feat(avm executor): kernel outputs & execution hints in executor #6769

Maddiaa0 · 2024-05-30T14:09:53Z

Please read contributing guidelines and remove this line.

Maddiaa0 · 2024-05-30T14:10:17Z

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @Maddiaa0 and the rest of your teammates on Graphite

AztecBot · 2024-05-30T16:45:26Z

Benchmark results

Metrics with a significant change:

proof_construction_time_poseidon_hash_ms (4): 47.0 (+34%)
protocol_circuit_witness_generation_time_in_ms (base-parity): 1,776 (+97%)

Detailed results

All benchmarks are run on txs on the Benchmarking contract on the repository. Each tx consists of a batch call to create_note and increment_balance, which guarantees that each tx has a private call, a nested private call, a public call, and a nested public call, as well as an emitted private note, an unencrypted log, and public storage read and write.

This benchmark source data is available in JSON format on S3 here.

Proof generation

Each column represents the number of threads used in proof generation.

Metric	1 threads	4 threads	16 threads	32 threads	64 threads
proof_construction_time_sha256_ms	5,705 (-2%)	1,546 (-1%)	695 (-2%)	756	776 (-1%)
proof_construction_time_sha256_30_ms	11,702 (-4%)	3,163 (-1%)	1,400 (-1%)	1,422 (-2%)	1,461 (-2%)
proof_construction_time_sha256_100_ms	43,636 (-5%)	11,734 (-5%)	5,411 (-6%)	5,389 (-4%)	5,343 (-4%)
proof_construction_time_poseidon_hash_ms	78.0 (-1%)	⚠️ 47.0 (+34%)	34.0	57.0 (-2%)	87.0
proof_construction_time_poseidon_hash_30_ms	1,516 (-1%)	417 (-2%)	201 (-1%)	231 (+2%)	267 (-1%)
proof_construction_time_poseidon_hash_100_ms	5,712 (-2%)	1,560 (-2%)	715 (-2%)	760 (-7%)	795 (-1%)

L2 block published to L1

Each column represents the number of txs on an L2 block published to L1.

Metric	8 txs	32 txs	64 txs
l1_rollup_calldata_size_in_bytes	1,412	1,412	1,412
l1_rollup_calldata_gas	9,452	9,464	9,476
l1_rollup_execution_gas	607,988	608,000	608,012
l2_block_processing_time_in_ms	1,296 (-1%)	4,861	9,657
l2_block_building_time_in_ms	44,330	175,342	350,313
l2_block_rollup_simulation_time_in_ms	44,135	174,614	348,874
l2_block_public_tx_process_time_in_ms	40,887	171,326	345,521

L2 chain processing

Each column represents the number of blocks on the L2 chain where each block has 16 txs.

Metric	3 blocks	5 blocks
node_history_sync_time_in_ms	9,433 (-3%)	14,391 (-2%)
node_database_size_in_bytes	14,495,824	21,389,392
pxe_database_size_in_bytes	18,071	29,868

Circuits stats

Stats on running time and I/O sizes collected for every kernel circuit run across all benchmarks.

