precompiles: Implement EIP-2537's bls12_g1mul
#994
Conversation
8223283 to
16f74d1
Compare
60dd68d to
fde4252
Compare
16f74d1 to
87008aa
Compare
fde4252 to
8135a50
Compare
87008aa to
118d6af
Compare
bls12_g1mul
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #994 +/- ##
==========================================
+ Coverage 93.84% 93.88% +0.03%
==========================================
Files 146 146
Lines 15460 15485 +25
==========================================
+ Hits 14509 14538 +29
+ Misses 951 947 -4
Flags with carried forward coverage won't be shown. Click here to find out more.
|
7bbddcf to
2ca9a94
Compare
118d6af to
525eb35
Compare
525eb35 to
acfc76d
Compare
lib/evmone_precompiles/bls.cpp
Outdated
| { | ||
| namespace | ||
| { | ||
| using namespace intx::literals; |
| const uint8_t _y[64], const uint8_t _c[32]) noexcept | ||
| { | ||
| blst_scalar scalar; | ||
| blst_scalar_from_bendian(&scalar, _c); |
There was a problem hiding this comment.
I wonder what it actually does. Because the later blst_p1_mult just uses the bytes. Are they the same bytes as the input?
There was a problem hiding this comment.
Answer: it seem blst expects bytes but in little-endian order.
lib/evmone_precompiles/bls.hpp
Outdated
| [[nodiscard]] bool g1_add(uint8_t _rx[64], uint8_t _ry[64], const uint8_t _x0[64], | ||
| const uint8_t _y0[64], const uint8_t _x1[64], const uint8_t _y1[64]) noexcept; | ||
|
|
||
| /// Scalar multiplication in BLS12-381 curve group. |
There was a problem hiding this comment.
Mention this is in the G1 subgroup and that the subgroup check is performed.
| blst_p1 p; | ||
| blst_p1_from_affine(&p, &*p_affine); | ||
|
|
||
| if (!blst_p1_in_g1(&p)) |
There was a problem hiding this comment.
The subgroup check is also a multiplication. Do you think can we combine both multiplication for better performance?
There was a problem hiding this comment.
I would be surprised if they implement group check just by multiplication by group order. There are planty of algorithms which make it faster. I will check the implementation of blst_p1_in_g1
There was a problem hiding this comment.
There are two different implementations of this check in blst. Non of them allows this kind of optimization.
acfc76d to
1d0caac
Compare
2f2076a to
34f483b
Compare
6c6d499 to
2a99823
Compare
Implementation of the bls12_g2add precompile: E2 affine points' addition from BLS12-381 curve according to the EIP-2537 spec https://eips.ethereum.org/EIPS/eip-2537#abi-for-g2-addition. Depends on #994
Implementation of the
bls12_g1mulprecompile: E1 affine point's multiplication from BLS12-381 curve by a scalar according to the EIP-2537 spec https://eips.ethereum.org/EIPS/eip-2537#abi-for-g1-multiplication.Depends on #982