/java/net/consensys/linea/zktracer/module/blake2fmodexpdata/LondonBlakeModexpOperation.java Outdated Show resolved Hide resolved

OlivierBBB added 2 commits November 14, 2025 09:07

stuff ...

56978e9

clean up

1bcb4a3

With Lorenzo and François

cursor bot reviewed Nov 14, 2025

View reviewed changes

arithmetization/src/main/java/net/consensys/linea/zktracer/ZkCounter.java Outdated Show resolved Hide resolved

OlivierBBB added 3 commits November 14, 2025 20:01

fix: 99% up to par + @EqualsAndHash(callSuper = true)

5be004c

One failing test (MODEXP limits)

fix: ModexpTests

de62eed

the old tests were overwriting mbs in memory !

feat: fix ZkCounter event detection

613c290

cursor bot reviewed Nov 14, 2025

View reviewed changes

spotless

7898299

cursor bot reviewed Nov 15, 2025

View reviewed changes

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java Outdated Show resolved Hide resolved

spotless ... again ?

09ba3c8

cursor bot reviewed Nov 15, 2025

View reviewed changes

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java Outdated Show resolved Hide resolved

...ensys/linea/zktracer/module/hub/section/call/precompileSubsection/OsakaModexpSubsection.java Outdated Show resolved Hide resolved

OlivierBBB added 3 commits November 16, 2025 20:06

feat: Osaka MODEXP pricing OOB call

799e5ae

stuff

b9351dc

spotless

448e1bc

cursor bot reviewed Nov 16, 2025

View reviewed changes

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java Show resolved Hide resolved

setting TOT in MODEXP's MMU instructions (DATA and ZERO)

516f1ef

cursor bot reviewed Nov 16, 2025

View reviewed changes

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java Show resolved Hide resolved

small stuff

c7374ee

cursor bot reviewed Nov 17, 2025

View reviewed changes

...nsys/linea/zktracer/module/hub/section/call/precompileSubsection/LondonModexpSubsection.java Show resolved Hide resolved

OlivierBBB requested review from amkCha and lorenzogentile404 November 18, 2025 16:29

OlivierBBB added 3 commits November 18, 2025 17:31

spotless

5fbf41e

fix: use fork appropriate gasLimit in LowGasStipendPrecompileCallTests

3c7db9d

Merge branch 'arith-dev' into EIP-7823-and-7883---Osaka-MODEXP-changes

1f0bb09

letypequividelespoubelles reviewed Nov 19, 2025

View reviewed changes

OlivierBBB added 2 commits November 19, 2025 18:31

fix: disable parallelism (enabled by mistake)

cc31318

wip: fix of huge gas consumption in 'preCallProgramGasMap'

af5ecfb

cursor bot reviewed Nov 19, 2025

View reviewed changes

lorenzogentile404 and others added 4 commits November 19, 2025 19:12

added meaningful input for call to POINT_EVALUATION precompile in Six…

fc8217f

…tyThreeSixtyFourthsPrecompileTests

fix: initialize KZG

54c8057

constraints update

18ea99c

ras: TracerTestBase now on OSAKA

cfa8e8e

cursor bot reviewed Nov 19, 2025

View reviewed changes

OlivierBBB mentioned this pull request Nov 20, 2025

Give name to "byte weight" constant (8/16) in ≈ log_2(exponent) expression for MODEXP Consensys/linea-specification#325

Open

review comments

8ce1623

OlivierBBB force-pushed the EIP-7823-and-7883---Osaka-MODEXP-changes branch from 9ffa4a4 to 8ce1623 Compare November 20, 2025 10:25

cursor bot reviewed Nov 20, 2025

View reviewed changes

amkCha reviewed Nov 20, 2025

View reviewed changes

letypequividelespoubelles mentioned this pull request Nov 20, 2025

Refactor the MmuCall to pass a ModexMetaData #2499

Open

update constraints

71c5055

Signed-off-by: F Bojarski <[email protected]>

letypequividelespoubelles approved these changes Nov 20, 2025

View reviewed changes

Merge branch 'arith-dev' into EIP-7823-and-7883---Osaka-MODEXP-changes

2fedec6

letypequividelespoubelles enabled auto-merge (squash) November 20, 2025 11:32

letypequividelespoubelles disabled auto-merge November 20, 2025 11:32

letypequividelespoubelles enabled auto-merge (squash) November 20, 2025 11:40

letypequividelespoubelles disabled auto-merge November 20, 2025 11:41

letypequividelespoubelles merged commit c2541bc into arith-dev Nov 20, 2025
14 checks passed

letypequividelespoubelles deleted the EIP-7823-and-7883---Osaka-MODEXP-changes branch November 20, 2025 12:12

EIP-7823 and EIP-7883 implementation --- Osaka MODEXP changes #2477

EIP-7823 and EIP-7883 implementation --- Osaka MODEXP changes #2477

Uh oh!

Conversation

OlivierBBB commented Nov 12, 2025 • edited by cursor bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor bot Nov 19, 2025

Choose a reason for hiding this comment

Bug: Missing maxNumberOfLines override in LondonModexpSubsection

Uh oh!

cursor bot Nov 19, 2025

Choose a reason for hiding this comment

Bug: Missing maxNumberOfLines override in LondonModexpSubsection

Uh oh!

cursor bot Nov 20, 2025

Choose a reason for hiding this comment

Bug: Mismatch between row count formula and actual indices

Uh oh!

amkCha Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

letypequividelespoubelles left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

`EIP-7823` and `EIP-7883` implementation --- Osaka `MODEXP` changes #2477

`EIP-7823` and `EIP-7883` implementation --- Osaka `MODEXP` changes #2477

OlivierBBB commented Nov 12, 2025 •

edited by cursor bot

Loading