Circuit	simulation_time_in_ms	witness_generation_time_in_ms	proving_time_in_ms	input_size_in_bytes	output_size_in_bytes	proof_size_in_bytes	num_public_inputs	size_in_gates
private-kernel-init	137 (+2%)	500 (+1%)	12,194 (-1%)	20,634	64,614	89,536	2,731	524,288
private-kernel-inner	409	1,087 (+8%)	45,014 (+1%)	92,326	64,614	89,536	2,731	2,097,152
private-kernel-tail	588	2,959 (+5%)	46,943 (-1%)	96,545	77,732	11,648	297	2,097,152
base-parity	6.50 (-1%)	⚠️ 1,776 (+97%)	2,575	128	64.0	2,208	2.00	131,072
root-parity	49.7 (-1%)	49.6 (-1%)	34,225 (-1%)	27,100	64.0	2,720	18.0	2,097,152
base-rollup	12,065 (+1%)	2,336 (-1%)	65,331 (+1%)	119,738	756	3,648	47.0	4,194,304
root-rollup	112	62.0 (-3%)	19,705 (+1%)	25,309	620	3,456	41.0	1,048,576
public-kernel-app-logic	575	3,507 (-1%)	38,745 (-2%)	108,073	86,550	116,768	3,582	2,097,152
public-kernel-tail	1,134	22,336 (-4%)	153,928 (-1%)	403,238	7,646	11,648	297	8,388,608
private-kernel-reset-small	592	2,167 (+3%)	40,789 (+1%)	120,737	64,614	89,536	2,731	2,097,152
public-kernel-setup	672 (-1%)	2,655 (-1%)	37,429	108,073	86,550	116,768	3,582	2,097,152
public-kernel-teardown	575 (-1%)	3,532 (-1%)	39,413	108,073	86,550	116,768	3,582	2,097,152
merge-rollup	29.8 (+3%)	N/A	N/A	16,542	756	N/A	N/A	N/A
private-kernel-tail-to-public	N/A	8,736	86,953 (+1%)	N/A	N/A	116,768	3,582	4,194,304

Stats on running time collected for app circuits

Function	input_size_in_bytes	output_size_in_bytes	witness_generation_time_in_ms	proof_size_in_bytes	proving_time_in_ms	size_in_gates	num_public_inputs
ContractClassRegisterer:register	1,344	9,944	464 (-1%)	N/A	N/A	N/A	N/A
ContractInstanceDeployer:deploy	1,408	9,944	42.2	N/A	N/A	N/A	N/A
MultiCallEntrypoint:entrypoint	1,920	9,944	1,800	N/A	N/A	N/A	N/A
SchnorrAccount:constructor	1,312	9,944	1,467 (+1%)	N/A	N/A	N/A	N/A
SchnorrAccount:entrypoint	2,304	9,944	2,751	16,768	50,077 (-1%)	2,097,152	457
Token:privately_mint_private_note	1,280	9,944	1,657 (+1%)	N/A	N/A	N/A	N/A
FPC:fee_entrypoint_public	1,344	9,944	1,152 (+1%)	16,768	10,127 (+1%)	524,288	457
Token:transfer	1,376	9,944	5,460	16,768	48,604 (+2%)	2,097,152	457
Benchmarking:create_note	1,344	9,944	1,389 (-1%)	N/A	N/A	N/A	N/A
SchnorrAccount:spend_private_authwit	1,280	9,944	76.3 (-7%)	N/A	N/A	N/A	N/A
Token:unshield	1,376	9,944	3,874 (-4%)	N/A	N/A	N/A	N/A
FPC:fee_entrypoint_private	1,376	9,944	4,771 (-5%)	N/A	N/A	N/A	N/A

Tree insertion stats

The duration to insert a fixed batch of leaves into each tree type.

Metric	1 leaves	16 leaves	64 leaves	128 leaves	512 leaves	1024 leaves	2048 leaves	4096 leaves	32 leaves
batch_insert_into_append_only_tree_16_depth_ms	10.5	17.4	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_count	16.7	31.8	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_ms	0.609	0.522	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	48.8 (-1%)	76.8 (+1%)	248	475 (-1%)	934	1,861	N/A
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	95.9	159	543	1,055	2,079	4,127	N/A
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	0.497	0.473 (+1%)	0.449	0.443 (-1%)	0.442	0.443	N/A
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	58.6	113	358	698 (-1%)	1,411 (+1%)	2,787	N/A
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	106	208	692	1,363	2,707	5,395	N/A
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.506	0.505	0.485	0.479 (-1%)	0.480	0.483	N/A
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	63.0
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	107
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	0.556

Miscellaneous

Transaction sizes based on how many contract classes are registered in the tx.

