Plutus V3 cost plan testing

[mesudip]

Task Details

• Check if the cost of new primitives matches with what is defined in plutus, through making transactions
• Check for each new primitives introduced in v3.

Approach

• List all the primitives and create plutus contracts for them.
• Create a base contract where operations is to be added
• Compose transaction for each of the primitive and calculate the cost
• Calculate the difference and compare with the costs provided in plutus.

Acceptance Criteria

• [ ] Tests are implemented
• [ ] A document or a report with all the primitives, cost and indication whether or not the value matches.

[reeshavacharya]

Attempted Tasks

• Created two validators: one using verifySchnorrSecp256k1Signature and another using alwaysPass as a base validator.
• Calculated the difference in execution units as follows:
  • Difference in memory = schnorr memory - base memory
  • Difference in steps = schnorr steps - base steps

Cases

• Parsing function parameters from datum and redeemer in both validators.
• Parsing function parameters directly from the redeemer in both validators.
• Using a script parameter as the function parameter without decoding the datum or redeemer.
• Wrote one contract to use two verifySchnorrSecp256k1Signature functions and another to use one; then calculated the difference in execution units.
• The differences in memory and steps did not match the Plutus-defined values in any of the cases.

Knowledge Block

• Difficulty understanding how to calculate the function's cost from the Plutus cost model.
• The model defines the cost for verifySchnorrSecp256k1Signature as follows:

  "verifySchnorrSecp256k1Signature": {
  "cpu": {
      "arguments": {
          "intercept": 39121781,
          "slope": 32260
      },
      "type": "linear_in_y"
  },
  "memory": {
      "arguments": 10,
      "type": "constant_cost"
  }
  }  

• tried calculating cost in the following way:
  • scriptExecutionCostCPU = cpu.argument.intercept + y * cpu.arguments.slope
  • calculated y in both base validator and test validator but did not relate.

[mesudip]

Approach is still not clear. Will require further investigation on this one.

[mesudip]

This is unclear because smart contract will contain extra code and not just single function call. The execution unit may change depending on

• script size
• transaction size
• fields used in transaction. It is not clear how we can test the cost of single function by making a full transaction.

[spannercode]

• Lets verify the cost plan for reference scripts.
• Lets check the changes on cardano-node-tests existing tests that cover the cost plan (including the lastest reduction in costs for conway era for Plutus v2) whether or not it exists.
• Lets compare the costs for same contract on v2 and v3 and compare the cost params and test v3 costs are lower than v2.
• For the new primitives cost plan, I'll reach out to Martin and Ziyang again.

[reeshavacharya]

Simple Market Efficiency Analysis

V3 Script Performance Comparison

Type	Script Bytes	Cost to Create Ref Script	Size of Tx to Create Ref Script	Cost to Withdraw Token Using Ref Script	Cost to Buy Token Using Ref Script
Normal	8118	530,657	8429	402,933	456,318
Lazy	7125	486,965	7436	376,888	417,008

---

Calculation Details

Note: Total cost due to reference script is calculated as `tx_bytes 15` since none of our reference scripts exceed 25,000 bytes.*

Withdraw Using Reference Script

• Script Size Difference:
  8118 - 7125 = 993 bytes

• Observed Cost Difference to Withdraw:
  402,933 - 376,888 = 26,045

• Estimated Cost Difference to Withdraw:
  993 * 15 = 14,895

• Difference Between Estimated and Observed Cost:
  26,045 - 14,895 = 11,150

Buy Using Reference Script

• Script Size Difference:
  8118 - 7125 = 993 bytes

• Observed Cost Difference to Buy:
  456,318 - 417,008 = 39,310

• Estimated Cost Difference to Buy:
  993 * 15 = 14,895

• Difference Between Estimated and Observed Cost:
  39,310 - 14,895 = 24,415

[reeshavacharya]

V3 Script Performance

Type	Script Bytes	Cost to Create Ref Script	Size of Tx to Create Ref Script	Cost to Withdraw Token Using Ref Script	Cost to Buy Token Using Ref Script
Normal	8118	530,657	8429	402,933	456,318
Super Lazy	2573	285,093	2848	258,164	297,826

V2 Script Performance

Type	Script Bytes	Cost to Create Ref Script	Size of Tx to Create Ref Script	Cost to Withdraw Token Using Ref Script	Cost to Buy Token Using Ref Script
Normal	4594	374,017	4869	378,849	438,889
Super Lazy	2848	297,193	3123	298,732	334,188

Efficiency of Fees in withdraw Tx

Normal V2 vs Normal V3

• (402933 - 378849) / 402933 = 0.0597 ~ 6%
• V2 is efficient by about 6%

Super Lazy V2 vs Super Lazy V3

• (298732 - 258164) / 298732 = 0.1358 ~ 13.6%
• V3 is effficient by about 13.6%

Efficiency of Fees in buy Tx

Normal V2 vs Normal V3

• (456318 - 438889) / 456318 = 0.03819 ~ 3.8%
• V2 is efficient by about 3.8%

Super Lazy V2 vs Super Lazy V3

• (334188 - 297826) / 334188 = 0.1088 ~ 10.9%
• V3 is effficient by about 10.9%