Skip to content
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

Jump to bottom

Some cleanup in n-body #816

Merged
merged 2 commits into from
Jan 2, 2024
Merged

Some cleanup in n-body #816

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3c37c5a
Use better names for inner/outer loop SIMD versions
bernhardmgruber Jan 1, 2024
1cd365b
Remove dead variables
bernhardmgruber Jan 1, 2024
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
48 changes: 23 additions & 25 deletions examples/nbody/nbody.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,9 +44,6 @@ constexpr auto rngSeed = 42;
constexpr auto referenceParticleIndex = 1338;
constexpr auto maxPosDiff = FP{0.001};

constexpr auto l1CacheSize = 32 * 1024;
constexpr auto l2CacheSize = 512 * 1024;

using namespace std::string_literals;

#ifdef HAVE_XSIMD
Expand Down Expand Up @@ -857,7 +854,7 @@ namespace manualAoSoAManualAVX
}

// update (read/write) 8 particles I based on the influence of 1 particle J
void update8(ParticleBlock* particles)
void updateOuterLoopSimd(ParticleBlock* particles)
{
PARALLEL_FOR
for(std::size_t bi = 0; bi < blocks; bi++)
Expand Down Expand Up @@ -902,8 +899,8 @@ namespace manualAoSoAManualAVX
// return a[0] + a[1] + a[2] + a[3] + a[4] + a[5] + a[6] + a[7];
}

// update (read/write) 1 particles I based on the influence of 8 particles J with accumulator
void update1(ParticleBlock* particles)
// update (read/write) 1 particles I based on the influence of 8 particles J
void updateInnerLoopSimd(ParticleBlock* particles)
{
PARALLEL_FOR
for(std::size_t bi = 0; bi < blocks; bi++)
Expand Down Expand Up @@ -951,9 +948,9 @@ namespace manualAoSoAManualAVX
}
}

auto main(std::ostream& plotFile, bool useUpdate1) -> Vec3
auto main(std::ostream& plotFile, bool innerLoopSimd) -> Vec3
{
auto title = "AoSoA" + std::to_string(lanes) + " AVX2 " + (useUpdate1 ? "w1r8" : "w8r1"); // NOLINT
auto title = "AoSoA" + std::to_string(lanes) + " AVX2 " + (innerLoopSimd ? "inner" : "outer"); // NOLINT
std::cout << title << '\n';
Stopwatch watch;

Expand Down Expand Up @@ -984,10 +981,10 @@ namespace manualAoSoAManualAVX
{
if constexpr(runUpdate)
{
if(useUpdate1)
update1(particles.data());
if(innerLoopSimd)
updateInnerLoopSimd(particles.data());
else
update8(particles.data());
updateOuterLoopSimd(particles.data());
statsUpdate(watch.printAndReset("update", '\t'));
}
move(particles.data());
Expand Down Expand Up @@ -1068,7 +1065,7 @@ namespace manualAoSoASIMD
}

template<typename Simd>
void update8(ParticleBlock<Simd>* particles)
void updateOuterLoopSimd(ParticleBlock<Simd>* particles)
{
constexpr auto lanes = Simd::size;
constexpr auto blocks = problemSize / lanes;
Expand Down Expand Up @@ -1105,7 +1102,7 @@ namespace manualAoSoASIMD
}

template<typename Simd>
void update8Tiled(ParticleBlock<Simd>* particles)
void updateOuterLoopSimdTiled(ParticleBlock<Simd>* particles)
{
constexpr auto lanes = Simd::size;
constexpr auto blocks = problemSize / lanes;
Expand Down Expand Up @@ -1166,7 +1163,7 @@ namespace manualAoSoASIMD
}

template<typename Simd>
void update1(ParticleBlock<Simd>* particles)
void updateInnerLoopSimd(ParticleBlock<Simd>* particles)
{
constexpr auto lanes = Simd::size;
constexpr auto blocks = problemSize / lanes;
Expand Down Expand Up @@ -1227,9 +1224,10 @@ namespace manualAoSoASIMD
}

template<typename Simd>
auto main(std::ostream& plotFile, bool useUpdate1, bool tiled = false) -> Vec3
auto main(std::ostream& plotFile, bool vectorizeInnerLoop, bool tiled = false) -> Vec3
{
auto title = "AoSoA" + std::to_string(Simd::size) + " SIMD" + (useUpdate1 ? " w1r8" : " w8r1"); // NOLINT
auto title
= "AoSoA" + std::to_string(Simd::size) + " SIMD" + (vectorizeInnerLoop ? " inner" : " outer"); // NOLINT
if(tiled)
title += " tiled";

Expand Down Expand Up @@ -1265,14 +1263,14 @@ namespace manualAoSoASIMD
{
if constexpr(runUpdate)
{
if(useUpdate1)
update1(particles.data());
if(vectorizeInnerLoop)
updateInnerLoopSimd(particles.data());
else
{
if(tiled)
update8Tiled(particles.data());
updateOuterLoopSimdTiled(particles.data());
else
update8(particles.data());
updateOuterLoopSimd(particles.data());
}
statsUpdate(watch.printAndReset("update", '\t'));
}
Expand Down Expand Up @@ -1603,18 +1601,18 @@ set y2tics auto
finalPositions.push_back(manualAoSoA::main<decltype(lanes)::value>(plotFile, false));
});
#if defined(__AVX2__) && defined(__FMA__)
// for (auto useUpdate1 : {false, true})
// r += manualAoSoA_manualAVX::main(plotFile, useUpdate1);
// for (auto innerLoopSimd : {false, true})
// r += manualAoSoA_manualAVX::main(plotFile, innerLoopSimd);
finalPositions.push_back(manualAoSoAManualAVX::main(plotFile, false));
#endif
#ifdef HAVE_XSIMD
using Simd = xsimd::batch<FP>;
// for (auto useUpdate1 : {false, true})
// for (auto innerLoopSimd : {false, true})
// for (auto tiled : {false, true})
// {
// if (useUpdate1 && tiled)
// if (innerLoopSimd && tiled)
// continue;
// r += manualAoSoA_SIMD::main<Simd>(plotFile, useUpdate1, tiled);
// r += manualAoSoA_SIMD::main<Simd>(plotFile, innerLoopSimd, tiled);
// }
finalPositions.push_back(manualAoSoASIMD::main<Simd>(plotFile, false, false));
// mp_for_each<mp_list_c<std::size_t, 1, 2, 4, 8, 16>>(
Expand Down
Loading