Metric	0 registered classes	1 registered classes
tx_size_in_bytes	84,053	665,267

Transaction size based on fee payment method

| Metric | |
| - | |

Maddiaa0 · 2024-05-31T13:04:29Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_common.hpp

@@ -35,4 +36,31 @@ static const size_t NUM_MEM_SPACES = 256;
 static const uint8_t INTERNAL_CALL_SPACE_ID = 255;
 static const uint32_t MAX_SIZE_INTERNAL_STACK = 1 << 16;

+// TODO: find a more suitable place for these
+/**
+ * Auxiliary hints are required when the executor is unable to work out a witness value


Maddiaa0 · 2024-05-31T13:18:11Z

barretenberg/cpp/pil/avm/avm_kernel.pil

-    pol START_L1_TO_L2_MSG_EXISTS_WRITE_OFFSET = 16;
-    pol START_EMIT_UNENCRYPTED_LOG_WRITE_OFFSET = 20;
-    pol START_EMIT_L2_TO_l1_MSG = 24;
+    pol START_NULLIFIER_EXISTS_OFFSET = 32; // START_NOTE_HASH_EXISTS_WRITE_OFFSET + MAX_NOTE_HASH_READ_REQUESTS_PER_CALL TODO: exists and non exists


Adjusted these values to line up with the true offsets when using the real PER_CALL limits, previously they were just small ranges as an MVP

Can you now autogen these from TS?

fcarreiro · 2024-05-31T13:31:49Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_common.hpp

+    ExecutionHints()
+    {
+        storage_values = std::unordered_map<uint32_t, FF>();
+        note_hash_exists = std::unordered_map<uint32_t, bool>();
+        nullifier_exists = std::unordered_map<uint32_t, bool>();
+        l1_to_l2_msg_exists = std::unordered_map<uint32_t, bool>();
+    }


This is not needed. They are objects, not pointers, so they get default-constructed. This is extra work.

dbanks12

I'm not in a position to "approve", but this mostly makes sense to me. Nice work! Left some comments to primarily confirm my understanding

dbanks12 · 2024-05-31T13:25:28Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_kernel_trace.cpp

+    info("output writing side effect counter: ", side_effect_counter);
+


Not related to your changes, but can we handle side-effect-counters across nested calls? So a nested call starts with parent's counter+1, and then after it completes, parent absorbs child's counter

Right now, it just starts from 0, but there is an issue to constrain it to be the same as the value provided within the public inputs

dbanks12 · 2024-05-31T13:25:45Z

barretenberg/cpp/pil/avm/avm_kernel.pil

+    // Exists checks
    pol START_NOTE_HASH_EXISTS_WRITE_OFFSET = 0;
-    pol START_EMIT_NOTE_HASH_WRITE_OFFSET = 4;
-    pol START_NULLIFIER_EXISTS_OFFSET = 8;
-    pol START_EMIT_NULLIFIER_WRITE_OFFSET = 12;
-    pol START_L1_TO_L2_MSG_EXISTS_WRITE_OFFSET = 16;
-    pol START_EMIT_UNENCRYPTED_LOG_WRITE_OFFSET = 20;
-    pol START_EMIT_L2_TO_l1_MSG = 24;
+    pol START_NULLIFIER_EXISTS_OFFSET = 32; // START_NOTE_HASH_EXISTS_WRITE_OFFSET + MAX_NOTE_HASH_READ_REQUESTS_PER_CALL TODO: exists and non exists
+    pol START_L1_TO_L2_MSG_EXISTS_WRITE_OFFSET = 96; //  START_NULLIFIER_EXISTS_OFFET + (MAX_NULLIFIER_READ_REQUESTS_PER_CALL + MAX_NULLIFIER_NON_EXISTENT_READ_REQUESTS_PER_CALL)
+
+    // Public storage requests
+    pol START_SLOAD_WRITE_OFFSET = 112; // START_L1_TO_L2_MSG_EXISTS_WRITE_OFFSET + MAX_L1_TO_L2_MSG_READ_REQUESTS_PER_CALL
+    pol START_SSTORE_WRITE_OFFSET = 144; // START_SSTORE_WRITE_OFFSET + MAX_PUBLIC_DATA_UPDATE_REQUESTS_PER_CALL
+
+    // Emit data 
+    pol START_EMIT_NOTE_HASH_WRITE_OFFSET = 160; // START_SLOAD_WRITE_OFFSET + MAX_PUBLIC_DATA_READS_PER_CALL
+    pol START_EMIT_NULLIFIER_WRITE_OFFSET = 176; // START_EMIT_NOTE_HASH_WRITE_OFFSET + MAX_NEW_NOTE_HASHES_PER_CALL
+    pol START_EMIT_L2_TO_l1_MSG = 192; // START_EMIT_NULLIFIER_WRITE_OFFSET + MAX_NEW_NULLIFIERS_PER_CALL
+    pol START_EMIT_UNENCRYPTED_LOG_WRITE_OFFSET = 194; // START_EMIT_L2_TO_L1_MSG + MAX_NEW_L2_TO_L1_MSGS_PER_CALL


