Skip to main content

Profiling Transactions

An Aztec transaction typically consists of a private and a public part. The private part is where the user executes contract logic within the PXE and generates a proof of execution, which is then sent to the sequencer.

Since proof generation is an expensive operation that needs to be done on the client side, it is important to optimize the private contract logic. It is desirable to keep the gate count of circuits representing the private contract logic as low as possible.

A private transaction can involve multiple function calls. It starts with an account entrypoint() which may call several private functions to execute the application logic, which in turn might call other functions. Moreover, every private function call has to go through a round of kernel circuits. Read more about the transaction lifecycle here.

In this guide, we will look at how to profile the private execution of a transaction, allowing you to get the gate count of each private function within the transaction, including the kernel circuits.

Prerequisites​

  • aztec-nargo installed (go to Sandbox section for installation instructions)
  • aztec-wallet installed (installed as part of the Sandbox)

Profiling using aztec-wallet​

The profiling tool is integrated into the aztec-wallet.

In this example, we will profile a simple "private token transfer" transaction which uses the transfer method in the token contract. Let us start by deploying the token contarct (included in the Sandbox) and minting some tokens to the test account.

# Import some test accounts included in cli-wallet
aztec-wallet import-test-accounts

# Deploy a token contract.
aztec-wallet deploy TokenContractArtifact --from accounts:test0 --args accounts:test0 TestToken TST 18 -a token

# Mint some tokens to the test0 account
aztec-wallet send mint_to_private -ca token --args accounts:test0 accounts:test0 100 -f test0

Now, the test0 account can transfer tokens by running:

# Send 40 tokens from test0 to test1
aztec-wallet send transfer -ca token --args accounts:test1 40 -f accounts:test0

Instead of sending the above transaction, you can profile it by running the profile command with the same parameters.

aztec-wallet profile transfer -ca token --args accounts:test1 40 -f accounts:test0

This will print the following results after some time:

Gate count per circuit:
   SchnorrAccount:entrypoint                          Gates: 21,724     Acc: 21,724
   private_kernel_init                                Gates: 45,351     Acc: 67,075
   Token:transfer                                     Gates: 31,559     Acc: 98,634
   private_kernel_inner                               Gates: 78,452     Acc: 177,086
   private_kernel_reset                               Gates: 91,444     Acc: 268,530
   private_kernel_tail                                Gates: 31,201     Acc: 299,731

Total gates: 299,731

Here you can see the gate count of each private function call in the transaction along with the kernel circuits needed in between, and the total gate count.

This will help you understand which parts of your transaction are bottlenecks and optimize the contract logic accordingly.

Flamegraph​

While the aztec-wallet provides a way to profile the gate count of each private function in a transaction, flamegraph tool lets you visualize the gate count of each operation within a private function.

You can run the flamegraph tool by running the following command:

aztec flamegraph <artifact_path> <function_name>

For example, if you want to flamegraph the cast_vote function aztec-starter, you can do

aztec-nargo compile
aztec flamegraph target/easy_private_voting_contract-EasyPrivateVoting.json cast_vote

This will generate a flamegraph of the cast_vote function and save the output svg to the target directory. You can open the svg file in your browser to visualize the flamegraph.

You can also run the same command with SERVE=1 to serve the flamegraph on a local server.

SERVE=1 aztec flamegraph target/easy_private_voting_contract-EasyPrivateVoting.json cast_vote

This will serve the flamegraph on http://localhost:8000.