EIP-7939 CLZ for Osaka #16122

Saw-mon-and-Natalie · 2025-07-16T03:41:12Z

Fixes #16118.

This PR addresses:

Add support for EIP-7939 CLZ #16118

By adding a new YUL builtin function clz for EVM versions greater than or equal to Osaka.

Followed the patterns from #14757 to create these modifications:

EIP 4844 - blobhash opcode (part 1) #14757

Tests

Currently the tests in semanticTests/inlineAssembly/clz.sol fail due to evmone not having implemented CLZ:

https://github.com/ethereum/solidity/blob/779daec70826225a447680d2030f5fdaafca541b/test/EVMHost.cpp#L420

In the above line m_vm.execute(...) returns EVMC_UNDEFINED_INSTRUCTION and the m_vm is:

{
    abi_version: 12,
    name: "evmone",
    version: "0.15.0",
    destroy: ... ,
    execute: ... ,
    get_capabilities: ... ,
    set_option: ...
}

References:

github-actions · 2025-07-16T03:41:22Z

Thank you for your contribution to the Solidity compiler! A team member will follow up shortly.

If you haven't read our contributing guidelines and our review checklist before, please do it now, this makes the reviewing process and accepting your contribution smoother.

If you have any questions or need our help, feel free to post them in the PR or talk to us directly on the #solidity-dev channel on Matrix.

Vectorized · 2025-07-19T02:48:10Z

libevmasm/Instruction.cpp

 	{Instruction::SHL,            {"SHL",             0,  2,   1,  false,      Tier::VeryLow}},
 	{Instruction::SHR,            {"SHR",             0,  2,   1,  false,      Tier::VeryLow}},
 	{Instruction::SAR,            {"SAR",             0,  2,   1,  false,      Tier::VeryLow}},
+	{Instruction::CLZ,            {"CLZ",             0,  1,   1,  false,      Tier::Low}},


CLZ has the same gas cost as ADD, which is in the Tier::VeryLow.

I think it was updated. Also it is 5 in geth. Plus in the EIP it mentions:

The cost of the opcode is 5, matching MUL (raised from 3 to avoid under-pricing DoS risk).

The linked EIP still mentions a gas cost similar to ADD, and as @Vectorized pointed out, it is indeed 3, i.e. Tier::VeryLow. Which is also the same used by evmone. Do you have the source for the change to 5, @Saw-mon-and-Natalie?

Which is also the same used by evmone.

Oh, I see that in the PR it was actually indeed changed to 5: https://github.com/ipsilon/evmone/pull/1264/files#diff-d47e4d7f522e531f3080bb99145762aef9857717b9e9218a0e90e09e58fa865eR176

this is fine

It's the execution spec that should be the source of truth for us here. count_leading_zeros() uses GAS_LOW, so that's the right value.

But note that it matters whether it's LOW or independently defined as 5. If it was the latter and it had its own constant, we would have to introduce a distinct tier for it. Technically, a future repricing could change the values assigned to those tiers. The cost of CLZ should change in such a case only if it is explicitly defined as a part of such a tier.

matheusaaguiar

Hi @Saw-mon-and-Natalie , thanks for the PR!
Looking good.

Changelog.md

docs/yul.rst

Saw-mon-and-Natalie · 2025-07-23T23:30:17Z

clz also got added to evmone which should solve the test issues (see PR description) here in this PR. But it has not assigned a new version I think yet. So in this repo, one might need to use the compiled binary from the HEAD of that repo.

EIP-7939: added CLZ opcode for Osaka ipsilon/evmone#1264

r0qs · 2025-10-14T12:34:58Z

Hi @Saw-mon-and-Natalie could you please rebase the PR? Or would you prefer if we take it over?

libevmasm/Instruction.h

clonker

generally lgtm

libyul/backends/evm/EVMDialect.cpp

test/libyul/yulSyntaxTests/clz_pre_osaka.yul

Changelog.md

cameel · 2025-10-15T12:57:17Z

libevmasm/Instruction.cpp

 	{Instruction::SHL,            {"SHL",             0,  2,   1,  false,      Tier::VeryLow}},
 	{Instruction::SHR,            {"SHR",             0,  2,   1,  false,      Tier::VeryLow}},
 	{Instruction::SAR,            {"SAR",             0,  2,   1,  false,      Tier::VeryLow}},
+	{Instruction::CLZ,            {"CLZ",             0,  1,   1,  false,      Tier::Low}},


It's the execution spec that should be the source of truth for us here. count_leading_zeros() uses GAS_LOW, so that's the right value.