Are these hand-set or code-generated?

Right now hand set, but they mirror values in cpp, so you just hover over the value in the LSP and copy it over, our pil does not compile time evaluate constants yet

dbanks12 · 2024-05-31T13:32:57Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_trace.cpp

+// TODO: fix the naming here - we need it to be different as we are writing a hint
+Row AvmTraceBuilder::create_sload(
+    uint32_t clk, uint32_t data_offset, FF const& data_value, FF const& slot_value, uint32_t slot_offset)
+{


Is this function just populating memory rows for slot and value? Is the row that is returned here, the sload's row in the main trace or is it a row dedicated just to loading the hint into memory?

This is temporary as sload had slightly differing behavior to the existing ones, it is called within op_sload which orchestrates it all

dbanks12 · 2024-05-31T14:18:32Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_trace.cpp

-void AvmTraceBuilder::op_sload(uint32_t slot_offset, uint32_t value_offset)
+void AvmTraceBuilder::op_sload(uint32_t slot_offset, uint32_t write_offset)
 {
    auto const clk = static_cast<uint32_t>(main_trace.size());

-    Row row =
-        create_kernel_output_opcode_with_metadata(clk, value_offset, AvmMemoryTag::FF, slot_offset, AvmMemoryTag::FF);
-    kernel_trace_builder.op_sload(clk, row.avm_main_ib, row.avm_main_ia);
+    // Read the slot
+    AvmMemTraceBuilder::MemRead slot_read = mem_trace_builder.read_and_load_from_memory(
+        call_ptr, clk, IntermRegister::IB, slot_offset, AvmMemoryTag::FF, AvmMemoryTag::FF);
+
+    // Get the data value from the execution_hints
+    // TODO: for now the hints are being offset by the offset - this will NOT fly, but i struggled to get the hash
+    // working for FF
+    FF value = execution_hints.storage_values.at(write_offset);
+    // TODO: throw error if the hint does not exist
+
+    Row row = create_sload(clk, write_offset, value, slot_read.val, slot_offset);
+    kernel_trace_builder.op_sload(clk, row.avm_main_ib, value);
+
    row.avm_main_sel_op_sload = FF(1);

    main_trace.push_back(row);


My attempt at understanding this:

Use the instruction's slot_offset operand and read slot from memory (adds row to memory trace)

Use that slot to retrieve a storage value from execution_hints.storage_values

Write the hinted value into memory (adds row to memory trace)

Add row to kernel trace for this storage slot->value "kernel output lookup"

Create a new row in the main trace for the sload, including mem read for slot, mem write for value, and flag to indicate presence of a "kernel output lookup"

Activate the selector for sload in this row

Yep this is pretty spot on

dbanks12 · 2024-05-31T14:24:33Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_trace.cpp

