SIMD-0339: Increase CPI Account Infos Limit by nbelenkov · Pull Request #8513 · anza-xyz/agave

nbelenkov · 2025-10-16T11:59:05Z

Problem

Adding SIMD-0339 functionality

Summary of Changes

Feature gate added called increase_cpi_info_account_limit
Max CPI account infos is now increased to 255 info accounts inclusive
Base cost of CPI invocation is lowered from 1000 to 946, to allow for per byte charging of account_info and ix_account_metadata.

Questions/still to do:

Integration tests have some info account sizes hardcoded, I couldn't work out how to pass a constant into the integration test

Lichtso · 2025-10-16T13:03:34Z

    solana_pubkey::declare_id!("8MhfKhoZEoiySpVe248bDkisyEcBA7JQLyUS94xoTSqN");
 }

+pub mod increase_cpi_info_account_limit {


You probably meant AccountInfo not "info account".

do you mean change it to increase_cpi_account_info_limit?

Lichtso · 2025-10-16T13:05:00Z

+        //sizeof(AccountInfo) is 80 bytes
+         let account_infos_bytes = account_infos
+            .len()
+            .saturating_mul(80);


Just insert std::mem::sizeof::<AccountInfo>() then

std::mem::size_of::<AccountInfo>() returns 48, because it contains references to the 2 Pubkey of owner and key, so that's why I didn't use it. The comment is a bit misleading I will update it

ended up using base type to allign with the other calculation and adjusting it to 80 bytes, its a bit messy though

Since 80 bytes is not really correlated with the size of AccountInfo, this should just be a constant.

Lichtso · 2025-10-16T13:07:07Z

+        consume_compute_meter(
+        invoke_context,
+        (account_infos_bytes as u64)
+            .checked_div(invoke_context.get_execution_cost().cpi_bytes_per_unit)


Would rather have this combined with the consume_compute_meter() call below.

Do you mean do the compute unit consumption inside translate_accounts_common()? Or did you mean to combine the 2 in translate_instruction_rust()?

Lichtso · 2025-10-16T13:09:51Z

+        if invoke_context.get_feature_set().increase_cpi_info_account_limit {
+            INVOKE_UNITS_COST_SIMD_0339
+        } else {
+            invoke_context.get_execution_cost().invoke_units


This should be feature gated in ComputeBudget not here. See

agave/compute-budget/src/compute_budget.rs

Line 131 in 8e696b9

pub fn new_with_defaults(simd_0268_active: bool) -> Self {

for an example of a similar feature gate there.

Makes sense, this should be in the right place now

Lichtso · 2025-10-16T13:15:57Z

 static const uint8_t TEST_DUPLICATE_PRIVILEGE_ESCALATION_SIGNER = 20;
 static const uint8_t TEST_DUPLICATE_PRIVILEGE_ESCALATION_WRITABLE = 21;
-static const uint8_t TEST_MAX_ACCOUNT_INFOS_EXCEEDED = 22;
+static const uint8_t TEST_MAX_ACCOUNT_INFOS_EXCEEDED_INCREASE_CPI_INFO = 22;


Generally we try to prepare the code base for an easy cleanup of the feature gates, so everything is named from the perspective of the "new normal". Thus, this should stay TEST_MAX_ACCOUNT_INFOS_EXCEEDED and the other one would be TEST_MAX_ACCOUNT_INFOS_EXCEEDED_BEFORE_INCREASE or something like that.

ohhhh I see what you mean, so now with feature gate enabled is the new norm, hence the previous tests become before the norm right? So we would have the following for example:

TEST_MAX_ACCOUNT_INFOS_SIMD_0339_OK -> new normal ok so TEST_MAX_ACCOUNT_INFOS_OK TEST_MAX_ACCOUNT_INFOS_OK -> the original ok test so, TEST_MAX_ACCOUNT_INFOS_OK_BEFORE_CPI_INCREASE_BEFORE_SIMD_0339

should be updated now

…v/agave into increase_cpi_info_account_limit

LucasSte · 2025-10-16T18:14:05Z

+    if invoke_context.get_feature_set().increase_cpi_info_account_limit{
+        // Each account meta is 34 bytes (32 for pubkey, 1 for is_signer, 1 for is_writable)
+        let account_meta_bytes = account_metas
+            .len()
+            .saturating_mul(34); 


I believe you can use size_of::<AccountMeta>() here.

good point, will update

LucasSte · 2025-10-17T14:23:02Z

Run cargo +nightly fmt --all to format your code after your changes.