But note that it matters whether it's LOW or independently defined as 5. If it was the latter and it had its own constant, we would have to introduce a distinct tier for it. Technically, a future repricing could change the values assigned to those tiers. The cost of CLZ should change in such a case only if it is explicitly defined as a part of such a tier.

also resolve a whitespace issue in one test

github-actions bot added the external contribution ⭐ label Jul 16, 2025

Vectorized reviewed Jul 19, 2025

View reviewed changes

matheusaaguiar reviewed Jul 23, 2025

View reviewed changes

Changelog.md Outdated Show resolved Hide resolved

docs/yul.rst Outdated Show resolved Hide resolved

docs/yul.rst Outdated Show resolved Hide resolved

cameel mentioned this pull request Oct 6, 2025

Update evmone in CI to 0.16.0 #16230

Closed

clonker reviewed Oct 15, 2025

View reviewed changes

libevmasm/Instruction.h Show resolved Hide resolved

clonker reviewed Oct 15, 2025

View reviewed changes

libyul/backends/evm/EVMDialect.cpp Outdated Show resolved Hide resolved

test/libyul/yulSyntaxTests/clz_pre_osaka.yul Outdated Show resolved Hide resolved

cameel reviewed Oct 15, 2025

View reviewed changes

clonker changed the base branch from develop to bump-evmone October 16, 2025 07:01

clonker added the has dependencies The PR depends on other PRs that must be merged first label Oct 16, 2025

clonker force-pushed the EIP-7939-CLZ branch from e6dace3 to 3024350 Compare October 16, 2025 07:03

clonker deleted the branch argotorg:develop October 16, 2025 07:34

clonker closed this Oct 16, 2025

clonker reopened this Oct 16, 2025

clonker changed the base branch from bump-evmone to develop October 16, 2025 07:47

clonker force-pushed the EIP-7939-CLZ branch from 3024350 to 05ad057 Compare October 16, 2025 07:47

cameel approved these changes Oct 16, 2025

View reviewed changes

Saw-mon-and-Natalie and others added 3 commits October 16, 2025 11:49

add CLZ to YUL build-in instruction for Osaka

049980b

some minor adjustments for Changelog and Yul docs for CLZ

4f5ea4a

Yul: Implement CLZ exception in EVMDialect with hasCLZ function

9242939

also resolve a whitespace issue in one test

clonker force-pushed the EIP-7939-CLZ branch from 05ad057 to 9242939 Compare October 16, 2025 09:50

clonker enabled auto-merge October 16, 2025 09:50

clonker merged commit 728ecd6 into argotorg:develop Oct 16, 2025
62 of 74 checks passed

EIP-7939 CLZ for Osaka #16122

EIP-7939 CLZ for Osaka #16122

Uh oh!

Conversation

Saw-mon-and-Natalie commented Jul 16, 2025 • edited by cameel Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Tests

References:

Uh oh!

github-actions bot commented Jul 16, 2025

Uh oh!

Vectorized Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Saw-mon-and-Natalie Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

r0qs Oct 14, 2025 • 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!

r0qs Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

clonker Oct 15, 2025

Choose a reason for hiding this comment

Uh oh!

cameel Oct 15, 2025

Choose a reason for hiding this comment

Uh oh!

matheusaaguiar left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Saw-mon-and-Natalie commented Jul 23, 2025

Uh oh!

r0qs commented Oct 14, 2025

Uh oh!

Uh oh!

clonker left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cameel Oct 15, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Saw-mon-and-Natalie commented Jul 16, 2025 •

edited by cameel

Loading

r0qs Oct 14, 2025 •

edited

Loading