+            // Increment the write offset counter for the following row
+            next.avm_kernel_note_hash_exist_write_offset =
+                curr.avm_kernel_note_hash_exist_write_offset + static_cast<const int>(src.op_note_hash_exists);
+            next.avm_kernel_emit_note_hash_write_offset =
+                curr.avm_kernel_emit_note_hash_write_offset + static_cast<const int>(src.op_emit_note_hash);
+            next.avm_kernel_emit_nullifier_write_offset =
+                curr.avm_kernel_emit_nullifier_write_offset + static_cast<const int>(src.op_emit_nullifier);
+            next.avm_kernel_nullifier_exists_write_offset =
+                curr.avm_kernel_nullifier_exists_write_offset + static_cast<const int>(src.op_nullifier_exists);
+            next.avm_kernel_l1_to_l2_msg_exists_write_offset =
+                curr.avm_kernel_l1_to_l2_msg_exists_write_offset + static_cast<const int>(src.op_l1_to_l2_msg_exists);
+            next.avm_kernel_emit_l2_to_l1_msg_write_offset =
+                curr.avm_kernel_emit_l2_to_l1_msg_write_offset + static_cast<const int>(src.op_emit_l2_to_l1_msg);
+            next.avm_kernel_emit_unencrypted_log_write_offset =
+                curr.avm_kernel_emit_unencrypted_log_write_offset + static_cast<const int>(src.op_emit_unencrypted_log);
+            next.avm_kernel_sload_write_offset =
+                curr.avm_kernel_sload_write_offset + static_cast<const int>(src.op_sload);
+            next.avm_kernel_sstore_write_offset =
+                curr.avm_kernel_sstore_write_offset + static_cast<const int>(src.op_sstore);
+


We increment the write offset counter so that the next kernel lookup in the main trace is constrained to the proper entry in the kernel trace?

noir-projects/noir-protocol-circuits/crates/types/src/constants.nr

fcarreiro · 2024-06-02T21:36:56Z

barretenberg/cpp/pil/avm/constants.pil

I mean these

Maddiaa0 · 2024-06-03T15:15:52Z

barretenberg/cpp/pil/avm/avm_kernel.pil


-    pol START_SLOAD_WRITE_OFFSET = 28;
-    pol START_SSTORE_WRITE_OFFSET = 32;

    // TODO(https://github.com/AztecProtocol/aztec-packages/issues/6465): Must constrain write_offset counters to be less than side effect MAX


just reordered

Maddiaa0 · 2024-06-03T15:17:34Z

barretenberg/cpp/src/barretenberg/vm/avm_trace/avm_kernel_trace.cpp

@@ -33,16 +33,17 @@ FF AvmKernelTraceBuilder::perform_kernel_input_lookup(uint32_t selector)
    return result;
 }

-void AvmKernelTraceBuilder::perform_kernel_output_lookup(uint32_t write_offset, const FF& value, const FF& metadata)
+void AvmKernelTraceBuilder::perform_kernel_output_lookup(uint32_t write_offset,


side effect counter has been moved into the main trace

This was referenced May 30, 2024

feat(vm): executor for public kernel inputs #6716

Merged

chore: add note hash read requests to public circuit public inputs #6754

Merged

Maddiaa0 mentioned this pull request May 30, 2024

chore: add l1 to l2 msg read requests to public circuit public inputs #6762

Merged

Maddiaa0 force-pushed the md/kernel-inputs-executor branch 2 times, most recently from 42c3d82 to d2355ee Compare May 30, 2024 15:28

Maddiaa0 force-pushed the md/kernel-outputs-executor branch from e28c940 to 5941499 Compare May 30, 2024 15:28

Base automatically changed from md/kernel-inputs-executor to md/05-30-chore_add_l1_to_l2_msg_read_requests_to_public_circuit_public_inputs May 30, 2024 15:31

Base automatically changed from md/05-30-chore_add_l1_to_l2_msg_read_requests_to_public_circuit_public_inputs to master May 30, 2024 16:25

Maddiaa0 force-pushed the md/kernel-outputs-executor branch from 5941499 to 27392a6 Compare May 31, 2024 13:03

Maddiaa0 commented May 31, 2024

View reviewed changes

Maddiaa0 changed the title ~~feat(avm executor): kernel outputs in executor~~ feat(avm executor): kernel outputs & execution hints in executor May 31, 2024

