Skip to content

Optimize uint256_t operations using WASM SIMD #1224

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Optimize uint256_t operations using WASM SIMD #1224

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
30 changes: 28 additions & 2 deletions cpp/src/barretenberg/numeric/uint256/uint256_impl.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,11 @@
#include "../bitop/get_msb.hpp"
#include "./uint256.hpp"
#include "barretenberg/common/assert.hpp"

#if defined(__wasm__)
#include <wasm_simd128.h>
#endif

namespace bb::numeric {

constexpr std::pair<uint64_t, uint64_t> uint256_t::mul_wide(const uint64_t a, const uint64_t b)
Expand Down Expand Up @@ -73,11 +78,13 @@ constexpr uint64_t uint256_t::mac_discard_hi(const uint64_t a,
{
return (b * c + a + carry_in);
}

#if defined(__wasm__) || !defined(__SIZEOF_INT128__)

/**
* @brief Multiply one limb by 9 limbs and add to resulting limbs
*
* @brief Optimized multiply-add operation using SIMD instructions when available
* This implementation provides better performance on WebAssembly platforms by utilizing SIMD
* instructions for parallel multiplication and addition operations.
*/
constexpr void uint256_t::wasm_madd(const uint64_t& left_limb,
const uint64_t* right_limbs,
Expand All @@ -91,6 +98,23 @@ constexpr void uint256_t::wasm_madd(const uint64_t& left_limb,
uint64_t& result_7,
uint64_t& result_8)
{
#if defined(__wasm__) && defined(__wasm_simd128__)
// Load 2 64-bit integers into a 128-bit SIMD vector
v128_t left = wasm_i64x2_splat(left_limb);

// Process 2 limbs at a time using SIMD
for (int i = 0; i < 8; i += 2) {
v128_t right = wasm_v128_load(right_limbs + i);
v128_t prod = wasm_i64x2_mul(left, right);
v128_t curr = wasm_v128_load(&result_0 + i);
v128_t sum = wasm_i64x2_add(curr, prod);
wasm_v128_store(&result_0 + i, sum);
}

// Handle the last limb separately
result_8 += left_limb * right_limbs[8];
#else
// Fallback implementation for non-SIMD platforms
result_0 += left_limb * right_limbs[0];
result_1 += left_limb * right_limbs[1];
result_2 += left_limb * right_limbs[2];
Expand All @@ -100,6 +124,7 @@ constexpr void uint256_t::wasm_madd(const uint64_t& left_limb,
result_6 += left_limb * right_limbs[6];
result_7 += left_limb * right_limbs[7];
result_8 += left_limb * right_limbs[8];
#endif
}

/**
Expand All @@ -119,6 +144,7 @@ constexpr std::array<uint64_t, WASM_NUM_LIMBS> uint256_t::wasm_convert(const uin
(data[3] >> 40) & 0x1fffffff };
}
#endif

constexpr std::pair<uint256_t, uint256_t> uint256_t::divmod(const uint256_t& b) const
{
if (*this == 0 || b == 0) {
Expand Down