Debug BLS
#773
Debug BLS
#773
Conversation
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
| (defconstraint pair-of-inputs-constancy () | ||
| (if-not-zero ACC_INPUTS | ||
| (if (will-remain-constant! ACC_INPUTS) | ||
| (if-zero (- (next ACC_INPUTS) ACC_INPUTS) |
There was a problem hiding this comment.
Is this rejected by go-corset ? The original version looked ok to me.
There was a problem hiding this comment.
no, they are probably equivalent, but I believe this way is more clear
| (vanishes! (shift [OUTGOING_DATA 1] 7)) | ||
| (eq! (shift [OUTGOING_DATA 2] 7) (prc---callee-gas)) | ||
| (vanishes! (shift [OUTGOING_DATA 3] 7)) | ||
| (eq! (shift [OUTGOING_DATA 4] 7) (prc-g1msm-prc-g2msm---precompile-cost))))) |
There was a problem hiding this comment.
Could we replace all the (shift [OUTGOING_DATA X] Y) with meaningful shorthand names ?
There was a problem hiding this comment.
yes, but I would prefer in a separated PR.
| (eq! (shift OUTGOING_INST 6) EVM_INST_DIV) | ||
| (vanishes! (shift [OUTGOING_DATA 1] 6)) | ||
| (eq! (shift [OUTGOING_DATA 2] 6) (prc---callee-gas)) | ||
| (eq! (* (shift [OUTGOING_DATA 2] 6) (msm-pair-size)) (prc-g1msm-prc-g2msm---msm-cost-numerator_msm-pair-size)) |
There was a problem hiding this comment.
Bug: EVM Division Precompile Constraint Error
The compute-precompile-cost-integer-division constraint misconfigures the EVM_INST_DIV operation. OUTGOING_DATA 2 is incorrectly constrained as a value related to the quotient, but it should be the dividend (numerator). The actual division result (the precompile cost) should come from OUTGOING_RES_LO.
| (vanishes! (shift [OUTGOING_DATA 3] 6)) | ||
| (eq! (* (shift [OUTGOING_DATA 4] 6) (msm-pair-size) PRC_BLS_MULTIPLICATION_MULTIPLIER) | ||
| (prc-g1msm-prc-g2msm---precompile-cost_msm-pair-size_PRC_BLS_MULTIPLICATION_MULTIPLIER))))) | ||
| (eq! (shift [OUTGOING_DATA 4] 6) PRC_BLS_MULTIPLICATION_MULTIPLIER)))) |
There was a problem hiding this comment.
Bug: Cost Calculation Error in Precompile Constraint
The prc-g1msm-prc-g2msm---compute-precompile-cost-integer-division constraint performs a division but doesn't store the result into (shift OUTGOING_RES_LO 6). As a result, prc-g1msm-prc-g2msm---precompile-cost will not reflect the computed value, leading to an incorrect precompile cost.
Note
Refactors BLS MSM precompile pricing/gas checks, raises MSM CT_MAX to 7, widens
OUTGOING_INST/PRC_NAMEto i16, marks OOB add/mod lookups as unchecked, and adjusts constancy condition logic.oob/*/precompiles/common/bls/bls_msm.lisp):reference-table-discount; computesdiscountconditionally onnum-inputs <= 128.EVM_INST_DIV,MOD_FLAG=1,WCP_FLAG=0); derivesprecompile-costat shift 6 and movesinsufficient-gasto shift 7.precompile-costin a separate LT step; simplifiesreturn-gastocallee-gas - precompile-cost(removes pair-size/multiplier scaling).OUTGOING_INSTto:i16inoob/*/columns.lisp.PRC_NAMEto:i16inreftables/prague/bls_reftable.lisp.CT_MAX_BLS_G1_MSM/CT_MAX_BLS_G2_MSMfrom 6 to 7 inoob/*/constants.lisp.oob-into-addandoob-into-modclookups as:uncheckedinoob/*/lookups/.blsdata/*/generalities/constancy_conditions.lisp):will-remain-constant! ACC_INPUTScheck withif-zero (- (next ACC_INPUTS) ACC_INPUTS)before enforcingNONTRIVIAL_POP_BITconstancy.Written by Cursor Bugbot for commit a9247c7. This will update automatically on new commits. Configure here.