Maddiaa0 commented May 31, 2024

View reviewed changes

fcarreiro reviewed May 31, 2024

View reviewed changes

dbanks12 reviewed May 31, 2024

View reviewed changes

Maddiaa0 mentioned this pull request Jun 2, 2024

chore: delete spike vm #6818

Merged

fcarreiro reviewed Jun 2, 2024

View reviewed changes

barretenberg/cpp/pil/avm/constants.pil Outdated

Copy link

Contributor

fcarreiro Jun 2, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I mean these

Maddiaa0 added 12 commits June 3, 2024 15:07

feat: add test that state getting opcodes are executed correctly

1ba1882

sweep

f955016

chore: add emit log execution tests

6da5d10

temp

c813a44

fix: merge

f59cb2e

fix: align offsets with real values

33c9c01

feat: mvp execution hints

494f966

feat: sload and sstore execution

fe02298

fix: update sstore side effect counter

69a0ce0

temp

18fef35

feat: kernel exists opcodes with hints

81d1ff5

feat: update execution hints to map side effect counter to value

e9c3acd

chore: rebase fix, update constants

aca948c

Maddiaa0 force-pushed the md/kernel-outputs-executor branch from a89221e to aca948c Compare June 3, 2024 15:15

Maddiaa0 commented Jun 3, 2024

View reviewed changes

Maddiaa0 added 3 commits June 3, 2024 15:39

fix: generalise sload

c4816e7

chore: cleanup kernel test

390ec2d

chore: revert bb main.cpp changes

c529842

Maddiaa0 marked this pull request as ready for review June 3, 2024 15:59

Maddiaa0 added 3 commits June 3, 2024 16:59

Merge branch 'master' into md/kernel-outputs-executor

dd72cf8

Merge branch 'master' into md/kernel-outputs-executor

c258dd4

fix: gcc build error - no int cast

36060f1

IlyasRidhuan approved these changes Jun 3, 2024

View reviewed changes

Maddiaa0 enabled auto-merge (squash) June 3, 2024 17:09

Maddiaa0 merged commit 6ab7360 into master Jun 3, 2024
111 checks passed

Maddiaa0 deleted the md/kernel-outputs-executor branch June 3, 2024 17:20

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(avm executor): kernel outputs & execution hints in executor #6769

feat(avm executor): kernel outputs & execution hints in executor #6769

Maddiaa0 commented May 30, 2024

Maddiaa0 commented May 30, 2024 •

edited

Loading

AztecBot commented May 30, 2024 •

edited

Loading

Proof generation

L2 block published to L1

L2 chain processing

Circuits stats

Tree insertion stats

Miscellaneous

Maddiaa0 May 31, 2024

Maddiaa0 May 31, 2024

fcarreiro Jun 2, 2024

fcarreiro May 31, 2024

dbanks12 left a comment

dbanks12 May 31, 2024

Maddiaa0 May 31, 2024

dbanks12 May 31, 2024

Maddiaa0 May 31, 2024 •

edited

Loading

dbanks12 May 31, 2024

Maddiaa0 May 31, 2024

dbanks12 May 31, 2024

Maddiaa0 May 31, 2024

dbanks12 May 31, 2024

Maddiaa0 May 31, 2024

fcarreiro Jun 2, 2024

Maddiaa0 Jun 3, 2024

Maddiaa0 Jun 3, 2024

		info("output writing side effect counter: ", side_effect_counter);

feat(avm executor): kernel outputs & execution hints in executor #6769

feat(avm executor): kernel outputs & execution hints in executor #6769

Conversation

Maddiaa0 commented May 30, 2024

Maddiaa0 commented May 30, 2024 • edited Loading

AztecBot commented May 30, 2024 • edited Loading

Benchmark results

Proof generation

L2 block published to L1

L2 chain processing

Circuits stats

Tree insertion stats

Miscellaneous

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dbanks12 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Maddiaa0 May 31, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Maddiaa0 commented May 30, 2024 •

edited

Loading

AztecBot commented May 30, 2024 •

edited

Loading

Maddiaa0 May 31, 2024 •

edited

Loading