Skip to content

Repository navigation

Accelerate BN254 CRT reconstruction and export #178

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
odraccir1 merged 2 commits into from
Oct 20, 2026
Merged

Accelerate BN254 CRT reconstruction and export #178

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b25f960
fuse normalization and implmemnt certified direct Montgomery reconstr…
odraccir1 Oct 20, 2026
e058e69
fix build
odraccir1 Oct 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 17 additions & 15 deletions libs/algebra/include/nil/crypto3/algebra/fields/detail/element/fp12_fast/limb_ops_x86.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -254,20 +254,20 @@
#define REDC_8_LIMBS_CANCEL_LOW(I) \
/* m = data[i] * p_dash */ \
"movq " T8(I, 0) ", %%rdx\n" \
"imulq %[p_dash], %%rdx\n" \
"imulq " PTR(constants, 4) ", %%rdx\n" \
"xor %[zero], %[zero]\n" \
/* multiply m by each modulus limb and add into the rotating window */ \
"mulxq %[p0], %[low], %[high]\n" \
"mulxq " PTR(constants, 0) ", %[low], %[high]\n" \
/* use dual carry chains */ \
"adcx %[low], " T8(I, 0) "\n" \
"adox %[high], " T8(I, 1) "\n" \
"mulxq %[p1], %[low], %[high]\n" \
"mulxq " PTR(constants, 1) ", %[low], %[high]\n" \
"adcx %[low], " T8(I, 1) "\n" \
"adox %[high], " T8(I, 2) "\n" \
"mulxq %[p2], %[low], %[high]\n" \
"mulxq " PTR(constants, 2) ", %[low], %[high]\n" \
"adcx %[low], " T8(I, 2) "\n" \
"adox %[high], " T8(I, 3) "\n" \
"mulxq %[p3], %[low], %[high]\n" \
"mulxq " PTR(constants, 3) ", %[low], %[high]\n" \
"adcx %[low], " T8(I, 3) "\n" \
/* merge OF, prior pending carry, and CF into the next live limb */ \
"adox %[zero], %[high]\n" \
Expand Down Expand Up @@ -697,7 +697,13 @@ namespace nil::crypto3::algebra::fields::detail::fp12_fast {
template<Fp12FastParams Params>
requires(Params::storage_limb_count == 8)
inline void montgomery_reduce_x86(limb *result, const limb *data) {
GET_MODULUS_4_LIMBS(Params::base_field_type);
constexpr auto mod_obj = Params::base_field_type::modulus_params.get_mod_obj();
// One explicit address operand keeps these constants from competing with scratch
// registers, including default-visibility Mach-O and unoptimized builds.
static constexpr std::array<limb, 5> constants = {
limb(mod_obj.get_mod().limbs()[0]), limb(mod_obj.get_mod().limbs()[1]),
limb(mod_obj.get_mod().limbs()[2]), limb(mod_obj.get_mod().limbs()[3]),
limb(mod_obj.get_p_dash())};
limb t0, t1, t2, t3, t4;
limb low, high, pending, zero;
asm volatile(
Expand All @@ -723,10 +729,10 @@ namespace nil::crypto3::algebra::fields::detail::fp12_fast {
"movq " T8(3, 4) ", %%rdx\n"

// Try q = t - p.
"subq %[p0], %[low]\n"
"sbbq %[p1], %[high]\n"
"sbbq %[p2], %[zero]\n"
"sbbq %[p3], %%rdx\n"
"subq " PTR(constants, 0) ", %[low]\n"
"sbbq " PTR(constants, 1) ", %[high]\n"
"sbbq " PTR(constants, 2) ", %[zero]\n"
"sbbq " PTR(constants, 3) ", %%rdx\n"

// If t - p did not borrow, keep q as the canonical result.
"cmovnc %[low], " T8(0, 4) "\n"
Expand All @@ -750,11 +756,7 @@ namespace nil::crypto3::algebra::fields::detail::fp12_fast {
[zero]"=&r"(zero)
: [data]"r"(data),
[result]"r"(result),
[p0]"m"(p0),
[p1]"m"(p1),
[p2]"m"(p2),
[p3]"m"(p3),
[p_dash]"m"(p_dash)
[constants]"r"(constants.data())
: "rdx", "cc", "memory"
);
}
Expand Down
Loading
Loading