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

Optimised shift left and shift right experimental implementations. #248

Open
wants to merge 20 commits into
base: master
Choose a base branch
from
Open

Optimised shift left and shift right experimental implementations. #248

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
d202dda
Add shl_c
chfast Feb 6, 2022
5fe1d3c
optimize sw
chfast Feb 6, 2022
6548dc5
[WIP] experimental shl tests added. exp shl implemented for all types.
rodiazet Feb 17, 2022
75a6ed3
[WIP] experimental shr test added. exp shr implemented for all types.
rodiazet Feb 17, 2022
8de0d87
Merge into shift.hpp
chfast Feb 17, 2022
1596e74
De-template
chfast Feb 17, 2022
494b42a
Extend benchmarks
chfast Feb 17, 2022
1042585
Add shl_e
chfast Feb 17, 2022
7c4f1fe
noinline
chfast Feb 17, 2022
e130635
Add shl_w()
chfast Feb 17, 2022
8ceecf1
fake
chfast Feb 17, 2022
38a80c2
New impl
chfast Feb 17, 2022
66c5702
Should be better
chfast Feb 17, 2022
517095f
Patch shl_c
chfast Feb 17, 2022
9c7a472
new shl_w
chfast Feb 17, 2022
7b9cacd
Expand more
chfast Feb 17, 2022
95c0cae
pass shift by value
chfast Feb 18, 2022
757e1d4
shl_bits
chfast Feb 18, 2022
7ce00d6
Just funkin multiply
chfast Feb 18, 2022
8550d74
Add AVX implementation
chfast Feb 21, 2022
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
57 changes: 42 additions & 15 deletions test/benchmarks/benchmarks.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
// Licensed under the Apache License, Version 2.0.

#include "../experimental/addmod.hpp"
#include "../experimental/shift.hpp"
#include <benchmark/benchmark.h>
#include <intx/intx.hpp>
#include <test/utils/gmp.hpp>
Expand Down Expand Up @@ -198,8 +199,7 @@ BENCHMARK_TEMPLATE(binop, uint512, uint512, public_mul);
BENCHMARK_TEMPLATE(binop, uint512, uint512, gmp::mul);

template <unsigned N>
[[gnu::noinline]] static intx::uint<N> shl_public(
const intx::uint<N>& x, const uint64_t& y) noexcept
[[gnu::noinline]] static intx::uint<N> shl_public(const intx::uint<N>& x, uint64_t y) noexcept
{
return x << y;
}
Expand All @@ -211,35 +211,52 @@ template <unsigned N>
return x << y;
}

[[gnu::noinline]] static intx::uint256 shl_halves(
const intx::uint256& x, const uint64_t& shift) noexcept
template <unsigned N>
[[gnu::noinline]] static intx::uint<N> shl_generic(const intx::uint<N>& x, uint64_t y) noexcept
{
return intx::operator<<<N>(x, y);
}

template <unsigned N>
[[gnu::noinline]] static intx::uint<N> shl_generic(
const intx::uint<N>& x, const intx::uint<N>& y) noexcept
{
return intx::operator<<<N>(x, y);
}

[[gnu::noinline]] static intx::uint256 shl_halves(const intx::uint256& x, uint64_t shift) noexcept
{
constexpr auto num_bits = 256;
constexpr auto half_bits = num_bits / 2;

const auto xlo = uint128{x[0], x[1]};
const auto xlo_ = uint128{x[0], x[1]};
const auto xlo = static_cast<builtin_uint128>(xlo_);

if (shift < half_bits)
{
const auto lo = xlo << shift;
const auto lo_ = xlo << shift;
const auto lo = uint128{lo_};

const auto xhi = uint128{x[2], x[3]};
const auto xhi_ = uint128{x[2], x[3]};
const auto xhi = static_cast<builtin_uint128>(xhi_);