LucasSte · 2025-10-17T14:26:33Z

    let max_cpi_account_infos = if invoke_context
        .get_feature_set()
-        .increase_tx_account_lock_limit
+        .increase_cpi_account_info_limit
+    {
+        MAX_CPI_ACCOUNT_INFOS_SIMD_0339
+    } else if invoke_context
+            .get_feature_set()
+            .increase_tx_account_lock_limit
    {
        MAX_CPI_ACCOUNT_INFOS
    } else {


These two feature gates conflict with each other. Is SIMD-0339 supposed to supersede increase_tx_account_lock_limit ?

That's my understanding of how this SIMD-0339 should behave, see here solana-foundation/solana-improvement-documents#339 (comment). But would leave it to @jstarry to confirm

Yes, SIMD-0339 will override the increase_tx_account_lock_limit feature here

LucasSte · 2025-10-17T14:27:21Z

+    if invoke_context.get_feature_set().increase_cpi_account_info_limit{
+        // Each account meta is 34 bytes (32 for pubkey, 1 for is_signer, 1 for is_writable)
+        let account_meta_bytes = ix_c.accounts_len 
+            .saturating_mul(34); 


size_of::<AccountMeta>()?

ah yes, good spot, I missed that one

codecov-commenter · 2025-10-21T14:26:57Z

Codecov Report

❌ Patch coverage is 76.56250% with 30 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.3%. Comparing base (e65003d) to head (7fa3a0d).
⚠️ Report is 1727 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #8513   +/-   ##
=======================================
  Coverage    83.3%    83.3%           
=======================================
  Files         855      855           
  Lines      372188   372286   +98     
=======================================
+ Hits       310272   310364   +92     
- Misses      61916    61922    +6

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

jstarry · 2025-10-22T04:14:43Z

+        //std::mem::size_of::<AccountInfo>() returns 48 bytes, which contains references to the 2 Pubkeys of owner and key,
+        //but we need the full size here so, need to add (32 + 32) bytes for Pubkey types and account for 8 + 8 bytes already existing for refence types.
+        //Hence adding 32 here due to 5 bytes being the padding and 11 bytes being the other data, see SIMD-0339 for calculations.


I don't think it makes sense to use std::mem::size_of::<AccountInfo>() here. The 80 bytes from the SIMD is just a sum of the sizes of pubkey, owner, data length, and lamports because these are the things that are translated from vm to host memory. I think it's better to define a constant with a comment explaining the calculation.

makes sense, I do agree, it looks very messy otherwise. It works nicely for size_of::<AccountMeta>() but not here

jstarry · 2025-10-22T04:15:40Z

        check_aligned,
    )?;

+    if invoke_context


Any particular reason you chose this location rather than putting the metering inside translate_account_infos after the check_account_infos call? I think that would be a better place because then we meter CU's before doing any allocation.

that is a very good point, it makes sense to move it there, it also somewhat mimics the behaviour in translate_instruction_rust/c()

jstarry · 2025-10-22T04:55:25Z

+        ),
+        (
+            increase_cpi_account_info_limit::id(),
+            "SIMD-0339: increase CPI info account limit",


This title doesn't match the SIMD

jstarry · 2025-10-22T04:57:15Z

 }

+pub mod increase_cpi_account_info_limit {
+    solana_pubkey::declare_id!("7wM2pdjwmSXviEdsorpcBY3T4YWUPQXDMepZudub7nGQ"); // Placeholder ID HAVE TO CHANGE BEFORE USE


Can you update this with a real key?

jstarry · 2025-10-25T15:32:02Z

 pub mod discard_unexpected_data_complete_shreds {
    solana_pubkey::declare_id!("8MhfKhoZEoiySpVe248bDkisyEcBA7JQLyUS94xoTSqN");
 }
-


this is very nit picky but this doesn't look intentional, can you undo this?

jstarry · 2025-10-25T15:41:11Z

    consume_compute_meter(
        invoke_context,
-        (data.len() as u64)
+        (total_cu_translation_cost)


Summing the total data size (ix data + account meta's) before dividing by "cpi bytes per unit" could reduce the impact of rounding error.. and that behavior doesn't match the SIMD. Combining the total cost is fine, just make sure you do the division first before summing the ix_data_cost with the ix_account_meta_cost

