baram2584/FESADev
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Integrate Refactor solver sources and dual MSVC CMake builds

Browse Source
This commit is contained in:
KOKO\Mimi
2026-10-18 02:10:51 +09:00
parent 7ee32e5837
commit 12d51480de
218 changed files with 17842 additions and 9689 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/fesa/elements/hex8_pair_simd_avx2.cpp
+6 -6
View File
@@ -144,11 +144,11 @@ bool TwoSum(__m256d left, __m256d right, VectorPair* output) {
return true;
}
/// @brief Routes exceptional packets to baseline checked Add in lane order.
VectorPair ScalarAdd(VectorPair left, VectorPair right, bool pair_shortcuts) {
VectorPair ScalarAdd(VectorPair left, VectorPair right) {
const auto a = Store(left);
const auto b = Store(right);
PairPacket result;
ScalarPacketAdd(a, b, pair_shortcuts, &result);
ScalarPacketAdd(a, b, &result);
return Load(result);
}
/// @brief Routes exceptional packets to baseline checked MultiplyPair.
@@ -160,7 +160,7 @@ VectorPair ScalarMultiply(const PairPacket& left, const PairPacket& right) {
/// @brief Retains normalized/zero branches and the checked per-lane Add graph.
VectorPair Add(VectorPair left, VectorPair right, bool pair_shortcuts) {
if (!NormalizedInputs(left, right))
return ScalarAdd(left, right, pair_shortcuts);
return ScalarAdd(left, right);
const int left_zero = _mm256_movemask_pd(
_mm256_and_pd(Zero(left.high), Zero(left.low)));
const int right_zero = _mm256_movemask_pd(
@@ -171,7 +171,7 @@ VectorPair Add(VectorPair left, VectorPair right, bool pair_shortcuts) {
if (right_zero == 15)
return {left.high, Select(Zero(left.low), _mm256_setzero_pd(), left.low)};
if (left_zero != 0 || right_zero != 0)
return ScalarAdd(left, right, true);
return ScalarAdd(left, right);
}
VectorPair head;
VectorPair tail;
@@ -181,7 +181,7 @@ VectorPair Add(VectorPair left, VectorPair right, bool pair_shortcuts) {
!TwoSum(left.low, right.low, &tail) ||
!TwoSum(head.low, tail.high, &middle) ||
!TwoSum(head.high, middle.high, &joined))
return ScalarAdd(left, right, pair_shortcuts);
return ScalarAdd(left, right);
const auto remainder = _mm256_add_pd(
_mm256_add_pd(joined.low, middle.low), tail.low);
const auto valid = _mm256_and_pd(Finite(remainder), _mm256_and_pd(
@@ -193,7 +193,7 @@ VectorPair Add(VectorPair left, VectorPair right, bool pair_shortcuts) {
VectorPair result;
if (!TwoSum(Select(valid, joined.high, _mm256_setzero_pd()),
Select(valid, remainder, _mm256_setzero_pd()), &result))
return ScalarAdd(left, right, pair_shortcuts);
return ScalarAdd(left, right);
return InvalidOutside(result, valid);
}
/// @brief Checks finite/zero/tail premises before products; neutralizes rejected