// Find the part moved from lo to hi.
// The shift right here can be invalid:
// for shift == 0 => lshift == half_bits.
// Split it into 2 valid shifts by (rshift - 1) and 1.
const auto rshift = half_bits - shift;
const auto lo_overflow = (xlo >> (rshift - 1)) >> 1;
const auto hi = (xhi << shift) | lo_overflow;
const auto hi_ = (xhi << shift) | lo_overflow;
const auto hi = uint128{hi_};
return {lo[0], lo[1], hi[0], hi[1]};
}

// This check is only needed if we want "defined" behavior for shifts
// larger than size of the Int.
if (shift < num_bits)
{
const auto hi = xlo << (shift - half_bits);
const auto hi_ = xlo << (shift - half_bits);
const auto hi = uint128{hi_};
return {0, 0, hi[0], hi[1]};
}

Expand Down Expand Up @@ -288,7 +305,7 @@ template <unsigned N>
}

#if INTX_HAS_EXTINT
[[gnu::noinline]] static intx::uint256 shl_llvm(const intx::uint256& x, const uint64_t& y) noexcept
[[gnu::noinline]] static intx::uint256 shl_llvm(const intx::uint256& x, uint64_t y) noexcept
{
unsigned _ExtInt(256) a;
std::memcpy(&a, &x, sizeof(a));
Expand All @@ -300,7 +317,7 @@ template <unsigned N>
#endif


template <typename ArgT, typename ShiftT, ArgT ShiftFn(const ArgT&, const ShiftT&)>
template <typename ArgT, typename ShiftT, ArgT ShiftFn(const ArgT&, ShiftT)>
static void shift(benchmark::State& state)
{
const auto& shift_samples_id = [&state]() noexcept {
Expand All @@ -324,7 +341,7 @@ static void shift(benchmark::State& state)

const auto& xs = test::get_samples<ArgT>(sizeof(ArgT) == sizeof(uint256) ? x_256 : x_512);
const auto& raw_shifts = test::get_samples<uint64_t>(shift_samples_id);
std::array<ShiftT, test::num_samples> shifts;
std::array<std::decay_t<ShiftT>, test::num_samples> shifts;
std::copy(std::cbegin(raw_shifts), std::cend(raw_shifts), std::begin(shifts));

while (state.KeepRunningBatch(xs.size()))
Expand All @@ -336,14 +353,24 @@ static void shift(benchmark::State& state)
}
}
}
BENCHMARK_TEMPLATE(shift, uint256, uint256, shl_public)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint256, shl_halves)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, const uint256&, shl_public)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, const uint256&, shl_generic)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, const uint256&, shl_halves)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, shl_public)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, shl_generic)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, shl_halves)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_c)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_e)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_w)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_avx)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_bits_1)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_bits_2)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_bits_3)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, experimental::shl_bits_4)->DenseRange(-1, 3);
#if INTX_HAS_EXTINT
BENCHMARK_TEMPLATE(shift, uint256, uint64_t, shl_llvm)->DenseRange(-1, 3);
#endif
BENCHMARK_TEMPLATE(shift, uint512, uint512, shl_public)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint512, const uint512&, shl_public)->DenseRange(-1, 3);
BENCHMARK_TEMPLATE(shift, uint512, uint64_t, shl_public)->DenseRange(-1, 3);

[[gnu::noinline]] static bool lt_public(const uint256& x, const uint256& y) noexcept
Expand Down
4 changes: 2 additions & 2 deletions test/experimental/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,8 @@
# intx: extended precision integer library.
# Copyright 2019-2020 Pawel Bylica.
# Copyright 2019 Pawel Bylica.
# Licensed under the Apache License, Version 2.0.

add_library(experimental STATIC add.cpp add.hpp)
add_library(experimental STATIC add.cpp add.hpp shift.hpp)
target_compile_definitions(experimental PRIVATE INTX_EXPERIMENTAL)
target_include_directories(experimental PUBLIC ${PROJECT_SOURCE_DIR}/test)
target_link_libraries(experimental PUBLIC intx::intx)
Expand Down
Loading