…st to match SIMDs expected behaviour

SIMD 0339 functionality + integration tests

23f8d97

nbelenkov requested a review from a team as a code owner October 16, 2025 11:59

Merge branch 'master' into increase_cpi_info_account_limit

1c9619b

Lichtso reviewed Oct 16, 2025

View reviewed changes

nbelenkov added 2 commits October 16, 2025 17:30

refactor feature gate to be in ComputeBudget

2f3446d

Merge branch 'increase_cpi_info_account_limit' of github.com:nbelenko…

12b7a09

…v/agave into increase_cpi_info_account_limit

LucasSte reviewed Oct 16, 2025

View reviewed changes

nbelenkov added 3 commits October 17, 2025 12:15

adding types for byte calculation and updated the comment

8a65c68

fix missed invocation with new function signature

0d31630

changed feature name to increase_cpi_account_info_limit

620ff0e

LucasSte reviewed Oct 17, 2025

View reviewed changes

nbelenkov added 4 commits October 20, 2025 12:51

test renaming for new normal prospective + whitespace format

917abad

changing missed hardcoded value to type size on c account translation

b0407e6

fixing CI issue with whitespaces in C program

01466de

adding execution cost initialization to bank creation

df4d61d

jstarry reviewed Oct 22, 2025

View reviewed changes

jstarry changed the title ~~[DRAFT] SIMD-0339: Increase CPI Account Info Limit~~ SIMD-0339: Increase CPI Account Infos Limit Oct 22, 2025

nbelenkov added 7 commits October 22, 2025 20:47

adding CU tests

8e3e3d1

cargo fmt didn't work on some files

3c81a81

c file whitespace

4629daa

using test-v3 for CU estimations

abc7e34

upper bound of CU test due to different sbpf versions

c51ec8c

panic error fix

9071ea7

constant for AccountInfo type size + account info charging relocation

004ac03

nbelenkov added 6 commits October 24, 2025 13:40

merge CU consuming calls into 1

c4b8563

cargo fmt

9076cff

mergin master into this PR

4db843a

cargo fmt the merge

ac84ebd

renamed tests + cargo.lock update

3f722bd

Merge branch 'master' into increase_cpi_info_account_limit

bcf5f2b

jstarry reviewed Oct 25, 2025

View reviewed changes

nbelenkov added 4 commits October 27, 2025 15:51

updated keypair for feature gate + refactored total_cu_translation_co…

634f923

…st to match SIMDs expected behaviour

Merge branch 'master' into increase_cpi_info_account_limit

f44b4c8

fmt after upstream sync

c9e055d

var renaming

171f59b

jstarry approved these changes Oct 27, 2025

View reviewed changes

Merge branch 'master' into increase_cpi_info_account_limit

3e71a15

jstarry previously approved these changes Oct 28, 2025

View reviewed changes

Merge branch 'master' into increase_cpi_info_account_limit

4f474f5

nbelenkov dismissed jstarry’s stale review via 4f474f5 October 28, 2025 12:04

Merge branch 'master' into increase_cpi_info_account_limit

7fa3a0d

jstarry approved these changes Oct 28, 2025

View reviewed changes

LucasSte approved these changes Oct 28, 2025

View reviewed changes

nbelenkov added this pull request to the merge queue Oct 28, 2025

Merged via the queue into anza-xyz:master with commit f12e46b Oct 28, 2025
44 checks passed

nbelenkov deleted the increase_cpi_info_account_limit branch October 28, 2025 16:33

Conversation

nbelenkov commented Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Summary of Changes

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!

nbelenkov Oct 16, 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!

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!

nbelenkov Oct 17, 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!

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!

LucasSte commented Oct 17, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nbelenkov Oct 20, 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nbelenkov Oct 20, 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!

codecov-commenter commented Oct 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nbelenkov Oct 22, 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!

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!

nbelenkov commented Oct 16, 2025 •

edited

Loading

nbelenkov Oct 16, 2025 •

edited

Loading

nbelenkov Oct 17, 2025 •

edited

Loading

nbelenkov Oct 20, 2025 •

edited

Loading

nbelenkov Oct 20, 2025 •

edited

Loading

codecov-commenter commented Oct 21, 2025 •

edited

Loading

nbelenkov Oct 22, 2025 •

edited

Loading