Constantine Integration #3587
Constantine Integration #3587
Conversation
|
Current issue eest_blockchain_test → processBlock → processTransactions
→ processTransaction → call_evm.txCallEvm → runComputation
→ interpreter_dispatch.executeOpcodes → precompiles.execPrecompile (line 842)
→ ecRecover
→ constantine/ethereum_evm_precompiles.nim(1360) ctt_eth_evm_ecrecover
→ …/signatures/ecdsa.nim recoverPubkeyImpl…
→ …/math/elliptic/ec_scalar_mul.nim(426) scalarMulGLV_m2w2 ⟵ hot loop
→ …/elliptic/ec_shortweierstrass_jacobian.nim(618) pluseq …So the test runner was executing a transaction that hit the ECRECOVER precompile; it was “stuck” (busy) inside Constantine’s secp256k1 scalar multiplication (GLV path), i.e., the inner loop that accumulates points for the pubkey recovery. NOTE: It is stuck for hours Exact input causing issue |
|
Issues found with TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_inputsFiller.yml::ecpairing_inputs[fork_Cancun-blockchain_test_from_state_test-d1-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_inputsFiller.yml::ecpairing_inputs[fork_Cancun-blockchain_test_from_state_test-d3-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_inputsFiller.yml::ecpairing_inputs[fork_Prague-blockchain_test_from_state_test-d1-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_inputsFiller.yml::ecpairing_inputs[fork_Prague-blockchain_test_from_state_test-d3-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_two_points_with_one_g2_zeroFiller.json::ecpairing_two_points_with_one_g2_zero[fork_Cancun-blockchain_test_from_state_test-d0-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_two_points_with_one_g2_zeroFiller.json::ecpairing_two_points_with_one_g2_zero[fork_Cancun-blockchain_test_from_state_test-d0-g3-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_two_points_with_one_g2_zeroFiller.json::ecpairing_two_points_with_one_g2_zero[fork_Prague-blockchain_test_from_state_test-d0-g0-v0] RunTest error: A bug? good block rejected: gasUsed mismatch
TestName: tests/static/state_tests/stZeroKnowledge/ecpairing_two_points_with_one_g2_zeroFiller.json::ecpairing_two_points_with_one_g2_zero[fork_Prague-blockchain_test_from_state_test-d0-g3-v0] RunTest error: A bug? good block rejected: gasUsed mismatchExact implementation issue still not investigated UPDATE : Fixed 8th Sept |
|
Regarding constantine, my recommendation is that the initial branch integrates sha256 only as a start - once that is working (ie sha256 is confirmed to be fast, constantine subrepo is present and so on), separate branches can be used for each precompile - in particular, this allows measuring (in benchmarks) and rolling back precompiles on an individual basis. |
|
Linux arm and windows arm issues |
| if defined(windows): | ||
| # Constantine | ||
| switch("passC", "-fomit-frame-pointer") | ||
| switch("passL", "-fomit-frame-pointer") |
There was a problem hiding this comment.
Constantine windows issue
Goal is to first integrate constantine and check performance metrics for all precompiles
Stats:-
Note
Need to create a error handler for constantine type errors